Research on the Algorithm of Tractor GNSS/INS Integrated Navigation System Assisted by CNN-BiLSTM

In farmland environments where GNSS signals are obstructed—such as forested areas or adverse weather conditions，the traditional GNSS/INS integrated navigation system is affected by the instability of satellite signals, which affects the positioning accuracy and navigation stability of the autonomous electric tractor. To address this, a model-assisted integrated navigation system is proposed, combining Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks. Trained under normal GNSS conditions, the CNN-BiLSTM model predicts positioning data during GNSS interruptions, ensuring continuous and stable navigation. The validity and reliability of the model is verified by the measured data of the tractor simulated field operation. The experimental results show that under the situation of GNSS signal denial for 100 seconds in the farmland scenario, the positioning accuracy of the neural network model-assisted INS output is highly close to that of the GNSS/INS output in normal conditions, and it exhibits comparable fitting accuracy to that of real GPS, which proves that the neural network model can be used to replace GNSS during GNSS denial scenarios. Moreover, the model-assisted integrated navigation significantly reduces the error of the pure INS navigation algorithm, provides effective support for the navigation of autonomous electric tractor.