Tank Turret Servo Predicted Control Based on Long Short Term Memory Neural Network

A control method based on LSTM-LADRC is proposed to address the issues of significant tracking errors and time delays in horizontal servo systems when tanks operate at high speed on complex road surfaces. Firstly, in Recurdyn software, a dynamic model of tank driving at high speed on a D class pavement has been established. D class pavement spectrum is generated by harmonic superposition method . The mathematical model of turret servo system is established based on Newton's law of motion . Secondly, the LSTM is constructed and trained to predict the turret angle position in multiple steps. It dynamically adjusts the expected tracking signal of the turret between the solution trajectory and the predicted trajectory, and sends the signal to LADRC to achieve predictive control of the turret servo system to reduce tracking errors and time delays . Finally, hybrid simulation is conducted to verify the correctness of the proposed algorithm .The hybrid simulation results of Simulin/Recurdyn indicate that, LSTM can effectively predict the input expected value of tank turret servo system and in comparison to the traditional method, the proposed approach can reduce the tracking error of the tank turret servo system by 6.12% under complex road conditions at high speeds.