Subscription-Based Integrated 5G-Radar Tracking System With Dynamic Resource Management Using Reinforcement Learning

By integrating radar services into existing 5G networks, we enable radar applications using 5G waveforms with minimal interference to existing cellular networks and no dedicated infrastructure. Prior work demonstrates that this "radar-as-a-subscriber technology" (RAST) approach achieves reasonable target detection and range estimation performance even with constraints on the number and allocation of time-frequency resource blocks. We previously demonstrated that the specific placement of 5G resource blocks can be manipulated to favor range estimation performance or sidelobe level of a RAST subsystem. In dynamic environments where target position, noise, interference, attenuation, and system priorities are dynamic, the joint detection-estimation performance of the RAST subsystem is significantly constrained if the resource allocation strategy is static or random. To optimize the RAST subsystem performance, we propose an adaptive resource manager that uses reinforcement learning to optimize the selection of 5G resource blocks in response to dynamic environmental conditions. We demonstrate that this strategy achieves performance improvements of up to 40% compared to a random resource allocation strategy in a simple MATLAB simulation environment.