Cellular self-organization architecture for wireless sensor networks

Cellular self-organization architecture for wireless sensor networks M.Asim, H.Mokhtar, and M.Merabti School of Computing and Mathematical Sciences Liverpool John Moores University [email protected] , [email protected], [email protected]

Abstract - Wireless sensor networks are composed of large number of sensor nodes, which are limited in resources i.e. memory, energy and computation power. Sensor network life time is directly related to nodes energy. Wireless sensor networks are expected to be capable of self-organization in an efficient, reliable and continues manner during the life time of the network. Self-organization of wireless sensor networks are usually involved in partitioning the network into connected groups or clusters and is challenging task because of limited bandwidth and energy resources available in these networks. In this paper we propose a new cellular self-organized architecture for wireless sensor networks that extends the network life by efficiently utilizing nodes energy and distribute management tasks to support the scalability of management system in a densely deployed sensor networks. In our solution the network is partitioned into a virtual grid of cells. A cell manager and a gateway node are chosen in each cell to perform management tasks. Cell manager and gateway nodes coordinate with each other to perform management with minimum energy consumption. We assume a homogenous network where all nodes are equal in resources.

I. INTRODUCTION Recent developments in wireless communication and electronics have made possible the development of small, inexpensive, low power, distributive devices. These devices are capable of local processing and wireless communication and are known as sensor nodes. Thus a sensor network can be described as a collection of sensor nodes which co-ordinate to perform some specific function. A sensor network ensures a wide range of application. Examples includes environmental monitoring- which involves monitoring air soil and water, condition based maintenance, habitat monitoring, seismic detection, military surveillance, inventory tracking, smart spaces etc. These sensor nodes are mainly in large numbers and are densely deployed either inside the phenomenon or very close to it. Also the position of sensor nodes not needs to be engineered or predetermine, which allows random deployment in inaccessible terrains or disaster relief operations. Another unique feature of sensor networks is the on board processing and co-ordination. Instead of sending the raw data to the nodes responsible for fusion, they use their processing abilities to carry out simple computation and transmit only the required

ISBN: 978-1-902560-19-9 © 2008 PGNet

and partially processed data. On the other hand, this also means that sensor networks protocols and algorithms must possess self-organizing capabilities. Self-organization is the process of autonomous formation of connectivity, addressing and routing structures. Self organization in wireless sensor networks is a significant research topic. A self organized wireless node can be clustered or grouped into an easily manageable network [1-6]. One of the crucial design challenges in wireless sensor network is energy efficiency. Since sensor nodes are operated on battery and keeping the nodes active all the time will limit the duration that battery last. Also, individual sensor nodes use a small battery as a power source and replacing or recharging of these batteries in remote locations are not practical. In some cases solar cells are also use as a source of energy but they provide limited power. Therefore, it is very important to tackle energy efficiently at all levels of sensor network infrastructure. Wireless radio is the major energy consumer in a sensor node and systematic management of network communication become critical. Network communication is involved in tasks like routing, gathering or forwarding sensing information to a nearby data sink or a remote base station. In order to achieve effective coordination among these activities, it is important to address the problems of sensor network organization and the subsequent reorganization and maintenance [7]. One of the research challenges for succeeding the vision of a self organized wireless sensor network is the scalability of management system. A sensor network may consist of hundreds, thousands, or even millions of inexpensive wireless sensor nodes that may be placed either regularly or irregularly. This is mainly concern network traffic and response delay as the result of self-managed maintenance upon resourceconstrained WSNs. Distribution of management tasks in sensor nodes is an energy efficient approach and utilizes node resources effectively in a large scale WSN [8]. Therefore, Sensor nodes will take more management responsibilities and decision making in order to achieve a self managed network. This study present a new cellular self-organization architecture for wireless sensor networks that extends the network life by efficiently utilizing nodes energy and distribute management tasks to support the scalability of management system in a densely deployed sensor networks. In our scheme, the whole network is divided into a virtual grid

