Improved GPF Algorithm For Indoor Positioning Based On Inertial Sensors

In the wireless indoor positioning research, the common method is the Location fingerprint algorithm, the algorithm is divided into parts: part one is Offline fingerprint database establishment phase, floor will be divided into grids, the center of grid as fingerprint point, collecting the RSSI value of all the fingerprint points that came from access point(AP) one by one, after collecting all the grid data, created the location fingerprint database of the floor; part two is the online positioning stage, according to the user’s mobile phone RSSI value of the current position acquired a number of APs, match the fingerprint database on the server, the highest similarity fingerprint point is the user’s current location. The most difficult to achieve location fingerprinting approach is the offline fingerprint database establishment phase, this stage requires collection the RSSI values of all the fingerprint points, due to multipath, attenuation, interference environment and many other factors, a single fingerprint point data of RSSI will not fixed, the RSSI value will be fluctuation, so when collection the fingerprint point RSSI data, usually collected multiple fingerprint RSSI values on a point, after processing the fingerprint RSSI values data, as RSSI value of the fingerprint stored in the fingerprint database point. Under normal circumstances, we need to locate the area of the coverage area generally larger, because each fingerprint point will take a few minutes, this phase will take a lot of time and resources.In order to reduce the offline phase fingerprint database collection work, proposed the use of smart phones built inertial sensors collect three-axis acceleration data, triaxial accelerometer data to calculate the motion path in the customer premises, record the RSSI value at point after point position fingerprint, automatically create off-line fingerprint library. If all of the built-in inertial sensors mobile phone users are involved, will basic coverage the area, to save the cost of establishing fingerprint database, and also solved the climate change, AP changes need to re-issue collection of fingerprint data.Because of phone’s built inertial sensor noise and the presence of random noise interference, cause measurement accuracy is low, result in the calculated path a large error, the farther the distance motion, the greater the accumulated error. This article proposes an improved Gaussian particle filter, first filter the acceleration value, and then calculate the path. The simulation results show that after the improved Gaussian particle filter filtering, the calculated path closer to the actual path of movement, positioning accuracy is consistent with the traditional method to establish the database. But it will save a lot of resources, and the algorithm has good real-time performance and robustness.