Research On Target Tracking Based On Cluster In Wireless Sensor Networks

In recent years, with the continuous development of modern science and technology,wireless sensor networks is increasingly being concerned by scientific researchers, and hasbeen applied in the military field, emergency situations, scientific research, business and otherfields. One of these applications is target tracking. Target tracking is an important applicationfor wireless sensor networks, with high tracking precision, good real-time, low cost, goodreliability and so on.This paper begins with an overview of wireless sensor networks, and gives anclassification of current researches on target tracking in wireless sensor networks. For theproblem of target tracking in wireless sensor networks, this paper focuses on target trackingnetwork model and target state estimation mechanism. In target tracking network model, thispaper uses the mechanism of cluster to manage and schedule sensor nodes. On the basis ofLeach cluster management mechanism and energy consumption model of wirelesscommunication, cluster management mechanism called EDC,based on energy and distance,is proposed. And in the case of the loss of target, a target recovery mechanism is proposed. Inthe target state estimation, this paper uses sound strength distance model to get informationabout the location of the target, and utilizes the idea of filter to estimate the state of the target.On the basis of traditional Kalman filter, this paper uses the method of Gradient-Newtonbased on likelihood function to calculate the preliminary location of the target. Then the valueof preliminary location will be intended as observation value for follow-up Kalman filter,which then estimate the state of the target.The EDC mechanism takes the residual energy of the node and the topology betweennodes as important factor to establish and manage the cluster. It selects the nodes appropriatein energy and position to join the cluster. A node which has big power and is nearest to othernodes in the cluster will be selected as cluster head, and nodes who have appropriate weightof energy and distance will become the members of the cluster. Then the cluster will beupdated in the way of cycle, which can avoid the same cluster or single node to work for longtime. From the simulation results, compared to traditional Leach algorithm, the EDCmechanism can effectively balance the energy load and reduce overall network energyconsumption. The state estimation of this paper intend the result of Gradient-Newton based onlikelihood function as the value of observation for Kalman filter. This may sacrifice theperformance of delay, but the final effect is also acceptable, while the benefit is able to effectively reduce the effect of noise in observation equation and Kalman filter is no longerover-reliance on state transition model. State transition error within a certain range can becorrected by observation model. Meanwhile, since the value of preliminary location has highconfidence, intending the value as observation value can make filter converged faster to thedesired result...