2010 An Efficient Data Gathering Routing Protocol in Sensor Networks

Contributed PaperManuscript received April 15, 2010 Current version published 06 29 2010;Electronic version published 07 06 2010. 0098 3063/10/$20.00 © 2010 IEEEAn Efficient Data Gathering Routing Protocol in Sensor NetworksUsing the Integrated Gateway NodeSung-Hwa Hong, and Byong-Kug KimAbstract — In the future ubiquitous home network, sensorswill collect various home environment data in the home. Since the sensor nodes are equipped with small, often irreplaceable, batteries with limited power capacity, it is essential that the network be energy-efficient in order to maximize its lifetime. Our study, described in this paper, was divided into two classes: the primary class was to set a gateway-selection level and the secondary class was to propose an offer of a home automation using a routing technique centered on a sensor network to set a flooding level. The proposed scheme is to increase the life time of the sensor network with the integrated gateway node. The IGN increases the life time of the networkto integrate the multiple gateway nodes 1.Index Terms — Home Network, Wireless Sensor Networks, Sensor Routing, Home Automation, Zigbee.I. INTRODUCTIONWith the rapid advance in semiconductor production and communications technologies, the variety of consumer electronics used at home has been and continues to become digital. Developers of these digital home appliances have pursued providing their users with better services in the manner of interworking based on data communications, rather than having these devices operate independently. In this context, home networking plays a key role in these evolutionary processes. Wireless home networking is coming into the spotlight, as a consequence of the inconvenience of in-home hard wiring, and the convenience of wireless’ innate mobility, in spite of the presence of several proposed home-networking techniques. Technologies intended to implement wireless home networking include: IEEE 802.11 wireless local area network (LAN), Bluetooth, IEEE 1394, UWB(Ultra WideBand)-based wireless network, and ZigBee used primarily to control all kinds of home appliances. Currently, there are several ongoing studies on ubiquitous sensor networks for context-aware-based home services. However, we expect that wireless home networking based on the IEEE 802.11 wireless LAN and Bluetooth, which have been adopted almost universally due to their technical maturity, will be predominant in the foreseeable future.1 Sung-Hwa Hong is with the Dep. of Software Engineering, Dongyang Technical College, Seoul, Korea. (email: shhong@dongyang.ac.kr)Byoung-Kug Kim is with the Dep. of Electronics and ComputerEngineering, Korea University, Seoul, Korea. (email: dearbk@final.korea.ac.kr)As mobile communication devices and communicationtechnology are developing, the demand for wireless communication network on a small scale besides existing network of infra-structure. Especially, the need for wireless network on a small scale is increasing in environments such as inside buildings and mountain area, where wired facilities are unavailable. Wireless sensor network is one example. Unlike network of infra-structure, there is no router that relays packet transport in wireless sensor, but each sensor node performs both the roles of host and router. Since the transport range of each node's wireless frequency is limited in wireless sensor network, there are times when event packets are unavailable tobe sent from the start node to the end node. In such case, eventpackets are transported via multiple sensor nodes that work asa router. Such method is called multi-hop and routing protocol is required for multi-hop method in wireless sensor network. The most important thing in wireless sensor network is an efficient energy use in network environment and there are efficient routing and MAC protocols introduced. Routing can be largely classified into flat structure and hierarchical structure. In hierarchical structure, although energy data-aggregation and data-aggregation is very efficient as well as in-network processing is easy, there are problems that the overhead to make hierarchical structure is large and hierarchical structure is not maintained and needs to be created again because energy use is focused on nodes like cluster head. In addition, it is energy-efficient because all sensor nodes are not used in communication but it intensively uses energy because of a large amount of data processing in header node as well as it is hard to synchronize between header node and other nodes. Flat-structure Routing Protocol has easier multi-hop communication than hierarchical structure and each sensor node uses even amount of energy because there are no nodes that intensively use energy like cluster header in hierarchical structure. However, it is energy-inefficient than hierarchical structure because there are frequent packet