FPGA-Based Moving Array Beamforming for Robust Mixed-Source MIMO Estimation

Abstract

Accurate estimation of mixed signal sources in MIMO arrays is critical for modern communication, radar, and sensing systems, yet remains challenging under steering vector uncertainties, source correlation, Doppler shifts, and dynamic platform motion. This paper presents an FPGA-based realization of a moving array beamforming framework for robust mixed-source estimation. The proposed framework integrates a min–max optimization criterion with an adaptive diagonal loading strategy derived via deep unfolding, explicitly modeling array manifold uncertainties and optimizing performance under worst-case conditions. The deep-unfolded loading mechanism adapts scenario-dependent regularization parameters, enabling fast convergence and consistent performance across diverse signal and motion conditions. Comprehensive software simulations and FPGA-oriented experiments demonstrate that the proposed framework outperforms existing methods, including RCB, RCBDL, RCB-INCM, and FIM-Capon, achieving output SINR gains of 2.5–4.5 dB, interference suppression improvements of up to 40.2 dB, and a 60% reduction in convergence iterations. The FPGA implementation achieves real-time processing, with computation times reduced to 16.3 ms for a 50-element array, significantly lower than the 39.4 ms observed with FIM-Capon. Incorporation of a coprime array structure further enhances spatial resolution and degrees of freedom, making the proposed framework highly suitable for practical, real-time mixed-source estimation in MIMO communication and sensing applications.

Keywords