Modeling And Motion Control Of AUV Hovering Tasks In Uncertain Current Environments

In recent years,autonomous underwater vehicles(AUVs)have become an important tool for the development of the world’s maritime powers due to their wide range of activities,high operational efficiency,and strong maneuverability.They are increasingly playing an irreplaceable role in marine environmental observation,resource development,and other fields.When performing certain ocean exploration,salvage and other tasks,it is required that the AUV maintain a fixed position of the target object,that is,have the ability to maintain hovering and counteract environmental disturbances.Due to the complexity of the marine environment,as well as the strong coupling and nonlinearity of the AUV itself,this poses a great challenge to the control system.Aiming at some issues arising from the application of traditional PID algorithm in AUV hover control,this thesis conducts research on the intelligent control algorithm of AUV and proposes an AUV hover control method based on deep reinforcement learning algorithm to resist the interference of uncertain ocean currents.The specific research contents are as follows:(1)According to the actual motion of AUV,its rigid body kinematics and dynamics were modeled respectively.Firstly,two reference coordinate systems were established,and on this basis,a dynamic mathematical model of the AUV was established using the Newton-Euler method.The hover and rotation simulation experiments of the model were conducted by using the PID control algorithm,and the problems of PID algorithm were analyzed in the case of uncertain ocean currents.(2)A deep deterministic policy gradient(DDPG)control method based on environmental optimal heading is proposed to address the difficulties of nonlinear dynamics and environmental uncertainty in the hovering operation of AUVs.In response to the interference of uncertain real-time ocean currents on AUVs,a reinforcement learning algorithm was adopted,utilizing the advantage of AUVs directly interacting with the environment to obtain experience.A Markov decision process for AUV hovering tasks was established,and a DDPG controller was designed.Simultaneously introducing an environmental optimal heading mechanism effectively reduces the energy consumption of AUVs during hovering.Simulation results show that the controller designed in this thesis can make the AUV resist interference in an environment with uncertain ocean currents and basically maintain a hovering state.(3)In response to the shortcomings of weak exploration ability,slow convergence speed,and large trajectory error of the DDPG algorithm in AUV hover control,a DDPG algorithm based on priority experience playback is proposed,which assigns different priorities to the sample data based on the absolute value of TD Error to improve the algorithm training speed and reduce errors during hover.At the same time,comparative experiments were conducted with the current cutting-edge SAC reinforcement learning algorithm to verify the effectiveness of the improved algorithm in the hovering process of AUVs.