An Improved Multi-dimensional Weighted K-nearest Neighbor Indoor Location Algorithm

Abstract

AbstractAiming at the defects of low positioning accuracy and poor stability of the traditional WiFi location fingerprint indoor positioning method based on the K-nearest neighbor (KNN) algorithm, an improved multi-dimensional weighted K-nearest neighbor (MDW-KNN) indoor positioning algorithm is proposed in this paper. The algorithm improves the three basic elements of KNN algorithm: distance metric, the number of nearest neighbors K and decision-making method, introduces multi-dimensional weight coefficients based on similarity, fingerprint distance and physical distance to reduce the spatial ambiguity in localization. Moreover, the number of nearest neighbors K is dynamically changed by the method of range restriction, which improves the stability of the algorithm. Simulation using MATLAB software, experiments show that under the same test environment, the location accuracy of the MDW-KNN is nearly 46% higher than that of the traditional KNN algorithm, and the MDW-KNN’s location estimation point trajectory is smooth, with good stability and fault tolerance.KeywordsIndoor positioningLocation fingerprintK-nearest neighbor (KNN)Multi-dimensional weight coefficients