collisions if event packets for Interest take place a lot and the number of re-transports increases due to the loss through multi-hop and all sensor nodes are always used tocommunicate. And there is also a problem that each node uses a lot of energy since sensor nodes installed in the same area all perform sensing as well as most flat-structure protocols have to have all sensor nodes in network routable for communication. Therefore, in this study, a method to make energy use efficient with router protocols of cluster concept and flat-structure that put a multiple sensor nodes in one group withoutchanging the purpose and function of existing network and performance analysis for the method is conducted.In Section II, we describe related works. In Section III, we describe our system model, and in Section IV, we present detailed operation of the proposed algorithm. Finally, we give concluding remarks in Section V.II.RELATED WORKSSupporting the mobility of sensor nodes is one of the most important factors to enable ubiquitous sensor networks, since mobile wireless sensor nodes can be attached to the human body, vehicles, and other mobile objects. So, the network layer should be implemented with the viewpoint of an efficient routing algorithm for wireless mobile sensor nodes [6]-[9].To efficiently maintain the routing path between a sink node and sensor nodes, various routing algorithms have been proposed. The hierarchical routing algorithm is one of them. Typical hierarchical routing algorithms are LEACH and LEACH-C.Direct communication is the simplest and the most intuitive way to send and collect sensor data. In a direct connection, each sensor sends data to the base station directly. It is simple, but may consume a large amount of energy for nodes further away from the base station. Based on the first order radio model, the energy drains more rapidly as the distance from the base station grows.Low-Energy Adaptive Clustering Hierarchy (LEACH) [3] divides the entire network into clusters. Each cluster has a cluster head. Nodes in each cluster send their data to the cluster head. The cluster head then collects the data and relays it to the base station. Each cluster uses different CDMA codes to avoid collisions.LEACH is a branch-based protocol, which is simple and scalable. However, this protocol is not very energy efficient for nodes. Besides, the radio of cluster heads must be turned on all the time to receive packets from the nodes in its cluster. This drains power more rapidly.In contrast to LEACH, Power-Efficient Gathering in Sensor Information Systems (PEGASIS) [4] organizes the sensor nodes by a single chain. Messages are sent hop-by-hop along the chain starting with the node farthest away from the base station. PEGASIS is often referred to as a chain-based protocol.The main advantage of PEGASIS protocol is the small total energy dissipation as nodes only need to communicate with their neighbors. However PEGASIS assumes every node has the global knowledge of the other nodes, which is not feasible. The delay also gets longer when the chain grows longer.III.THE SYSTEM MODELIn most cases, it is reasonable to assume that the sensor nodes have fixed and relatively short transmission range. In this case, an energy-efficient multi-hop routing mechanism is essential, and cluster organization becomes more complex than in the single-hop condition stated above. Efficient clustering algorithms for WSN have to satisfy several requirements, such as:1. Clusters should cover entire sensor field.2. Average cluster size should be as large as possible tomaximize data aggregation efficiency.3. The clusters should be repeatedly reorganized to balanceenergy consumption among the nodes.4. Clustering overhead should be small.5. Clustering algorithm should be simple enough to beperformed by low performance processor with small available memory space.Clustered structure of a network is very beneficial to energy conservation as shown in [10]. The benefit comes from the data aggregation of cluster heads. Aggregation efficiency increases as more data packets are aggregated. This benefit, however, is limited in multi-hop networks since cluster size is limited by the radio transmission range of the nodes. On the other hand, clustering overhead increases since clustering becomes more complex. The complexity comes mainly from the following two reasons: First, in multi-hop networks, it is difficult to re- cluster in a synchronized way as in single-hop networks. Second, when one cluster is reorganized, i.e. the role of a cluster head shift from one node to another, physical region the cluster head covers is also changed. This may necessitate reorganization of other clusters to satisfy the above requirements 1 and 2. These two requirements are for efficiency of the clusters. To better satisfy these requirements, however, more signaling and processing is required. But, requirements 4 and 5 prevent increasing the