Modeling train movements through complex rail networks

合集下载

地铁站疏散Modeling the pedestrian’s movement and simulating evacuation

Modeling the pedestrian’s movement and simulating evacuation dynamics onstairsYunchao Qu,Ziyou Gao ⇑,Yao Xiao,Xingang LiSchool of Trafﬁc and Transportation,Beijing Jiaotong University,Beijing 100044,Chinaa r t i c l e i n f o Article history:Received 23December 2013Received in revised form 15May 2014Accepted 22May 2014Available online 26June 2014Keywords:Pedestrian ﬂowStaircase movement Dynamic characteristics Social force modela b s t r a c tThis paper presents an enhanced social force model to describe the pedestrian’s movement and evacua-tion dynamics on pared with original models that described the pedestrian’s planar motion,our model introduces some mechanisms of the staircase movement,such as the inﬂuence of staircase geometry,the restriction of the step size and the optimal velocity selection.The body shape of each pedestrian is regarded as a set of three circles to precisely quantify the movement.In addition,the rotation dynamics are included into the model to describe the congestion effect.The improved model can obtain individual velocity under different staircase geometries and the ﬂow characteristics of the evacuation dynamics.Some empirical data and a series of observations captured in two subway stations in Beijing are applied to study the characteristics and further validate the model.The results show that our model performs well consistent with the observed data.At last,simulations are implemented to ﬁnd the solutions of estimating the evacuation time and evaluating the capacity of stair.Ó2014Elsevier Ltd.All rights reserved.1.IntroductionStairs are widely used in all kinds of buildings,especially in large scale public places,i.e.,subway stations,shopping malls and ofﬁce buildings.Walking on stairs is very common and important in our daily lives,and scientiﬁc design and effective utilization of stairs are urgently needed for designers and managers (Peacock et al.,2009).In emergency,such as power failure,ﬁre,earthquake or other hazards,the elevators may be out of commission,and the stairs become the primary escape routes.If there are too many peo-ple crowded on stairs,they will pack closer together or even lead to some dangerous situations (Shields and Boyce,2009).Knowing the ﬂow characteristics and predicting the egress time are the key points to grasp the evacuation dynamics and make emergency response plans on stairs (Graat et al.,1999;Oven and Cakici,2009).The characteristics of pedestrian staircase movement are deter-mined by organizational,constructional and behavioral factors:the organizational factors,i.e.,preparation for emergencies;the constructional factors,i.e.,the staircase geometry including riser height,tread depth and step width (Graat et al.,1999;Fujiyama and Tyler,2010);the behavioral factors,i.e.,responses and move-ment characteristics of pedestrians (Yang et al.,2012;Ma et al.,2012).The study of the staircase movement is an interdisciplinaryﬁeld with different focuses,such as biomechanics,physics,physiol-ogy,phycology,computer science,safety science (i.e.,Hankin and Wright,1958;Fruin,1971;Predtechenskii and Milinskii,1978;Templer,1992;Batty,1997;Helbing et al.,2000;Hase and Yamazaki,2002;Nelson et al.,2002;Hoskin,2004;Pauls,2005;Trew,2005;Casburn et al.,2007;Hostikka et al.,2007;Galea et al.,2008;Kretz et al.,2008;Seer,2008;Xu and Song,2009;Fujiyama and Tyler,2010;Galea et al.,2010;Hoskins,2011;Halsey et al.,2012;Yang et al.,2012;Peacock et al.,2012;Burghardt et al.,2013).To make quantitative analyses and detailed descriptions of staircase movement,many researches have carried out a lot of sur-veys,experiments and evacuation drills on stairs,and have collected large amounts of experimental and observational data of staircase movement (please see Table 1for details).In these studies,the ﬂow characteristics of staircase movement are described in individual level and collective level.Pedestrian ﬂow in low densities reﬂects the characteristics in the individual level,and walking speed is inﬂu-enced by physiological feature and body function,such as gender,age,height,weight,heart rate,rate of oxygen consumption and rate of energy expenditure (Irvine et al.,1990;Teh and Aziz,2002;Halsey et al.,2012).It is also inﬂuenced by stairway geometries and movement direction.Pedestrian ﬂow in high densities reﬂects the characteristics in the collective level.The collective behaviors include pedestrians’self-organized behaviors and optimal route choice behaviors (i.e.,Helbing et al.,2000,2005;Moussaid et al.,2011).The researches of pedestrian ﬂow in the collective level focus on three aspects,(1)evaluating evacuation time,(2)reproducing/10.1016/j.ssci.2014.05.0160925-7535/Ó2014Elsevier Ltd.All rights reserved.⇑Corresponding author.Tel.:+861051688193.E-mail address:zygao@ (Z.Gao).fundamental diagram,and(3)describingﬂow characteristics,i.e., inﬂow,outﬂow,capacity.Staircase movement is a complicated three-dimensional move-ment,and modeling the movement is a quite challenging work. Nowadays,researches have integrated behavioral and construc-tional factors,and have established many models to analyze the ﬂow characteristics and simulate evacuation processes in both sin-gle-story and multi-story buildings(Table1).In our work,we mainly focus on the case of single-story staircases.These models are classiﬁed into two categories:macroscopic model and micro-scopic model(Zheng et al.,2009).The macroscopic models regard the crowd as a single entity,and focus onﬁtting the expression of fundamental diagram.Linear,piecewise linear and non-linear functions(i.e.,Fruin,1971;Warren,1984;Tanaboriboon et al., 1986;Weidmann,1993;Lam and Cheung,2000;Proulx,2002; Peacock et al.,2012;Hoskins and Milke,2012)have been applied to describe relationship between velocity and density under differ-ent stair geometries.Compared with macroscopic models,the microscopic models are able to precisely describe the individual behavior,qualitatively explain the evacuation dynamics and reproduce some self-orga-nized phenomena(Helbing et al.,2000).These microscopic models are spatial-discrete models(cellular automation model,i.e., Kirchner et al.,2004;Huang and Guo,2008;Schadschneider and Seyfried,2009)and spatial-continuous models(social force model, i.e.,Helbing et al.,2000).These models have been applied to reveal two-dimensional planar movement,but few of them have described the three-dimensional staircase movement(Song et al., 2006;Pelechano and Malkawi,2008;Xu and Song,2009;Ma et al.,2012).In addition,the spatial-discrete models are some restricted to describe the staircase movement,such as grid size,fatigue factor,route selection,and uneven use of stairs (Pelechano and Malkawi,2008).Although the spatial-continuous models are advantageous to solve most of the aforementioned problems,these models are quite rare.Social force model(Helbing and Molnar,1995)is a well-known spatial-continuous model in theﬁeld of pedestrianﬂow.The model can reproduce several self-organized phenomena,such as lane forming,arching queue,shock waves and clogging effects (Helbing et al.,2005,2007).Moussaid et al.(2011)have proposed a heuristics-based model to replace the social force with a heuris-tics intelligent optimum function.Based on the heuristic social force model,this paper introduced some special rules and estab-lished an enhanced model to describe the mechanisms of pedes-trian movement and evacuation dynamics on stairs.Firstly,the body shape of each pedestrian is regarded as a set of three circles (Thompson and Marchant,1995).Compared with traditional sin-gle-circle shape(i.e.,Helbing et al.,2000),the three-circle shape precisely represents the projection of human body and describes the rotation movement when two pedestrians collide with others. Secondly,pedestrians usually walk more carefully on stairs than on planar,so two‘safety rules’are proposed to describe staircase movement behavior.Theﬁrst rule is that a pedestrian wants to walk upstairs/downstairs with integral steps at a time,and the step-size is restricted by the staircase geometry,such as tread depth,riser height and step width.The second rule is that a pedes-trian tends to walk along the sides,i.e.,holding handrails,propping up against walls.Thirdly,the relaxation time is extended to a var-iable in our model.The relaxation time is deﬁned that a pedestrian tends to correspondingly adapt his/her actual velocity to desired velocity with a certain characteristic time s(i.e.,Helbing et al., 2000;Moussaid et al.,2011).The relaxation time is mostlyTable1The state-of-the-art of staircase movement.Author(s)Year Method Model EvacuationprocessWalking speed NoteDynamics FD Geometry Direction Hankin and Wright1958Data analysis–s d s d–Fruin1971Data analysis–s d s s Planning methodPredtechenskii and Milinskii1978Data analysis Planning model s d s s Planning methodTanaboriboon et al.1986Macro Linear function s d s s Fundamental diagram Weidmann1993Macro Non-linear function s d d d–Frantzich1996Data analysis–s d d d–Graat et al.1999Data analysis–s d s s Capacity estimation Lam and Cheung2000Macro BPR function s d d d Fundamental diagram,capacityestimation Proulx et al.2002Data analysis Non-linear function d d s s SFPENelson and Mowrer2002Data analysis Non-linear function d d s s SFPEHoskin2004Softwaresimulation Coordinate-basedmodeld d d d Simulex32Pauls2005Data analysis–d s s s–Hostikka et al.2007Data analysis–d d s s–Kretz et al.2008Data analysis–s d d d Pedestrian movement on long stairs Seer et al.2008Data analysis–d d s s Flow characteristicsPelechano and Malkawi2008SoftwaresimulationGrid based model d s s s Literature review(STEPS,EXODUS)Galea et al.2008Software Evacuation model d s s s Merging behavior at interactions Xu and Song2009Micro Multi-grid model d s s s Flow characteristics,such as in and outﬂow Fujiyama and Tyler2010Macro Linear function s s d d Individual walking speed Galea et al.2010Data analysis–d d s s Evacuation softwareHoskins2011Macro Linear function d d d d Fundamental diagramYang et al.2012Data analysis–d d d s Evacuation drillLei et al.2012Softwaresimulation–d s s s Software(FDS,EVAC) Hoskins and Milke2012Data analysis–s d s s NISTPeacock2012Data analysis–d d d s NIST,different measurement methods Ma et al.2012Data analysis CA d s s s SimulationBurghardt et al.2013Data analysis–s d s s Fundamental diagramd Represents the factor is included and s represents the factor is not included.190Y.Qu et al./Safety Science70(2014)189–201assumed to be a constant in previous models,but it is not inade-quate to describe the staircase movement.As mentioned before, the individual walking speed on stairs is inﬂuenced by many fac-tors,such as staircase size,movement direction,and physical char-acteristics.For example,pedestrians spend more energy on walking upstairs than downstairs,spend more time on walking steep stairs than gentle stairs.Pedestrians with different age, weight or gender may require different relaxation times when walking on different stairs.Considering the inﬂuence factors,the relaxation time is formulated as a linear function of individual weight,moving height,and the slope of stairs.The linear function is similar to the model(Fujiyama and Tyler,2010),and some parameters are introduced to distinguish the upstairs and down-stairs movement.In this paper,some empirical data in literatures are collected and new observations from subway stations are conducted.In Sec-tion2,the characteristics of pedestrianﬂow on the stairs are dis-cussed.Based on the data and analysis,the constructional and behavioral factors are introduced to the social force model to pre-cisely describe the individual staircase moment in Section3.To validate the model,a series of simulations are implemented,and the simulation results are compared with the observational and empirical data in Section4.Simulations are implemented to ana-lyze theﬂow characteristics and the evacuation process in subway stations in Section5.Finally,the conclusions and the further work are given in Section6.2.Data collection andﬂow characteristics on stairs2.1.Empirical data of pedestrian speedOccupant speed is a very important element of pedestrianﬂow, and pedestrian speed on stairs is mainly affected by the slope of stairway,depth of tread,height of riser,and presence and location of handrails(Gwynne et al.,2009).Graat et al.(1999)had found that speed and capacity on stairs were higher with a normal (30°)slope than a steep(38°)slope.Kretz et al.(2008)had found that some pedestrian accelerate when walking upward a short stairway,and the mean upward walking speed on the short stair-way was found to be roughly twice as large as the one on the long stairway.Fujiyama and Tyler(2010)had proposed a model to pre-dict the walking speed based on the weight,leg power and the gra-dient of the stairs.The evacuation process of a large number of people is another major concern for researchers and designers. During the evacuation,the evacuation process something likes a queuing system that contains the processes of congestion forming, propagation and dissipation(Ma et al.,2012).Besides,the stair width and capacity will affect the route/exit choice behaviors and the evacuation efﬁciency(Lei et al.,2012).Researches have obtained many observational and experimen-tal data of staircase movement.However,this study here is not intended to be an exhaustive review of all researches.The refer-ences which mentioned both staircase geometry and individual speed are taken into consideration.Ten instances of staircase are included in our paper,and more detailed data can be found in the literatures(i.e.,Weidmann,1993;Frantzich,1996;Fujiyama and Tyler,2010;Hoskins,2011;Peacock et al.,2012).From Table2, it is found that walking speeds listed in the studies are different. This may be caused by natural variation of individual capability, staircase geometry,density of crowd and other factors.Besides, different measurement methods of calculating travel distances and areas on stairs may lead to different results(Hoskins and Milke,2012).Fujiyama and Tyler(2010)have made some experi-ments and found that average upstairs and downstairs speeds were0.58m/s and0.67m/s,respectively.Peacock et al.(2012) have mentioned that average downstairs speeds in their study of 0.48±0.16m/s were observed to be quite similar to the range of literature values.Kretz et al.(2008)have pointed that the density also affected the individual speed.Ma et al.(2012)have made a series of evacuation drills to obtain the average downstairs speed of0.547m/s.Even though different researchers have come up with different values for movement on stairs,most give a maximum for density of4.5–5.5pedestrians/m2,a maximum for speed of0.7–1.2m/s,and maximum for capacity of0.8–1.5pedestrians/(m s).2.2.Observations in subway stationsSomeﬁre drill evacuations of ofﬁce buildings have been imple-mented by National Institute of Standards and Technology(NIST), and the collected staircase movement data have included a range of stair geometries and occupant densities(Peacock et al.,2012; Hoskins and Milke,2012).However,in those studies,the local speed and the density were inaccurately estimated according to the collected data.It was because the cameras did not fully record the whole of staircase movement.In the drills,cameras were set every twoﬂoors to record an overhead view of occupant move-ment.The view of each camera only covered the main landing area plus tread depth area for about4–6steps of one story.Between every two cameras,there was a mid-landing,where pedestrian movement was a planar movement,but not a staircase movement. It was impossible to dissociate the planar movement on mid-land-ing from the video,so the calculations of the travel distance and travel time of staircase movement were inaccurate.To precisely investigate the pedestrian movement characteristics on stairs,we improved the method of video recording,and conducted observa-tions of whole staircase movement in two stations of Beijing sub-way Line1.The two stations are Sihui East Station(ascendingﬂow)and Xidan Station(descendingﬂow).The Sihui East Station is a termi-nal station and all the passengers should get off the train and go to the transfer hall;therefore,the pedestrianﬂow on the observed stair is an ascendingﬂow.The Xidan Station is also a transferTable2Individual horizontal speeds in ten instances of staircases with different geometries.ID Riser height(mm)Tread depth(mm)Gradient(°)Horizontal speed(m/s)Source Note#H120021043.60.361("),0.509(;)a Frantzich(1996)Narrow stair #H215030519.00.427("),0.601(;)Lam(2000)MTR #H316327131.00.417("),0.569(;)Lam(2000)KCR #H419027035.10.423(")Kretz(2008)Long stairs #H515029027.30.538("),0.581(;)Fujiyama(2010)Elder people #H615726730.50.590("),0.721(;)Fujiyama(2010)Young people #H718623838.00.488(;)a Peacock(2012)11-Floors #H819125436.90.440(;)a Peacock(2012)18-Floors #H915028028.20.547(;)Ma(2012)SWFC #H1014028026.50.53(;)Yang(2012)Stair No.2#a The speeds were converted to horizontal speeds.Y.Qu et al./Safety Science70(2014)189–201191station and the passengers can get on or off the train by the stair, and there is an escalator on one side of the stair to relieve the coun-tering passengerﬂows.Therefore,the pedestrianﬂow is a descend-ingﬂow.The schematic diagram of the observation stations is shown in Fig.1,and the information of the stairs is shown in Table3.A HD camera was set on the transfer hall to record the trail of each passenger.The observations in the subway stations were made during the afternoon rush hours of weekend(17:30–18:30,Sunday)on May 12,2013.In our observations,383pedestrians(216males,167 females)were collected at the selected staircases in Xidan Station, and221pedestrians(129males,92females)were collected in Sihui East Station.Most of the pedestrians were young and mid-dle-aged people,and their ages mostly ranged from25to55years old.The proportion of children and elderly was very low.In our observations,most of the pedestrians carried light bags and walked in a normal speed on staircases.Because the camera was not right above on the observed region, the passengers were sometimes overlapped in the video.Each pedestrian was recorded by individual characteristics,such as gen-der,age,body size,hair,shirt and pants.Then,the pedestrians were recognized by their features,entering and leaving time.The speed of each pedestrian was calculated by the travel time(leaving time minus entering time)dividing the travel distance on the stairs.The video recordings were processed semi-manually,and the dynamic evacuation characteristics,such as average density,speed andﬂow,were analyzed.Take Xidan station for example,the evac-uation dynamic characteristics and a snapshot were illustrated in Fig.2a–c).It was found that the curve of time-varying density was divided into several segments,and each segment represented a stream of pedestrians entering and leaving the stairs.During the observation,some measures,such asﬂow restriction and guidance, had been adopted to avoid crowdedness,so the density of pedestri-ans on stairs was in a normal(low)level.There were eight local maximum points exceeding1.0pedestrians/m2,and the maximal density was about1.6(1/m2).Velocity showed an opposite trend of density.The velocityﬂuc-tuated between0.4m/s and1.0m/s,and the average velocity was about0.57m/s.The volatility of individual velocities might be caused by different individual capability and desired velocity.In a low density,the pedestrians who walked fast would overtake front pedestrians who walked slowly and blocked them.When the density became larger,the pedestrians began to slow down and follow with others,and then queues might form on stairs.In Fig.2d),the acquainted or familiar people might walk abreast, which is regarded as‘subgroup behavior’(Yang et al.,2012).If they walked slower than the surrounding people,they would form a dynamic bottleneck.Additionally,lane-forming phenomenon was also found.In Fig.3a),the distributions of the speeds during the observa-tions followed normal distributions,which were similar with the reference(Peacock et al.,2012).The average velocities of walking upstairs and downstairs were0.55m/s and0.63m/s,respectively. Affected by gravity,going upstairs was slower than going down-stairs.By gathering the observed data of unidirectionalﬂow,the relationships between the velocity and the density were shown in Fig.3(b).The velocity decreased as the density increased.It should be noted that,in a low density,the velocity of going upstairs was a little higher than downstairs.It was because some of the pedestrians were hurried out of station and ran more than one steps at one time.3.Modeling the pedestrian’s movement on stairs3.1.Body shapesIn the existing models,the projection of a pedestrian’s body shape is usually regarded as a square(i.e.,Kirchner et al.,2004), a rectangular(i.e.,Song et al.,2006;Weng et al.,2007),a circle (i.e.,Helbing et al.,2000),an ellipse(i.e.,Chraibi et al.,2010)or a set of three circles(Thompson and Marchant,1995).In these geo-metrical shapes,the three-circle shape has some geometrical and computational advantages on modeling the staircase movement. First of all,the three-circle shape is a better alternative to describe the pedestrian’s body shape.It is because the occupied space of one pedestrian is restricted by the stairs,and the shoulder width of the pedestrian is larger than lateral width(Xu and Song,2009).In addi-tion,a pedestrian walks with a relative slow speed on stairs,and his/her space requirement keeps almost constant.Secondly,in social force model,the distance of closest approach of two pedes-trians is a key parameter when calculating the self-driven force, the repulsive force and the contact force.The closet distance between two single-circle or three-circle shapes can be easily cal-culated(Thompson and Marchant,1995);however,the calculation of the closest distance of two ellipses is surprisingly difﬁcult (Zheng and Palffy-Muhoray,2007).For the convenience of calcula-tion,the three-circle shape is a better alternative than ellipse shape.Therefore,the three-circle shape is chosen to describe pedestrian’s body shape,and the schematic diagram is shown in Fig.4.3.2.Modiﬁed social force modelThe well-known social force model(Helbing and Molnar,1995; Helbing et al.,2000)is a microscopic force-based model that can reproduce several self-organized phenomena,such as lane form-ing,arching queue,shock waves and clogging effects(Helbing et al.,2005,2007;Moussaid et al.,2011).The model describes pedestrians’movement behavior by introducing the self-driven force~f Di,the contact force with pedestrians~f Cijand walls(obstacles) ~f Ciw.The self-driven force can be calculated by Eq.(1),and the total force~f exerted on pedestrian i can be formulated as Eq.(2).192Y.Qu et al./Safety Science70(2014)189–201~f D i ¼m ~m desiÀ~m isð1Þ~f i ¼~f D i þXj~f C ij þXw~f C iwð2ÞMoussaid et al.(2011)have proposed a heuristics-based modelto replace the social force with a heuristics intelligent optimum function,which can be regarded as a so-called collision prediction process.The model can overcome some difﬁculties in the original versions.Based on the model,we use the three-circle shape,intro-duce some special rules and establish an enhanced model todescribe the staircase movement.In our model,the modiﬁcations are concentrated on the calculations of optimal direction selection,self-driven force and contact force.3.2.1.Selecting optimal direction In Eq.(1),the desired velocity ~m des i can be obtained by the mag-nitude m des i multiplies by the direction ~e des iof desired velocity.The calculation of ~e desi is called ‘optimal direction selection’,which isTable 3Detailed step sizes.ID StationStep number Width (mm)Depth (mm)Height (mm)Gradient (°)Flow direction#O1Xidan Station 16240030014025.0Descending ﬂow #O2Sihui East Station15190033015726.1Ascending ﬂow(b) Change of pedestrian density with time inthe observation of staircase #O1(c) Change of pedestrian speed with time inthe observation of staircase #O1(d) A snapshot of the observations in staircase #O2(a) Change of pedestrian flow with time inthe observation of staircase #O1 Fig.2.Processed data and a snapshot.Y.Qu et al./Safety Science 70(2014)189–201193an important component of the model(Moussaid et al.,2011).In our work,the body shape is extended to three-circle shape,and the calculation becomes a little complex.To make a clear state-ment,some notation and deﬁnitions are given as follow:for pedes-trian i,the large circle’s radius is r i1,the small circle’s radius is r i2, the mass is M i,the maximum velocity is v0i,the location is~l i,the velocity is~m i and the desired destination is~D i.Assume that pedestrian i moves at the velocity v0i along the direction of direction a,and will contact with pedestrian j after D t time.The i0and j0represent the locations of i and j at time t+D t.Then,fðaÞ¼v0iÁD t is the distance to theﬁrst collision with other pedestrian or obstacle in the direction a.If no collision is expected to occur,f(a)is set to a default value d max,which repre-sents the‘maximum horizon distance’of pedestrian i.The calcula-tion of furthest distance without collision f(a)can be improved as follows:l ixm ðtþD tÞ¼l ixmðtÞþv ix D t;l iy mðtþD tÞ¼l iy mðtÞþv iy D tðm;n2f1;2;3gÞl jxn ðtþD tÞ¼l jxnðtÞþv jx D t;l jy nðtþD tÞ¼l jy nðtÞþv jy D tðl ixm ðtþD tÞÀl jxnðtþD tÞÞ2þðl iymðtþD tÞÀl jynðtþD tÞÞ2¼ðr imþr jnÞ2ð3ÞPut theﬁrst two items into the third item,we can get a qua-dratic equation with moving time D t.Given the locations and velocities of pedestrians i and j,the D t can be easily solved.Then, the value fða i m j nÞand dðaÞcan be calculated asfða i m j nÞ¼min f v i Deltat;d max g;fðaÞ¼min f fða i m j nÞg;aÃ¼argmin f dðaÞg dðaÞ¼d2maxþfðaÞÀ2d max fðaÞcosða0ÀaÞð4ÞThe optimal velocity direction~e¼ðcos aÃ;sin aÃÞ;here,aÃ¼argmin f dðaÞg is the optimal direction.Fig.5illustrates the calculation.3.2.2.Self-driven forces and contact forcesThe pedestrian’s staircase movement is a three-dimensional motion,which contains horizontal and vertical motion.Pedestrian should change his/her center of gravity to ascend or descend the stairs,which are shown in Fig.6(a)and(b).The complicated move-ment includes the pedestrian’s physiological activity and energy transformation.In our work,we mainly focus on the pedestrians’horizontal optimal choice and crowding behaviors,so the vertical motion is approximately regarded as a linear motion,which is shown in Fig.6(c).In horizontal motion,pedestrians not only over-take front people with slower speeds but also have to notice the steps and prevent themselves from falling down from stairs.To mathematically depict horizontal movement,some assumptions and rules are introduced to simplify the movement,which is illus-trated in Fig.6(d).Theﬁrst assumption is that a pedestrian wants to move forward within n steps(n is integer).And the pedestrian’s desired destina-tion of next footstep is the center of the forward step.For example, in Fig.6d),pedestrian often moves forward with integer steps,i.e., one step(point A)or two steps(point B).When he/she moves for-ward with non-integer step,i.e.,2.5steps(point C),he/she will move to the edge of the step,and may feel unstable,unsafe or even fall down from the stairs.Therefore,point C is not considered in our model.The pedestrian’s horizontal footstep length is deﬁned as d h¼nDcos b,and is restricted by the tread depths D and riser heights H of the step.Here,b is the included angle between the optimal direction aÃand the x-coordinate.If the pedestrian is obstructed by other pedestrians or obstacles,he/she will slow down and avoid collision,and the footstep length does not exceed the maximal dis-tance fðaÃÞ(Eq.(4))in the optimal direction aÃ.Finally,the footstep length is expressed as:d h¼minnDcos b;fðaÃÞ&'ð5ÞAccording to the model(Moussaid et al.,2011),a pedestrian maintains a distance from theﬁrst obstacle in the chosen walking direction that ensures a time to collision of at least a relaxation time s.In other words,desired velocity of pedestrian i is formu-lated as v des i¼d h s.Additionally,pedestrian’s speed is assumed to not exceed a maximum velocity v max.Then,the horizontal maxi-mum velocity is v max cos h,and the angle h represents the slope of the stair tan h¼HÀÁ.The horizontal desired velocity v des i can be formulated as Eq.(6):v desi¼mind hs;vmaxcos hð6Þ(a) Probability distribution(b) Fundamental diagramFig.3.Speed and fundamental diagram of theobservations.194Y.Qu et al./Safety Science70(2014)189–201。

