Fingerprint-Based Visible Light Positioning Using Adaptive WKNN Algorithm With Reliable Reference Nodes

In indoor visible light positioning (VLP) systems, fingerprint-based techniques have gained significant attention due to their high precision and wide applicability. However, conventional fingerprint-based localization relying on a fixed number of unreliable reference nodes (RNs) often leads to elevated positioning errors. To mitigate this limitation, an adaptive weighted K-nearest neighbors with reliable RNs (AWKNN-RRNs) algorithm is proposed in this article. The proposed algorithm initially identifies candidate reliable RNs based on the closest RN. Subsequently, a reliability coefficient is calculated to evaluate each candidate RN. Finally, artificial RNs are generated through optical channel modeling to adaptively optimize the number of reliable RNs. Software simulations show that, when the signal-to-noise ratio (SNR) is set to 15 dB and the interval of the fingerprint database is 1 m, the AWKNN–RRN algorithm achieves a reduced average positioning error when compared with weighted K-nearest neighbor (WKNN) (23.4%), K-nearest neighbor (KNN) (49.96%), adaptive residual WKNN (10.03%), multilayer perceptron (MLP) (57.41%), and extreme learning machine (13.04%) methods. Furthermore, hardware-in-the-loop simulation (HILS) results are also provided to verify the algorithm’s practical effectiveness.