A cooperative radar system for collision avoidance and communications between vehicles

Abstract — In order to improve reliability, security on roadsor exploitation of automatic guided transports, many sensors are being developed to equip vehicles. Research has been carried out at INRETS and IEMN on low cost microwave collision avoidance sensors for metro guided transports and for road vehicles. For this goal, an original radar system has been designed in ours laboratories. It is based on cooperative collision avoidance radar which uses a transponder inside targets. The proposed system is made of a couple of microwave transmitting and receiving equipment fixed on each vehicle, one ahead and the other behind. In order to avoid interferences between the transmitted and the received signals, two duplex frequencies of 2.2 and 2.4 GHz are used respectively for downlink and for uplink. The system is based on pseudo - random sequences combined with a correlation receiver and has a broadband of about 50 - 100 MHz. This bandwidth has been exploited to establish a high data flow communication between sensors.The aim of this paper is to describe the mock-up realized and technical solutions developed using spreading spectrum techniques to allow multi-user access and to combine the two desired main functionalities, localization and communication, inside a unique cooperative radar system. The system performances have been evaluated in terms of BER using simulations and main results will be presented. I.I NTRODUCTION To enhance safety on road, many studies have been carried out on on - board collision avoidance system. The goal of these systems is to keep the vehicle at a safe distance from any obstacle in front of it. In this field, many radar sensors have been developed inside our laboratories [1] [2] [3]. We can cite FM/CW radar - Frequency Modulated Continuous Wave -, pulsed radar, Bimode radar which is a combination of FM/CW and pulsed systems. These systems, especially useful also in bad weather conditions, are able to Manuscript submitted on March 17, 2006.Dr C. Tatkeu, French National Institute for Transport and safetyResearch – INRETS -, email : charles.tatkeu@inrets.fr, phone + 33 3 20 4384 01, fax: + 33 3 20 43 83 59 Dipl. Ing. P. Deloof, French National Institute for Transport and safety Research – INRETS -, email : pascal.deloof@inrets.fr, phone + 33 3 20 4383 16, fax + 33 3 20 43 83 59Dr Y. Elhillali University of Valenciennes, IEMN/DOAE email : yassin.elhillali@univ-valenciennes.fr , phone + 33 3 27 26 51 13 66, fax + 33 3 27 51 11 89Dr A. RIVENQ, University of Valenciennes, IEMN/DOAE, email : Atika.Menhaj@univ-valenciennes.fr, phone + 33 3 27 26 51 13 66, fax + 333 27 51 11 89 Pr Jean Michel ROUVAEN University of Valenciennes, IEMN/DOAE,email : jean-michel.rouvaen@univ-valenciennes.fr, phone + 33 3 27 26 5113 65, fax + 33 3 27 51 11 89provide information about obstacles or vehicles ahead. However, in some environments such as subways, multipath propagation occurs and it is then difficult, with these previous sensors, to distinguish the right target among echoes. For that reason, a radar sensor, called DIBT for Detection and Identification of Broken - down Trains, has been developed to facilitate the exploitation of automatic guided transports. This kind of radar is helpful when a critical problem of exploitation, due to a broken down train, occurs on the line. Due to security reasons, all the trains of the line are stopped immediately and this situation generates traffic perturbation. The train must be removed out from the line to a maintenance workshop and the procedure consists in pushing it with the following coach. With this intention, it is necessary to accost, in safety conditions, this broken - down train by another. At the moment, in the subway network of Lille - France, the accosting procedure is realized at a very low speed, called coupling speed, which is less than 2 mph and corresponds to security speed limit. This process, at a very low speed, wastes a lot of time and this trading loss has also a very high cost for the operator. The main objective is to reduce accosting phase duration between the two vehicles in various environments amongwhich subways, viaduct and free space. With the DIBTsensor, the process will be shortened. Indeed, the sensor is able to identify the broken - down train, to read its state of functioning and to provide, in real time, the distance to the preceding vehicle. This allows the next train to run faster, and to slow down close to the blocked train. DIBT was improved also by adding a communication function in order to allow a high data rate exchange between vehicles. It was called CODIBT for Communications Detection and Identification of Broken - down Trains. ThisCODIBT sensor uses also as DIBT an adapted digital correlation receiver in order to detect the position andcompute the distance towards the preceding vehicle.Moreover, it also uses a multiplexing technique to extract data such as identification or the state of the vehicle, or any information exchanged between sensors. It has a broad band of about 50 - 100 MHz which has been exploited to combine the two main functionalities: