Research And Improvement Of Directed Diffusion Algorithm For WSN

Wireless sensor networks(WSN) are composed of lots of sensor nodes, which are deployed randomly over the WSN fields. Sensor nodes, the main components of WSN, are tightly constrained in terms of storage capacities, data processing, and energy which is provided by limited unchangeable battery. The network lifetime is determined by the remaining energy of nodes. Therefore, high efficient use of energy is the key point of wireless sensor networks. Directed Diffusion(DD) algorithm applied to small scale networks with large amount of information per query. This algorithm establishes a reinforcement path from source to sink, though interest diffusion and data propagation. In practical applications, there is more than a source node, thus will leads to more than one reinforcement path. In order to solve this problem, Directed Diffusion algorithm adopts a data integration strategy though intermediate nodes to reduce the unnecessary energy waste due to the redundant link.Regarding to the defect of multi-sources in Directed Diffusion algorithm, a cluster method is applied among the sources nodes, and a Directed Diffusion algorithm based on the cluster (DD-CS) is proposed. DD-CS employs the directed diffusion algorithm to find the source nodes by flooding the interest messages layer of layer. Once the source nodes are uncovered, they formed a temporary cluster and choose a cluster head by the centripetal degree of node. Later, a reinforcement path between the cluster head and sink node is developed. Finally, the cluster is dismissed after the monitor task. The marked characteristic of DD-CS is that all the sources nodes are gathered in a cluster, where the cluster head in charge of the data integration and send the data to the sink node. As a result, network only have to maintain one path from source to sink, which reduce the redundant path and energy consumption of the network. The centripetal degree of node is calculated by the Sum of Squares between the other source nodes and the cluster head, which is linear proportional to the energy consumption consumed by the data integration. The minimum centripetal degree of node guarantees the least energy consumption on data fusion.It was shown via matlab that DD-CS reduce energy consumption compared to regular DD in the multi-sources network.