Accuracy Enhancement of Indoor Visible Light Positioning Using Improved WKNN Algorithm

Recently, weighted K-nearest neighbors (WKNN) algorithm was used in fingerprint-based visible light positioning (VLP) systems and gained significant attention due to its low complexity. However, the standard WKNN algorithm only utilizes the distance or cosine similarity (angle) between the power vectors received by the target and by the reference points (RPs) in the fingerprint database to determine the weights. That means it does not fully exploit the similarity between the target and RPs which depends on their distance and angle. In order to enhance the positioning accuracy, an improved WKNN (IWKNN) algorithm is proposed in this paper, in which both the distance and angle measurements are fully exploited. Firstly, the K nearest RPs are obtained using the distance measurements. Then, the angles between each RP and the target are calculated, and the weights of K nearest RPs are determined based on the reciprocal of the product of distance and angle measurements. Finally, the estimated position of the target is obtained through the weighted average of the positions of selected K RPs. Simulation results demonstrate that the proposed IWKNN algorithm outperforms the standard WKNN algorithm in terms of positioning accuracy, making it a more effective approach for fingerprint-based VLP systems.