Model Test 2参考答案[优质文档]

Model Test 2参考答案Part ⅡReading Comprehension (Skimming and Scanning)1. 答案B解析：根据题干，考生可锁定文章的第三段。

段首讲，关于解梦，没有人比Sigmund Freud 给出的答案让人更满意。

2. 答案C解析：根据选项的内容，考生可锁定文章的第四段。

第四段一开始就讲到Sigmund Freud 的生平。

选项B是个干扰性，文中只说到end his days(终老)，并没有说end his life(自杀)。

3. 答案A解析：根据题干，考生可关注文章的第六段。

第六段中讲：他了解人身体如何运行，却越来越关注人类心理。

从中可以判断，他对人类心理更感兴趣。

4. 答案D解析：根据题干，考生可锁定文章的第八段。

段首讲，那个年代，基本上没有医生对这个话题感兴趣。

考生可回到第七段，确定“这个话题”指代的是“想法、意见和梦境”。

5. 答案A解析：根据题干中的人名Dr Josef Breuer ，考生可锁定文章的第九段。

Josef Breuer 医生给Freud讲了自己一个病人的故事，这给Freud带来了灵感。

由此可以判断，在Freud 学说中，Josef Breuer医生提供了一些帮助。

6. 答案D解析：根据题干，考生可锁定文章的第十段。

此段主要讲了什么是心理分析，也就是所谓的“谈心疗法”，即病人可以自由地讨论自己遇到的困扰。

7. 答案C解析：根据题干，考生可锁定文章的第十一段。

段中的原话为，He discovered that the feelings of very young children are not so different from those of their parents，而选项C是其同义句。

8. 答案the human mind解析：根据题干，考生可锁定文章的第十二段。

很多人肯定，Freud找到了打开人类心理的一把钥匙。

雅思g类阅读真题回忆解析汇总

雅思g类阅读真题回忆解析汇总雅思的阅读备考可以采用题海战术，下面小编给大家整理了雅思g类阅读真题回忆解析汇总，希望大家喜欢。

雅思g类阅读真题回忆解析1篇章介绍体裁：记叙文结构：第一段鹰击长空情愫不灭第二段动力滑翔存在缺陷第三段遭遇险情才知培训第四段特技飞行魅力无限第五段 Rossy改行亲身体验第六段借助翅膀飞行稳健第七段即便梦圆恐不多见试题解析·题目类型：MULTIPLE CHOICE·题目解析：题号：28定位词：Vandenbulcke, paragraph 3文中对应点：第三段：Patrick Vandenbulcke答案解析：题目：以下哪项关于Vandenbulcke的信息出现在第三段?分析：解题的关键在于与此人相关的来自第三段的原文信息。

选项A“他险些未能避免一次危险情况”与原文中Another keen paramotorist recently experienced a close call when in the air以及这句话之后的关于事情经过的描述相对应。

选项B“他不懂得自己使用的装备”在该段中没有出现。

选项C“他没有对当时的情况作出迅速的反应”与原文中I realized I had to get to the ground fast意思相反。

选项D“他幸运地得到了所需的帮助”在该段中没有提及。

因此，本题答案为A。

题号：29定位词：second-hand, equipment, sale中文对应点：第三段：equipment secondhand, pre-used kit, sale答案分析：题目：当作者提到一些有待出售的二手动力滑翔设备时，他在强调。

分析：选项A“动力滑翔设备供不应求”在原文中没有提到。

选项B“动力滑翔设备需要认真测试”在原文中也没有对应的内容。

选项C“动力滑翔运动是一项昂贵的兴趣爱好”与本话题无关。

选项D“动力滑翔运动是一项可能带来危险的娱乐消遣活动”与第三段倒数第四句However he warns:‘Although it seems cheaper to try to teach yourself, you will regret it later a s you won’t have a good technique.’以及最后一句‘Scared myself to death,’the seller reported,‘hence the reason for this sale.’对应，构成同义表述。

