Collaborative Target Tracking in WSNs Based on Maximum Likelihood Estimation and Kalman Filter

Target tracking using wireless sensor networks requires efficient collaboration among sensors. Existing collaborative tracking approaches based on the extended Kalman filter, as addressed in many previous papers, suffer from low tracking accuracy. In this paper, we present a new collaborative target tracking approach in wireless sensor networks based on the combination of maximum likelihood estimation and Kalman filtering. The leader firstly converts the nonlinear measurements collected from the scheduled sensors into a linear observation model in target state using maximum likelihood estimation-based localization, then applies a standard Kalman filter to recursively update the current target state and predict the future target location. Lastly, an information measure based on the Fisher information matrix (FIM) is proposed to select the most informative sensors and one of them is designated as the leader for the next time tracking so as to achieve more tracking accuracy under the energy constraint.