Channel and Location Estimation Enabled a Novel BDCNP Network for Massive MIMO VLCP Systems

Massive multiple-input multiple-output (m-MIMO) integrated visible light communication and positioning (VLCP) systems are of paramount importance to achieve the vision of next generation wireless communication networks since they can simultaneously enhance communication and positioning performance. However, since the computational complexity of channel and location estimation increases with the number of transceivers, it becomes intractable to obtain accurate channel state and location information. In turn, this significantly limits the performance gain of m-MIMO VLCP systems. To tackle this issue, a blind denoising convolutional neural and positioning (BDCNP) network is proposed to estimate both channel state information (CSI) and location in the m-MIMO VLCP system. Specifically, the BDCNP is firstly adopted to denoise the channel matrix with unknown and time-varying superimposed noise to obtain accurate CSI. Then, the location information is extracted from the estimated CSI based on the proposed network. Simulation results demonstrated that the proposed BDCNP network can simultaneously obtain accurate channel state and location information, which outperforms the existing benchmark methods with superior convergence and acceptable computational complexity.