Learning-Based Joint Waveform Optimization and Receiver Design for Dual-Functional MIMO Radar and Communications

For the joint multiple-input multiple-output (MIMO) radar and MIMO communications system with widely separated antennas, designing the waveforms properly plays a pivotal role, especially when dual-functional is requested. Considering target detection and decoding as the main radar and communications tasks, for the purpose of improving both performance, in this paper the dual-functional waveforms are designed jointly with the radar and communications receivers. Supervised learning is employed for the radar detector and communications decoder, while reinforcement learning is employed for the waveform design. We propose a joint training framework, where the detection and decoding performance metrics are used as the loss function of the supervised learning and the reward of the reinforcement learning. The reinforcement learning-based transmitter and supervised learning-based receivers are trained in an iterative way to meet the desired performance. Numerical examples are provided to show the efficiency and effectiveness of the proposed method.