Scaling Wideband Massive MIMO Radar via Tiled Beamspace Processing (2512.06536v1)

Abstract: We present a coordinated tiled architecture for scalable wideband digital beamforming in massive MIMO radar systems. As aperture size increases, conventional full-array MVDR processing becomes prohibitive due to the cubic complexity of covariance estimation and inversion. Building on the principle of energy concentration in beamspace, we introduce a tiled windowed-beamspace MVDR framework that distributes spatial FFT processing across subarrays while performing joint beamforming in a compact global beamspace domain. Each tile applies a 2D spatial DFT followed by an angle-of-arrival dependent beamspace window, producing a reduced-dimensional representation that preserves the dominant spatial structure of the received signal. The windowed outputs from the tiles are concatenated and processed by a centralized MVDR beamformer, enabling coherent full-aperture processing with drastically reduced dimensionality. Our numerical results demonstrate that the proposed architecture achieves detection and interference suppression performance comparable to full-dimensional processing, while substantially lowering computational cost, memory usage, and training requirements. The framework also reveals tradeoffs among tile size, window size, and beamspace resolution that govern overall system scalability.