Simplified Neural Network With Physics-Informed Module in MIMO Visible Light Communication Systems

Visible light communication (VLC) has emerged as a valuable addition to the existing spectrum for wireless communication, complementing traditional radio frequency (RF) communication. By utilizing the untapped spectrum of visible light, VLC has the potential to play a substantial role in advancing wireless communication technologies. Nonetheless, it is worth noting that the predominant research focused in LED-based visible light communication has primarily revolved around point-to-point communication or the more ideal multiple-input multiple-output (MIMO) system. In this paper, a transmission matrix-assisted neural network (TMANN) is proposed for a 2×4 MIMO visible light communication system to implement MIMO demultiplexing. The 2×2 transmission matrix is estimated by three methods, including the preamble method, the least mean-squares MIMO (LMS-MIMO) method, and the proposed attention-based weight neural network (AttWNN) method. The performance of transmission matrix estimation methods has been thoroughly investigated through both simulation and experimental evaluations. During the experiments conducted, it is observed that without employing any MIMO demultiplexing techniques, the total transmission rate can only achieve 2.787 Gbps, while by utilizing the classical least mean-squares MIMO (LMS-MIMO) method, the transmission rate can significantly increase to 5.961 Gbps. Compared with a multi-branch hybrid neural network (MBNN), the proposed TMANN has a full 59.23% reduction in computational complexity, although there is only a slight increase in data rate (from 6.643 Gbps to 6.689 Gbps). To the best of our knowledge, this is the highest transmission rate achieved in a LED-based MIMO visible light communication system in the simultaneous transmission of two data channels. These results showcase the promising applications of our proposed TMANN in low-cost MIMO visible light communication systems.