Bayesian Learning-Based ISAC in RIS-Aided MmWave Systems with Superimposed Symbols

This paper investigates the integrated sensing and communication (ISAC) dual functionalities in a reconfigurable intelligent surface (RIS)-aided millimeter wave (mm Wave) system. In the considered system, both sensing and communication signals occupying the same time-frequency resources are fused in superimposed symbols, which eliminates the assignment of separate time/frequency resource to the sensing function, thereby improving the spectral efficiency. To realize the ISAC dual functionalities performance, we propose an expectation-maximization (EM) based Bayesian learning algorithm by leveraging the angular sparsity of the mm Wave channels and inherent properties of superimposed symbols to simultaneously estimate channel angles and detect data. Simulation results verify the convergence of the proposed algorithm and demonstrate the effect of the information symbols' power scaling factor
$\xi$
on channel angular perception and data recovery performance, in which equal power allocation, i.e.,
$\xi=0.5$
, is recommended. Moreover, it also indicates that the proposed superimposed symbols strategy at a high signal-to-noise ratio (SNR) can significantly increase data transmission throughput on the premise of little sacrifice of ISAC performance.