overhead and complexity.Requirement 3 says that the clustering overhead should be continually generated for fair energy consumption among the nodes. Moreover, if the nodes have mobility, clustering overheads will be far more increased. Thus, the benefit of clustering can be cancelled by the clustering overhead. In single-hop networks node mobility does not affect any network operation as long as the node does not move out of the transmission range of any other node. Among many of the previous researches, the network models for hierarchical protocols for WSNs are single-hop networks in [10-12] and those for flat routing protocols are multi-hop networks [13-14]. Clustering complexity in multi-hop networks can be one explanation for this research trend in sensor networks.We consider for the following network and application model.1. A lot of sensor nodes are dispersed randomly on aninterested region.2. Sink nodes are placed at some convenient places in ornear the sensor field. The users can get theinformation from the sensor field and control itthrough the sink nodes by direct or remote access toit. Thus the sink nodes should have user interface orcapabilities to communicate with remote users withhigh powered radio or wired connection. The numberof sink nodes is very small compared with thenumber of sensor nodes. Thus they can have specialcapability, battery with larger capacity or externalpower supply.3. The sensor nodes have limited processing andcommunication capabilities in order to satisfy thelow-cost condition. Thus very complex and/or energyconsuming algorithm is difficult to be adopted.4. All the sensor nodes have the same constanttransmission ranges.5. Users request data from the sensor network bydisseminating query packets through the sink nodes.And, the data sensed from each node is gathered bysink nodes through cluster heads so that users canaccess it through the sink nodes.IV.IGCP—A I NTEGRATED G ATEWAY-NODE C ONTROLP ROTOCOLAlthough the hierarchical structure is energy-efficient and great in data-aggregation as well as in-network processing is easy, the hierarchical-structure cannot be maintained for a long time and needs to be created again because of intense energy use in nodes like cluster head. The flat-structure has easier multi-hop communication that the hierarchical-structure and allows even energy use of each node. However, the most important aspect of wireless sensor network, energy use, is very inefficient because all sensor nodes should always be used for communication. This study suggests the IGN (Integrated Gateway Node) Algorithm to compensate the vulnerability of two structures. It is an Algorithm in which virtual gateway nodes consisting of several nodes like the cluster of hierarchical-structure routing protocol and allows the flat-structure routing protocols between virtual nodes.A. The initial flooding process:Routing information is flooded from the sink nodes. The procedure of each node to set the routing information for each sink node is similar to the distance vector algorithm. In the routing information packet, the number of hops to a specific sink node and the address of transmitting node are included. When a node receives routing information from a neighbor node, it increases the number of hops by one and uses the number as its own number of hops to the sink node, and then, retransmit this information with its own address. When different number of hops is received from different neighbor nodes, the smallest number of them is used. If a node receives smaller number after it has retransmitted routing information, the smaller number should be again retransmitted to correct the propagated errors. Through this procedure, each node can know its own number of hops to a specific sink and the address of the next hop node to the sink.B.Clustering:The initial clustering occurs during the initial routing information distribution. In a routing information packet, energy state information of the transmitting node should be included. When a node has transmitted routing information, every neighbor of the node, except those who have previously transmitted the information, will retransmit the information. The node can gather information about the energy states of every neighbor node with the routing information packet, and compare them with its own energy state. When a node has found that it has the local maximum amount of energy, it becomes cluster head and broadcast a cluster head advertisement (CHAD) message to its neighbors. Before a node decides to be a cluster head, it has to wait for a sufficient time to gather the energy state information from all the neighbor nodes. The nodes that decided not to be a cluster head wait for a CHAD message from any other node. If a node waiting for a CHAD message cannot receive one for a pre-determined time period, it repeats the exchange procedure of energy state information with other nonaffiliated nodes. This procedure is