A Joint Radar and Communication Approach for 5G NR using Reinforcement Learning

Radar operation partly overlaps and thus interferes with 5G spectrum bands. Examples are vehicular radar, long-range air traffic control, terminal air traffic control, marine radar, airport surveillance, or also operations in the mmWave band. We propose a mechanism to efficiently share the spectrum resources for communication and radar operation using a Reinforcement Learning (RL)-based approach. Unlike the state-of-the-art, our approach enables both systems to keep their own waveforms. Compared to the use of a single waveform for joint radar and communication, this results in less complex signal processing and improved sensing resolution. Our approach is compatible with existing radar systems and requires software modification only for the communication system. We demonstrate how both systems can work simultaneously, thereby eliminating the need for time sharing. The effectiveness of the approach is studied through a comprehensive set of experiments implemented in an open source simulation environment. It is shown that in the presence of interference, the radars can still achieve a high accuracy for range and velocity estimation of targets. The system achieves high spectrum utilization and is on-demand adjustable to realize any desired level of trade-off between communication and sensing.