Online Antenna Selection for Adaptive Beamforming in MIMO Radar

Antenna selection offers a low-cost solution for adaptive beamforming in large-scale multiple-input multiple-output (MIMO) radar as it alleviates the high cost of digital signal processing (DSP), and radio frequency (RF) stages. However, solving the selection problem is computationally expensive and scenario dependent. In this paper, we propose an online method for antenna selection to harness the power of unsupervised learning for reducing the computational cost. We first formulate the problem as an online game in the context of online convex optimization (OCO), and employ a gradient-based method that makes a decision for each new realization of the signal environment to minimize the total loss after T steps. We study the performance of the proposed method through a numerical example and compare the results with that of the global solution obtained by exhaustive search and an approximated solution provided by offline maximization. We demonstrate that the proposed online method is able to offer a low complexity solution that tracks the global solution, converging to the offline approximation method but with significantly reduced computational complexity.