上海工程技术大学城市轨道交通车辆专业英语复习要点2

第一章概述第三章电力与电子技术第四章气动系统和制动系统/Automatic Train Control,ATC列车自动控制/Automatic Train Operation,ATO 列车自动操作/Auxiliary Inverter,AI辅助逆变器/Brake Control Electronics,BCE制动控制电子元件/Brake Control Module ,BCM制动控制模块/Brake Control Units，BCU制动控制单元/Brake Control Panel，BCP制动控制板/Brake Electronic Control Unit,BECU制动控制电子单元/Capacitor Charging Contactor,CCC电容充电接触器/Central Control Functions，CCF 中央控制功能/Central Control Units，CCU中央控制单元/ Compact I/O，CIO 紧密式I/O/ Door Control Unit，DCU门控单元/ Driver Display Units，DDU 司机显示单元/Emergency Brake，EB紧急制动/Friction Braking，FB摩擦制动/ High Voltage，HV高压/Heating，Ventilation，Air Conditioning，HVAC 加热，通风，空调/High Speed Circuit Breaker，HSCB高速回路断路器/Human Machine Interface，HMI人机接口/Insulated Gate Bipolar Transisitor，IGBT 绝缘栅双机晶体管/Intelligent Display Unit，IDU智能显示单元/ Interior Display Unit，IDU内部显示单元/Intermediate Voltage，IV中压/Line Contactor Relay，LCR线路接触器继电器/Line Contactor，LC线路接触器/Low Voltage，LV低压/ Operation Control Centre，OCC操作控制中心/ Parking Brake，PB停车制动/Passenger Emergency Communication Unit，PECU 乘客紧急通信单元/Pulse Width Modulation，PWM脉冲宽度调制/Remote Input Output Modules，RIOM远方输入输出模块/ Traction Control Functions，TCF 牵引控制功能/Traction Control Unit，TCU牵引控制单元/Train Information and Management System，TIMS列车信息与管理系统/Variable Voltages and Variable Frequencies，VVVF，变压变频/Vehicle Control Unit，VCU车辆控制单元/ Vehicle Mounting Plate，VMP车辆安装盘/ATO作用：①station-to-station automatic driving ②speed control ③accurate station stop ④station departure/arrival management ⑤automatic station name announcements to the passengers ⑥control of opening/closing of door.ATP作用：①ensuring train spacing ②monitoring train speed against limit conditions ③ensuring that imperative stopping points are not overrun ④avoiding uncontrolled movement⑤detecting track occupancy ⑥measuring train speed ⑦localizing the train in the net work ⑧triggering the Emergency Brake if necessary ⑨monitoring train doors opening and closing ⑩applying temporary speed restrictions.转向架组成：①wheelsets,comprising wheels,axle,axle boxes,earthing brush,slip-slide generators and speed generators.②two spring suspension systems: primary suspension and second suspension.③bogie frame.④brakes.⑤traction drive units.⑥connection of bogie to the car body.RIOM可执行的任务：①reading digital inputs and setting digital outputs.②communicating via serial links.③filtering and suppressing contacting bounce of acquired data.④returning to a default state in the absence of network communication ,including extracting standard protocol framing for serial links.⑤self-testing inputs and outputs in a continuous way,performing local hardware self-tests and software control.DDU可执行的任务：①the control of the preparation of the train.②the control of the status of the trainset.③the control of the display of the faults which have occured in the trainset.④the control of the visual passenger information system.⑤the control of the broadcast of pre-recorded audio message.。

高铁技术的英语作文初一

Highspeed rail,commonly known as bullet trains,has revolutionized the way we travel.It is a testament to the advancements in modern transportation technology.Heres an essay on highspeed rail technology suitable for a junior high school student:The Marvel of Modern Transportation:HighSpeed RailIn the realm of transportation,the advent of highspeed rail has been nothing short of a marvel.It has transformed the way we travel,making long distances seem shorter and more accessible than ever before.The concept of highspeed rail is not new,but its recent advancements have made it a preferred mode of travel for many.Introduction to HighSpeed RailHighspeed rail is a type of rail transport that operates significantly faster than traditional rail traffic.It is characterized by its highspeed train sets and dedicated tracks,which allow for speeds exceeding200kilometers per hour.The technology behind highspeed rail is a combination of aerodynamics,advanced materials,and sophisticated control systems.History and DevelopmentThe idea of highspeed rail was first realized in Japan with the Shinkansen,which began operations in1964.Since then,countries like France,Germany,and China have developed their own highspeed rail networks.The development of highspeed rail has been driven by the need for efficient,environmentally friendly,and comfortable transportation options.Technological InnovationsThe technology behind highspeed rail is continually evolving.Key innovations include:1.Aerodynamics:Highspeed trains are designed with aerodynamic shapes to reduce air resistance,allowing them to travel at high speeds with minimal energy loss.2.Maglev Technology:Some highspeed trains use magnetic levitation,which reduces friction by levitating the train above the tracks,further increasing speed and efficiency.3.Regenerative Braking:This system captures the energy generated during braking andreuses it to power the train,making highspeed rail more energyefficient.4.Advanced Materials:The use of lightweight materials in the construction of highspeed trains reduces the overall weight,allowing for higher speeds and lower energy consumption.5.Control Systems:Sophisticated control systems ensure the safety and precision of highspeed rail operations,including automatic train control and realtime monitoring of train performance.Benefits of HighSpeed RailThe benefits of highspeed rail are numerous and include:1.Time Efficiency:Highspeed rail significantly reduces travel time between cities, making it a viable alternative to air travel for shorter distances.2.Environmental Impact:Compared to other forms of transportation,highspeed rail produces less carbon dioxide and other pollutants,contributing to a cleaner environment.3.Economic Growth:The development of highspeed rail networks stimulates economic growth by improving connectivity between regions,encouraging tourism,and facilitating business activities.4.Safety:Highspeed rail is considered one of the safest modes of transportation,with a low rate of accidents and fatalities.Challenges and the FutureDespite its many advantages,highspeed rail faces challenges such as high initial infrastructure costs,land acquisition issues,and competition from other modes of transportation.However,with ongoing technological advancements and increasing environmental concerns,the future of highspeed rail looks promising.In conclusion,highspeed rail is a remarkable achievement in the field of transportation technology.It offers a fast,efficient,and environmentally friendly way to travel,and as technology continues to advance,we can expect even greater improvements in speed, comfort,and safety.This essay provides a comprehensive overview of highspeed rail technology,its history, technological innovations,benefits,and future prospects,making it suitable for a junior high school students understanding and writing ability.。

英语习题(新教材新高考人教版)Unit3SportsandFitness

必修第一册Unit 3Sports and Fitness高考题型组合练Ⅰ.阅读理解AThe days of staring at the computer screen pretending to be interested in an assignment even though you are bored out of your mind may soon be coming to an ’s because if Dr Harry Witchel，Discipline Leader in Physiology at England’s Brighton and Sussex Medical School，has his way，computers of the future will be able to detect boredom and even react to it real-time.But before you get concerned，the machine is not reading your is just keeping track of the constant involuntary(无意识的)movements that people exhibit when in front of a computer or even a are not the bigger instrumental actions like moving a mouse or using the remote，but barely noticeable movements like scratching，fidgeting，or says the level of movement is directly linked to how absorbed the person is in what he or she is reading or higher the interest level，the less the movement!To test the theory，Witchel and his team invited 27 people and exposed them to a variety of digital contents for three minutes at a activities ranged from playing online games to reading documents like the banking regulations that most people would find boring.A video motion tracker monitored their movements as they powered through each as the researchers had expected，the involuntary actions decreased dramatically，by as much as 42%，when the participants were totally absorbed in what they were reading or seeing.Fortunately，the scientists are not planning to use the findings to create machines that report students who are not focusing at ，they believe that combining the motion-detecting technology with future computers will help enhance the digital learning experience.The scientists say that being able to measure the students’interest level will enable educators to adjust the materials real-time and re-engage the also believes that the technology can provide filmmakers with honest audience opinions.1．According to Dr Harry，what will future computers be able to do?A．Keep a learner from distraction.B．Help a learner with his assignments.C．Read a learner’s mind exactly real-time.D．Identify dullness of a learner and respond to it.答案D解析细节理解题。

2020年高考英语真题重点语法和题型分类汇编精讲第09题阅读理解(解析版)

第09题阅读理解A number of Americans predict that driverless cars will revolutionize the form of travelling in cities and on highways However, recent experiments have shown that autonomous vehicles also have the potential to improve the quality of life for millions of Americans, especially the elderly and disabled, so long as the government and lawmakers carry out smart policiesA retirement community in Alabama which has been transformed by a small group of driverless taxis shows the potential of self-driving cars to change people's lives in America. Although the modified Ford Fusions are currently limited to a two-mile road, residents are already having the benefits of them to take part in socialactivities which they would otherwise be unable to enjoy simply because they could not get to them.When the experimental run finally reaches 15 miles of road, these residents whose average age is 77 will also have a convenient and reliable new way to keep their appointments. Because these cars continue to serve residents there, it is not difficult to understand why California is gradually simplifying regulations for the business.In New Jersey, ahead-thinking policies have the potential to unlock other hidden benefits of autonomous vehicles, especially for those with physical disabilities . The New Jersey Disability Righted for the development of this technology , saying that it could give people with disabilities greater opportunities in the workforce and enable them to lead more satisfying and independent lives.Many Americans admit that autonomous vehicles will be the future of transportation, but it is too often overlooked that this future cannot arrive fast enough for millions of Americans on others for day-to-day travel. The policymakers should follow the lead of places like California and New Jersey, and pass regulations to unlock these hidden benefits of driverless cars.1. What's the attitude of most American people to the future of autonomous vehicles?A. Uncertain.B. Doubtful.C. Indifferent.D. Optimistic.【答案】D【解析】推理判断题。

(NEW)北京航空航天大学外国语学院211翻译硕士英语[专业硕士]历年考研真题及详解

9. If you _____ something, such as food or drink, you reduce its quality or make it weaker, for example by adding water to it.
A. adulterate B. moor C. vaccinate D. sue 【答案】A 【解析】句意：如果你往食物或饮品之类的东西里掺假，例如往里面兑水，就会降低它们的质量。adulterate掺杂。moor停泊；固定。 vaccinate注射疫苗。sue控告；起诉。
10. The orphanage is just one of her _____ causes. A. phonetic B. philanthropic C. prevalent D. lunatic 【答案】B 【解析】句意：这座孤儿院只是她的慈善事业之一。philanthropic仁慈的；慈善的。phonetic语音的。prevalent盛行的，流行的。lunatic精神
2010年北京航空航天大学211翻译硕士英语考研真题及详解
Part Ⅰ. Vocabulary (30 points) Directions: There are 30 incomplete sentences in this part. For each sentence there are four choices marked A, B, C and D. Choose the ONE answer that best completes the sentence. 1. The _____ is used by astrologers to help calculate the influence of the planets on people’s lives. A. zephyr B. zodiac C. zyme D. zest 【答案】B 【解析】句意：天文学家通过占星术中的黄道十二宫来计算星球对人类生活的影响。zodiac黄道十二宫（用于占星术）。zephyr和风，微风。zyme酶。zest热情；热心。

关于交通工具的英语作文

Transportation plays a pivotal role in our daily lives,connecting us to various places and people.It is an integral part of modern society,facilitating the movement of goods and services,and enabling individuals to travel for work,leisure,or education.Here is an essay discussing the various modes of transportation and their impact on our lives.Introduction:The evolution of transportation has been a testament to human ingenuity and the desire to explore and connect.From the invention of the wheel to the development of the automobile,airplane,and highspeed train,each advancement has revolutionized the way we travel.Land Transportation:Automobiles:Cars have become the most common mode of personal transportation. They offer convenience,flexibility,and the freedom to travel at ones own pace.However, they also contribute to traffic congestion and environmental pollution.Bicycles:As a healthier and ecofriendly alternative,bicycles are gaining popularity, especially in urban areas.They are ideal for short commutes and promote physical fitness. Public Buses and Trains:These are essential for mass transit,providing affordable and efficient means to move large numbers of people.They help reduce traffic and are often powered by cleaner energy sources.Air Transportation:Airplanes:They have made the world a smaller place by enabling people to travel long distances in a short amount of time.The aviation industry has grown exponentially, connecting remote regions and facilitating international trade and tourism. Helicopters:Used for quick,shortrange travel,they are particularly useful in emergency services,sightseeing,and accessing difficult terrains.Water Transportation:Boats and Ships:Historically,water has been a primary mode of transportation for trade and exploration.Modern ships carry a significant portion of global trade,while ferries and cruise ships serve as means of travel and leisure.Submarines:Though not a common mode of transportation,submarines offer unique opportunities for underwater exploration and travel.Future of Transportation:Electric Vehicles:With a focus on sustainability,electric cars,buses,and bikes are becoming more prevalent,reducing our reliance on fossil fuels and decreasing carbon emissions.Hyperloop:A proposed mode of passenger and freight transportation,the hyperloopcould revolutionize travel by reducing transit times significantly through vacuumsealed tubes.Conclusion:Transportation is not just about getting from point A to point B its about the experience, the connection,and the impact it has on our environment and society.As we continue to innovate and develop new technologies,the future of transportation promises to be more efficient,sustainable,and accessible.Reflection:Reflecting on the various modes of transportation,one can appreciate the progress humanity has made.It is essential to consider the environmental and social implications of our choices in transportation,striving for a balance between convenience and sustainability.The future holds exciting possibilities,and it is up to us to shape it responsibly.。

25A subway train timetable optimization approach based on energy-ef

R
Manuscript received May 7, 2012; revised October 28, 2012 and January 22, 2013; accepted January 30, 2013. Date of publication March 15, 2013; date of current version May 29, 2013. This work was supported in part by the National Natural Science Foundation of China under Grant 71101007, by the National High Technology Research and Development Program of China under Grant 2011AA110502, by the National Basic Research Program of China under Grant 2012CB725401, by the State Key Laboratory of Rail Trafﬁc Control and Safety under Grant RCS2012ZT002, and by the Fundamental Research Funds for the Central Universities under Grant 2011JBZ014, 2013YJS019. The Associate Editor for this paper was M. Chowdhury. (Corresponding author: X. Li). S. Su, T. Tang, and Z. Gao are with the State Key Laboratory of Rail Trafﬁc Control and Safety, Beijing Jiaotong University, Beijing 100044, China. X. Li is with the School of Economics and Management, Beijing University of Chemical Technology, Beijing 100029, China (e-mail: xiang-li04@ ). Color versions of one or more of the ﬁgures in this paper are available online at . Digital Object Identiﬁer 10.1109/TITS.2013.2244885

