Pedestrian Positioning Based on Dual Inertial Sensors and Foot Geometric Constraints

Abstract

Foot-mounted inertial navigation with zero velocity update (ZUPT) is commonly used in pedestrian positioning. To suppress the accumulating position error owing to the low-cost inertial sensor, classical dual-foot pedestrian navigation method provides the compensation using the two feet's relative step length, whose values can be set an upper bound constraint or measured by ranging sensors. However, it suffers from low improvement and poor wearability. In this study, the two sensors are distributed on the toe and ankle of the same leg. The sensor-to-sensor position is unchangeable with the lower limb segments. This makes the uncoupled two subsystems be related. Through gait analysis, the step is divided into the stance and swing phases. For each phase, positions, velocities, and the related characteristics of the toe and the ankle are analyzed and modeled, respectively. A Kalman filter is constructed to fuse the position and velocity observations according to the geometric constraints and the foot vector. The Mahony filter by fusing the inertial sensor and magnetometer is adopted for the attitude improvement. Several experiments at low and high walking velocities on the plane and upstairs are carried out to verify the proposed method. The results show that compared with the conventional single-inertial measurement unit (IMU) ZUPT and dual-IMU two-foot constraint methods, the position and heading errors have been well suppressed.