UAV-Assisted NOMA for Enhancing ISAC: A Deep Reinforcement Learning Solution

This letter examines a NOMA downlink scenario where the UAV is deployed to concurrently assist in communication and sensing functionalities, empowering ISAC technology. In this regard and with the goal of maximizing the average achievable rate, we formulate an optimization problem to determine the UAV trajectory, and beamforming vectors at the transmitting base station and UAV, while at the same time satisfy the quality of service constraints for communication users and sensing for a moving target in terms of Cramer Rao bound (CRB) metric. The formulated problem showed to be non-convex and hard to be solved because of the high coupling between the variables as well as the randomness in the environment due to the channels variations and the mobility of the target. For that reason, we adopted a reinforcement learning algorithm, namely, deep deterministic policy gradient (DDPG) approach to deal with the aforementioned problems. Numerical results proved the superiority of the presented model over traditional UAV trajectory benchmarks and the ability to gain knowledge from the environment.