Working Day Movement Model

Working Day Movement ModelFrans Ekman,Ari Keränen,Jouni Karvo and Jörg Ott Helsinki University of T echnology TKK,Dept.of Communications and Networking {frans.ekman|ari.keranen|jouni.karvo|joerg.ott}@tkk.ﬁABSTRACTAbstract movement models,such as Random Waypoint,do not capture reliably the properties of movement in the real life scenarios.We present and analyse a movement model for delay-tolerant network simulations that is able to produce inter-contact time and contact time distributions that follow closely the ones found in the traces from the real-world mea-surement experiments.We validate the movement model using the ONE simulator.Categories and Subject DescriptorsC.2.1[Network Architecture and Design]:Store and forward networksGeneral TermsDesign,Experimentation,Measurement,Veriﬁcation KeywordsMovement Model,Simulation,DTN,Delay-Tolerant Net-working,Mobility Models,Routing1.INTRODUCTIONMovement of the network nodes is essential for the per-formance of delay-tolerant networks(DTN).A movement model that captures the behaviour of the nodes in the real usage scenarios is thus needed for a reliable assessment of a new protocol.There are two types of movement models that have been proposed for these analyses—generic high level models that aim to produce movement accurate enough with statisti-cal measures,and models that describe incidental scenarios, hoping for a more accurate depiction of single devices. While eﬃcient to use in simulations,the high level mod-els,such as Random Waypoint(R WP)[9],often imply that the scenarios for which the protocols are simulated have huge numbers of nodes,so that the relevant protocol fea-tures are given statistically realistic distributions of events. Permission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provided that copies are not made or distributed for proﬁt or commercial advantage and that copies bear this notice and the full citation on theﬁrst page.To copy otherwise,to republish,to post on servers or to redistribute to lists,requires prior speciﬁc permission and/or a fee.MobilityModels’08,May26,2008,Hong Kong SAR,China. Copyright2008ACM978-1-60558-111-8/08/05...$5.00.For scenarios with few nodes,the diﬀerences between diﬀer-ent usage scenarios become more signiﬁcant.Thus,move-ment models that depict more precisely some speciﬁc types of movement are needed.We present a new movement model to be used in DTN simulations,called Working Day Movement Model.The model presents the everyday life of average people that go to work in the morning,spend their day at work,and commute back to their homes at evenings.The model intuitively de-picts the movement pattern of people,but we also verify the model by simulating it and compare the statistical features of the model to real-world traces.This paper is organised as follows.Section2reviews re-lated work on mobility models based upon which our model is presented in Section3.The simulation tool we use for evaluation in introduced in Section4.Section5shows the evaluation results,andﬁnally,Section6concludes the paper.2.RELATED WORKMobility models have been under active research recently, see[2].Inter-contact times and contact durations are typi-cal metrics for characterising mobility in sparsely populated DTNs.An inter-contact time,or sometimes referred to as an inter-meeting time,is the time interval between contacts for a node pair.It is deﬁned as the time interval a node pair is not in contact with each other.The contact time or con-tact duration is the time a contact between two mobile nodes lasts.Inter-contact times correspond to how often nodes will have an opportunity to send packets to each other,while the contact durations limit the amount of data that can be sent. Usually,inter-contact time distributions and contact time distributions are used in comparisons.Musolesi et al.[14]show that simple mobility models have very diﬀerent properties in terms of inter-contact time and contact durations compared to real user traces.The dis-tribution of inter-contact times does have practical impli-cations.Chaintreau et al.[4]show that if the distribution of inter-contact times is power-law distributed with expo-nent less than one,any possible routing algorithm in a delay tolerant network will produce an inﬁnite average delay for packet delivery.They also analysed four diﬀerent traces of real people moving,and concluded that the inter-contact times are power-law distributed with the power-law expo-nent less than one.Karagiannis et al.[10]show that the inter-contact times are only power-law distributed up to12hours,and have an exponential cut-oﬀafter that.A possible cause for the phenomenon is the daily routines people have.Han Cai et al.[1]show that simple models on a boundless area can produce a power-law distribution of inter-contact times.Additionally,they show that the exponential cut-oﬀis in many cases a side-eﬀect of the bounded area.The motivation behind this is that nodes that would move over the edges on an inﬁnite area are forced to stay within the area,thereby meeting other nodes sooner than they other-wise would and the number of long inter-contact times will be smaller.Kim et al.[12]extracted various parameters such as speed and pause time distributions from real user traces.These parameters were then used in a synthetic mobility model. Their model is validated based on the number of nodes within diﬀerent regions on the map at diﬀerent hours.This might not be a suﬃcient criterion alone to determine the suitability of a model,since the performance of protocols and applications in DTN is highly related to the nature of the contacts.Rhee et al.[15]uses Levy Walks to generate movement traces.The model is very similar to random walk,except that theﬂight lengths and pause times are drawn from a power law distribution.They manage to produce similar inter-contact time distributions as many real world traces, but the model does not capture characteristics as hetero-geneity among nodes,repetitiveness,group mobility or any relationships between nodes.The community based mobility model[14]is based on the idea that nodes favour squares with higher social attractiv-ity.The social attractivity is based on how many friends are in the same square.Changing friends depending on the time of day results in periodic patterns like for example people meeting their work colleagues in the day and their family in the evening.This model lacks group movement and the movement is relatively homogeneous.The paper does not show the inter-contact time distribution behaviour for more than up to roughly one third of a day.The time-variant mobility model[8]is somewhat similar. In this model,nodes move to diﬀerent squares at diﬀerent times of day in a periodic manner,thereby creating some heterogeneity in both time and space.Nodes do not move in groups and the movement is homogeneous in the sense that every node follows the same instructions.Little work has been made on indoor movement,espe-cially combined with outdoor movement.Some models and thoughts of oﬃce scenarios can be found in[13]and[5],but in a context not really practical for delay tolerant network simulations.Hsu and Helmy[7]show by studying real user traces that nodes are very often turned on/oﬀand only visit a small por-tion of the WLAN access points in campus areas.Moreover, theyﬁnd that node mobility while using network is very low and one node only meets a small portion of all other nodes in the area.Furthermore,they reveal repetitive patterns with a period of one day and heterogeneity among nodes.Accord-ing to them,the biggest issue with most synthetic models is that they are not capturing such characteristics as hetero-geneous behaviour,switching devices on/oﬀor relationships between users.Various group mobility models exist and analysis of the impact of group mobility[6]has been made.However,to the best of our knowledge,group mobility has never been a com-ponent of a larger model covering many other aspects like community,daily routines,heterogeneity,etc.The same ap-plies to most of the ideas presented above;they only model one aspect of mobility.Our approach is to combine these diﬀerent elements to create a new movement model.3.WORKING DAY MOVEMENT MODEL We have developed a new mobility model by combining diﬀerent movement model elements together.These models are called submodels.The model consists of three diﬀerent major activities that the nodes can be doing.They are be-ing at home,working and some evening activity with friends. These activities are the most common and capture most of a working day for the majority of people.More subtle varia-tions and many other activities exist but,for now,we assume that they are reasonably well captured by the activities we have modeled or their overall impact is small and leave in-troducing further diversity(by means of more submodels) to future work.On a more detailed level,the activities diﬀer from each other.These submodels repeat every day,resulting in peri-odic repetitive movement.Their parametrisation and adding further submodels as needed allowsﬁne-tuning the model to meet the needs of the target scenarios.Communities and social relationships are formed when a set of nodes are doing the same activity in the same location. For example,nodes with the same home are family members, while nodes with the same oﬃce location are colleagues from work.Nodes are doing the activities on a daily basis starting from home in the morning.Each node is assigned a wakeup time,which determines when the node should start from home.This value is drawn from a normal distribution with mean0and conﬁgurable standard deviation.The node uses the same wakeup time every morning during the whole sim-ulation.The variance in the wakeup time models the diﬀer-ences in rhythms in real life.At the wakeup time,nodes leave their homes,and use diﬀerent transport methods to travel to work.Nodes travel between activities either by car or by bus,which are both dif-ferent submodels.The working time is conﬁgurable.After the working hours,the nodes decide,by drawing,whether they go out for the evening activity,or return home.Again, diﬀerent submodels are used for transitions between the lo-cations.Diﬀerent user groups have diﬀerent locations where the activities take place.3.1Home Activity SubmodelThe home activity submodel is used for the evenings and nights.Each node is initially assigned a map point as its home location.Having reached this location,the node walks a short distance away and stays still until the wakeup time. We do not model any movement inside homes.Node activ-ities at home can consist of the device lying on some table until the next day,people watching TV,cooking,sleeping etc.,where the movements within the house are not rele-vant.3.2Ofﬁce Activity SubmodelThe oﬃce activity submodel is a2-dimensional model for movement inside an oﬃce where the employee has a desk and sometimes needs to walk to other places for meetings or just to quickly talk to someone.Minder et al.[13]present a model for meetings where organisation structure is taken into account.We do not use such a model because we areactually interested in the contacts of nodes,due to the ap-plication to delay tolerant networking.Habetha et al.[5]used a more detailed oﬃce movement model,where employees are moving in rooms and corridors. The walls will have a signiﬁcant eﬀect on the path-loss.For a simpler modelling,we do not model the signal attenuation on walls.The model adopted is as follows.The oﬃce is entered from a speciﬁc map point,called a door.The oﬃce is a square where the upper left hand corner is the door.Each node is assigned a coordinate inside the building where the node’s desk is located.The movement inside the oﬃce starts immediately when the node reaches the door;the node starts walking towards the desk with the walking speed deﬁned in the settings. When it reaches its desk,it stops for an amount time,drawn from a Pareto distribution.When the node wakes up from the pause,it selects a new random coordinate inside the of-ﬁce,walks there and waits for an amount of time also drawn from the same Pareto distribution.The movement between the desk and randomly selected coordinates repeats until the work day is over.The purpose of nodes moving be-tween their desk and random coordinates is that nodes hav-ing their desks close to each other will meet each other more frequently and nodes with their desks located next to each other will be in contact most of the time.Earlier research suggests that the length of meetings at an oﬃce follow a log-normal distribution[13].However,the study covers only team meetings,which does not necessarily correlate with pause times in movement.A truncated Pareto distribution is suggested in[15]for general movement inside buildings.We choose the Pareto distribution for our pause times inside the oﬃce.We also added parameters to turn oﬀthe pausing completely and to have an inﬁnite pause time, in which case nodes stay at their desk for the whole workday. Obviously,there are a variety of diﬀerent jobs and build-ings where people move accoring to diﬀerent patterns.Our model is aﬁrst level abstraction,but we are working on fur-ther parameterizations and extensions of the submodel to introduce broader diversity(and determine how this diver-sity impacts the mobility metrics).3.3Evening Activity SubmodelThe evening activity submodel models the activities that nodes can do in the evening,i.e.after work.This activity is done in groups.The evening activity model can be inter-preted as shopping,walking around the streets or going to a restaurant or a bar.Each node is in the beginning of the simulation assigned a favourite meeting spot.Immediately when a node ends its working day,it is assigned to a group based on its favourite meeting spot.If all groups for a given favourite meeting spot are full,a new one is created with a randomly selected and uniformly distributed size with min-imum and maximum values deﬁned in settings.The node then uses the transport submodel to move to the meeting spot.The node waits at the meeting spot until all the nodes of the group are present.Then they start moving according to the map based movement model,which is actually a ran-dom walk on streets.They all walk in a group along roads a certain distance deﬁned in settings,and then they pause for a longer time deﬁned in settings,andﬁnally split up and walk back to their homes.3.4Transport SubmodelNodes move between home,oﬃce and evening activity us-ing the transport submodel.During the initialisation,a con-ﬁgurable percentage of nodes in each group are set to use a car for transportation between activities.Nodes not moving by car will use the bus or walking submodel.Nodes mov-ing by car only use the car submodel for all transportations. Supporting diﬀerent types of transport models adds addi-tional heterogeneity and has impact on the performance of routing protocols,since quicker nodes,like cars for instance, can transfer packets longer distances quickly.•Walking submodelNodes that walk use streets to advance with a constant speed towards the destination.Dijkstra’s algorithm is used forﬁnding the shortest path to the destination.•Car submodelNodes owning a car can travel at a higher speed be-tween diﬀerent locations.Otherwise it does not diﬀer from walking.Within an activity submodel,car own-ers behave as the other nodes.•Bus submodelNodes without a car can use buses for travelling faster.There are pre-deﬁned bus routes on the city map.The buses run these routes according to a schedule.Buses can carry more than one node at a time.Each node that does not own a car knows one bus route. It can use any bus driving that route.The nodes make the decision of taking the bus if the Euclidean distance from the node’s location to the nearest bus stop summed with the Euclidean distance from the destination to the nearest bus stop is shorter than the Euclidean distance between the node’s location and the destination.Otherwise,it walks the whole distance.If the node decides to take the bus,it uses the walking submodel to the closest bus stop and waits for the bus.When the bus arrives,the node enters it and travels until the bus comes to the bus stop nearest the destination. Then it switches back to the walking submodel to reach the destination.3.5The MapAll nodes move on a map.The map deﬁnes the space and routes in which the nodes can move;it contains all the in-formation of the locations of the houses,oﬃces and meeting spots,as well as the bus routes with bus stops.The design of the map is an important part of the mobility model.Since all the movement of the nodes is determined by activities with speciﬁc locations,the placement of these locations de-ﬁne how nodes are moving on a larger scale,i.e.,in which areas of the map nodes will be doing diﬀerent activities. The positions of these locations can be node group speciﬁc, which makes it possible to create small districts within the map.Therefore,the map can be used to limit node move-ment to small areas,which we refer to as increasing the locality.On one hand,houses,oﬃces and meeting spots can be spread randomly on the map,thereby,having very little locality and nodes meeting easily.On the other hand,it is possible to restrict node movement to very small areas by creating lots of small districts,thereby increasing the local-ity.These may also be combined:diﬀerent sized districtwhere some overlap others allows to have high locality but also some movement between districts,which corresponds to nodes coming to some district to work or meet friends,while others are leaving their district for similar reasons.Nodes moving between districts,not located next to each other,will have to pass through other districts,thereby appearing as drive through traﬃc in the intermediary districts.Figure 1shows an example setup.The most suitable conﬁguration of districts is environmentspeciﬁc.Figure 1:A map of Helsinki city’s central areas di-vided into 4artiﬁcial districts4.EXPERIMENTAL SETUPWe implemented our movement model as an extension to the Opportunistic Network Environment (ONE)simula-tor [11].The ONE is a highly customisable communication network simulator for delay tolerant networking that has several movement models implemented,from simple Ran-dom Waypoint to more realistic Map Based Movement mod-els that can import map data and constrain node movement to the streets and roads of the imported data.ONE can also visualise the imported data and node movement using a GUI which helps on validating the model in an intuitive way.Based on the node movement and nodes’radio device’s range,ONE generates contact information reports that can be used for detailed analysis.This contact information also feeds the simulation engine embedded in ONE which sup-ports multiple DTN routing protocols [11].4.1Mobility ModellingONE can import mobility data from real-world traces or other mobility generators.Movement models and report modules are loaded dynamically based on the given conﬁgu-ration so that the simulator can be easily extended with new modules and the modules used in diﬀerent scenarios can be changed as needed.In a simulation setting,any number of types of mobile nodes—referred to as a node group —may be deﬁned.A node group shares a common set of simulation parameters like speed and pause time distributions,message buﬀer size,and radio range,among others.Diﬀerent node groups can also use diﬀerent movement model modules.The basic version of ONE supports the Random Way-point [9]mobility model,arbitrary mobility models by us-ing externally generated movement data,and diﬀerent map-based movement models.All map-based movement models obtain their conﬁguration data using ﬁles formatted with a subset of the Well Known Text (WKT)format.WKT ﬁles can be edited and generated from real world map data us-ing Geographic Information System (GIS)programs such as OpenJUMP 1.With map-based movement models,the nodes move using roads and walkways from the map data.In ad-dition,diﬀerent node groups can be set to use only certain parts of the map,thus allowing to distinguish between cars and pedestrians so that the former do not drive on pedes-trian paths or inside buildings.The simple random Map-Based Movement model (MBM)is a derivative of the Random Walk model,where nodes move to randomly determined directions on the map fol-lowing the roads as deﬁned by the map data.The Shortest Path Map-Based Movement model (SPMBM)is a derivative of the Random Waypoint model,where nodes use Dijkstra’s shortest path algorithm to calculate shortest paths from the current location to a randomly selected destination,by using the roads or paths.Finally,some nodes may have pre-determined routes in the map that they follow.This Route-Based Movement model (RBM)uses the same map data but nodes always select the next destination on the route they are currently travelling.This mode of movement is useful for modelling e.g.,bus and tram routes.4.2ImplementationThe Working Day Movement model was added to the ONE as a combination of many mobility models.One move-ment model implements the main model controlling the move-ment of the nodes going to work,to their homes and meeting their friends.The main model passes the responsibility to lower level models handling diﬀerent activities and trans-portation.Additional information about the destination is passed to the transportation models by the main model so that the nodes can ﬁnd the way to the right place.The main movement model also decides whether to travel by bus or by walking between activities,and whether to do some evening activity or not.Buses are an extension of the Route-Based Movement model,using bus routes deﬁned in WKT ﬁles.Buses in-teract with passengers through a bus control system.The bus control system acts as a mediator between buses and passengers,informing passengers when the bus stops and buses when passengers enter them.Each bus control sys-tem has a unique ID,which is used to link bus node groups together with normal node groups in the settings ﬁle.The evening activity makes use of a similar control system,deﬁned in the settings for each node group,to facilitate the group movement.The locations of oﬃces,meeting spots and homes are listed in WKT ﬁles deﬁned in settings for each group,or randomly selected by ONE if no WKT ﬁles are provided.4.3Measurements in ONECreation of reports for various events with the ONE is implemented with the help of diﬀerent event listeners.The default package of the ONE contains some commonly usedreporting tools and some new ones were developed.In the case of inter-contact times and contact times,the ONE gives as output a list with event lengths and observation counts, from which it is easy to calculate a CCDF.We have imple-mented a contacts per hour report generator,which counts the number of contacts happening each hour.Finally,we have implemented two diﬀerent reporting tools to measure encounters.Theﬁrst one counts the number of other nodes a node has encountered(unique encounters),and provides a distribution for the fraction of user population nodes have encountered.The second one counts the number of total encounters and the number of unique encounters for each node separately,and provides a list with all the nodes and their measurements.5.SIMULATIONWe validated our model by comparing it to real user traces in terms of inter-contact times,contact durations and con-tacts per hour.The inter-contact times and contact dura-tions are commonly used in the literature for characterizing connectivity in DTNs,while the contacts per hour metric has been used to measure activity at diﬀerent hours of the day.Data from real world measurements is available for all three metrics,thus allowing us to validate our model.We had over1000nodes moving on a map of the Helsinki centre area with the surrounding districts with the size of roughly7000×8500m2.The area was divided into4main districts,see Figure1.Additionally,3overlapping districts were created to simulate movements between the centre and other districts,and one district to cover the whole simulation area.See Table1for details about the assignments of nodes, oﬃces and meeting spots.Every district,except the one covering the whole map,was assigned its own bus route and 2buses.The district covering the whole map has4buses driving on one route.Table1:The assignment of nodes,oﬃces and meet-ing spots to the diﬀerent districtsDistrict OﬃcesA30501C201002E(A and B)201504G(A and D)302005Half of all the nodes were set to travel by car.The walking speed for nodes was set to0.8–1.4m/s and for buses7–10m/s with a10–30s waiting at each stop.The probability to do some evening activity after work was set to0.5with the group size1–3.The working day length was28800s and the pause times inside the oﬃce were drawn from a Pareto dis-tribution with coeﬃcient0.5and minimum value10s.The oﬃce size was set to a100m×100m square.The size of the oﬃce was chosen so that it would compensate for the lack ofﬂoors,walls,etc.The diﬀerences in schedules of nodes were drawn from a normal distribution with a standard de-viation of7200s.We also added10nodes moving according to the SPMBM model in the background to simulate taxis, delivery of goods,etc.The transmit range of all nodes was set to10m.The nodes were considered to be in contact when they were closer to each other than the transmit range.In other words,imme-diately when two nodes were in reach of each other,a con-nection was established.In the real world,there is usually a connection setup delay and the frequency of scanning for other devices is usually limited to keep energy consumption low(leading to a detection delay).It is worth noting that this phenomenon has probably aﬀected real world contact traces used for comparison.A study about optimal prob-ing of contacts can be found in[18].Considering scanning intervals is subject to ongoing work.We used a warmup period of half a day,which is suﬃcient due to the periodic nature of the mobility model.For comparison,we simulated a R WP scenario on a same sized simulation area with1000nodes,moving with speed 0.5–5m/s and pause time1–3600s,both uniformly distributed. We estimated the inter-contact and contact time distri-butions by sampling them from simulation runs of length T=5·105s.Due to theﬁnite simulation time,the longer events are less likely to get observed.This is because a larger fraction of them has the beginning or end outside the sim-ulation time.This leads to a systematic error so that it is not easy to say whether there is an exponential decay in an empirical distribution or just this systematic error.To avoid this uncertainty,we adjust the experimental distributions as follows.We assume that the events are uniformly distributed over a longer period of time.Then,consider the probability of an event of length x,p(x).Only events that begin during the time interval[0,T−x]will get recorded.To compensate this,the estimated probability density functionˆp(x)isˆp(x)=Thow much heterogeneity the model has compared to R WP.A dot indicates how many total encounters and unique en-counters a node has.If nodes move according to diﬀerent patterns and speeds within diﬀerent sized areas,nodes will meet diﬀerent amounts of other nodes.Additionally,a node restricted to a small area will keep meeting the same nodes over and over again,while a scout node exploring the whole simulation area will mostly meet new nodes.Therefore,the scatter diagram can be used to measure heterogeneity.))We experimented with three scenarios with diﬀerent num-bers of background SPMBM nodes,to better understand how mixing diﬀerent movement models aﬀects the inter-contact time distribution.Figure 6shows that more SPMBM nodes smoothen the inter-contact times distribution and the exponential cutoﬀgets less sharp.This is due to the expo-nential nature of the SPMBM movement.To investigate the eﬀect of the evening activity,we con-ducted two experiments with diﬀerent values for the evening activity probability;one where nodes always do evening ac-tivity and one where nodes never do.The ﬁrst scenario)Total encountersU n i q u e e n c o u n t e r s(b)R WPFigure 5:Total-vs.unique encounters for each node plotted as a dotcreated a peak in the contacts per hour graph at the time when most nodes were doing their evening activity,while the second scenario lowered the activity graph to almost the same state as during night.We concluded that the proba-bility of evening activity parameter can be used to adjust the model to diﬀerent environments where there is more or less night-life.To determine whether the way nodes move between dif-ferent activities makes a diﬀerence,we experimented with two scenarios:one where the all the nodes move by car and another where all use the bus.A high percentage of nodes travelling by car seems to reduce signiﬁcantly the fraction of the population a node has encountered.Cars move fast and always along the shortest path,therefore the probability of meeting new nodes gets smaller.The bus travels longer routes and nodes have much more opportunities to meet oth-ers at bus stops or in the bus.We also get slightly longer contact durations with buses,but it seems as if most of the contact durations originate from other things.To understand the impact of the map,we created a Man-hattan like map where one block is a square with the side 180m.We compared the results to the results from the Helsinki map,and surprisingly,there were not many diﬀer-ences.To make the comparison easier we decided to remove the districts and randomly select the locations of the homes,。

