An effective Pedestrian Dead Reckoning algorithm using a unified heading error model

Nowadays, navigation is an important application in mobile phones. However, locating a mobile user anytime anywhere is still a demanding task, because the GPS signal is easily corrupted or unavailable in urban canyons and indoor environments. Integrating GPS and self-contained dead reckoning sensors is an autonomous method to obtain a seamless positioning solution by means of Pedestrian Dead Reckoning (PDR) algorithms. A low-cost Multi-Sensor Positioning (MSP) platform has been developed by the Finnish Geodetic Institute, which includes a GPS receiver, a 2-axis digital compass and a 3-axis accelerometer. To construct a trajectory in GPS degraded environments, step length and the heading of each step are two key issues in PDR. In this paper, three typical estimation models of step length are presented and compared to demonstrate that in most cases, step length is not as critical as the determination of heading. Therefore, a unified heading error model is proposed, which includes all predictable errors from the navigation platform and the pedestrian's walking behavior, and applies to calibrating 2-axis magnetic compasses without tedious and complicated calibration procedures. Then the corresponding PDR algorithm is introduced, which integrates the step length estimated from a nonlinear model and the heading compensated by the unified model suggested through an Extended Kalman Filter (EKF). Several tests were conducted to validate the effectiveness of the heading error model and evaluate the positioning performance of this PDR algorithm. The results demonstrated that the heading error model is applicable for calibrating the 2-axis compass, and based on the PDR algorithm, the typical positioning performance of MSP can reach an accuracy of below 1.5% of the travelled distance during 10 minutes of continuous walking when GPS outages occur.