The application of deep learning technology in trajectory control systems
Research the application of deep learning technology in trajectory control systems to achieve intelligent and precise trajectory control,in order to adapt to complex and ever-changing environments and task requirements.Using the LOS algorithm to determine the expected heading angle and trajectory error of ship navigation,after fusing the ship state information collected by the ship model,the state variables of trajectory tracking control are obtained,which are input into the trajectory controller based on the MDP model.The highest long-term cumulative return is set as the target,and the controller is trained using convolutional neural networks to learn the mapping relationship between the given state and the executed action,in order to obtain the optimal rudder angle action value that enables the ship to travel along the predete-rmined trajectory and achieve precise trajectory tracking control.The experimental results show that the controller used in the system can achieve precise capture of trajectory data patterns after 150 rounds of training,and has outstanding learning ability;Under interference conditions,the system can also enable the ship to navigate along the desired trajectory,and the trajectory control effect is significant.
deep learningtrack controlLOS algorithmMDP modellong term cumulative returnsconvolutional neural network
NETL
NSTL
万方数据