211133177_全自动限速运行模式方案研究

全自动限速运行模式方案研究常峰（中铁二院工程集团有限责任公司，成都 610031）摘要：列车在全自动运行驾驶模式下会出现丢失定位的故障情况，需快速获取定位或者运行至就近站台区域。

结合列车在正线区间、折返轨、车辆段运行等场景，对全自动限速运行模式的申请流程、移动授权距离、场景处置等进行全面分析研究，并提出解决方案。

关键词：全自动运行系统；信号系统；限速运行；移动授权中图分类号：U284.48 文献标志码：A 文章编号：1673-4440(2023)04-0063-05Research on Fully Automatic Speed-restriction Operation ModeChang Feng(China Railway Eryuan Engineering Group Co., Ltd., Chengdu 610031, China)Abstract: In view of the fault that a train loses positioning in the fully automatic operation mode, it is necessary to quickly regain the positioning or operate the train to the nearest platform area. In this paper, the application process, movement authority distance and the handling of scenarios in the fully automatic speed-restriction operation mode (FRM) are comprehensively analyzed and studied, and the solutions are put forward. The following scenarios are discussed: operation in a section on a main line, the turning back of a train, and train movements in a depot, etc.Keywords: fully automatic operation system; signal system; speed-restriction operation; movementauthorityDOI: 10.3969/j.issn.1673-4440.2023.04.012收稿日期：2022-03-29；修回日期：2023-04-05作者简介：常峰（1989—），男，工程师，硕士，主要研究方向：交通信息工程及控制，邮箱：***************。

火车训练英语作文

火车训练英语作文英文回答：Train travel is an experience that offers a unique blend of comfort, convenience, and scenic beauty. As a seasoned traveler, I have had the privilege of embarking on numerous train journeys across diverse landscapes, each providing its own set of memorable experiences.One of the most notable advantages of train travel is the comfort it affords. Unlike airplanes, where space is often cramped and legroom is limited, trains offer spacious seating and ample room to stretch out and relax. This is particularly beneficial on long-distance journeys, where being able to move around and avoid muscle stiffness is essential. Moreover, train carriages are often equipped with amenities such as power outlets, Wi-Fi, and dining areas, making it possible to work, stay connected, and enjoy refreshments without having to leave your seat.Another significant benefit of train travel is its convenience. Train stations are typically located incentral areas, making it easy to access the city or town you wish to visit. This eliminates the need for long transfers or additional expenses associated with airport parking or transportation. Additionally, train schedules are generally reliable, and delays are often less frequent than those experienced with air travel, providing peace of mind and allowing for better planning.Beyond the practical advantages, train travel also offers a unique opportunity to appreciate the scenic beauty of the surrounding landscape. As the train glides through rolling hills, lush forests, or along sparkling coastlines, passengers are treated to breathtaking views that would otherwise be missed when traveling by car or plane. The panoramic windows of train carriages provide an immersive experience, allowing travelers to fully soak in the natural wonders and create lasting memories.Of course, no mode of transportation is without its drawbacks. One potential downside of train travel is thelimited flexibility it offers compared to driving or flying. Trains operate on predetermined routes and schedules, which may not always align perfectly with your desired itinerary. Additionally, train fares can sometimes be more expensive than other modes of transportation, especially for long-distance journeys. However, when considering the overall benefits and experiences that train travel provides, these drawbacks often pale in comparison.In conclusion, train travel offers a compelling combination of comfort, convenience, and scenic beauty that makes it an ideal option for both leisure and business travelers. Whether you are planning a short day trip to a nearby destination or an extended journey across the country, a train ride promises a unique and unforgettable experience.中文回答：火车旅行是一种体验，它提供了一种独特的舒适、便利和美丽的风景。

当代中国交通英语作文

Living in the heart of contemporary China, Ive witnessed the remarkable evolution of the countrys transportation system. Its a transformation that has not only reshaped the landscape but also the daily lives of its citizens. The rapid development of transportation in China is a testament to the nations commitment to modernization and progress.Growing up in a bustling city, Ive always been surrounded by the hum of traffic and the constant movement of people. The first mode of transportation that captured my attention was the bicycle. It was a common sight to see families riding together, weaving through the streets with ease. The bicycle was not just a means of transport it was a symbol of freedom and mobility for the average Chinese citizen.As I grew older, the cityscape began to change. The introduction of public buses and trams brought a new dimension to urban travel. The expansion of public transportation was a response to the growing population and the need for efficient movement within the city. The buses, with their distinct routes and schedules, became an integral part of my daily commute to school.The most significant leap in transportation, however, came with the advent of highspeed trains. The Chinese highspeed rail system, known for its speed and efficiency, has revolutionized longdistance travel. I remember the first time I boarded a highspeed train, the sense of excitement was palpable. The train glided smoothly over the tracks, covering vast distances in a matter of hours. It was a stark contrast to the slow, rumbling trains of the past.The highspeed train has not only made travel more convenient but has also connected cities and regions in ways previously unimaginable. It has fostered economic growth and cultural exchange, bringing people closer together. The network of highspeed trains is a marvel of engineering, a symbol of Chinas technological prowess.In recent years, the focus has shifted towards more sustainable forms of transportation. The rise of electric vehicles and the expansion of subway systems reflect a commitment to reducing pollution and creating a greener environment. The subway, with its extensive network and punctuality, has become the preferred mode of transport for many city dwellers, including myself.The subway system is a testament to the ingenuity and foresight of urban planners. It has alleviated traffic congestion and provided a reliable alternative to road transportation. The experience of navigating the subway, with its labyrinth of tunnels and stations, has become a part of my urban life.Moreover, the emergence of ridesharing apps and bikesharing services has added another layer to the transportation landscape. These services have made it easier for people to move around the city, offering flexibility and convenience. The sight of shared bikes lined up on the streets has become a common feature of the urban environment.The transformation of transportation in China is not just about the vehiclesand infrastructure its about the people and their stories. Its about the elderly couple who now have the opportunity to visit their children in distant cities, or the young student who can travel to school without the hassle of traffic. Its about the connections being made and the lives being improved.In conclusion, the evolution of transportation in contemporary China is a story of progress and innovation. Its a narrative that reflects the countrys ambition to create a more connected, efficient, and sustainable society. As a young person growing up in this era, I feel a sense of pride and excitement about the future of transportation in China. The journey has been remarkable, and I look forward to seeing where it leads next.。

WL080103CommunicationsinNonlinearScienceandNumericalSimulation(476)

