Low Overhead Beamtraining for Millimeter-Wave MIMO Systems: Machine Learning Approach Based on Path Parameters

Machine learning has been widely used as a solution to deal with beam management training overhead for 5th generation wireless communication systems. However, the types of information adopted for the training base have not been sufficient to achieve a robust intelligent system. Channel path parameters provide valuable information that can increase the accuracy of this type of solution. In this work, we present initial results of the application of the Kalman filter to estimate path parameters aiming at a robust training base. Simulated results in 3D ray-tracing show promising results on obstruction conditions.