Underwater robot attitude control method based on reinforcement learning

Abstract

The invention discloses an underwater robot attitude control method based on reinforcement learning. A strategy gradient reinforcement learning algorithm of a support vector machine is combined to achieve hovering fixed-point operation of an underwater robot underwater. The method comprises the steps of: defining a Markov decision process, utilizing a priori knowledge to obtain some sample pointsand generate an initial strategy by utilizing SVM according to the sample points, improving an initial strategy by using a strategy gradient algorithm, generating a new sample point according to the improved strategy, generating a strategy by using an SVM again, adjusting parameters on the basis of the strategy, cycling the process to obtain an optimal strategy, and finally, applying a controllerthrough reinforcement learning to an actual underwater robot system. According to the method, a strategy gradient reinforcement learning algorithm combined with a support vector machine is adopted, the problem that an underwater robot kinetic model is difficult to establish is solved, various uncertainties can be overcome in the learning process, the optimal strategy is better approached, and thecontrol precision of the system is improved.

Description

technical field [0001] The invention relates to an attitude control method of an underwater robot, in particular to an attitude control method of an underwater robot based on reinforcement learning. Background technique [0002] The focus of underwater rescue operations is underwater search and underwater rescue operations. The use of human search and rescue is limited, and these tasks can be completely completed by underwater robots. The biggest feature of the underwater robot is its strong deep-water operation ability and easy operation. The operator can remotely control the robot to perform difficult underwater operations through a simple button on the console in the ground control room. Underwater robots can complete high-intensity and heavy-load underwater rescue operations in depths and unsafe waters that divers cannot reach. When the search and rescue underwater robot performs underwater hovering operations, due to the influence of interference factors such as the ma...

AI Technical Summary

Problems solved by technology

However, the simulation experiments conducted on URIS and GARBI underwater robots show that if the optimal solution is to be found, the convergence speed will be very slow

Images