Finite-Horizon Adaptive Dynamic Programming for Collaborative Target Tracking in Energy Harvesting Wireless Sensor Networks

Abstract

Collaborative target tracking is a typical application of wireless sensor networks (WSNs), in which sensors shall be scheduled for trading off between the tracking performance and the energy utilization. Energy harvesting in wireless sensor networks is attractive for the continuous operation of the network, but poses new challenges for collaborative target tracking due to the limited energy harvesting capabilities of the nodes. In this paper, we will propose a novel finite-horizon adaptive dynamic programming (FHADP) based sensor scheduling for collaborative target tracking in an energy harvesting WSN. Unscented Kalman filter (UKF) is used for prediction and estimation of the tracking performance, and multiple-step sensor scheduling is performed using ADP based on the predictive harvested energy of the nodes and tracking performance. Simulation results show that FHADP based sensor scheduling scheme can obtain superior tracking performance compared with ADP.