Optimized MIMO channel estimation with deep neural networks for hybrid optical-wireless smart city networks

With the rapid expansion of smart cities and the exponential growth of connected devices, future wireless communication systems face increasing demands for higher data rates, reliability, and energy efficiency. In this context, the integration of Multiple Input Multiple Output (MIMO) and Optical Wireless Communication (OWC) technologies provides promising pathways to enhance system capacity. This paper proposes a deep learning-based channel estimation approach for MIMO systems, with potential application in hybrid radio-optical communication architectures for smart city deployments. The method utilizes a two-stage neural network structure: a Convolutional Neural Network (CNN) reconstructs incomplete channel information, followed by a Residual Deep Neural Network (Res-DNN) for accurate estimation. Multi-loss function optimization is employed to refine performance. Simulation results demonstrate that the proposed model outperforms traditional estimation methods, particularly in scenarios lacking prior statistical channel information. These findings are significant for real-time, adaptive communication systems, such as intelligent traffic networks and infrastructure monitoring in smart cities. The proposed deep learning model demonstrates strong potential in advancing robust, low-latency channel estimation for future wireless and optical-integrated smart urban environments.