localization and high data rate communication. An original multiplexing method called Sequential Spreading Spectrum by 2 codes technique - SSS2- has been developed. It allows multi-user access and communications at a rate of 1.6 Mbps. In this paper, we show the principle and the utility of the CODIBT radar system. Afterwards, the main components A cooperative radar system for collision avoidance andcommunications between vehiclesC. Tatkeu, P. Deloof, Y. Elhillali, A. Rivenq, J.M. RouvaenCONFIDENTIAL. Limited circulation. For review only.Proceedings of the IEEE ITSC 20062006 IEEE Intelligent Transportation Systems Conference Toronto, Canada, September 17-20, 2006TB6.3are defined, starting from the system framework. The complete system based on radar transmitter and responder will be described. Ideas for its extension and adaptation to road applications will be given.II.T HE PRINCIPLE OF THE PROPOSED CODIBT SYSTEMThe proposed radar collision avoidance system developed in our laboratories is called CODIBT for Communications, Detection and Identification of broken-down trains [1]. It includes DIBT functionalities and a new data communication feature. With this sensor, the distance to thetargets is measured, in real time, with a resolution between 1.5 and 3 meters in a range of 800 meters in tunnel and 700 meters in free space [4]. The implemented DIBT has a broad band of about 50 -100 MHz that remains unexploited. We propose to develop a new system to exploit this band in order to establish high data rate communications between trains and operators center. The new system, called, CODIBT sensor is able: t o detect the position, get the identification and the status of the train, t o compute, in real time, the distance towards the preceding vehicle, t o allow a high data rate communications for exchangingdata information between trains.Its principle relies on a transponder system using an Interrogator/Responder pair (see Figure 1.a) which equips respectively the front and the rear of vehicle. As shown on Figure 1.b, the first vehicle (Interrogator) sends a signal at a frequency of 2.2 GHz, towards the preceding vehicle (Responder). This signal, which has its own radar code, is a P.N. (Pseudo-Noise) sequence. It is received by the second vehicle ahead. The sensor of this vehicle ahead process and sends a replica of the received signal that is amplified, filtered and filled out with data at the same time. These data contain information about its identification (or identity), its working mode or state (broken-down or not, failure status) etc. The new signal sent at 2.4 GHz frequency is received by the Interrogator that is able to deduce the inter-train distance and to recover the data sent by the Responder (identification, status (broken-down or not)). The frequency choice is an important item, because it depends on the line configuration and the possibility of resolving both effects of masking and multipath, which strongly affect the resulting signal. The present choice issettled in the range of 1-10 GHz band. For a low poweremitter/consumption, we choose ISM band for our sensor.Fig. 1.a The CODIBT radar sensorFig. 1.b The CODIBT transmitter/receiver design architectureSuch a co-operative radar system for which the target becomes active like in a transponder system has great advantages among others: I t works in each kind of environment free space and especially subway tunnel or viaducts areas, where conventional RADAR systems based on distance measurement using signal echoes on obstacles proves inefficient. M oreover, the pseudo-random sequences used, combined with correlation receiver, are very adapted to the detection of signals over noisy communication channels and can be generated easily. On the following paragraphs, this paper will present characteristics and performances in terms of BER and data rate of the system. We focus on developing appropriate coding methods for this sensor to favor high data rate and the robustness according to SNR (Signal Noise Ratio). III.P RESENTATION OF THE MULTIPLEXING TECHNIQUE In this paragraph, we focus on technical solutions to develop the new communication feature and optimize the combination of the two main functionalities: localization and high data rate communication. In order to provide this combination with high speed data flow, different coding methods were tested and one of them is presented hereafter. Indeed, this method allows a continuous refreshing of the measurement of distance and also ensures a sufficient flow rate for communication with a suitable BER.The technique is inspired from the DS-CDMA and usesfamilies of orthogonal codes (Binary Pseudo-Random Sequence – BPRS) with two different lengths. The first one has code length of 1023 bits (C1023) intended for the localization and the second is constituted by shorts codes of 31 bits long (C31) dedicated to the communication.The code families usable in CODIBT: Different codes families (mlfsr codes, Gold codes, Kasami codes) were studied for use in CODIBT and were compared according to the number, the length and the maximum of their cross-correlation. These sequences look like a noise and so have a spread spectrum. The