Application of system fault detection and intelligent reconstruction method based on machine learning in micro inertial pedestrian navigation system

Pedestrian navigation assisted by human body characteristics is a new research branch in indoor and outdoor positioning field in recent years. Aiming at the problem that the pedestrian navigation system of inertial measurement unit foot installation cannot effectively measure the information when the foot inertial measurement unit fails, and as a result the navigation function malfunctions, this paper presents a pedestrian navigation method with installation of inertial measurement units at the other parts of the human body beside foot. The output data of foot inertial measurement unit are simulated by machine learning methods such as neural networks. With fault detection and system intelligent reconstruction principle, the reliability and the performance index of pedestrian navigation system under complex gait conditions can be improved. The experiment results show that, while different BP neural networks are used under different gaits the complexity of neural network model can be reduced. Part of the positioning performance of the pedestrian navigation system based on this method, is equal to that of the foot inertial navigation system with same sensor precision under the condition of no sensor fault. While the foot inertial measurement unit fails, the pedestrian navigation system can also realize the navigation and positioning function of certain accuracy.