Professional Documents
Culture Documents
information or is having more than one directional antenna. Using a location information system or directional antenna is a very costly approach for wireless sensor networks. LORD [5] is a localized, reactive and distributed network coverage protocol. It conserves overall system energy by minimizing the number of active nodes, localizing the execution to the dying sensor(s), and minimizing the frequency of execution by reacting only to the occurrence of a sensing hole. It consists of two phases, the set-up phase and the steady phase. When the network is deployed initially, the setup phase runs, at each node, to determine its initial state, which is ON or OFF. Once a sensor node chooses to be ON, it runs still it dies or receives an intimation to be turned OFF. So fairness is not assured in this protocol. The steady phase algorithm is invoked before a sensor si dies to find a set of replacement nodes, si.replace. This set consisting of a minimum number of sleeping neighbors, which can cover si. To take such a decision each node should maintain the list of its sleeping neighbors. Moreover, it assumed a perfect and reliable communication layer and location information is available. CCP [6] is an integrated coverage and connectivity protocol (CCP), which can configure the network to the required degree of coverage. Each node takes a local decision to be active or not based on location information of it and its neighbors. Location information of neighbors is maintained in a neighbor table with the help of Hello messages. Each node runs an eligibility algorithm to decide to be active or not. The eligibility algorithm checks whether all intersection points1 inside its sensing circle are covered by some neighboring sensor or not. If so, it goes to sleep state otherwise it announces itself as active sensor by sending a Hello message to all its neighbors. In CCP, Rc > 2 Rs is a sufficient condition for coverage to imply connectivity. But for the case Rc < 2 Rs, it uses Span [2] to provide network connectivity. The eligibility algorithm assumes each node to know its geographical location, which is a costlier approach. The existing protocols, on node scheduling, treat coverage and connectivity separately. Moreover, all the proposed solutions are dependant on the position of the nodes. Therefore, we need a protocol with the following requirements: first, self configuration is mandatory because it is inconvenient or impossible to manually configure sensors after they have been deployed in hostile or remote working environments. Second, the design has to be fully distributed and localized, because a centralized algorithm needs global synchronization and is not scalable to large networks. Third, the algorithm should allow as many nodes as possible to be in sleep most of the time. At the same time, it should preserve the application defined QoS. Fourth, the protocol should be position independent. We consider these design issues and proposed scalable energy efficient position independent (SEEPI) node scheduling algorithm. III. SEEPI NODE SCHEDULING In this section we present a scalable energy efficient position independent (SEEPI) node scheduling in WSNs. Our protocol makes the following design assumptions:
Nodes in the network are static and data delivery model to the sink is event driven. Node deployment is nondeterministic and the density of nodes is high in every region. All sensor nodes are homogeneous. That is, having equal communication, computation, and sensing capability. And node failure is possible only when it used up all its energy. We denote the communication range and sensing range of a node by Rc and Rs respectively. The SEEPI is mainly divided into three phases. Set of nodes selected in the first two phases ensures network connectivity. The third phase nodes helps in maintaining the coverage. A. Node Election for Maintaining Connectivity A set of nodes to ensure network connectivity can be elected either using the SPAN [2] or LECA [7]. SEEPI uses LECA, since it selects less number of nodes compare to SPAN, to conserve power. For the sake of completeness we briefly discuss the LECA node selection algorithm in the following paragraphs. Each node periodically broadcasts Hello messages that contain the nodes status (active or not), its current active neighbors, and its current neighbors. From these Hello messages, each node maintains a list of neighbors and active neighbors, and for each neighbor, a list of its neighbors and active neighbors. Nodes in the network wake up periodically and collect this information by listening for a brief period before participating in the election algorithm.
