Section 20.5. Case Study: Simulation of a Sensor Network


20.5. Case Study: Simulation of a Sensor Network

This section presents a case study that shows the implementation of DEEP and ICR for a wireless sensor network spread over an area. The network is used for monitoring and protecting the area. The basic objective is to deploy a large number of low-cost and self- powered sensor nodes, each of which acquires and processes data from a hazardous event and alerts a base station to take necessary action. In this scenario, 3,000 sensor nodes are randomly distributed in a field of 550 m x 550 m. Therefore, the density of sensor nodes is about one per 10 m x 10 m area, which is the maximum detection range for the hazard sensors.

MAC assigns a unique channel for every node and prevents possible collisions. With this assumption, we extracted the MAC layer from our simulations, and data packets were sent directly from the network layer of one node to the network layer of the neighbor. We simulated the DEEP algorithm, using parameters d r , d r 1 , d r 2 , and m, and put the initial cluster head at the center of the field.

20.5.1. Cluster-Head Constellation and Distribution of Load

Figure 20.12 shows the result of the simulation with parameters d r = 30 m, d r 2 = 80 m, d r 1 = 78 m, m = 14. Based on the results obtained from Section 20.2, the distance of 30 meters is an initial choice for d r . In order to avoid overlapping between clusters, the value of d r 1 and d r 2 should be more than twice the value of d r . Since the average distance between sensor nodes in this application is 10 m, 80 m is a fair choice for d r 2 . The width of the ( d r 1 , d r 2 ) ring should be large enough to accommodate new cluster-head candidates and small enough to avoid cluster-head candidates that are too close to each other. We chose an initial value 2 m for the ring width.

Figure 20.12. Simulation results on distributed clusters whose sensor nodes are directly connected to their associated cluster heads. The initial cluster head is put in the center of the sensor field, the simulation starts by advertising its candidacy, and cluster heads are gradually dispersed across the network.


In order to balance the load among cluster heads, DEEP controls the cluster-head distribution rather than the number of cluster members. Although cluster heads that manage more members should execute more signal processing for the sake of data aggregation, digital processing consumes much less energy than wireless transmission, and no overutilized cluster head is using this protocol.

Figure 20.13 demonstrates the cluster-head distribution achieved using LEACH and DEEP. Because of the random selection of cluster heads in LEACH, some of the cluster heads are too close to each other; others, too far. This type of cluster-head selection causes a lot of burden on some cluster heads and quickly drains their batteries. It can be shown that compared with LEACH, DEEP is capable of minimizing energy consumption associated with reclustering overheads more efficiently by reducing the number of necessary rounds.

Figure 20.13. Comparison of cluster-head constellation between (a) LEACH and (b) DEEP. DEEP generates well-distributed cluster heads across the network.

20.5.2. Optimum Percentage of Cluster Heads

In order to determine the optimum cluster-head density and compare the performance of the routing protocol on both DEEP and LEACH, we used a 1,600-node network.

Nodes were randomly distributed in a field 400 m x 400 m. In this scenario, sensor nodes send the information directly to their associated cluster head. Each cluster head compresses the data and waits for neighbor cluster-heads' data packets. Then, the cluster head compresses all the received data packets into a packet with fixed length and sends it to the relay neighbor. The relay neighbor address has been saved in node memory through the propagation of the interest signal. In-network data aggregation performed by cluster heads helps to reduce the amount of data dispersed in the network.



Computer and Communication Networks
Computer and Communication Networks (paperback)
ISBN: 0131389106
EAN: 2147483647
Year: 2007
Pages: 211
Authors: Nader F. Mir

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net