Factored Multi-Agent Soft Actor-Critic for Cooperative Multi-Target Tracking of UAV Swarms

In recent years, significant progress has been made in the multi-target tracking (MTT) of unmanned aerial vehicle (UAV) swarms. Most existing MTT approaches rely on ideal assumption a pre-set target trajectory. However, practice, trajectory moving cannot be known by UAV advance, which poses great challenge for realizing real-time tracking. Meanwhile, state-of-the-art multi-agent value-based methods have achieved cooperative tasks. contrast, actor-critic (MAAC) face high variance and credit assignment issues. To address aforementioned issues, this paper proposes learning-based factored soft (FMASAC) scheme under maximum entropy framework, where swarm is able to learn an unknown environment. This method introduces idea value decomposition into MAAC setting reduce policy updates efficient assignment. Moreover, further increase detection coverage swarm, spatial reward (SER), inspired concept, proposed scheme. Experiments demonstrated that FMASAC can significantly improve performance it outperforms baselines terms mean success rates. Additionally, scales more successfully as number UAVs targets increases.