repeated until every node is affiliated with one cluster head. Any nonaffiliated node affiliates with the node whose CHAD massage it first receives.C.Gateway-Selection:One cluster head has one gateway node to a sink node. A gateway node is selected by cluster head among the nodes which are one hop closer to the sink node right after the cluster head is elected. The gateway node may be or not be a cluster member of the cluster head which selects it as a gateway. A cluster head sends a gateway selection (GWS) message to a selected gateway node, and thus the selected node can know whose gateway it is. Query dissemination from a sink node and data gathering to a sink node is performed through cluster heads and gateway nodes.D.Integrated Gateway nodeAlthough the hierarchical structure is energy-efficient and great in data-aggregation as well as in-network processing is easy, the hierarchical-structure cannot be maintained for a long time and needs to be created again because of intense energy use in nodes like cluster head. The flat-structure has easier multi-hop communication than the hierarchical-structure and allows even energy use of each node. However, the most important aspect of wireless sensor network, energy use, is very inefficient because all sensor nodes should always be used for communication. This study suggests the IGN (Integrated Gateway Node) Algorithm to compensate the vulnerabilities of two structures. It is an Algorithm in which virtual gateway nodes consisting of several nodes like the cluster of hierarchical-structure routing protocol and allows the flat-structure routing protocols between virtual nodes.E. Link setting Cluster head advertisement (CHAD) is largely classified into the busy status and static status. The busy status is a status in which IGN is operating as IGNH(Integrated Gateway Node Head) and the static status is a status in which IGN is operating as IGNM(Integrated Gateway Node Member) because there is no need of the IGNH role. The busy status occurs when different IGN's are created, adjacent IGN's are newly created, or the routing table of adjacent IGN's is changed. It also becomes the busy status when it is short of IGNM nodes and needs to borrow IGNM nodes from IGN with many IGNM nodes. In the busy status, IGNM does not play the role of IGNM node but plays the IGN role only and always stays awaken. IGNH in the busy status has many roles thus it does not play any role for data communication but only works for Control Signal Communication. It is to even out the energy use of all sensor nodes within virtual nodes. The staticstatus occurs when IGNH does not have any controlcommunication signals between IGNH's during thecommunication period of IGN. The communication period isas shown Picture 4.5 and consists of Frame and Sync. Frameis the time for virtual nodes to communicate and Sync is thetime for schedule re-configuration and error process. Sync is alot shorter than Frame. IGNH does not have many roles in thestatic status and IGNH works as IGNM node.Once the initialization level is over and virtual nodes arecreated, it becomes the busy status of Gradient Level. Theoperations that require IGNM transport between virtual nodes orIGNH information take place. In the first busy status, it does notmove on to the static status until a routing table is created.Although Routing Table Configuration take place whensending query or broadcasting just like other flat-structureprotocols, there is a bit of difference in the configurationmethod. If virtual nodes receive query or packets inbroadcasting, it first checks that received nodes are receivedwith less than one-half power of the transport distance. Arouting table is created if the nodes are received with morepower and it is ignored otherwise. Also, routing table information is saved up to 2 hops of virtual nodes. The reason that virtual nodes require receiving power is because of the communication limit mentioned earlier. In fig.1, the process of the integrated gateway node initialization shows. Node B operates as IGNH, node A and C are operates as IGNM.Node A Node BNode C4F. Flooding level Above all, once cluster formation has been completed, the flooding process is not executed. Based on the level acquiredthrough flooding, nodes piggyback their own levels in the data transmitted to update them. This level-updating process is seen in Fig. 2.