Reinforcement Learning for FDA-MIMO Radar Power Allocation in Congested Spectral Environments

Abstract

This paper studies the coexistence problem of frequency diverse array multiple-input multiple-output (FDA-MIMO) radar and communication systems. The communication system and radar share the same frequency band, and the overlap of the frequency spectrum will interfere with each other's performance, which constitutes a dynamic non-cooperative coexistence problem. We model the FDA-MIMO radar interference suppression process as a Markov decision process (MDP). The interference signals received by FDA-MIMO radar are generated by the communication system, including constant interference, frequency-hopping interference, and intermittent interference. FDA-MIMO radar realizes flexible transmission waveform spectrum control through transmission power allocation. Reinforcement learning (RL) is applied to obtain the optimal power allocation strategy of FDA-MIMO radar, and it is compared with the Sense and Avoid (SSA) algorithm. The simulation results show that the FDA-MIMO radar power allocation method based on reinforcement learning can effectively suppress the communication system's interference and improve the radar's signal-to-interference-and-noise ratio (SINR) and range resolution.