selected codes have the same length : 2n-1 [1][2]. These pseudo-random sequences have an autocorrelation peak and a low level only for the cross-correlation. The mlfsr, also called m-sequences, presents an autocorrelation with a peak at 2n-1 and a –1 level elsewhere. They have good performances even when the signal to noise ratio is very low. Their implementation is simple. They could be easily generated using shift registers with XOR feedback. The number of these codes per familly is a function of their length : 2 for n=3 or n=4, 6 for n=5 or n=6 : 6, 18for n=7, 16for n=8, 48 for n=9, 60 for n=10. These families are considered as the reference in this study.As we can see on Figure 2, the method consists of sending periodically the code of localization to ensure a regular renewal of the measurement of distance. One proposes to insert between two codes of localization a variable structure of coded data burst. Between two localization codes we insert 1023 bits, which can be dividedinto several short codes.Fig. 2: General structure of a frame sent with the coding technique.The coding technique entitled SSS2 for Sequential Spectrum Spreading using by 2 codes is inspired from the CDMA technique.The spreading with the C1023 is used to assume localization function. The second one is used to code data communications with the C31 in the classical DS-CDMA (Direct Sequence CDMA) [4] [5] [6] [7]. This techniqueallows us to send 33 bits of data between two codes of localization. The length of the first code is chosen to reach the required distance (about one kilometer) and due to important number of reply codes (60). The length of thesecond code affects the rate of communication: If we choosea shorter one, we will have a higher rate but the robustnesswill decrease significantly. Multiple simulations have been done and the length of 31 bits seems to be a good trade-off between the data rate and the robustness to noise. Figure 3 shows the standard structure of the frame transmitted by this method. Fig. 3: Detailed structure of the Frame sent by the SSS2 technique.To calculate the distance, the correlation between the received signal and the reference codes (C1023) is computed. The correlation peak allows the synchronization process. Then, to recover data, a second correlation between the received signal and the C31 code is used.IV.P ERFORMANCESThe SSS2 technique has been simulated in AWGN (Additive White Gaussian Noise) channel in order to evaluate its performances in terms of data flow rate and Bit Error Rate (BER).The radar antennas used are a 2*2 patches for each link with a beam aperture of 20° to operate in curves.Concerning the localization characteristics, it gives the same resolution in distance, which is between 1.5 meter and 3 meters depending of the clock frequency used (50 MHz or 100 MHz). The maximal range obtained is about 800 meters in tunnels and 700 meters in free space.Moreover, the radar detection is physically limited in low range, under 10 or 15m, due to the recovery time of the sensor.To evaluate the robustness of this technique to noise, some simulations are performed in AWGN channel. On Figure 4, the Bit Error Rate corresponding to several Signal to Noise Ratio values, obtained by simulations is given forthis technique.Fig. 4: The BER obtained with SSS2 technique. f L N c *2Time sentbits of number flow Data == (1) were N is the number of data bits sent periodically, Lc is the length of the localization code, f is the signal frequency. Furthermore, the maximum length of data is 1023 (as localization code); due to the choice of code of 31 bits tospread out data, the maximum numbers of bits which could be sent in a periodic interval is 33 bits.Mbpsf L N c 6.110*1001023*233*2flow Data 6≈==In this case, the data flow which could be reached is about 1.6 Mbps for a clock of 100 MHz. Simulation results show that, in AWGN channel, SSS2 technique is robust to noisy environments (i.e SNR less than -2 dB). Moreover, a BER of 10-5 can be reached with SNR equals to –2 dB with SSS2 method. The actual laboratory mock-up integrates a multiplexing SSS2 technique using flexible components like FPGAs[4]. TABLE I P ERFORMANCES OF CODIBTCoding SSS2Maximum flow1.6 Mbps SNR for BER = 10 -5-2 dB Range in free space 700 m Range in subway800 mResolution at 100MHZ 1.5m Sensor CharacteristicsAntenna aperture 15° Antenna type 2x2 patches Antenna size (cm) 12x12Table 1 gives a summary of performances. The resolution and range in free space and tunnel are the same. They are the same values than those obtained with the previous sensor DIBT i.e 1.5 meters for a clock frequency of 100 MHz, 700 meters maximum range in free space and 800 meters maximal range in tunnels.The preliminary results of simulations confirm that performance of the SSS2 technique offers a weak BER and a sufficient high-speed information exchange.V.E XTRAPOLATION OF COOPERATIVE SYSTEMS TO THEROAD DOMAINA simple idea is to make certain massive objects cooperative, by providing them with an active responder. This solution is considered more and more in the USA for obstacle detection on ordinary roads. These systems can be conceived only on one radar which functions firstly