Research On WiFi Indoor Localization Algorithm Based On Location Fingerprint

With the development of Internet,the application of location services has become more widespread.Outdoor positioning technology has met the needs of society,but indoor positioning technology still faces challenges.This paper studies WiFi indoor localization algorithm based on location fingerprint,which is an indoor positioning technology with low cost,wide application range and easy implementation.In order to reduce workload of fingerprint acquisition and improve fingerprint matching accuracy of WKNN algorithm,this paper proposed a fingerprint localization algorithm based on improved Gaussian process regression and chi-square distance(ISOSGPR-CAIWKNN).In the offline phase,the fingerprint data was sparsely collected in the positioning area and input to Gaussian process regression(GPR)model as training samples.The improved symbiotic organisms search algorithm was used to optimize GPR parameters,improving generalization ability of GPR and obtaining ISOSGPR fingerprint prediction model.Then,ISOSGPR model was used to predict the fingerprint data of other locations in the positioning area to form a dense fingerprint database.In the online phase,to improve fingerprint matching accuracy and positioning accuracy,the distance between target and fingerprint data of fingerprint database was calculated by using chi-square distance,and AP weighting was performed when calculating weights.The experimental results showed that,when the number of training samples is only 50% of total number of samples collected,the fingerprint database constructed by ISOSGPR model can achieve the positioning effect of whole fingerprint collection database;compared with WKNN and CIWKNN algorithm,the positioning accuracy of ISOSGPR-CAIWKNN algorithm has increased by 16.6% and 11.2%,respectively.In order to improve fingerprint matching efficiency and solve the problem on fixed K value and non-optimal weights of WKNN algorithm,this paper proposed a fingerprint localization algorithm based on improved affinity propagation clustering and weight optimization(AIAPC-AIWKNN).In the offline phase,affinity propagation clustering algorithm(IAPC)was improved by fingerprint location and pairwise point constraint expansion method to divide the positioning area into several sub-areas to improve fingerprint matching efficiency.In the online positioning stage,to avoid positioning target located at the edge of sub-area was misjudged when selecting sub-area,adaptive selection method(AIAPC)was used to determine the sub-area which it belonged to.Then,the number of nearest neighbor reference points was determined by adaptive K-value selection method,and weights were optimized by Adam algorithm.The experimental results showed that AIAPC algorithm improved fingerprint matching efficiency while ensuring positioning accuracy.Compared with WKNN and IWKNN algorithms,the positioning accuracy of AIAPC-AIWKNN algorithm was increased by 26.5% and 8.4%,respectively.The paper has 34 figures,18 tables and 60 references.