Radar Network Time Scheduling for Multi-Target ISAR Task With Game Theory and Multiagent Reinforcement Learning

Abstract

In this paper, contrapose the impendency for multi-target high-resolution imaging with the limited resources, a radar network time scheduling is proposed based on game theory and reinforcement learning for inverse synthetic aperture radar (ISAR) imaging task regarding the targets in different radar beams. According to the demand for using the least amount of time to achieve the expected imaging resolution, the radar observation time scheduling problem is formulated. The game behaviour in the optimization problem is analyzed, and a time scheduling game is constructed to acquire the time scheduling strategy. For the purpose of finding out the optimal strategy profile, an equilibrium-based multiagent reinforcement learning (MARL) for the time scheduling game is proposed. Simulation results demonstrate that the time scheduling game belongs to exact potential game and can converge to the optimal strategy profile of the radar observation time scheduling problem by the proposed equilibrium-based MARL. Besides, the learning ability of the equilibrium-based MARL is proved.