Research And Application Of Localization Database Construction Based On Crowdsourcing

With the rapid proliferation of wireless network and smartphone,the Received Signal Strength Indicator(RSSI)based WLAN localization is considered to be the preferred technology for indoor localization an d navigation and LBS since it exploits the existing WLAN infrastructure.The Wi-Fi location fingerprint positioning system needs to collect a large amount of RSSI data to establish a Radio Map.The RSSI data is highly uncertain due to the complex indoor electromagnetic environment.To ensure the positioning accuracy,the Radio Map needs to be continuously updated,which makes the construction and maintenance of Radio Map extremely costly High,limiting the large-scale application of Wi-Fi location fingerprint positioning system.Crowdsourcing technology through a large number of volunteers to complete the construction and update of Radio Map,effectively reducing the Wi-Fi location fingerprint positioning system Radio Map operation and maintenance costs.This article focuses on issues such as differences in RSSI collected values of heterogeneous devices with Wi-Fi location fingerprints,RSSI variability at the same location,PDR location estimation and cumulative errors,and the large amount of location fingerprint data,which reduces the real-time online positioning.The main research contents and innovations are as follows:Firstly,to address the problems of heterogeneous and time-varying RSSI data devices that reduce the quality of the location fingerprint database and the accuracy of online positioning,a standardized location fingerprint based on AP sequences is proposed.This position fingerprint combines the AP sequence with the normalized RSSI of Prouc,which effectively reduces the impact of device heterogeneity and time variability of RSSI,and solves the symmetry problem in the positioning algorithm.Experiments show that when the positioning error is less than 2m,the positioning accuracy of the fingerprint is 13.2% higher than that of the SSD.Secondly,to solve the problem that the PDR cannot obtain the absolute position of the user and the accumulated error is serious,a map matching algorithm based on indoor road network semantics is proposed.This algorithm converts indoor maps and user movement trajectories into indoor road network semantic maps and semantic sequences,and constructs a hidden Markov model.The hidden Markov model is solved by the Viterbi algorithm to obtain the hidden state sequence.Introduce semantic information to effectively filter hidden states,improve map matching accuracy,and reduce algorithm complexity.Experiments show that the similarity between the location fingerprint and the manually collected fingerprint is 95.97%,for this experimental scenario,the algorithm complexity is reduced by an average of 24.7% compared to ordinary indoor map matching algorithms.Thirdly,in view of the problem that the location fingerprint database constructed by the crowdsourcing data model greatly reduces the real-time performance of online positioning,an AP-GA(Affinity Propagation-Group average)clustering algorithm is proposed.The algorithm takes the self-similarity of the grid as the reference degree of the AP clustering,and then combines the AP clustering results by the GA algorithm.Not only skip the tedious parameter adjustment process,but also effectively solve the problem that AP clustering is not ideal for clustering complex structure data.Experiments show that the Purity index of the clustering algorithm reaches 96%,the number of location fingerprints is reduced to 36.5% of the number of location fingerprints collected by crowdsourcing.The location fingerprints after clustering are used for positioning.The cumulative probability when the positioning error is 2m reaches 44.1%.In comparison,the positioning accuracy did not decrease significantly.