Channel Estimation for Intelligent Reflecting Surface in 6G Wireless Network via Deep Learning Technique

Channel estimation for the wireless link has several challenges the hardest challenge is the randomness in the real channel. In the 6G wireless networks, the Intelligent Reflecting Surface (IRS) mitigate the problems in massive multiple input multiple output (mMIMO) in 5G like high cost, low coverage and high-power consumption. Communication network with (mMIMO) and IRS must approach smart network to enhance quality of service and reduce path loss. In this paper, simulation of the direct channel and the cascade channel was implemented with multipath for different users as point to multipoint transmission to improve robustness of the estimation. Channel estimation for the network makes base station (BS) and IRS work with each other adaptively by using deep learning. The output performance of the estimation is checked by root mean square error (RMSE), training loss, complexity and time delay for channel training. The validation RMSE in the direct channels arrives to 0.375 while it arrives to 1.116 in the cascade channels. Normalize mean square error (NMSE) is studied with respect to signal to noise ratio (SNR) to show the more stable channel with SNR.