Underwater robot three-dimensional path visual tracking method

Abstract

The invention relates to an underwater robot three-dimensional path visual tracking method. The invention belongs to the technical field of underwater robot three-dimensional path planning. Accordingto the method, a geodetic coordinate system, a carrier coordinate system and a curve coordinate system are established, a six-degree-of-freedom model of an underwater robot is established according tothe coordinate systems, a path tracking error model is established according to the established six-degree-of-freedom model of the underwater robot, and course angle deviation and submergence angle deviation are determined; three-dimensional path tracking is performed on the established six-degree-of-freedom model of the underwater robot by adopting a back-stepping sliding mode control method; and three-dimensional path tracking is trained by adopting a deep reinforcement learning method to complete three-dimensional path visual tracking of the underwater robot. According to the invention, the tracking error is continuous, and the path tracking stability is improved; an integral term is added into the conductivity of the sight method to introduce the influence of time; and a boundary reward function is added to accelerate the convergence of path tracking, the overshoot is reduced, and the precision is improved.

Description

technical field [0001] The invention relates to the technical field of three-dimensional path planning of an underwater robot, and relates to a visual tracking method for a three-dimensional path of an underwater robot. Background technique [0002] The ocean is the cradle of life on the earth. The ocean area accounts for 71% of the earth's surface area. It stores abundant water resources, biological resources and mineral resources. With the decrease of land mineral resources and the shortage of water resources gradually revealed, all countries in the world are aware of the ocean. The importance of resource development, the development and utilization of marine resources has become a necessary way for sustainable development and a new field of cooperation and competition among countries. At the same time, the world economy is inseparable from shipping, and ocean transportation is an important channel for the circulation of bulk commodities. Therefore, protecting the safety o...

AI Technical Summary

Problems solved by technology

Such as line-of-sight method, virtual target method, time delay estimation technology, virtual control quantity, energy dissipation theory, etc., but most of them are complex and poorly adaptable. In the study of controller design using deep reinforcement learning, The controller has good adaptive ability, but there are some problems to be solved when applying deep reinforcement learning to the 3D path tracking of underactuated AUVs. First of all, the deep neural network is the core of the controller, which cannot understand the underactuation in essence. The control characteristics and complex sports environment can only obtain certain control strategy training through the learning process, and have the ability to choose different actions under different conditions.

Therefore, deep neural networks often only output actions with a relatively single target, while ignoring the manipulation characteristics of underactuated AUVs.

Secondly, the controller using reinforcement learning may not be sensitive enough to small errors in path tracking, which affects the further improvement of tracking accuracy

Images