An Improved Q-Learning-Based Sensor-Scheduling Algorithm for Multi-Target Tracking

Target tracking is an essential issue in wireless sensor networks (WSNs). Compared with single-target tracking, how to guarantee the performance of multi-target tracking is more challenging because the system needs to balance the tracking resource for each target according to different target properties and network status. However, the balance of tracking task allocation is rarely considered in those prior sensor-scheduling algorithms, which may result in the degradation of tracking accuracy for some targets and additional system energy consumption. To address this issue, we propose in this paper an improved Q-learning-based sensor-scheduling algorithm for multi-target tracking (MTT-SS). First, we devise an entropy weight method (EWM)-based strategy to evaluate the priority of targets being tracked according to target properties and network status. Moreover, we develop a Q-learning-based task allocation mechanism to obtain a balanced resource scheduling result in multi-target-tracking scenarios. Simulation results demonstrate that our proposed algorithm can obtain a significant enhancement in terms of tracking accuracy and energy efficiency compared with the existing sensor-scheduling algorithms.