Deep reinforcement learning control method for vertical surface path following of intelligent underwater robot

Abstract

The invention provides a deep reinforcement learning control method for vertical surface path following of an intelligent underwater robot. Step 1, according to the path-following control requirements of the intelligent underwater robot, establish an intelligent underwater robot environment that interacts with the agent; Step 2, establish a set of agents; Step 3, establish an experience cache pool; Step 4, establish a learner; Step 5, using distributed deterministic policy gradients for path-following control of the intelligent underwater vehicle. Aiming at the phenomenon that the ocean environment where the intelligent underwater robot is located is complex and changeable, and the traditional control method cannot actively interact with the environment, the invention designs a deep reinforcement learning control method for the vertical plane path following of the intelligent underwater robot. Using the deterministic policy gradient to complete the path following control task of the intelligent underwater robot through a distributed method has the advantages of self-learning, high precision, good adaptability and stable learning process.

Description

technical field [0001] The invention relates to a control method of an underwater vehicle, in particular to a deep reinforcement learning control method for vertical surface path following of an intelligent underwater robot. Background technique [0002] With the deepening of marine development, intelligent underwater robots have been widely used in marine environmental protection and marine resource development due to their flexible movement, convenient portability, and autonomous operation, and their status has become increasingly important. In addition, the precise control of intelligent underwater robots makes some extremely dangerous tasks safe, such as exploring subsea oil, repairing subsea pipelines, and tracking and recording the location of explosive materials. [0003] Traditional path-following control methods such as fuzzy logic control, PID control, and S-plane control require artificial adjustment of control parameters, and the control effect depends on human e...

AI Technical Summary

Problems solved by technology

[0003] Traditional path-following control methods such as fuzzy logic control, PID control, and S-plane control require artificial adjustment of control parameters, and the control effect depends on human experience. Intelligent underwater robots cannot actively interact with the environment

Images