Adaptive Receiver Processing Control for Sensing-Assisted LMMSE Channel Estimation in Bistatic OFDM ISAC Systems
Abstract
In bistatic orthogonal frequency division multiplexing (OFDM) integrated sensing and communication (ISAC) receivers, sensing-assisted linear minimum mean square error (LMMSE) channel estimation builds tolerance-aware correlation matrices from pilot-derived delay–Doppler descriptors. This paper proposes a hedge-algebra-based adaptive tolerance and update (HA-ATU) controller for a single-input single-output (SISO) bistatic OFDM ISAC receiver. HA-ATU separates sensing invocation from coefficient updating and selects among coefficient reuse, diagonal delay–Doppler phase retracking without matrix inversion, and full correlation rebuilding under a residual-drift test. Over SNR <inline-formula> <tex-math notation="LaTeX">$\in \{0,10,20,30\}$ </tex-math></inline-formula> dB and receiver processing-capacity index <inline-formula> <tex-math notation="LaTeX">$\kappa \in \{0.2,0.5,0.8,1.0\}$ </tex-math></inline-formula>, sensing every two and four coherent intervals reduces the complex-multiplication-equivalent (CME) count by 34.8% and 57.8%, respectively, at average normalized mean square error (NMSE) losses of 0.179 and 0.363 dB. Against a continuous non-HA controller, the NMSE differences are small and change sign across the tested settings for <inline-formula> <tex-math notation="LaTeX">$\kappa \le 0.8$ </tex-math></inline-formula>; at <inline-formula> <tex-math notation="LaTeX">$\kappa =1$ </tex-math></inline-formula>, HA-ATU reduces NMSE by 0.83 dB and CME by 4.6%. In profile-based tests using the 3GPP TR 38.901 TDL-C and TDL-D power–delay profiles with independent tap evolution, a sensing period of four intervals reduces CME by 72.3% and 71.7%, respectively, at mean NMSE changes of 0.04 and 1.01 dB. Retracking rates are 1.0% and 2.0% for TDL-C and TDL-D, respectively, while rebuilding rates are 56.9% and 50.9%, indicating that the retained CME reduction is primarily attributable to sensing gating and that phase retracking is confined to coherent path displacement. MIMO, multi-user, and hardware-level performance remain unvalidated.