Developing novel channel estimation and hybrid precoding in millimeter-wave communication system using heuristic-based deep learning

The digital or analog precoders are not carried the optimal energy efficiency in the mm-wave massive MIMO and so, every antenna is needed with one radio frequency chain in the system. Based on this view, a cost-efficient technique is developed for hybrid precoding. From this technique, the short dimensional precoding is obtained from the high dimensional beam-formers in the digital domain for steering antenna elements. Therefore, the main intention of this paper is to develop a channel estimation and hybrid pre-coding model for the mm-wave massive MIMO communication system. The channel estimation phase is performed by the Adaptive Deep Convolutional Neural Network (A-Deep CNN), which covers both channel estimation and channel reconstruction. The introduction of a hybrid meta-heuristic algorithm with Forest-Tunicate Swarm Algorithm (F-TSA) is used for enhancing A-Deep CNN that is adaptable for efficient channel estimation. Once the channel estimation is done, the deployment of Optimized Recurrent Neural Networks (O-RNN) is used for hybrid precoding. Simulation results demonstrate that the proposed A-Deep CNN-based channel estimation scheme outperforms the existing schemes in terms of the Normalized Mean-Squared Error (NMSE) and spectral efficiency, while the O-RNN-hybrid precoder design method has better spectral efficiency performance than other methods.