<![CDATA[Multiple-input multiple-output (MIMO) system relies on a feedback signal which holds channel state information (CSI) from receiver to the transmitter to do pre-coding for achieving better performance. However, sending CSI feedback at each time stamp for long duration is an overhead in the communication system. We introduce a deep reinforcement learning based channel estimation at receiver end for single user MIMO communication without CSI feedback. In this paper we propose to train the receiver with known pilot signals to analyse the stochastic behaviour of the wireless channel. The simulation on MIMO channel with additive white Gaussian noise (AWGN) shows that our proposed method can learn the different characteristics affecting the channel with limited number of pilot signals. Extensive experiments show that the proposed method was able to outperform the existing state-of-the-art end to end reinforcement learning method. The results demonstrate that the proposed method learns and predicts the stochastic time varying channel characteristic accurately at receiver’s end.]]>
Copyrights © 2019