where each cell consists of a group of nodes. A cell manager and a gateway node are chosen in each cell to perform management tasks. These cells combine to form various groups and each group promote one of their cell managers to a group manager. A group of cells are than managed by their group manager. In our scheme, both cell manager and gateway node mutually co-ordinate with each other to perform management tasks. We assume a homogenous network where all nodes are equal in resources. We considered energyefficiency and scalability of management systems as primary challenges in succeeding the vision of self-organized WSNs. Our architecture can be considered as a special kind of clustering architectures. However, it is more systematic, more robust and more scalable. This paper is organized as follows: Section 3 provides a brief review of related work in the literature. In section 4, we define the architecture model of our proposed solution Section 5 describes the management process Section 6 discusses our architecture by highlighting a few significant features. II. RELATED WORK Self organization (or self configuration) has been a significant research topic in wireless networks. Self organization involves abstracting the communicating entities into an easily controllable network infrastructure. Clustered or connected dominating set (CDS), grid, tree, or mesh based organization are key terms in self organization. A selforganized wireless node can be grouped or clustered into an easily manageable network infrastructure [7]. Clustering has been used to address various issues i.e. routing, energy efficiency, management and huge-scale control. Therefore clustering can be formed in several ways. Nodes generally form a cluster in two stages: (1) a header is selected among the nodes through election algorithm, randomized election, degree of connectivity or pre-definition, and (2) the headers and the nodes interact to form a group or a cluster [11]. Clustering is an efficient approach for building scalable and energybalanced applications. The scheme proposed in [12] is based on cluster formation. It divided the network into different clusters and a cluster head is appointed for each cluster. A cluster head has more resources than other cluster members. This cluster head perform major tasks and management operations. The data can only be transmitted to other clusters through cluster heads. Failure of a cluster head limits accessibility to the nodes under its supervision. Load balance clustering has been proposed to balance the load on cluster heads [13]. It deploys some less energy constrained nodes (called gateway) as cluster headers. This approach may not generate the shortest communication route. The wireless link between the header and the member nodes is asymmetric. In this scheme the header can reach the member nodes but the member nodes cannot necessarily reach the reply link. It verifies the problem of the density of clusters but does not show how to determine the ratio of headers to member nodes [11]. The architecture proposed in [15] is based on

hierarchical management of sensor nodes. This study presents an algorithm for self-organization mechanism of high-level nodes, contesting member nodes by multi-hop to form hierarchical clusters, and applying the ‘20/80 rule’ to determine the ratio of headers to member nodes. MANNA [19] is a policy based self managed architecture for wireless sensor networks that collects dynamic management information, map this into WSN models, and execute management functions and services based on WSN model. Thus, network can manage them with out human intervention. Management functions in MANNA represent the lowest granularity functional of a management service. A function can be shared by different services. MANNA is based on an agent based model that distributes the functionality in the hierarchical network management. in this way, energy efficiency and increased accuracy of management decision can be achieved. In this architecture, sensor nodes maintain their management role through out the network life time. However, this type of static approach is impractical in a highly dynamic sensor network [20]. Our proposed architecture enables sensor nodes to autonomously reconfigure their management role according to node real time capability such as energy. III. PROPOSED MECHANISM In this section we present our new cellular self-organized architecture for WSNs and an overall working flow of the system. A. Cellular architecture In this architecture we divide the network into a virtual grid of cells. A cell can be considered as a special kind of clustering. However it is more systematic and scalable. Cells can merge together to produce large cells that would be managed using the same process. We extended the architecture proposed in [14]. In our proposed solution the whole network is divided into a virtual grid using some type of virtual coordinate system. We consider sensor networks where each sensor node is aware of its own location. The network can use location services such as [16] and [17] to estimate the locations of the individual nodes, and no GPS receiver is required at each node.

Fig. 1. Sensor nodes For example, Figure 1 overlays virtual cells on Figure 2, creating three virtual cells A, B, and C. according to our definition of virtual grid, nodes in Cell A can communicate with all the nodes in Cell B and nodes in Cell B can reach all the nodes in Cell C.

