Font Size: 
Intelligent Robust Control based on Reinforcement Learning for a kind of Continuum Manipulator
Da Jiang, Zhiqin Cai, Xiaolu Qiu, Haijun Peng, Zhigang Wu

Last modified: 2020-07-17


Designing a controller with strong coupling dynamic model is extremely challenging and is more serious in continuum manipulator control problems. In this article, a dynamic control method based on deep reinforcement learning(RL) is used for the robust control of a continuum manipulator. According to the dynamic equation of the continuum manipulator, a model-based reinforcement learning optimized sliding mode controller is proposed. According to the limited observation state of the system, the agent learns and corrects the control parameters online to optimize the slope of the sliding surface, and realize the dynamic adaptive tracking control under the uncertainty of the model. Under the dynamic adaptive tracking control, the results show that the strategy network can give the appropriate control parameters according to the system state and output mechanism saturation problem. The efficacy of the approach on a difficult tracking control problem with highly nonlinear dynamics is demonstrated. Results indicate that the RL optimized sliding mode method is faster than the sliding mode method and only requires a small control increment.


Robust Control, Reinforcement Learning, Continuum Manipulator

An account with this site is required in order to view papers. Click here to create an account.