Research On Fusion Algorithm Of 3D Pedestrian Autonomous Navigation

Service software based on location track user's position accurately and continuously, which make people's life style more and more convenient. The satellite signal can't be used indoor due to the rapid decay caused by obstructions from buildings. Location is achieved through pedestrian navigation methods based on technologies, such as Wi-Fi, ZigBee and RF, which are limited to external signal transmitting apparatus. Therefore, pedestrian autonomous navigation technology based on multi-sensors has attracted widespread attention. At present, magnetometer and barometer are used in some autonomous pedestrian navigation technologies are relatively mature usually, and the accuracy and stability are not ideal due to magnetic interference and pressure instability. So, in-depth research of pedestrian autonomous navigation technology is conducted. The fusion algorithm which can realize 3D autonomous navigation only adopting accelerometers and gyroscopes is firstly proposed in this thesis.Firstly, the features of pedestrian gait are analyzed in this thesis. A complete gait cycle consists of four parts, including stance phase, lifting, swing and dropping. During per pedestrian gait, the zero velocity state and movement state can be estimated through hidden Markov model, and the number of steps can be estimated according to state periodic.The cumulative error caused by gyroscope bias leads to navigation algorithm divergence. The least squares method is used to estimate gyroscope bias for fixing angular rate. During zero velocity state, the attitude angle of inertial measurement module is estimated by using acceleration. In the process of toe movement, heading and attitude angles are estimated by using angle rate. The trajectory of pedestrian gait is estimated according to step length estimating algorithm proposed by this thesis.3D pedestrian dead reckoning is realized by improved traditional pedestrian dead reckoning algorithm, which uses horizontal distance, relative height and heading during every step. The fusion algorithm of 3D pedestrian autonomous navigation including gait detection, heading and attitude estimation, step length estimation and 3D trajectory estimation are simulated in MATLAB. Testing results show that the proposed fusion algorithm of 3D pedestrian autonomous navigation can provide accurate pedestrian position information continuously. In a loop route, the maximum distance error between the starting point and end point is less than 4%.