An Online Virtual Gyroscope Technique Using Convolutional Neural Network

摘要

Inertial device is the core of the SINS and the gyroscope senses the angular velocity of the carrier. The failure of the gyroscope causes the navigation system to be unable to resolve the attitude of the carrier. In order to improve the fault-tolerance of SINS, this paper proposes an online virtual gyroscope algorithm based on convolutional neural network using IMU's own realtime data and analyzes the feasibility of online virtual algorithm. First, the equations for calculating the angular velocity of the carrier using the information contained in the accelerometer are analyzed to determine the input data and output data of the convolutional neural network. Then, the online training convolutional neural network model is established, and a four-layer neural network with two convolutional layers, one fully connected layer, and an output layer is proposed as the predicting model. Finally, the feasibility of the proposed virtual algorithm is verified by mathematical simulation.