Weather Optimal Position Control for AUV in Presence of Unknown Ocean Currents Based on Sliding Mode and Reinforcement Learning
A sliding mode control method based on reinforcement learning compensation is proposed for the environ-mental optimal heading positioning control of under actuated AUVs.Firstly,using the relationship between the fixed coordi-nate system and the on-board coordinate system,establish a three degree of freedom model of under actuated AUVs with consideration of the role of ocean current velocity.Secondly,based on the principle of weather optimal heading control(WOHC),the sliding model position controller and attitude controller of the AUV are designed separately.Thirdly,in order to overcome the uncertainty of AUV hydrodynamic model and ocean current,a reinforcement learning neural network was de-signed using the deep deterministic policy gradient(DDPG)algorithm to adaptive estimate and compensate for the serious interference caused by the above two uncertainties in sliding mode control.Finally,simulation is conducted under ocean current conditions,and the results show that the proposed method can effectively achieve AUV WOHC with much better ro-bustness against external disturbances and much higher accuracy compared with the classic sliding mode control.
AUVsliding model controlreinforcement learningweather optimal controlocean current interference
NETL
NSTL
万方数据
