WIFI Indoor Positioning Method Based on Global Search K-means Clustering and Improved WKNN Algorithm

Due to the impact of database expansion and fluctuations in Received Signal Strength (RSS), the accuracy of WIFI indoor fingerprinting localization has decreased. To address this issue, this paper proposes a WIFI indoor fingerprinting localization algorithm based on Global Search K-means Clustering and Modified Weighted K-nearest neighbors (GSKCMW). Specifically, in the offline phase, we propose a novel K-means clustering algorithm that utilizes an improved Particle Swarm Optimization (PSO) technique and dual distances between RSS indicators and location coordinates for clustering, thus avoiding the problem of converging to local minima. Subsequently, an improved Weighted K-nearest Neighbors (WKNN) online matching algorithm by incorporating distance weights is improved to enhance the Mahalanobis distance, approximating the true distance between reference points and test points, thereby refining the accuracy of nearest neighbor fingerprints. Additionally, we employ dynamic K-values in WKNN to enable different optimal neighbor selections for different test points at each location, further enhancing localization accuracy. Simulation results indicate a significant improvement in localization precision achieved by this algorithm.