Computer Physics Communications143(2002)1–10Computer simulation and modeling in railway applications✩T.K.Ho a,∗,B.H.Mao b,Z.Z.Yuan b,H.D.Liu b,Y.F.Fung aa Department of Electrical Engineering,Hong Kong Polytechnic University,Kowloon,Hong Kongb School of Trafﬁc and Transportation,Northern Jiaotong University,People’s Republic of ChinaAbstractAn electriﬁed railway system includes complex interconnections and interactions of several puter simulation is the only viable means for system evaluation and analysis.This paper discusses the difﬁculties and requirements of effective simulation models for this specialized industrial application;and the development of a general-purpose multi-train simulator.PACS:89.40.+k;02.60.Cb;89.20.+aKeywords:Computer simulation and modeling;Software development;Computer-aided engineering tool;Industrial applications1.IntroductionRailway is one of the most popular means of mass transportation.Every railway system is unique in terms of its technical speciﬁcations and available ser-vices as it is built in accordance with its own social de-mands,ﬁnancial resources,geographical restrictions and sometimes political considerations.An electriﬁed railway system is a complicated system with a num-ber of interacting sub-systems.To further perplex the design,the sub-systems are full of variety.For in-stances,DC motors with DC–DC chopper converter or inverter-fed three-phase induction motors can be used for traction drive;signaling system may apply the ﬁxed-block or moving-block concepts.Because of the high cost of building a new railway system or rede-veloping an existing one,it is necessary to assess all ✩Paper presented of the Conference on Computational Physics 2000,“New Challenges for the New Millenium”,Gold Coast, Queensland,Australia,December3–8,2000.*Corresponding author.E-mail address:eetkho@.hk(T.K.Ho).functions and features of the system in depth in order to justify the cost;to specify the system parameters, to identify possible hazards and safety issues in op-eration and to evaluate the options for improvement. Computer simulation is now conceived to be the most ﬂexible and cost-effective tool to serve these purposes.A complete railway system simulation must consist of a hierarchical structure consistent with the opera-tion requirements,system conﬁgurations and compo-nent design.In a railway simulator,each train has to satisfy the equations of motion,for which the traction characteristics,resistance to motion and line geome-try are considered.Train control requirements are usu-ally communicated to the trains by track-train data link and the trains interact through both the signaling and power supply systems.The diversity of the tasks to satisfy the above features imposes strict constraints on the development of whole-system simulators.The variations and characteristics of the major com-ponents of a railway system,which requires careful and substantial modeling in the simulation,will be dis-cussed here.A whole-system simulator has been de-2T.K.Ho et al./Computer Physics Communications143(2002)1–10veloped for general studies of multi-train operation in a railway system.It is designed to cater for various types of evaluation and analysis on railway systems of a wide range of speciﬁcations and operation condi-tions,but not tailor-made for one particular system or application.The models,structure and functions of the simulator will then be presented.2.Railway systemAn electriﬁed railway system is a closely-knitted in-tegration of a number of sub-systems which interact continuously with each other and inﬂuence the train performance directly.The major sub-systems are sig-naling,power supply and traction drives.They have their own speciﬁcations and variations in different railway systems to meet their corresponding require-ments.2.1.SignalingSignaling is concerned with the safety of train movement and regulation of trafﬁcﬂow.There are two basic principles in signaling,the block safety signaling theory to maintain a safe separation between two successive trains;and the interlocking of points and signals to protect trains at track stations and junctions. With the more commonly usedﬁxed-block signal-ing[1],the track is divided into blocks ofﬁxed length and the trains are detected occupying one or more blocks.A train is allowed to enter a block if that block is clear of other trains and if its speed is low enough for the train to stop before the entrance to an occupied block.In order to improve headway(train frequency), moving-block signaling has been introduced so that a hypothetical,instead of physical,block exists between two successive trains[2].The block length is deﬁned by the current speed and braking rate of the train be-hind.Train movement is regulated by the signaling system and hence train movement simulation relies on appropriate modeling of the signaling functions. Despite the relatively simple operation principles, the implementation of a signaling system is anything but simple.Because of the requirements and con-straints on different railway lines,such as speed re-strictions,line capacity,mixed trafﬁc and even cli-matic conditions,no two signaling systems are imple-mented in the same way.There are always some spe-cial features within the signaling system.For exam-ple,an overlap distance is introduced in the British Rail to provide extra protection to trains[3];point-machine defrosters have to be deployed in Sweden to keep the mechanical parts moving smoothly under se-vere weather etc.2.2.Power systemThe power supply in electriﬁed railway system is characterized by the unique requirements on power transmission,harmonics content and movement of the load,and electromagnetic compatibility in the vicinity. DC power supply is mainly used in metro systems where the lines are shorter,trains are lighter and train speed is lower.DC power is of course obtained from the utility AC supply through transformer/rectiﬁer substations.However,only the behaviour of the DC side imposes an inﬂuence on train movement.Hence, simulation can be conﬁned within the DC side,which is relatively simpler.On the other hand,AC sup-ply technology encounters problems of transmitting high level of power required by trains over long dis-tance and electromagnetic compatibility with the sur-roundings.As a result,catenary feeding systems with booster transformers or auto-transformers have been developed to improve transmission efﬁciency and sys-tem regulation and to reduce earth-current leakage and electromagnetic interference.Inevitably,the feeding system is complicated by the introduction of addi-tional overhead conductors[4].2.3.Traction equipmentRailway traction drives must allow forﬂexible speed control as the train speed varies over a wide range.DC traction motors meet this requirement easily with DC–DC chopper or AC rectiﬁer drives and hence they have been used in most railway systems.AC induction motors enable saving on maintenance and overhaul cost,operate at high maximum speeds and provide inherent regenerative braking capability.How-ever,speed control involves variable-voltage,variable-frequency input,which was not viable until the devel-opment of advanced high-rating thyristors a couple of decades ago.Hence,AC traction drives have become more popular in recent years[5].T.K.Ho et al./Computer Physics Communications143(2002)1–103As DC and AC traction drives operate on different principles and the supply can be in DC or AC,power electronics circuits of different conﬁgurations are re-quired to provide the necessary speed control.To at-tain a thorough model of a drive system,it is neces-sary to begin the analysis from the secondary side of the step-down transformer which is directly connected to the input of the power electronics circuits.3.Simulation models3.1.Train movementTrain movement is the calculation of the speed and distance proﬁles when a train is traveling from one point to another according to the limitations imposed by the signaling system and traction drive character-istics.As a train runs on the track,its movement is also under the constraints of track geometry and speed restrictions.The data structure representing the track geometry and signaling system modeling are there-fore critical to achieve effective simulation.Besides, the models to denote the evolution of train movement also play an important role in theﬂexibility of apply-ing the simulator on different studies.Asﬁxed-block signaling was solely used until a decade or so ago,data representation and storage are mostly in the shape of two-dimensional arrays with the rows of the arrays denoting the signaling blocks with ﬁxed block signaling.The track-based data include block identity,gradients,speed restrictions,coasting points and signal aspects etc.within the signaling block.The major advantage of this structure is easy referencing.However,only a single data type(e.g.,ﬂoating-point numbers)is allowed within the array, which is inﬂexible for accommodating the diversiﬁed nature of data.Besides,the array structure does not offer any representation of track layout,except that the adjacent rows in an array may depict a sequence of adjacent signaling blocks.When it is necessary to describe the track connections within a complicated railway network,the array structure must be enhanced or additional data structures are needed.Further,the size of the arrays required vary with applications, a number of‘supposedly’large enough arrays are usually deﬁned.In other words,excessive amount of memory space is reserved in the simulator in order to meet the demands of most applications.Object-oriented approach provides the solution for the above deﬁciency.It has been successfully applied in railway network and signaling modeling[6].The network structure is represented by a number of‘node’objects jointed by‘link’objects.The nodes can be sta-tions,junctions,points and termini whilst the links are the tracks connecting the above features and they are directional.They have their own data structures char-acterizing their properties,and the data structures ac-cept mixed data types.Fig.1illustrates how a section of track is represented by the nodes and links.Train movement is realized by moving the train from one node to the next through a permitted link,which con-tains the necessary information for the movement cal-culation.Similarly,trains and signals can be encap-sulated in their respectively deﬁned classes.Various types of trains and signals are further deﬁned by sub-classes through inheritance and the functions operat-ing on them are allowed to be re-used through poly-morphism.The object-oriented approach is particu-larly useful for the moving-block signaling because the signaling blocks do not exist physically and the modeling of continuous communication among trains (objects)can be made possible.Depending upon the level of details required,there are two major approaches to simulate train move-ment,time-based and event-based models.In time-based models,the time span is divided into evenly-spaced intervals and train movement is evaluated at each interval.This approach is conceptually analo-gous to how the trains move along the track in real-ity,hence it is easier to design.Despite its simplicity,a time-based model requires a highly computational de-mand as a signiﬁcant amount of information has to be produced during every update.This demand can only be justiﬁed in applications where full details of train movement are needed.Energy consumption and sig-naling layout design[7]are typical examples.Event-based models,on the other hand,express the progress of train movement by the occurrence of a sequence of pre-deﬁned events,such as arrival at and departure from stations[8].Since the events are linked together through the interactions among trains via signaling or power system,one event,as the consequence of a pre-vious one,will trigger another to happen.As a result,a chain of events reﬂects the progress of the -putational effort can be substantially reduced because the calculation of exact train movement between two4T.K.Ho et al./Computer Physics Communications 143(2002)1–10Fig.1.Node-link model of railway track.events is skipped.Nevertheless,this apparent advan-tage may be overshadowed by the fact that the updates of train movement are not carried out synchronously.It is possible that the processing of an event is postponed because the event to trigger it has not yet occurred,or it is processed with certain assumptions and it will be re-processed if the assumptions are found invalid later.3.2.Traction power supplyAn electriﬁed railway line is a huge electrical circuit with feeding substations as sources and trains as moving loads.The voltage seen by a train may vary with time and it then determines the traction performance,which in turn affects train movement.Thus,it is necessary to attain the voltages at certain nodes of this circuit at consecutive time intervals,which requires circuit analysis and the solution of a matrix equation.Efﬁcient algorithms are essential for the power network simulator to produce accurate results within reasonable computation time.DC supply systems may use overhead lines as sup-ply and rails as the return paths (a third rail is used as supply in some cases).Modeling is quite straight-forward as all the conductive components are safely assumed linear.The feeder substations are modeled as a voltage source (with nominal line voltage)in series with a source resistance (i.e.a V-R model).The only complication is to denote the capability of a substa-tion on power reception when regenerative braking is allowed,the V-R model is replaced by a simple resis-tive load when the current is proved to be ﬂowing in the opposite direction.Power supply for AC traction is obtained from the utility supply through traction feeder substations.Traction currents are often so contaminated by har-monic content that the current return-path has to be conditioned by return conductors,booster transform-ers or auto-transformers in order to avoid interfer-ence with communication systems and electrolytic corrosion.The complete conductor arrangement of the feeder system then becomes tedious.The V-R model isT.K.Ho et al./Computer Physics Communications143(2002)1–105still valid for the substations,but the feeder-conductors along the line carry different portions of the traction current when a train is moving in the vicinity.Care-ful modeling is required to take into account of the electromagnetic interference effects on different har-monics of the power supply,as well as other frequen-cies used in the signaling system.The physical sizes, relative resistivity,permeability and geometry of the conductors,relative permittivity of the medium and earthing conditions are essential for the feeder system model.Further,as the rail is of unique shape,calcu-lation of impedance is far from easy.It can be treated as a single cylindrical steel conductor with different cross-section areas for a wide range of frequencies[9].3.3.Traction drivesThe basic function of a traction equipment model is to provide traction effort output and current/power de-mands according to the given input parameters for the train movement and power network calculations.The DC supply voltage with respect to the rolling stock can vary from−30%to+20%of the nominal value.A voltage sensitive drive model is,therefore,essential in achieving accurate electrical and mechanical rep-resentations for the conventional DC traction equip-ment.However,for the modern three-phase induction motor drive,the voltageﬂuctuation on the train pan-tograph(or collecting shoe)is less signiﬁcant with the advanced pre-conditioning front–end technology.The voltage at the DC link can remain at a fairly constant level.Two voltage-sensitive drive-modeling approaches, namely,the detailed[10]and simpliﬁed[11],have been widely used.When the information of motor terminal characteristics,winding resistance and reac-tance is available,it is often desirable to model a drive with the detailed approach.However,at a feasibility study or preliminary engineering stage,it is not al-ways possible to gather all the necessary information to model a new drive comprehensively.The simpliﬁed drive modeling approach based on dataﬁtting and nu-merical techniques provides a much more practical al-ternative as it only requires the high level information, such as traction effort vs.train speed curves,which are generally much easier to obtain.To integrate into the same electrical circuit with the power system,the traction systems are also rep-resented by the V-R model.However,the values of V and R vary with the speed and operational mode of the train.Either the detailed or simpliﬁed methods can therefore be employed to synthesize,usually off-line, the look-up tables of V and R over the whole range of operation conditions.3.4.Power network solutionWith the models of power supply and traction equipment available,the circuit of the electriﬁed rail-way line is established for the calculation of voltages and currents at various points of interest.Because of the size of the circuit,the power network solution is the most time-consuming step of railway simulation. The V-R models for substations and traction equip-ment are thus crucial to the simpliﬁed structure of the circuit,which in turn has an impact on the computa-tional demand and hence the simulation speed. There are two major approaches to solve the power network.In theﬁrst approach,loadﬂow calcula-tion[12],the power network simulator is quite of-ten separated from the train movement simulator.The traction power network simulator,as a stand-alone module,takes in the train movement results,such as train locations and train power demands etc.from data ﬁles or intermediate storage.This approach,from a programming viewpoint,provides a much easier inter-face between the train movement and traction power simulators.It does,however,provide no direct reﬂec-tions of voltage variations back to the train movement calculations.With modern three-phase drives,the trac-tion drive is less dependent on supply voltage than is the case with DC motors,but all drives have a designed ‘graceful degradation’response to reduced traction voltage.Thus,if the performance limits of the total system are to be properly examined,feedback of the power network solution to the traction performance calculations becomes essential.For the second approach,the direct matrix method [13],a matrix equation is formulated from the elec-trical circuit using either mesh or nodal analysis.The latter is more suitable complex networks,because it is often easier to identify nodes than loops in non-planer networks.From a programming perspective, automatic network set-up procedures and advanced network graph techniques are easier to implement with the use of the nodal approach.Fig.2shows the circuit6T.K.Ho et al./Computer Physics Communications 143(2002)1–10Fig.2.Simpliﬁed power network with branches.representation of a typical DC traction power network with branches.It is not difﬁcult to see that the railway traction power network is characterized by sections of ladder type networks infrequently cross-connected.For this characteristic topology,the sparse matrix tech-nique coupled with efﬁcient matrix elimination meth-ods leads to an expeditious network solution,since it does not suffer from the ﬁll-ins that the direct inver-sion of the coefﬁcient matrix requires.The coefﬁcient matrix of the traction power network equation is usually of positive deﬁnite (PD),sym-metric and sparse type.Many sparse matrix elimina-tion techniques tailored for PD matrices,such as LL T decomposition and Cholesky decomposition,are ap-plicable.For the sparse techniques,the essence is the equation-ordering.There are generally two types of or-dering method namely,static and dynamic ordering.Typical examples of static ordering for long and thin ladder type networks include the Cuthill and McKee algorithm [14]and reverse Cuthill and McKee algo-rithm [15],which make use of the property that the zero elements situated before the ﬁrst non-zero ele-ment on any row always remain zero.For dynamic or-dering,the minimum degree algorithm provides an ef-ﬁcient alternative for solving the network matrix.Thisis actually a heuristic algorithm for ﬁnding an order-ing for the coefﬁcient matrix which suffers low ﬁll-ins when it is factored.Therefore,this scheme requires a simulation of the effects on the accumulation of non-zeros of the elimination process.In order to avoid di-rect elimination with actual values,a symbolic factor-ization is usually adopted to obtain the zero and non-zero structures of the factored matrix.As the numeri-cal values of the matrix components are of no signiﬁ-cance in this connection,the problem could be studied using a graph approach,instead of using an actual ma-trix factorization.The minimum degree algorithm is particularly suitable for solving medium to large net-works,in which there are 200nodes and more.For smaller networks,the minimum degree algorithm be-comes a less efﬁcient option,since a signiﬁcant por-tion of the overall processing time is used in the sym-bolic factorization process.4.General-purpose railway system simulator Most of the existing railway simulators were devel-oped speciﬁcally to solve one particular problem only or to study the performance of a particular part of theT.K.Ho et al./Computer Physics Communications 143(2002)1–107system.Different applications lead to different speci-ﬁcations of simulators.For a railway research group,it is impractical to develop a number of simulators to carry out various studies.A general-purpose multi-train simulator suite is therefore in great demand to deal with studies of various kinds and levels.This sec-tion brieﬂy reviews the development of a multi-train simulator based on the models described,and its ap-plications.4.1.StructureThe simulator consists of a library of software mod-ules modeling the features and variations of the sub-systems within a railway system.To cater for a par-ticular application,appropriate modules are selected from the library and bound together to attain the sim-ulator required.This simulator suite contains a graphical input front–end,a number of categories of functions rep-resenting the sub-systems of a railway system and a common data-structure shared among the categories.Under the heading of each category,there are modules modeling functions of the sub-systems.To enhance the applicability of the simulation suite,different concepts and approaches of modeling the sub-systems can be incorporated.For example,in the signaling category,there are modules for both ﬁxed and moving-block signaling schemes.In the power-system category,dif-ferent modules are certainly needed for AC and DC railways.Such ﬂexibility however requires intelligent supervision to ensure that the selections of modules within one sub-system category are compatible with those for others.As illustrated in Fig.3,the vast amount of data re-quired for simulator is supplied through the input in-terface.It allows easy input with dialogue boxes and,in some cases,graphical input with typical mouse ac-tions.Data integrity check is also incorporated to pre-vent the simulator from operating with foul data.A set of databases is necessary to systematically store data of different purposes.Function modules of different features and variations of the sub-systems are available within the simulator.They are deﬁned anddevelopedFig.3.Structure of the general-purpose railway system simulator.8T.K.Ho et al./Computer Physics Communications 143(2002)1–10Fig.4.An input interface.independently so that they can be inserted,deleted or modiﬁed without affecting others.The core of the sim-ulator is a Simulator Manager which selects,according to the application speciﬁcations,the appropriate func-tion modules and arranges the necessary initialization for the simulation.Modules of the same function but different approaches should have the same protocol in order to maintain consistent interfaces with other function modules.The modularity is essential for the adaptability of this simulation suite.4.2.Progress and further developmentThe development of the simulator has been divided into a number of milestones and the latest version contains the train movement under ﬁxed-block and moving block signaling and their minor features,DC power supply systems and DC motors with chopper drives.The input interface now allows fully interactive data acquisition and the tedious data input process can be divided into a number of stages and stored in separate application-oriented ﬁles.One of the input interfaces is given in Fig.4.The simulation results may be presented in graphical,textual and AutoCAD formats.The graphical output can be displayed as the simulation proceeds,an example is shown in Fig.5.To save computational effort and to reduce simulation time,the real-time graphical output can be turned off and the analysis of output data is carried out upon the completion of simulation.AC power supply network is now being imple-mented.The complexity of the model,which gener-ally involves equations of very large matrices (typi-cally,200×200to 1000×1000)of complex numbers,T.K.Ho et al./Computer Physics Communications143(2002)1–109Fig.5.Train speed proﬁles(simulation results).leads to such a computationally demanding solution process that parallel computing by Streaming SIMD Extensions(SSE)[16]on a single processor or par-allel algorithms on multi-processor platform has been contemplated.4.3.ApplicationsBecause of its generic nature,the simulation suite can be applied in literally most of the studies related to railway systems.To date,it has been used in the stud-ies of the trafﬁc control at conﬂict areas;energy con-sumption for trains re-starting under moving block sig-naling;train control optimization at inter-station runs and train service scheduling in both Hong Kong and China.The list will certainly go on when the cur-rent development is completed and all the modules are fully functional.This simulator is a pragmatic and vi-able tool for every railway operator who is responsible for the daily operation,disturbance recovery and risk assessment and management.5.ConclusionsRailway system is a specialized multi-disciplinary engineering application.Its unique characteristics,re-quirements and social impact on transportation call for the imminent needs of a reliable,accurate andﬂexible tool to assess system performance in various aspects. This paper reviews the difﬁculties of the model devel-opment for electriﬁed railway simulator.Variations of railway system functions and their implications on ap-propriate modeling have been brieﬂy discussed.The10T.K.Ho et al./Computer Physics Communications143(2002)1–10accuracy and resolution of simulation results,com-putational time,ﬂexibility for use,applicability and transportability of the simulator are largely determined by the models of various components of the railway system.Since a simulator is often employed when a railway line is still at the design stage,the simula-tion results cannot be veriﬁed by actual measurements. Hence,the nearly over-cautious concern on adequate and efﬁcient modeling within the simulator is never an overstatement from simulator developers.A general-purpose multi-train simulator has been developed with the appropriate modeling.It provides a generic computer-aided engineering tool for the railway operators and researchers to carry out studies without any modiﬁcations on the program-code level. It enables the users to focus on the concerns of the studies but without losing the realism of a total system model and thus preserving knowledge of the system implication.AcknowledgementThis work has been supported by the Research Committee of the Hong Kong Polytechnic University. References[1]R.J.Hill,Electric railway traction,Part4,Signalling andinterlocking,Power Engrg.J.(1995)201–206.[2]M.J.Dockyear,Changing track—moving-block railwaysignalling,IEE Rev.(1996)21–25.[3]O.S.Nock,Railway signalling,A&C Black Publishers,1980.[4]R.J.Hill,Electric railway traction,Part3,Traction powersupplies,Power Engrg.J.(1994)275–286.[5]R.J.Kemp,Developments in electric traction,Power Engrg.J.(1989)71–82.[6]L.K.Siu, C.J.Goodman,An object-oriented concept forsimulation of railway signalling and train movements,in: COMPRAIL’92,V ol.2,1992,pp.545–556.[7]D.C.Gill,C.J.Goodman,Computer-based optimisation tech-niques for mass transit railway signalling design,IEE Proc.B139(3)(1992)261–275.[8]T.K.Ho,C.J.Goodman,J.P.Norton,An event-based trafﬁcﬂow model for trafﬁc control at railway junctions,in:As-pect’95,International Conference on Advanced Railway Con-trol,1995,pp.163–170.[9]J.C.Brown,J.Allan,B.Mellitt,Calculation and measurementof rail impedence applicable to remote short circuit fault currents,IEE Proc.B139(4)(1992)295–302.[10]C.J.Goodman,B.Mellitt,N.B.Rambukwella,CAE for theelectrical design of urban rail transit systems,in:COM-PRAIL’87,1987,pp.173–193.[11]J.G.Yu,R.W.Struand,L.R.Denning,A general motormodelling method for transit system simulation studies,in: COMPRAIL’94,1994.[12]S.N.Talukdar,R.L.Koo,The analysis of electriﬁed groundtransportation network,IEEE Trans.Power Appl.Sys-tems76(1)(1977)240–247.[13]B.Mellitt,C.J.Goodman,R.I.M.Arthurton,Simulator forstudying operational and power-supply conditions in rapid-transit railways,Proc.IEE125(4)(1978)298–303.[14]E.Cuthill,J.McKee,Reducing the bandwidth of sparsesymmetric matrices,in:Proc.24th put.Mach,ACM Publishers,1969,pp.157–172.[15]A.George,J.W.H.Liu,Computer Solution of Large SparsePositive Deﬁnite Systems,Prentice-Hall,Englewood Cliffs, 1981.[16]G.Conte,S.Tommesani,F.Zanichelli,The long and windingroad to high-performance image processing with MMX/SSE, in:Proc.5th IEEE International Workshop on Computer Architectures for Machine Perception,2000,pp.302–310.。

