Beam Tracking with Deep Learning Models for Mobile Beamforming Systems

Highly directional signals between devices help reduce interference to others and improve overall network bandwidth. However, one major design challenge in high throughput directional communication system is in achieving fast and accurate beam alignment and tracking in high mobility environments. Traditional signal processing algorithms for beam tracking are typically derived based on assumptions regarding signal, antenna, and device mobility. In this paper, we present a deep learning model-based method for beam tracking, which could be conveniently implemented without the need for accurate modeling of system or device mobility. The performance of such a method has been evaluated for mobile beamforming systems. Effectiveness of the proposed beam tracking method has been demonstrated through simulations.