Adaptive Dynamic Programming-Based Sensor Scheduling for Multi-Target Tracking in Energy Harvesting Wireless Sensor Networks

This paper investigates sensor scheduling strategies for sustainable multi-target tracking in energyharvesting wireless sensor networks (EH-WSNs). Addressing the fundamental trade-off for tracking accuracy and energy efficiency in perpetually-operated networks, we develop an adaptive optimization framework combining reinforcement learning with nonlinear estimation. The proposed solution features: 1) an Extended Kalman Filtering (EKF) predictor for target state estimation, 2) an Adaptive Dynamic Programming (ADP) based approach for sensor activation scheduling, and 3) an energy-aware target selection mechanism. Simulation results validate the effectiveness of our approach for long-term surveillance applications in the EHWSNs.