Design Of Indoor Positioning Method Based On Fusion Particle Filter

With the rapid development of mobile communication technology,Location Based Service(LBS)is becoming more and more popular.At present,the most commonly used indoor positioning methods are WIFI fingerprint positioning method and Pedestrian Dead Reckoning(PDR)positioning method.However,the positioning error of PDR will increase because of the accumulation with time passed.This can be corrected by using the location information obtained by WIFI.So the indoor positioning fusion method based on PDR and RSSI has received widespread attention.In the selection of filtering algorithm,particle filter algorithm is most widely used because of its characteristic that is suitable for nonlinear non Gauss,but particle filter algorithm also has the problems of particle degradation,diversity and so on.In this thesis,we improve the positioning accuracy by improving the resampling algorithm to avoid the problem of particle degradation and the loss of diversity,and use the ray tracing model to replace the original propagation model to reduce the time of the establishment of the fingerprint database.The resampling algorithm solves the problem of particle degradation,but the new problem of particle diversity appears.Therefore,this thesis adds preprocessing and weight optimization,the particle of small weight is moved to the high likelihood region,and the problem of particle diversity is improved.The former used in the particle filter for fusion location is the wireless signal propagation model or the fingerprint model.The former does not consider the multipath effect in the complex indoor environment,and the latter need a great time to establish the database.Therefore,we use ray tracing algorithm to improve the intensity of the signal obtained by the observation model,and reduce the establishment time of the fingerprint database.The experimental results show that when the positioning error of WIFI and PDR is reduced to 2.81 m and 1.62 m,the positioning error of the fusion positioning method by standard particle filter algorithm is reduced to 0.96 m.The error of the improved particle filter algorithm is reduced to 0.79 m,which is increased by 12.2%.