Phase1 Sensor
Fig. 1. Scenario after Phase 1 1) Phase-1 (Independent Set Construction): The first phase is a process to select the best set of nodes that constitute the minimal independent set for the network at hand. This selection is done in a distributed manner using
the neighbor information collected through Hello packets during listen period. Each node decides to be a active or not based on the following heuristics: Nodes that have higher number of neighbors without a phase-1 sensor are given more preference. Because, such a nodes can help a larger number of neighbors in communication. Nodes that have power greater than certain threshold or higher amount of percentage of remaining power are given more preference over others to be elected as phase-1 sensors. This introduces the required fairness in the protocol by ensuring proper rotation in the selection of phase-1 sensors. These selected phase-1 sensors go back to sleep after they have used up a fixed percentage of their power to ensure fairness by allowing other nodes to become phase-1 sensors. Announcement contention occurs when multiple nodes discover the lack of a phase-1 sensor at the same time, and all decide to become a phase-1 sensor. It resolves with a randomized back-off delay. That is, each node chooses a random delay value, and delays the Hello message that announces the nodes volunteering as a phase-1 sensor for that amount of time. The delay value depends on the extent to which the node is satisfies heuristics. At the end of the delay, the node reevaluates its eligibility based on Hello messages recently received, and makes its announcement if and only if the eligibility rule still holds. Fig. 1 shows a scenario after phase-1.
Nodes having higher amount of percentage of remaining energy. Nodes having more number of phase-1 sensors in the 1-hop neighborhood. This rule helps in selecting a sensor which can connect more phase-1 nodes. Contention, if any, is also resolved using the back off mechanism discussed earlier.
Phase1 Sensor
Phase2 Sensor
Phase3 Sensor
Fig. 3. Scenario after Phase-3 B. Election for Maintaining Coverage (Phase-3) First two phases of the election algorithm ensures connective. The third phase of the algorithm ensures coverage by electing more sensors. Moreover, it elects a small number of sensors to avoid any large redundant coverage. We have to choose nodes which are going to cover large uncovered area. An inactive node need not to be selected if it is very close to an active node, because, such a nodes coverage area largely overlap with the near by active node. Another observation is, if an inactive node is far away from the active node, no need to consider such far away active nodes for deciding the status of inactive node, because intersection of the coverage area of these nodes is very less. Experimentally, we have fixed two parameters dmin = 0.2 Rs and dmax = 1.4 Rs to classify very close and far away neighbor respectively. We need to estimate the distance of each 1-hop neighbor to make use of dmin and dmax. This can be done using received signal strength. Distance of each 1hop neighbor is maintained in the neighbor table and sends along the Hello messages. Sensors which are not selected in first two phases are participating for deciding their status using the following heuristics. If there is an active sensor within range dmin then node does not become active. This heuristic is based on the intuition that if such sensor will become active then it will not help in covering much area but will increase redundancy.
Fig. 2. Scenario after Phase-2 2) Phase-2 (Connecting the Independent Set): In the second phase, more nodes are selected to connect the phase1 sensors and make connected network. The node selection is based on the following criteria:
If there is no active sensor or only one active sensor with in the range dmax then the node is eligible to become active in this phase. Such node uses a random back-off delay for announcing its status as active node. After this delay period, it will again check its eligibility and will make an announcement if eligible. If there is more than one active sensor with in the range dmax then it will check whether there is any pair of nodes whose distance between them is less than or equal to dmax. If yes then the node is not eligible. This is because if such node will become active then it will not help in covering much area but increase redundancy. Nodes, which are not selected in these three phases, are go to sleep to conserve energy. After spending a fixed amount of time in the sleep state, each node comes up and participates in the election process, by checking eligibility criteria of three phases discussed above. C. Node Withdrawal In order to rotate the node active status among all nodes fairly, each node will withdraw after some fixed interval of time or when it has used up some percentage of its energy. Each node marks itself as a tentatively active and announces its withdrawal by sending Hello message. A tentatively active node still forwards packets and sense the environment. However, the election algorithm described above treats a tentatively active node as a non-active. A node stays tentatively active for some fixed amount of time, and then goes to sleep mode. IV. SIMULATION ENVIRONMENT AND RESULTS We simulated our algorithm in the NS-2.28 network simulator with CMU wireless extension and compared with CCP. The CCP code for NS-2.1b1 is obtained from [6] and ported to NS-2.28. We have used the following simulation scenario.
side of the region. Each CBR flow sends 128 byte packets with 3Kbps rate. We have used AODV routing because it suits the event driven application environment where each node upon detecting an event sends data to sink node and route is established on the fly. Nodes in our simulation use radio with a 2Mbps bandwidth and a sensing range of 50m. We used TwoRayGround radio propagation model in all NS-2 simulations.
