Abstract: With the rapid development of science and technology, positioning in the ideal visible channel model can no longer meet the practical application requirements. The positioning system model established in this paper considers the direct link of visible light and a reflection of irregular walls.Based on this single LED lamp model, a bald eagle search algorithm is proposed to optimize of radial basis network (BES-RBF). The positioning system first optimizes the initial weights and thresholds of the radial basis function (RBF) neural network through the bald eagle search algorithm to make the structure of the optimized RBF neural network more stable. Then, the fingerprint database data consisting of the light power values received by each PDs of the receiver is introduced into the optimized RBF network model for training, and the positioning model is established. After the location information is obtained, some points are corrected by WKNN to obtain the final location information. In 3 m×3 m×3 m space, the simulation results show that, the average positioning error of the algorithm proposed in this paper is 5.54 cm, and 80% of the positioning error is within 4.5 cm. Compared with other positioning algorithms, the positioning accuracy has been significantly improved.

Key words: visible light positioning, received optical power, radial basis function neural network, bald eagle search algorithm