Blind OFDM-ISAC Relying on Asymmetric Modem Constellations

TL;DR

This paper introduces a blind OFDM-ISAC method that leverages asymmetries and higher-order statistics to enable joint radar sensing and communication under severe interference conditions.

Contribution

It proposes a novel higher-order-statistics based approach for blind joint radar and communication estimation using OFDM signals with constellation asymmetry.

Findings

Reliable radar parameter estimation achieved in interference scenarios.

Effective communication demodulation demonstrated despite severe TF interference.

Performance approaches theoretical bounds in simulations and experiments.

Abstract

Integrated sensing and communication (ISAC) is increasingly expected to operate under aggressive spectrum reuse, where co-channel orthogonal frequency division multiplexing (OFDM) interference can be catastrophic for data recovery on the time-frequency (TF) grid. We show that supporting blind ISAC is feasible by exploiting a fundamental asymmetry in the impact of co-channel OFDM interference: while communication is fragile on the TF grid, sensing depends on structured physical parameters whose signatures remain identifiable by relying on higher-order statistics. Based on this observation, we construct a fourth-order measurement tensor from the received OFDM signal whose coherent component preserves the delay-, Doppler-, and angle-dependent phase evolution of each source. We then develop a three-dimensional higher-order-statistics (HOS) based periodogram for iterative peak search and refinement to jointly estimate both range, velocity, and angle in the presence of unknown co-channel interferers. We further exploit constellation asymmetry to resolve the remaining phase ambiguities of blind recovery, enabling blind coherent demodulation via minimum constellation fitting. We also benchmark the performance through matched data-aided and stochastic Cramer-Rao lower bounds. We then quantify the cost of signal blindness. Simulations and experimental validations demonstrate reliable radar parameter estimation together with effective communication demodulation even when the TF-domain link is severely interfered with.