Multi-Agent Reinforcement Learning Assisted Multi-Task Decision Making for UAV Swarm

The rapid development of mobile communication has led to a shortage of wireless resources, while the unmanned aerial vehicles (UAVs) has created a growing demand for these resources. This paper investigates the resource allocation problem for multiple UAVs in a radar communication integrated scenario. We propose a joint efficiency metric for evaluating the performance of radar communication, and use a deep reinforcement learning algorithm to establish bit-level end-to-end communication between UAVs for the sake of efficiently and cooperatively carrying out the tasks. In the training process, parameter sharing is conducted to accelerate the algorithm convergence rate. Simulation results show that the proposed algorithm has good convergence and can maximize the performance metric of UAVs.