CRB Optimization for Integrated Sensing and Communication Systems Using Hybrid Linear-Nonlinear Precoding

This paper proposes a waveform design technique for integrated sensing and communication (ISAC) systems based on hybrid linear-nonlinear precoding (HLNP). To obtain accurate direction of arrival (DOA) estimation and satisfactory waveform ambiguity properties, we optimize the weighted sum of the Cramer-Rao bound (CRB) of DOA estimation and waveform similarity, subject to constraints on the SINR of each communication user. In addition to constraints on the total power and per antenna power of the transmitted signal, we also constrain the peak to average power ratio (PAPR) on each antenna. We deploy successive convex approximation (SCA) to solve the resultant nonconvex problem while leveraging feasible point pursuit SCA (FPP-SCA) to provide a feasible initial point for the SCA algorithm. To reduce the computational cost of waveform design, we introduce a sub-block design technique. Simulation results verify the effectiveness of the HLNP algorithm and its extension, and validate their superiority over the conventional nonlinear precoding (NLP) scheme.