Near-Optimal Detection of Zak-OTFS Signals

Orthogonal time frequency space (OTFS) modulation is emerging as a promising alternative to currently deployed multicarrier modulation schemes, particularly in high-Doppler wireless channels. The multicarrier (MC) version of OTFS (MC-OTFS) has dominated the OTFS research literature so far due to its compatibility with the prevailing MC modulation schemes. Recently, Zak transform based OTFS (Zak-OTFS) has gained prominence on account of its superior performance and low complexity. However, unlike in its MC counterpart, the transceiver signal processing aspects like equalization/detection and channel estimation in Zak-OTFS remain vastly unexplored. In this paper, we present an early investigation of low-complexity detection algorithms for Zak-OTFS that can scale efficiently for large frame sizes and achieve near-optimal performance. For this, we resort to efficient local search algorithms, namely, likelihood ascent search (LAS) and reactive tabu search (RTS) algorithms, that scale well for large dimensions. To assess the closeness of the performance of these algorithms to the optimum maximum likelihood (ML) performance, we obtain a lower bound on ML performance via RTS simulation using the transmit vector as the initial vector and defining a suitable neighborhood. Our simulation results indicate that while the performance of the popularly known minimum mean square error (MMSE) and message passing (MP) algorithms are far from the ML performance bound, LAS and RTS algorithms initialized with MMSE solution achieve close to ML performance bound within about 1 dB at a bit error rate of 10
⁻⁵
.