(a) (b) (c)Fig. 2 The level-updating procedure.In this figure, (a) indicates the levels of child nodes ifthe root (or parent) node equals level 1. This figure showsavailable connections between individual nodes and rootor child nodes. However, in the event that the node withlevel 3 malfunctions, of child nodes detecting that, thenode whose upstream connection has relied only on aconnection with the faulty node finds out a node with thehighest-layer level, i.e., with the lowest level value amongits adjacent nodes. This node, then, sets its own level tothe result of adding 1 to the existing level. Theseprocesses are happen for a level update. The processes(a), (b), and (c) represent level-dependent datatransmission from usual low-layer sensor nodes to a high-layer gateway node.G. Data transmission Once a cluster has been formed, as a rule, all of its operations are controlled by the gateway node. The information, detected and obtained in the cluster, is transmitted to the gateway node. Then, the gateway node transmits the data via data aggregation to the relay node. Next, the relay node simply transmits the data to a selected upper link. This facilitates significantly in setting upper links. Mostinterestingly, the upstream nodes of a cluster go towards thesink node.IV. A Performance Evaluation We create a simulator with C++and make sensor nodes as one class object. Each class contains a handler that processeseach event and operates differently according to the type of the handler. It is also possible for the class which represents sensor nodes to create virtual nodes proposed in this study andapply operation algorithm. Packets are also a class object and inherit events to be configured. There are various types depending on packet's function and packets are processed in the node object. An arbitrary node object is selected, and if the object is in inactive status or is already a member of other virtual node then next object is selected. If the selected object is not a member of virtual node or in inactive status, algorithm defined in the node object is performed. With a repetition of such process, virtual node algorithm is applied till all node objects are selected. Data of sensor field with virtual nodes created are set to take place in Sync node. The packet type can be either 'interest' or 'query' depending on the flat=structure routing protocol and the size can be changed to be suitable for application. Also, packets can be transported in a certain period of time in CBR method. We will look at the applied parameter values for simulation.Fig. 3 Our simulator program.This simulation analyzes the energy amount used when existing routing protocols are operating and the energy amount used after the algorithm suggested in this study isapplied. This simulation does not consider the energy amountused in the initialization level to create virtual nodes. It is because it is a very small energy amount comparing to the total energy amount used.D i s s i p a t e d e n e r g y i n t h e s e n s o r n e t w o rk (J o u l e )Simulation time(sec)Fig. 4 The simulation results of the energy comparison between GeneralFlooding.D i s s i p a t e d e n e r g y i n t h e s e n s o r n e t w o r k (J o u l e )Simulation time (sec)Fig. 5 The simulation results of the energy comparison between SPIN.In this simulation, one 1024-byte packet is created in Syncand, this packet is sent to entire network and the energy used in the process is analyzed. The routing protocols used are Flooding and Spin. Parameters used in Verse 5.1.2 are used to experiment under the same condition as other simulations. The figure 4 shows the energy comparison between General Flooding in which virtual node algorithm is not used and Flooding via Virtual Node in which virtual node algorithm is applied. As shown in the figure 4, there is up to 55% of energy saving effect when the IGN algorithm is applied. Also, the energy efficiency is increasing in a long run. The figure 5 is when Spin algorithm is applied in the same way. Theenergy used is different by the size of meta-data because Spinhas a lot of meta-data. This simulation sets the size of meta-data to 32 bytes. Like Flooding, Spin protocol (Spin via virtual node) [15]with virtual node algorithm suggested in this study shows up to 32% of energy saving effect. V. C ONCLUSION Many ways of energy-efficient routing for wireless sensor network are suggested and are largely divided into the hierarchical-structure algorithm and flat-structure algorithm.Each algorithm has own advantages and disadvantages.This study suggests a mixed algorithm with virtual gateway nodes, which includes both the advantages ofexisting hierarchical-structure algorithm and flat-structure algorithm in wireless sensor network. The suggestedalgorithm communicates with application of flat-structure typeprotocols after bundling up several nodes like the hierarchical-structure cluster and making them work as one node. Virtualgateway node algorithm allows efficient energy managementsince it not only increases the energy efficiency which isconsidered important in wireless sensor network but alsocreates virtual nodes. In addition, various applications can berealized. It also makes up the disadvantages of existinghierarchical-structure routing protocols by allowing even energy use of each node as well as performing dataaggregation and in-network that are characterized in existing sensor network routing protocols.The algorithm suggested in this study allows easily to create virtual nodes