Message Passing Neural Network-Enhanced Optical OTFS Iterative Symbol Detection

Optical orthogonal time frequency space (O-OTFS) has been proven to exhibit spectral and power efficiency superiority than previous optical orthogonal frequency-division multiplexing (O-OFDM) in visible light communication (VLC) systems. However, in prior related works, the efficient symbol detection method has not been fully investigated yet. In this paper, we devise an effective delay-Doppler (DD) domain detection scheme based on message passing neural network (MPNN) iterations, which utilizes the zero-forcing (ZF) equalized symbols as additional prior information. Numerical results have demonstrated that the proposed ZF-aided MPNN-iterated symbol detection scheme (MPNN-ZF) provides notable enhancement in terms of the bit-error-rate (BER) and convergence performance compared with other conventional methods.