Research On Multi-sensor Assisted RSS Fingerprint Indoor Positioning Technology

With the rapid development of the Internet and GPS positioning technology, location based services are becoming more and more prosperous and have a broad market prospect. However, GPS can not be used in the indoor environment, which leads to an urgent demand on accurate positioning that can adapt to various indoor environment. Nowadays, smart devices like smart phone and tablet are more and more popular, and WiFi hotspots are widely deployed. WiFi based indoor positioning develop quickly. Among all the existing WiFi based methods, the received signal strength(RSS) fingerprint positioning technology is the most promising.The RSS fingerprinting is an empirical method, including offline phrase and online phrase. During the offline phrase the RSS information from all access points at all reference points are sampled and sent to the location-fingerprint database. During the online phrase the RSS information received from the access points is utilized to search the database to match the location.A multi-sensor assisted RSS fingerprinting technology is proposed to improve the performance of the classical RSS fingerprinting technology. Now the smart devices are equipped with multi-sensor such as accelerometer and gyroscope. Based on the data samples from these sensors, the users' trajectory can be estimated. The confidential degree of estimate trajectory can also be obtained and used to evaluate the trajectory's accuracy. A new and more comprehensive probability model is established to combine the RSS fingerprint and users' trajectory to implement fingerprint-location match to improve locating accuracy. To cope with various indoor environment, the RSS information are modeled as multi-state Gaussian distribution and stored in the database. And a compensation method is introduced to reduce the effect of the abnormal RSS components on the positioning accuracy.Compared with the classical fingerprint positioning, multi-sensors assisted method reduces the average position error from 1.18 meters to 0.97 meters. When the indoor environment changes, the average positioning error increases only 0.17 meters, while the errors of the classic positioning method, Horus and RADAR increase 0.27 and 0.31 meters respectively. The results demonstrate multi-sensor assisted methods can improve the position accuracy and ease the effects of environment.