Deep Learning Based Channel Prediction and CSI Feedback for Wireless Communication in High Mobility Environment

With the development of high speed train and vehicular network, improving the communications in the high mobility environment has become a urgent mission for the wireless mobile network. Nonetheless, Due to the high mobility of the user device, the channel between the base station and the user is fast changing. The user devices have to estimate the channel more frequently, which leads to high overhead in the communication process. Recently, Deep Learning has been introduced as a good solution to reduce the communication overhead of the Multiple-Input-Multiple-Output (MIMO) system. Many Deep Learning based schemes such as channel state information (CSI) prediction and CSI feedback are proposed. However, the existing schemes have two shortcomings. The first one is that the current CSI prediction schemes do not support the Orthogonal-Time-Frequency-Space (OTFS) multiplexing. The outstanding performance of OTFS in high mobility environment has been demonstrated by many researchers. The second shortcoming is that the existing CSI feedback schemes are not efficient enough for communications in high mobility environment. To tackle the issues, the objective of this proposed project is to reduce the communication overhead of MIMO-OTFS system and improve CSI feedback efficiency. Therefor, we propose to develop a Deep Learning based CSI prediction scheme for the MIMO-OTFS system and develop a low-complexity CSI feedback scheme.