with simple MAC protocols since control signals that create virtual nodes are not complicated and also allows smooth packet flows since not many sensor nodes are used for communication and there are less packet collisions and less re-transports. But, it is not possible to use the algorithm in environments with sensor nodes in which RSSI is not supported because RSSI is needed to create virtual nodes and form routing tables.As analyzed in the simulation, it shows a great performance in energy use and does not use a lot of energy to create virtual nodes. It is more efficient in long-distance transport environments.The algorithm suggested in this study is expected to be used in a number of application fields of sensor network since it has many characteristics and advantages. However, it will be efficient only in the environments suggested in this study, which is only a small part of many application fields of sensor network. In the future, it will be necessary to have more studies on many types of algorithm that can work in many more application fields with more efficient energy use if we look at infinite possibilities of sensor network used in numerous fields with numerous environment and application factors.R EFERENCES[1]I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, "Wirelesssensor networks: a survey," Computer Networks 38 (2002)[2]Praveen Rentala, Ravi Musunnuri, Shashidhar Gandham, Udit Saxena,"Survey on Sensor Networks"[3]W. R. Heinzelman, A. Chandrakasan, H. Balakrishnan, "Energy-Efficient Communication Protocol for Wireless Microsensor Networks"Proceedings of the Hawaii International Conference on System Science, ,page 1-10, Jan, 2000.[4]S. Lindsey and C. S. Raghavendra, "Pegasis: Power-efficient gatheringin sensor information systems," in IEEE Aerospace Conference, March 2002.[5]K. Du, J. Wu, and D. Zhou, "Chain-based protocols for databroadcasting and gathering in sensor networks," in Proceedings of Workshop on Parallel and Distributed Scientic and Engineering Computing with Applications (in conjunction with IPDPS), April 2003. [6] C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directed Diffusion: AScalable and Robust Communication Paradigm for Sensor Networks,” in Proceedings of the ACM/IEEE International Conference on Mobile Computing and Networking (MOBICOM), 2000. [7]J. Kulik, W. R. Heinzelman, and H. Balakrishnan, “Negotiation-BasedProtocols for Disseminating Information in Wireless Sensor Networks,”ACM Wireless Networks, vol. 8, no. 2-3, pp. 169–185, 2002.[8]K. Sohrabi, J. Gao, V. Ailawadhi, and G. J. Pottie, “Protocols for Self-Organization of a Wireless Sensor Network,” IEEE Personal Comm.Mag., vol. 7, no. 5, Oct. 2000.[9]Z. Haas and S. Tabrizi, “On Some Challenges and Design Choices inAd-Hoc Communications,” in IEEE MILCOM’98, October 1998.[10]Huseyin Ozgur Tan, Ibrahim Korpeoglu, "Power Efficient DataGathering and Aggregation in Wireless Sensor Networks", ACM SIGMOD Record, Vol. 32, No. 4, Pages 66-71, December, 2003.[11] A. Manjeshwar, D. P. Agrawal, "TEEN: a routing protocols forenhanced efficiency on wireless sensor networks." International Proceedings of 15th Parallel and Distributed Processing Symposium, page 2009-2015, 2001[12] A. Manjeshwar, D. P. Agrawal, "APTEEN: A Protocol for EfficientRouting and Comprehensive Information Retrieval in Wireless Sensor Networks," Proceedings of the 2nd Int. Workshop on Parallel and Distributed Computing Issues in Wireless Networks and Mobile Computing, April 2002[13] C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directed Diffusion: AScalable and Robust Communication Paradigm for Sensor Networks,” in Proceedings of the ACM/IEEE International Conference on Mobile Computing and Networking (MOBICOM), 2000.[14]J. Kulik, W. R. Heinzelman, and H. Balakrishnan, “Negotiation-BasedProtocols for Disseminating Information in Wireless Sensor Networks,”ACM Wireless Networks, vol. 8, no. 2-3, pp. 169–185, 2002.[15]W. R. Heinzelman, J. Kulik, and H. Balakrishnan, "Adaptive Protocolsfor Information Dissemination in Wireless Sensor Networks," Proc.ACM MobiCom '99, Seattle, WA, 1999, pp. 174-85.BIOGRAPHIESSung-Hwa Hong received his B.S.degree in ComputerScience from Seoul, Korea University, 1996 and his M.S.degrees in Information and Communication Engineeringfrom Hankuk Aviation University, in 2002. Since 2002, hehas been in the Ph.D program in Electronics and ComputerEngineering at Korea University, Seoul, Korea. Hisresearch interests include the Home Network, WLANs, ad-hoc networks.Byoungguk Kim received his B.S. degree in ComputerEngineering from Won Kwang University, Iksan, Korea,2002, and his M.S. degree in Electronics and ComputerEngineering from Korea University, Seoul, Korea, 2004.Since 2004, he has been in the Ph.D program in Electronicsand Computer Engineering at Korea University, Seoul,Korea. His research interests include Bluetooth, embedded-system, ad-hoc and sensor networking, and ubiquitous computing.。