Fig. 6. Coverage Vs Rc/Rs To measure the performance of both CCP and SEEPI under different ratio of communication range/sensing range, we varied the communication range by setting appropriate value of transmission power. The results are average of five different randomly chosen scenarios for 500 seconds. Fig. 4, 5, 6, and 7 represent number of active nodes, average energy consumption, coverage percentage, and delivery ratio respectively of the protocols SEEPI and CCP with varying ratio of Rc/Rs. The ratio of successful events reported to sink and total events detected defined as delivery ratio. Fig. 4 shows that number of active nodes decreases as Rc/Rs increases for both the protocols. This is because as communication range increases the nodes required to make
Fig.4. Number of active nodes Vs Rc/Rs Both the protocols were run on top of the 802.11 MAC layer in 400 X 400 m2 coverage region with 160 randomly distributed stationary nodes. Two sources and two sinks are placed in opposite sides one at each corner. Each of the sources sends a CBR flow to sink node located on the other
network connected decreases. Moreover, it show that our protocol results in less number of active nodes. As our protocol activates less number of nodes its average energy consumption per node is less than that of CCP and that is represented in Fig. 5. The CCP uses node location information for activating sensors, where as SEEPI does without it. SEEPI achieves near 100\% coverage for all cases as shown in Fig. 6.
V. CONCLUSION AND FURTHER REMARKS This paper explores the problem of energy conservation in wireless sensor network while maintaining both connectivity and acceptable level of sensing coverage. Our simulation results show that the coverage and delivery ratio achieved by SEEPI are nearer to that of CCP, while average energy consumption of node is less than that of CCP without using location information. Our design uses local information stored in neighbor table, so it is efficiently scalable. Moreover, our protocol works under node mobility also. It would be interesting to study how SEEPT works for query driven applications and target tracking applications. REFERENCES
[1] [2] Kemal Akkaya and Mohamed Younis, A survey on routing protocols for wireless sensor networks, Ad Hoc Networks, vol. 3, no. 3. pp. 325349, 2005. B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris, Span: An energy-efficient coordination algorithm for topology maintenance in ad hoc wireless networks, Wirel. Netw., Vol. 8, no. 5, pp. 481494, 2002. A. Cerpa and D. Estrin, ASCENT: Adaptive Self-Configuring sensor Networks Topologies, IEEE Transactions on Mobile Computing, vol. 3, no. 3, pp. 272285, 2004. D. Tian and N. Georganas, "A Coverage-preserving Node Scheduling Scheme for Large Wireless Sensor Networks", 2002 [online] citeseer.ist.psu.edu/tian02coveragepreserving.html. A. Ghosh and T. Givargis, LORD: A Localized, Reactive and Distributed Protocol for Node Scheduling in Wireless Sensor Networks, in DATE, pp. 190-195, 2005. X. Wang, G. Xing, Y. Zhang, C. Lu, R. Pless, and C. D. Gill, Integrated coverage and connectivity configuration in wireless sensor networks, in SenSys, pp. 2839, 2003. R. Hegde, Power aware mobile ad hoc networks", M.Tech Thesis, Indian Institute of Technology Guwahati, 2005.
[3] [4]
[5]
Fig. 7. Packet Delivery Ratio Vs Rc/Rs The simulation results shows that SEEPIs performs is at par with CPP with less power consumption and with out using location information.
[6] [7]