internal training management procedure 内部培训管理规定

Approved by:Purpose:To develop appropriate in-service curricula based on the identified training needs using departmental resources and guide the process management of internal training.目的：利用部门资源开发适当的在职课程；引导内部培训的规范管理．1.Approaches to internal training内部培训的方式Approaches to internal training include apprentice training,vestibule training,on-the-job training, off-the job training.内部培训包括学徒培训、岗前培训、在职培训、职外培训．1.1The apprentice commits to a period of training and learning that involves both formal classroomlearning and practical on-the-job experience.During this period,the pay is less than that for the master workers.学徒培训包括课堂学习和现场实践．在此期间的薪水低于熟练工人的．1.2In vestibule training,the trainee learns the job in an environment that simulates the real workingenvironment.A machine operator trainee might run a machine under the supervision of a trainer until he/she learns how to use it properly.Only then is the trainee sent to the shop floor.在岗前培训中,参训者在模拟真实的环境下，在培训者的监督下学习机器的操作直到能完全独立地使用,然后才被安排到现场工作．1.3On-the-job training means the employee is placed into the real work situation and shown the job andthe tricks of the trade by an experience employee or the supervisor.在职培训安排在工作现场,由主管或经验丰富的员工来示范工作技巧．1.4Off-the-job training includes lecture-discussion,programmed instruction,and computer-assistedinstruction.Lecture-discussion means that a trainer gives a lecture and involves the trainee in a discussion of the material to be learned.Frequently,these lectures are supplemented with audiovisual aids,videotapes or audiotapes.Programmed instruction is the technique for instructing without the presence or intervention of a human instructor.Material can be presented on teaching machines or in text form.职外培训包括课堂教学、程序教学与计算机辅助教学．课堂教学是由培训者进行讲授，参训者进行问题讨论的一种经常使用的学习方式,这种方式还可借助视听设备、录像带或录音磁带来进行．程序教学是在没有讲师的环境下进行学习的方式,学习资料通过教学设备或资料表格方式呈现给参训人员．2.Process of internal training内部培训程序：2.1Determine training needs and set objectives for these needs and get approval of the departmentmanager;确定培训需求,设定培训目标并得到部门经理的确认．2.2Choose trainer,trainees and training time and get joint approval of the department manager and HR&Admin manager;确定培训讲师、培训对象和培训时间后并得到部门经理与人事行政经理的确认．2.3Develop training material and get approval by the department manager;开发培训教材并到部门经理的确认．2.4Plan and implement the training program,coordinated by HR&Admin manager;Approved by:人事行政部经理统筹培训计划的规划与实施．2.5Evaluate the training;评估培训效果．2.6File the training records.将培训资料归档保存．3.The guideline of on-the-job training is as bellows:在职培训政策如下：3.1Decide what the learner must be taught in order to do the job efficiently,safely,economically,andintelligently.确定培训对象的学习内容以使之能高效、安全、经济、聪慧的工作．3.2Preparation of the trainer:put the trainer at ease;find out what he/she already knows about the job;getthe learner interested and desirous of learning the job;讲师准备:确认培训对象已经知道的工作内容，了解参训者感兴趣的和希望了解的内容．3.3Presentation of the operations and knowledge:tell,show,illustrate,and question in order to put overthe knowledge and operations;instruct slowly,clearly,completely,and patiently,one point at a time;check,question and repeat;make sure the learner really knows;理论与实践知识的讲授：通过宣讲、演示、案例与提问使培训对象能掌握理论知识和实践操作．每次讲授一个知识点，在讲授时要缓慢、清晰、完整和有耐心的．并且通过检查、提问与反复而使培训对象能真正的学到所学的．3.4Performance tryout:test the learner by having him/her perform the job;ask questions beginning withwhy,how,when,or where;observe performance,correct errors,and repeat instructions if necessary;continue until you know that the learner knows;成效评估：可以通过参训者现场实践来测试；可以通过提问的方式(为什么，何种方式，何时或何地)来测验；如果有必要也可以通过观察，纠正错误和重复教学的方式来测试．通过这些测试方式你可以知道培训对象的学习状况．3.5Follow-up:put the employee on his or her own;check frequently to be sure the learner followsinstructions;taper off extra supervision and close follow-up until the person is qualified to work with normal supervision.跟进：参训者重回工作岗位后，要经常的去检查他是否遵循教导行事；逐渐减少监督；密切跟进直到参训者能按照标准工作．4.The guideline of training needs analysis is as bellows:培训需求分析政策如下：4.1The training needs may be identified through analyzing the company’s goals and objectives;performance evaluation;quality assurance processes;the knowledge,skill,and ability needed to perform the job;and the person or job holder’s needs;培训需求必须通过分析公司的目标、绩效评估、质量方针、知识、技能、能力和个人需要等因素确定．4.2If the performance analysis indicates that behaviors need to be altered,training becomes a viableApproved by:consideration.Any training approaches available should be weighed and considered to find the one best suited to correct the behavior discrepancy;如果绩效分析表明行为方式需要改变，培训可以成为可行的选择方式．任何的培训方式都要以最能纠正行为差异为考量标准．4.3The following factors might indicate training needs:new employees;performance’s,production’s,orsafety’s problems;new technology,equipment or programs;managerial competency assessments;new or existing legal requirements；audit or inspection deficiencies；modernization of equipments;下列的因素也能挖掘培训需求：新入职员工；绩效；生产或安全问题；新技术、新装备或新工艺；管理水平；现行法规；稽核或检查出的不足；装备的现代化程度等因素．5.Internal lecture-discussion procedure内部课堂教学程序：5.1Give notice:internal lecture-discussion notice should be given at least one day in advance in writtenform or verbally by HR department;cancel or postpone notice should be sent in time by HR department.通知：内部课堂教学由人事行政部至少在培训前提前1天以书面或口头的方式通知；取消或推迟培训的由人事行政部及时知会相关人员．5.2Prepare classroom,training equipments,teaching material and other documents;准备教室、培训器械、教材和其它资料．5.3Trainees should arrive at the classroom on time and sign in the attendee list;参训者要准时到制定地点培训并在培训签到表上签名．5.4Evaluate the trainees’score,file the scores and give feedback to the department manager the traineereporting to.评估培训效果，整理归档并且将培训效果反馈给部门经理．5.5Trainee should follow the discipline that leave should be approved in advance;being late,leavingearly and absenteeism are forbidden.参训者有事不能参加的需要提前请假，不允许迟到、早退和旷课．。