Fig. 2. Division of the network into a virtual grid After the division of the network into small virtual cells as shown in figure 2, a cell manager is appointed in each cell. The cell manager then selects a gateway node and decides which common node will perform sensing and which will go to sleep. Cell manager receive updates from its cell member on regular basis. Both cell manager and gateway node mutually coordinate to perform management tasks. B. Distributed cell formation In order to elect cell managers and gateway nodes, any node can send a discovery message that consist of its node ID, Cell ID, and energy level and only nodes with higher energy would respond. The combination of Cell manager with gateway node guarantees a connected network. 1) Cell manager selection The node with the highest life time or energy will be appointed as a cell manager. Cell manager keep changing in each cell in order to extend network life time. 2) Gateway node selection Cell manager will select the gateway node on the basis of Maximum energy. Communication between cells takes place through gateway nodes. The cell manager broadcast a message to its cell members which in return send their updates including their node ID and energy level. Upon receiving updates from cell members, the cell manager appoints a gateway node. Both cell manager and gateway node stores node ID’s of their cell members. 3) Group manager selection After the selection of cell managers and gateway nodes, cells combine to form various virtual groups. Each group of cells then selects a group manager with mutual co ordination. A group manager is a cell manager which performs its normal tasks for its own cell but at the same time act as a group manager for a group of cells. This is shown in figure 3.

Fig. 3. Virtual cells in the form of a grid Cells 1,2,4,5 combine to form a group and with mutual co ordination they promote cell 9 cell manager to a group manager. Cell 5 manager will now perform two roles one as a cell manager to perform management tasks for its own cell and second as a group manager for a group of cells. The main goal of introducing this group manager is to perform high level management tasks and predict future faults. The selection of group manager is based on the available energy and it keeps changing in the group to balance the energy. Each cell maintains its health status in terms of energy. It can be High, Medium or Low. These health statuses are then sent out to there associate group managers. Upon receiving these health statuses, group manager predict and avoid future faults IV. OUR SELF-ORGANIZED ALGORITHM In this section, we describe our proposed algorithm which helps in self-organizing a set of sensor nodes, randomly scattered in an area. The algorithm consists of four phases and performs the following operation in the order they are mentioned. A. Discovery phase In discovery phase, each node turns on its radio and exchange discovery messages to find other nodes within the same cell. This discovery message is the combination of node id, cell id and node state. A node uses its location and size of the grid to determine the cell id. B. Organizational phase During this phase the network is organized and performs the following operations: 1) Nodes organized themselves into cells and cell combines to form virtual groups. 2) Each node is allocated with an address i.e. combination of node id and cell id. 3) The node with lowest coordinates starts acting as a cell manager. This is a criteria to select the cell manager for the first time in each cell.

4) All cell members send there updates to the cell manager, which in return appoint a gateway node for its region. 5) Cell manager also decide which node will perform sensing and will go to low computational mode to conserve energy. 6) Gateway node is responsible for routing information to other gateway nodes. 7) A group of cells can form a bigger cell to perform high level management task. A group manager is required for each group of cells and this can be achieved by promoting any cell manager in a group to a group manager. C. Maintenance phase In the maintenance phase the following operations are performed. 1) In active monitoring, every node keeps track of its energy level and sends regular updates to there cell managers. This is called in-cycle updates 2) Each cell manager aggregates its cell energy and sends to its group manager. This is less frequent than in-cycle update and called health cycle updates. This is to identify those cells, which no longer can participate in network operation due to insufficient energy. D. Self-organization phase In this phase a cell manager send a low energy node to sleep before it completely shut down. Cell manager ask common nodes and gateway nodes on regular bases to send there updates. If the cell manager does not receive an update from any node then it send an instant message to the node and acquire about its status. If cell manager does not receive the acknowledgement in a given time then declare it as a faulty node and pass this information to the rest of the network. The gateway node is like another common node and sends updates regularly to its cell manager. If the gateway node is low in energy then cell manager appoint another node to act as a gate way node and send the existence gateway node to low computational mode or sleeping mode. The gateway node can also be detected as faulty, if the cell manager does not hear from it during updates cycle. Also, a cell manager can be replaced by any other cell member as if it is low in energy. Each cell manager sends its health status information to its group manager through gateway nodes. This is less frequent than in-cell update cycle. If group manager doesn’t receive health status from a particular cell then it waits for the second health status update. If a group manager doesn’t receive the health status updates during the second update cycle then it informs the whole network about the occurrence of a faulty cell. A group manager can be replaced by

