Integrated Navigation on Vehicle Based on Low-cost SINS/GNSS Using Deep Learning

摘要

The high-precision integrated navigation time for the MEMS SINS/GNSS combination system is too long, and it is difficult to ensure the flexible navigation of existing low-cost vehicles. A deep learning assisted integrated navigation method is proposed, which uses MEMS gyroscopes, accelerometers and dual-antenna satellite receivers to achieve fast alignment under MEMS SINS/GNSS. Based on the analysis of traditional coordinates and strapdown inertial navigation, the dynamic error model is given. The integrated navigation frame of CNN-LSTM is proposed. At the same time, the EKF filter for training is designed and the motion characteristics of the car body are selected. The state equations and observations, using the EKF to train the CNN-LSTM model, ultimately achieve integrated navigation. Finally, the proposed method is simulated and tested. The final attitude accuracy is better than 0.2 degrees, the alignment time is 10 s and position accuracy is better than 3 m. Compared with the EKF navigation method, the navigation accuracy and the alignment time are significantly improved, which can meet the requirements of low-cost vehicle flexibility.