Adaptive Karhunen–Loève Expansion for Robust and Energy-Efficient Signal Detection in 6G V2X Networks

Abstract

High-mobility 6G vehicle-to-everything (V2X) networks face rapidly time-varying doubly selective channels and structured co-channel interference, which can limit conventional receivers and produce error floors. This paper proposes Adaptive KLE-based Interference Mitigation (AKIM), an interpretable covariance-subspace projection stage preceding OTFS detection. AKIM estimates the received-signal covariance from a sliding window of 100 observations, selects the dominant interference-subspace dimension using the Minimum Description Length criterion, and projects the received vectors onto its orthogonal complement. The evaluation considers 4×4 MIMO, 16-QAM, a 128×32 OTFS grid, vehicle speed of 500 km/h, and carrier frequencies of 28 and 100 GHz. At 16 dB SNR, AKIM achieves a reported BER of approximately 10⁻⁵, compared with nearly 10⁻³ for MMSE. At 10 dB input interference, the output SINR reaches 14 dB. For N=32, AKIM reduces the plotted processing-complexity metric by 83.3% relative to MMSE. Its peak reported energy efficiency is 3.2 bits/J under the stated normalized power model. These results support AKIM as a transparent and computationally efficient front end for structured-interference mitigation in high-mobility MIMO–OTFS receivers. Further validation using standardized channels, reproducible Monte Carlo trials, and calibrated hardware measurements remains necessary.