another cell manager in case of a fault or energy reason. V. MANAGEMENT PROCESS In most centralized management systems the base station acts as a central manager and controls the entire network. It collects information from all nodes and performs complex management tasks. The central station becomes a single point of data traffic for processing all the messages and making management decisions. But, this approach incurs a high message overhead (bandwidth and energy) from data polling, and this consequently limit network scalability. Some proposed solutions shows that as the central manager have the global knowledge of the network it can provide accurate management decision [9]. However, we believe that this cause network traffic overhead and limit network scalability, especially in large-scale sensor networks. Therefore, localized decision in certain degree reduces the network traffic overhead and has a quicker response to events that occurs in the network. Our proposed solution is based on homogeneous paradigm, where all nodes are equal in resources. Common nodes periodically send out their status (i.e. current energy levels) only on request from cell managers. Both cell manager and gateway node perform management task with mutual coordination. They are responsible for the members of their own cell. The main emphasize is to perform management functions locally. Both cell manager and gateway node are at the same management level. Cell manager and gateway nodes can be replaced by common nodes in case of backup and recovery. Another layer of management comprises of group managers, which are responsible for operation like detecting faulty cells and predict future faults. VI. DISCUSSION Energy-efficiency is an important research challenge to succeed the vision of a self organized wireless sensor network. Our approach addresses this challenge by employing a load balancing strategy so that all nodes remain up and running together for as long as possible. Therefore, we consider that all the nodes in the network are equal in resources and no node should be more resourceful than any other node. The optimal role assignment and reconfiguration scheme support the network management system to utilize the network nodes in the most efficient manner. Our approach does not rely on specific nodes with extra resources but assign tasks due to there optimal capabilities. Nodes are ranked according to their available energy. Therefore, the selection of the gateway node and a cell manager is based on the available energy. The basis idea of this design is to encourage nodes to be more selfmanageable and extend the network life time for as long as possible. In centralized approaches, information flow is towards a single point, which limits the scalability of the network and cause traffic overhead. Therefore, the more local decision a node can make, the less information is required to

be delivered to the central manager or base station. The proposed architecture efficiently distributes management tasks across the virtual grid to reduce the valuable network traffic and node energy. Fault tolerance is the ability to maintain sensor networks functionalities without any interruption due to sensor nodes failure. Sensor node may fail due to lack of energy, communication problem, physical damage or interference from environment. The network should be able to identify faults and organize itself to reconfigure and recover from failure without losing any information [18]. The proposed architecture performs fault detection and recovery locally with minimum energy consumption. Both cell manager and gateway node detect faulty node through mutual coordination. If the cell manager does not receive an update from any node then it send an instant message to the node and acquire about its status. If cell manager does not receive the acknowledgement in a given time then declare it as a faulty node and pass this information to the rest of the network. Specifically, the grid based architecture permits the implementation of fault detection in a distributed manner and allows the failure report to be forwarded across cells. It is a simple grid base architecture which supports lightweight operations. The main idea to propose a model that doesn’t consume much energy or memory. It is based upon the virtual coordination due to which cells are built automatically and there is no need to exchange too much messages to build the clusters. The cell member remains on the same cell regardless of the cell manager. The cellular architecture is for management purpose only so they can be merged into clusters for routing or any other purpose if needed. VII. CONCLUSION WSN has presented various research challenges and one major challenge is to design an efficient self-organized architecture of wireless sensor network. In this paper we described a new cellular self-organized architecture for wireless sensor networks. We divide the whole network into a virtual grid, where each cell comprises of a group of nodes. Decomposition of the network in to a virtual grid supports the distribution of management tasks to scale the management system in a densely deployed sensor network. Unlike centralized system, traffic is not directed toward the single point and this conserve valuable network traffic and energy. By employing nodes with equal resources adopt the load balancing strategy which enables the nodes to remain up and running for as long as possible. This extends network life time and encourages network nodes to be more self-manageable.