1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。

Modeling Train Movements Through Complex Rail NetworksQUAN LU and MAGED DESSOUKYUniversity of Southern CaliforniaandROBERT C.LEACHMANUniversity of California,BerkeleyTrains operating in densely populated metropolitan areas typically encounter complex trackage conﬁgurations.To make optimal use of the available rail capacity,some portions of the rail network may consist of single-track lines while other locations may consist of double-or triple-track lines. Because of varying local conditions,different points in the rail network may have different speed limits.We formulate a graphical technique for modeling such complex rail networks;and we use this technique to develop a deadlock-free algorithm for dispatching each train to its destination with nearly minimal travel time while(a)abiding by the speed limits at each point on each train’s route, and(b)maintaining adequate headways between trains.We implemented this train-dispatching algorithm in a simulation model of the movements of passenger and freight trains in Los Angeles County,and we validated the simulation as yielding an adequate approximation to the current system performance.Categories and Subject Descriptors:I.6.3[Simulation and Modeling]:Applications;I.6.5 [Simulation and Modeling]:Modeling Development—Modeling methodologiesGeneral Terms:Algorithms,PerformanceAdditional Key Words and Phrases:Trains,modeling,dispatching,deadlock1.INTRODUCTIONWorldwide container trade is growing at a9.5%annual rate,and the U.S.rate is around6%(Vickerman[1998]).For example,the Ports of Los Angeles and Long Beach(San Pedro Bay Ports)are anticipated to double and possibly triple their cargo by2020.The growth in the number of containers has already introduced congestion and threatened the accessibility and capacity of the rail network system in the Los Angeles area.The research reported in this paper was partially supported by the National Science Foundation under grant DMI-0307500.Authors’addresses:Q.Lu and M.Dessouky,Department of Industrial and Systems Engineer-ing,University of Southern California,Los Angeles,CA90089-0193;email:qlu@, maged@;R.C.Leachman,Department of Industrial and Operations Research,University of California,Berkeley,Berkeley,CA94720;email:leachman@.Permission to make digital/hard copy of part or all of this work for personal or classroom use is granted without fee provided that the copies are not made or distributed for proﬁt or commercial advantage,the copyright notice,the title of the publication,and its date appear,and notice is given that copying is by permission of ACM,Inc.To copy otherwise,to republish,to post on servers,or to redistribute to lists requires prior speciﬁc permision and/or a fee.C 2004ACM1049-3301/04/0100-0048$5.00ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004,Pages48–75.Modeling T rain Movements•49 To partially address this rail trafﬁc growth,the Alameda Corridor is being developed in Los Angeles County.The Alameda Corridor is a high-speed double-track line from the Ports of Long Beach and Los Angeles to Downtown Los Angeles(Leachman[1991]).However,the Alameda Corridor does not address the transcontinental rail trafﬁc from Downtown Los Angeles to the Eastern Inland area.The rail network in this area is rather complex.Some high-trafﬁc portions of the rail network in this area consist of double-or triple-track lines.The control logic of rail networks that consist of multiple trackage can be very complicated in order to avoid potential train deadlock movements.For example,a rail net-work consisting strictly of double-track lines is simple to control since it is similar to vehicle trafﬁc movement.That is,a control logic of keeping the traf-ﬁc moving only on the right-hand side ensures there will be no deadlocks.No similar simple rules exist when there is a mix of trackage conﬁgurations in the rail network.In fact,for a general trackage network,to determine the optimal dispatch times that minimize total train delays is an NP-hard problem,because it can be reduced to the deadlock avoidance problem for sequential resource al-location,which has been proven to be NP-hard by Lawley and Reveliotis[2001].Another complicating factor of rail networks in metropolitan areas is the existence of multiple speed limits at different points in the network because of physical contours,crossovers,or other safety considerations.In this case,the issue is to determine the fastest speed the train can travel at each instant of time without violating any speed limit considering the train’s acceleration and deceleration rates.If the acceleration and deceleration rates are assumed to be inﬁnite,the speed of the train at each instant of time is simply set to the constraining speed limit.When they areﬁnite,we show that the fastest speed the train can travel at each instant of time is a complex combination of the acceleration and deceleration rates and the uniform motion of the train.In order to analyze the capacity of the rail network in the area from Down-town Los Angeles to the Eastern Inland,we developed a deadlock-free sim-ulation methodology to model rail networks that consist of a multiple track-age conﬁgurations and speed limits.The purpose of this paper is to present this methodology.Our primary contributions are the development of a network framework that describes the trains and their travel paths,the development of an algorithm that determines the fastest a train can travel while observ-ing multiple speed limits,and the development of a deadlock-free dispatching heuristic.There has been some prior work in simulation modeling of rail networks. Dessouky and Leachman[1995]developed a simulation modeling methodology for either strictly single-track or double-track rail networks consisting of a sin-gle speed limit without considering deceleration rates.This prior model was suitable for simulating the Alameda Corridor,which is a high-speed double-track network.Since there was a single speed limit and a small number of locations for potential train conﬂicts(e.g.,crossover junctions),there was no need to take into account the train deceleration process.Petersen and Taylor [1982]presented a structured model for rail line simulation.They divided the line into track segments representing the stretches of track between adjacent ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.50•Q.Lu et al.switches and developed algebraic relationships to represent the model logic. Higgins and Kozan[1998]proposed an analytical model to quantify the posi-tive delay for individual passenger trains on the track links in an urban rail network.They also developed a schedule time-driven simulation model with ﬁxed routing to validate their analytical model.Cheng[1998]applied a hybrid approach consisting of a network-based simulation and an event-driven simu-lation model of rail networks with tracks dedicated to the trains running in the same direction.Carey[1994]developed a mathematical model for routing trains under differ-ent speed limits.He assumed that the departure schedule of the trains from the initial station is deterministic.Komaya[1991]used a knowledge-based model to simulate train movements for large and complicated rail systems.Lewellen and Tumay[1998]described a model used to simulate train movements for Union Paciﬁc.However,the authors did not present any details of the model logic.The modeling methodology presented in this paper differs from the previous work by considering a multiple trackage conﬁgurations in the same rail network with multiple speed limits while taking into account the trains’acceleration and deceleration rates.Furthermore,we do not assume that the initial departure schedule of the trains is known.We model it as a stochastic process for freight trains and as aﬁxed schedule for passenger trains.Imbedded in our modeling methodology is a central dispatching algorithm that decides the movement of each train in the network considering whether to continue moving at the same speed,to accelerate or decelerate,or to stop. The algorithm also selects the next track to seize among multiple alternative tracks.Our central dispatching heuristic guarantees that no deadlock occurs while attempting to keep the train delays to a minimum.Even though the pro-posed simulation modeling methodology was applied to the rail network from Downtown Los Angeles to the Eastern Inland area,it can be applied to various situations to simulate rail networks with any kinds of topology,crossovers,and speed limits.The rest of the paper is organized as follows.Section2gives the details of the system model and the train movement control logic.Section3formulates and solves the problem of determining the fastest speed a train can travel under multiple speed limits while considering the train’s acceleration and decelera-tion rates.Section4describes the issues of deadlock avoidance and efﬁcient routing.Finally,in Section5we present the overall integrated model and show its effectiveness in modeling train movement in Los Angeles County.2.DESCRIPTION OF SYSTEM MODELThis section presents a simulation modeling methodology used to analyze generic complex rail networks.The modeling methodology does not depend on the size of the rail network and is insensitive to the trackage conﬁguration. That is,it can be used to model rail networks consisting of single-track lines, double-track lines,or any number of tracks.Note that trains operate quite differently depending on the number of tracks.For example,in a single-track ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.Modeling T rain Movements•51 system,trains moving in opposite directions compete for the same track and there must exist sufﬁcient buffer trackage for trains to wait for opposing direc-tion trains to move in order to prevent potential deadlocks.In a double-track system,the common routing logic is to dedicate each track for trains moving in only one direction.A triple-track system exhibits features similar to both single-track and double-track systems.We willﬁrst discuss how to construct the simulation network from the ac-tual physical system.Then we will discuss the train movement process in our simulation network.Although the train movement process is a continuous pro-cess,our modeling approach is a discrete-event methodology.We approximate the continuous motion of train movement by dividing the movement into small discrete steps.A continuous simulation modeling approach would require the development of motion equations that represent train movement,which are dif-ﬁcult to represent mathematically in the context of train conﬂicts and deadlock-free movement.2.1Simulation Network ConstructionThe physical resources that we model are:rail junctions and track segments.A rail junction is typically used for train crossover movement in a rail network. One idea behind the modeling approach is to divide the physical track into seg-ments,as in Dessouky and Leachman[1995].A segment is the minimum unit in the simulation model and each segment is represented as a unique resource with capacity one.Junctions are also represented as a resource with capacity one in our simulation model.A track segment has the following characteristics:(1)Travel in each segment is restricted by one speed limit.(2)No junction exists within the segment.Junctions can only be located at thebeginning or end of a segment.(3)The length of the segment is no longer than the maximum train length.Theﬁrst two characteristics are not restrictive since there is no limit on the minimum length of the segment.Hence,the deﬁnition of the segment is sufﬁ-ciently generic to model any physical trackage conﬁguration.However,having many small track segments will increase the number of resources in the simu-lation model and the computational run time of the model.On the other hand, since we restrict the capacity of each segment to be one,too large of a segment deﬁnition will increase the headway between trains,needlessly decreasing the capacity of the network.Thus,the third characteristic restricts the maximum size of the segment to be the maximum train length.The above three charac-teristics can also be considered as a rule to divide the segments.We refer it as the segment divide rule.As an illustration of the segment deﬁnition,Figure1diagrams a small por-tion of a rail network near Downtown Los Angeles.There are six junctions in the network at points B,F,H,I,K,and L.Note that junctions F and K consist of a double-direction crossover.In Figure1,we represent this type of crossover with two cross connections.Junctions K and L consist of a single-direction crossover. In Figure1,we represent this type of crossover with a one cross connection.ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.52•Q.Lu et al.Fig.1.A sample rail network.The speed limit at all these junctions is assumed to be20miles/hour.Assuming the maximum train length is1.5miles,we deﬁne eleven track segments listed in Figure1.Each segment has a single speed limit and is a resource.Only one train can be in a track segment at each instant of time.Furthermore,a train at segment E-F crossing over the track to move to segment K-M will seize junction resources F and K,thus blocking all the trafﬁc that needs junction resources F and K(e.g.,trafﬁc from segment F-G to segment I-K).We next translate the segment deﬁnition of the physical system into a net-work architecture.Each node in our network deﬁnes a combination of one or more contiguous segments.Each node has two ports:port0and port1.Port0 indicates the starting point of travel for a train moving in the node from one direction.Port1indicates the starting point of travel in the opposite direction of port0.Two distances locate each segment in a node:(1)the distance from the end of the segment to port1of this node,and(2)the length of the segment itself.Note that the length of the node equals the sum of all the segments’lengths within this node.The nodes are connected by arcs,which represent movement from one node to another.Arcs may include junctions or not.All the arcs in our network are undirected and have zero length.Therefore,the total travel distance of a train in the network equals the sum of the lengths of the nodes it visits.As we will describe in detail in Section2.2,in order to move to a successor node,all the resources associated with the track segments of the successor node,as well as any connecting junction resources,must be available.Since the capacities of the track segment and junction resources are unity,only one train can be at ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.Modeling T rain Movements•53work architecture for the sample rail network.a node at any instant of time.Figure2illustrates the network concept for the example given in Figure1.In each node,the two numbers within the parentheses next to the segment’s name denote the distance from the end of this segment to port1of this node, and the segment’s length,respectively.For instance,node8contains two track segments,D-E and E-F,from port0to port1.Segment D-E is1mile long.Since segment E-F is between segment D-E and port1of node8,the length from the end of segment D-E to port1of node8equals the length of segment E-F,0.5 miles.Similarly,segment E-F is0.5miles long and,because there are no other segments between it and port1,the distance from the end of segment E-F to port1of node8is0.One characteristic of our network deﬁnition is the existence of multiple paths for each origin and destination node pair.For instance,a train starting from point H and ending at the station has two alternatives in Figure1.Alternative1 consists of moving to node2,then to node6,and then to node11(i.e.,H-I to I-K and crossover at junctions K and L to the station).Alternative2consists of moving to node3,then to node7,and then to node11(i.e.,H-J to J-L and to the station).Besides the different speed limits,the total distances of these two paths are also different.Our network structure allows forﬂexible routing once the train enters the network.That is,trains with the same origin and destination may run through different paths according to the routing logic discussed in Section4.ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.54•Q.Lu et al.Arcs connect nodes.Some arcs include junction resources.These arc-included junctions are necessary since they represent the junction resources the train needs in order to travel from one node to another.For instance,a train that is trying to enter the station from node8needs to cross at junctions F,K, and L.In the network,the train moving from port1of node8,crossing the arc connected between node8and node11,andﬁnally entering node11,represents this motion.The resources that need to be seized in this movement in sequence are junction F,junction K,junction L,and the station.Sometimes,we also separate segments in different nodes and connect them with an arc containing no junction resources.Including both of these two segments in a node may result in an unnecessarily high spacing between trains,since only one train can occupy a node at a time.For example,segment H-J(node3)is connected to segment J-L(node7)by an arc including no junction resource.The speed limits for the segments are applied to all trains traveling across these segments.Unlike the speed limit for the segment,the speed limit for the junction is only used to restrict the trains that cross over at the junction.If a train only bypasses the junction,the speed limit for the junction will not be applied.For instance,a train moving from segment B-I to segment I-K needs to reduce its speed to less than20miles/hour while passing through junction I.But a train moving from segment H-I to segment I-K does not need to be restricted by junction I’s speed limit.To distinguish this difference,in our network we use an“*”following a junction name to denote that this junc-tion is needed for the train passing it,but no junction speed limit needs to be applied.Figure2shows the advantages of deﬁning the ports for a node.First,they indicate the direction of train movement when it enters a node.Second,in reality,a train in segment H-I cannot turn around and move to segment B-I at junction I.But from the network point of view,a train could move from node2 to node5through node6,since node2,node5,and node6are all connected. The deﬁnition of ports solves this conﬂict.We add a restriction that if a train enters a node from port0(respectively,port1),then the train must leave from port1(respectively,port0).2.2Train Movement ProcessThe entities in the simulation model are the trains.To support the ability of ﬂexibly routing train entities,we deﬁne a routing table at each node that stores a sorted node list for each route,which lists all the successor nodes that have at least one path to the train’s destination node.The successor nodes in a list are sorted in ascending order of the minimum travel time from the current node to the destination node.The minimum travel time is computed assuming there is no downstream conﬂicting trafﬁc ahead of the current train.Note that the shortest path in terms of travel distance may not be the fastest path in terms of travel time,because of the possible existence of different speed limits for the different paths.The routing tables at nodes are set automatically before the simulation starts and remain unchanged through the entire simulation. Note that,in practice,the routing tables have to be updated in real time to ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.Modeling T rain Movements•55 reﬂect congestion,accidents,and mechanical malfunctions that may render some segments in the network unusable.A successor node is considered available if the following two conditions are met:(1)All track segments’resources of the successor node must be currently avail-able as well as any possible connecting junction resources.(2)The movement of a train to this successor node must not create a deadlock.When a train enters a node,a routing algorithm is applied to check the availability of each successor node and locate the list of the available successor nodes.Then,one of the available successor nodes will be selected for the train’s next move based on some heuristic criteria.The complete details of the routing algorithm and successor node selection criteria are presented in Section4.If no successor node is currently available,the train will have to stop to wait for an available node.In this paper,we assume that each train has a constant acceleration rate a1and deceleration rate a2.Note that different trains may have different a1 and a2.We cannot wait until the head of the train reaches the end of the node to make a decision on whether a train should stop or not since we need to account for decelerating time,and some distance is necessary for a running train to fully stop.A routing algorithm(see the details in Section4)is applied before the head of the train reaches the end of each node it passes.The point of applying the routing algorithm to decide whether a train should stop or not is referred to as the stop checking point(SCP)of the node.Let S i be the distance from the SCP of node i to the end of node i and V i be the train’s velocity at the SCP of node i. The relationship between S i and V i is(Bueche et.al.[1997])V i=2a2S i(2.1)Figure3illustrates the concept of a stop checking point.Suppose a train is moving from node8to node10with the head of the train currently at position E in node8.The velocity of the train at position E is20miles/hour.The SCP at node10is set to be the intersection of the two dotted lines.Each point on the thick dotted line represents the fastest speed the train can travel at that location,while the train is moving from node8to node10without stopping. Each point on the thin dotted line represents the fastest speed the train can travel if the train needs to fully stop at the end of node10(point M).Note that the train has to adhere to a speed limit of20miles/hour even after its head has passed point K and is in the zone with a40miles/hour speed limit,because the maximum speed of the train at each instant of time is always restricted by the minimum speed limit of all the tracks it is occupying.In this case,the speed limit at segment E-F still restricts the train before its tail passes point F(its head arrives at point P).We next deﬁne the different types of events used in our simulation system, the logic of the train movement process,and the method for seizing and freeing the resources within the nodes(track segments)and arcs(junctions).Four types ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.56•Q.Lu et al.Fig.3.Stop checking point example.of events are deﬁned:(1)arrival of a train to a SCP,(2)freeing of a resource,(3)a train coming to a full stop,and(4)arrival of a train to a terminal station.If train m’s head reaches the SCP of node i,the arrival SCP event is invoked. Then the central dispatching algorithm is called to determine train m’s next movement.If the central dispatching algorithmﬁnds that there exists one or more available successor nodes for train m to move forward from the current node i,the algorithm selects one of them,let it be j,and commands train m to move forward to node j.(Details on the selection criteria if more than one successor node is available are provided in Section4.)Then,train m performs the following operations in sequence:(1)Let train m seize all the resources belonging to node j as well the arcconnecting node i and node j.(2)Determine S j and V j of the SCP of node j.(3)Calculate the minimum travel time t subject to the speed-limit constraints,while train m’s head moves from the SCP of node i to the SCP of node j.This step is based on the Velocity Augmenting algorithm described in Section3.Schedule an arrival event of train m at the SCP of node j after t time units from the current time.(4)Schedule the resource-freeing events to free all the junction and segmentresources that train m has currently seized and will pass through during the movement from the SCP of node i to the SCP of node j.The exact times to signal these events are computed based on the algorithm in Section3. ACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January2004.Modeling T rain Movements •57If the central dispatching algorithm ﬁnds that there is no available successor node for train m to move forward,it returns a command to decelerate train m immediately to prepare to stop at the end of node i .However,this does not imply that train m will necessarily stop,since a successor node may become available before train m fully stops at the end of node i .We use two separate queues to distinguish trains in the deceleration process from the trains that have come to a complete stop.Queue 1stores all the train entities that are in the deceleration process between the SCP and the end of the node.Queue 2stores all the train entities that have come to a complete stop at the end of the nodes.In order to minimize train stoppage,we give priority to trains in queue 1over trains in queue 2when a resource is freed.If train m is noti ﬁed to prepare to stop at the SCP of node i ,the following three operations are performed.(1)Store train m in Queue 1.(2)Schedule an event for train m to come to a complete stop at the end of nodei after 2S i /a 2time units from the current time.(3)Schedule all the resource-freeing events of all the resources that can befreed while train m moves from the SCP of node i to the end of node i .When a resource-freeing event occurs,a routing algorithm is invoked to ﬁnd a decelerating train in queue 1,which can stop decelerating and start accelerating after seizing the currently freed resource.If such a train is found,let it be train m .First,the scheduled event of train m coming to a complete stop at the end of node i is removed from the event calendar.Also,the scheduled events of freeing the resources that train m will release before its arrival at the end of node i will be rescheduled,since train m has changed its speed.Finally ,the entity of train m is removed from queue 1.Then,train m seizes all the resources it needs and starts moving to the SCP of the new available successor node.If no such train m is identi ﬁed in queue 1that can stop decelerating,the routing algorithm checks if a train in queue 2can begin moving from the stop state.The train entities in queue 2are sorted by some priority based on the trains ’characteristics (e.g.,a passenger train has higher priority than a freight train).If a train in queue 2is identi ﬁed and can begin moving to its successor node,the train will be removed from queue 2,and the train will seize all the resources it needs and start moving to the SCP of the new available successor node.If an event associated with a train coming to a full stop occurs,its associated entity in queue 1is moved to queue 2.When a train arrives at a terminal station,the train entity is terminated and the statistical information regarding trip time and delay is collected.3.MINIMUM RUN TIMES UNDER MULTIPLE SPEED LIMITSIn rural areas,a single speed limit may apply for long stretches of travel.In contrast,many rail networks in metropolitan areas consist of multiple speed limits because of physical contours,crossovers,or other safety considerations.In this section,we develop an algorithm that determines the minimum run times of trains traveling on the segments and junctions with multiple speed limits considering maximum train speed,and acceleration and deceleration rates.InACM Transactions on Modeling and Computer Simulation,Vol.14,No.1,January 2004.。