Intelligent Waveform Design for Integrated Sensing and Communication

Integrated sensing and communication (ISAC) represents a unified paradigm to enhance spectral efficiency and reduce hardware costs by enabling the coexistence of communication and radar functionalities using the same spectral and hardware resources. Dual-functional (DF) waveform design, as an essential component of ISAC, often entails high-complexity, non-convex optimization algorithms, hindering its practical online deployment. Leveraging the robust predictive capabilities of deep learning (DL) and deep reinforcement learning (DRL), these technologies have emerged as viable and streamlined approaches for the online design of DF waveforms more suitable for the dynamic environment. Thus, in this article, we first provide a comprehensive over-view of ISAC, with a particular focused examination of its waveform design. Then, we introduce DL/DRL and highlight its important roles in ISAC, especially in intelligent waveform design. Moreover, we develop DL- and DRL-based algorithms tailored for conventional DF waveform design, and simultaneously transmitting and reflecting (STAR) reconfigurable intelligent surface (RIS)-aided DF waveform design, respectively. Simulation results verify the effectiveness of the developed algorithms, with emerging research directions presented.