Robust Attitude Controller Designation of Launch Vehicle under Actuator Failure Condition via Deep Reinforcement Learning Algorithm
- Resource Type
- Conference
- Authors
- Jia, Chenhui; Liu, Xiaodong; Wang, Zhaolei; Gong, Qinghai; Huang, Xu
- Source
- 2023 35th Chinese Control and Decision Conference (CCDC) Control and Decision Conference (CCDC), 2023 35th Chinese. :3223-3228 May, 2023
- Subject
- General Topics for Engineers
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Deep learning
Actuators
Attitude control
Simulation
Process control
Reinforcement learning
Position measurement
launch vehicle
thrust drop failure of attitude control nozzle
reinforcement learning
TD3 algorithm
end to end and decouple free
strong adaptive attitude control
- Language
- ISSN
- 1948-9447
This paper focuses on the robust and adaptive attitude control algorithm of the launch vehicle under the condition of the thrust drop failure of the attitude control nozzle during the flight process. To solve this problem, we build a three-channel end to end attitude controller with TD3 (Twin-delayed Deep Deterministic Policy Gradient) reinforcement learning algorithm which performs outstanding on solving the problems of continuous action space reinforcement learning. The simulation results indicate that the attitude controller designed with the algorithm can automatically adapt to the thrust drop condition of the launch vehicle nozzle and quickly adjust the output of the actuator to correct the attitude control error. This method shows strong adaptability and robustness to the fault conditions and can be used as a potential algorithm in solving the problem of controller reconfiguration under the fault conditions of the launch vehicle.