Sensor management method based on TD3 in multi-target tracking

Abstract This paper proposes an innovative sensor management method utilizing deep reinforcement learning to optimize the performances of multi-target filters based on random finite sets. Within the theoretical framework of partially observable Markov decision processes, the twin delayed deep deterministic policy gradient reinforcement learning algorithm is employed to decide optimal sensor management strategies in continuous space. For each control command, predicted ideal measurement set is generated to update posterior multi-target density. The divergence between the predicted multi-target posterior density and the updated posterior multi-target density is used as a reward function to guide the decision of sensor management strategies. The simulation results demonstrate that, compared to other algorithms, the proposed method is more effective in enhancing the performance of multi-target tracking.