REFERENCES [1] A. Bharathidasas, and V. Anand, “Sensor networks: An overview”, Technical report, Dept. of Computer Science, University of California at Davis, 2002 [2] D. Estrin, R. Govindan, J. Heidemann, and S. Kumar, "Next century challenges: Scalable coordination in sensor networks", in Proceedings of ACM Mobicom, Seattle, Washington, USA, August 1999, pp. 263-270, ACM. [3] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam and E. Cayirci, "A Survey on Sensor Networks", IEEE Communications Magazine, pp. 102--114, August 2002. [4] D. Estrin, L. Girod, G. Pottie, M. Srivastava, “Instrumenting the world with wireless sensor networks”, In Proceedings of the International Conference on Acoustics, Speech and Signal Processing (ICASSP 2001. [5] E. S. Biagioni and G. Sasaki, “Wireless sensor placement for reliable and efficient data collection”, in the 36th International Conference on Systems Sciences, Hawaii, January 2003. [6] R. Krishnan and D. Starobinski, “Efficient clustering algorithms for self-organizing wireless sensor networks”, Ad Hoc Networks, 2005. In print. [7] M. Kochhal, L. Schwiebert, and S. Gupta, “Rolebased hierarchical self organization for wireless ad hoc sensor networks”, In Proc. of the 2nd ACM Intl. Conf. on Wireless Sensor Networks and Appl., pages 98--107, 2003. [8] M. Yu, H. Mokhtar, M. Merabti, “A survey of network management architecture in wireless sensor network”, in: Proceedings of the Sixth Annual PostGraduate Symposium on The Convergence of Telecommunications, Networking and Broadcasting, 2006. [9] W. L. Lee, A. Datta, and R. Cardell-Oliver, “Network Management in Wireless Sensor Networks”, to appear in Handbook on Mobile Ad Hoc and Pervasive Communications, edited by M. K. Denko and L. T. Yang, American Scientific Publishers. [10] M. Younis, M. Youssef and K. Arisha, "EnergyAware Routing in Cluster-Based Sensor Networks," Proceedings of the 10th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS2002. [11] J.L Chen , H.F Lu and C.A Lee, “Autonomic selforganization architecture for wireless sensor communications”, International Journal of Network Management, v.17 n.3, p.197-208, June 2007. [12] A. T. Tai, K.S.T., William H. Sanders, “ClusterBased Failure Detection Service for Large-Scale Ad Hoc Wireless Network Applications in Dependable Systems and Networks”, DSN ' 04. 2004.

[13] G. Gupta and M. Younis, “Load-Balanced Clustering in Wireless Sensor Networks”, in the Proceedings of International Conference on Communication (ICC 2003), Anchorage, AK, May 2003. [14] Y. Xu, J. Heidemann, and D. Estrin, “Geographyinformed energy conservation for ad hoc routing”, In Proceedings of the ACM/IEEE International Conference on Mobile Computing and Networking, pages 70--84, Rome, Italy, July 2001. [15] C.-A. Lee, Y.-C. Chang, and J.-L. Chen, "Intelligent self-organization management mechanism for wireless sensor networks," Proc. 6th International Conference on Wireless Communications, pp.47–54, July 2004. [16] A. Savvides, C. Han, and M. Strivastava, “Dynamic finegrained localization in ad-hoc networks of sensors”, In Proc. Of ACM MOBICOM’01, pg 166– 179. ACM Press, 2001.

[17] L. Doherty, L. El Ghaoui, and K. S. J. Pister, “Convex position estimation in wireless sensor networks”, In Proc. Of IEEE INFOCOM 2001, Anchorage, AK, April 2001. [18] L. B. Ruiz, I. G.Siqueira, L. B. Oliveira, H. C. Wong, J.M. S. Nogueira, and A. A. F. Loureiro. “Fault management in event-driven wireless sensor networks”, MSWiM’04, October 4-6, 2004, Venezia, Italy [19] L. B. Ruiz, J. M. S. Nogueira, and A. A. Loureiro, “MANNA: A management architecture for wireless sensor networks,” IEEE Communications Magazine,41(2):116{125, Feb 2003. [20] M. Yu, H. Mokhtar, M. Merabti, “A self-organized middleware architecture for wireless sensor networks”, Int. J. Ad Hoc and Ubiquitous Computing, Vol. 3, No. 3, 2008