Multi-Robot Enhanced Intelligent Multi-User Millimeter-Wave MIMO Systems under Uncertain Environment

This paper investigates how to maximize the practical communication quality of multi-user millimeter-wave (mmWave) MIMO systems with uncertain environment through effectively using the mobility from multi-robots as dynamic relays and adopting machine learning techniques. mmWave MIMO has been considered as a promising wireless communication technology due to the high frequency usage efficiency from beamforming. However, the uncertain environment could seriously affect the effectiveness and practicality of beamforming since wireless channels may have a more complicated structure, and the coordination among multiple nodes could be more difficult. For instance, the uncertain distribution of mobile users could significantly affect the performance of wireless channels, and then significantly degrade practical communication quality. Therefore, this paper presents a novel Multi-Robot Enhanced Intelligent Multi-User Millimeter-Wave MIMO (MREI-MU-MIMO) system that adopt both dynamic codebook based beam training protocol and online reinforcement learning to supervise the mobility of multi-robot-relay as well as handle the serious effects form the uncertain environment. Firstly, a novel dynamic codebook development is presented that cannot only lower the complexity of existing beamforming codebooks and also handle the complicated channel structure under uncertainty during multi-user beam training. Then, a decentralized Deep Q-Network (DQN) rein-forcement learning algorithm has been developed to intelligently manage multi-robot mobility and further effectively assign the optimal MIMO to handle the uncertainty from environment. The effectiveness of the proposed design has been demonstrated through real-time simulation and experiment.