autonomously, then gains in reliability when it receives a signed response from the object. Thus, certain infrastructures, such as the piles or bridge decks, causing significant echoes could be equipped with responders whichwould bring a signature to the radar and would allow a filtering such as “the edges of the road”. Indeed, these objects are dangerous only if the vehicle runs out of its normal trajectory, which is often materialized on the ground by white tapes. Nowadays, video systems exist to supervise these white tapes and to detect any dangerous variation.Now, the CODIBT system allows the dynamictransmission of information by coding the signal. Such a system could reflect towards the back, the behavior of the preceding vehicle (braking, acceleration) and thus avoid the pile-up. CODIBT could transmit data directly between thecomputers of two consecutive vehicles, taking account of the car behavior and even the driver's behavior, when he (orshe) is being to overdrive or to slow down. This kind ofsystem also resolves the lack of detection in the rear mirror about other vehicles present in the same area (blind spots), since each vehicle can be recognized by its identity, its distance from the interrogator and its behavior. In the autonomous mode, the range calculation is basedon a transmission coded by pseudo-random sequences andthe recognition of this code in the received signal, by using a correlation process.VI.C ONCLUSIONIn this paper, a new cooperative radar system designed forautomatic guided transports is presented. This sensorincludes different functionalities among which: localization, identification and communications with high data rate. Toinclude the new communication feature, we developed technical solutions using spreading spectrum techniques to allow multi-user access. 1023-bit Pseudo-Noise sequences was used for localization in order to reach a large range; itprovides a sufficient number of reply codes, necessary to identify each train and to evaluate its functioning status. 31-bit Pseudo-Noise was used for communications in order to provide a good trade-off between the data rate and the robustness to noise. Algorithms and solutions developedwere implemented using flexible components such as FPGA devices inside CODIBT sensor. A mock-up was realized, described and tested in real conditions. The mainperformances obtained with less than 1 Watt of radiated power are: 700 m range in free space or 800 m in confined area (tunnel), with a resolution of 1.5 m for a 100 MHz frequency clock. The system has been also evaluated in terms of BER and data flow rate. The BER obtained is about 10 -5 for a SNR of about -2 dB. A data flow rate of about 1.6 Mbps could be reached with the SSS2 technique described and developed. The CODIBT system could also be an interesting opportunity to apply a car warning radar concept to a train management system and maintenance operations - for example an intervention when a train light breakdown occurs, causing a service interruption -.The extensions of the use of this sensor to the road domain were also studied. Using the CODIBT, the transponder system could be a good relay of road information. The driver could be informed, in real time, when the safety margin distance to the preceding vehicle is insufficient - in case of bad weather conditions such as fogwhere risks of collision between vehicles are increased significantly -.R EFERENCES[1] A. Menhaj, J. Assaad, P. Deloof, M. Heddebaut “Enhancing safety onroad with on-board microwave anticollision system by combining tworadar sensors” 3rd World Congress On Intelligent Transport Systems,Orlando, October 1997.[2]P. Deloof, A. Menhaj, J. Assaad, "Collision avoidance radar systembased on numerical correlation receiver". 7p, Paper N°2041, Session AV8 : AHS and Pre-AHS, 4th World Congress on Intelligent Transport Systems, Berlin 21-24/10/1997.[3] B. Frémont, A. Menhaj, P. Deloof, M. Heddebaut, « A cooperativecollision avoidance and communication system for railway transports», MP4-6, 3rd IEEE Conference on intelligent transportation systems,Dearborn, 1-3 octobre 2000.[4] C. Tatkeu, Y. Elhillali, A. Rivenq, J-M Rouvaen. « Evaluation ofcoding’s methods for the development of a Radar sensor for localization and communication dedicated to guided transport IEEE VTC 2004 26-29 Septembre 2004, Los Angeles, USA.[5]Z. Moe Win and R.A. Scholtz, “Impulse Radio: How it works”, IEEECommunications Letters, Vol 2, No. 1, Jan 1998.[6] B. Fremont , « Etude et realisation d’un systeme radar cooperatifdestine aux systemes de transport guides ». PhD Thesis of Universityof Valenciennes –France November 2003.[7]R. Dixon, ed., Spread Spectrum Techniques, New York, IEEE Press,1976.[8]Y. Elhillali « Etude et realisation d’un systeme de communication etde localisation, base sur les techniques d’etalement de spectre, dedie aux transports guides », pHD thesis of University of Valenciennes France, order number 05-27.[9]Y. Elhillali, A. Rivenq, C. Tatkeu, J-M. Rouvaen, « Implementation ofa localisation and communication system dedicated to guidedtransports », the 5th International Conference on ITS Telecommunications ITST 27-29 June 2005 Brest, paper n° 65-2445203603.。