Improved average consensus scalable algorithm of target tracking for wireless sensor network

Target tracking is one of the most important applications for wireless sensor networks. In this paper, a scalable dynamical average consensus algorithm of target tracking is proposed. Kalman prediction is employed to help setting tracking cluster through nearest neighbor criterion. Selected sensors, called Tasking sensor nodes, are working locally for target estimate through observation or innovation based consensus distributed kalman filter. In the algorithm, Cluster head is not fusion center, but a ordinary node for choose tasking sensor of next step and forwards current state estimate and corresponding error covariance. Each tasking sensor only requires information from neighboring nodes that will decrease communication bandwidth of overall network. Simulation results show that, compared with central kalman filter algorithm, observation and innovation based consensus distributed kalman filter will get comparable accuracy while it will be more robust due to distributed manner.