Indoor positioning system for single LED light based on deep residual shrinkage network

To enhance the indoor visible light positioning (VLP) system's performance, this paper proposes a deep residual shrinkage network (DRSN)-based VLP system utilizing a single LED and multiple photodetectors. This neural network effectively resolves issues prevalent in the training phase of conventional neural networks, such as vanishing or exploding gradients, noise interference, and indistinct feature characterization, by incorporating residual learning, soft threshold functions, and an attention mechanism. This approach contributes to a more stable and efficient training process. Simulation results illustrate that the DRSN algorithm produces an average localization error of 1.02 cm, with a maximum error of 4.35 cm in a 3.6 m × 3.6 m × 3 m space. During the experimental phase, the average localization error remains below 6 cm, with 90% of the errors below 10 cm. Comparative analysis with alternative localization algorithms reveals the proposed method's ability to achieve precise localization accuracy.