The Convergence Reserch Of Reinforcement Potential-field Algorithm

Robotics research has entered a new stage of development, In recent years, with growing application requirements, robot technology has been sustainable development. This extension of the many new areas of technology research, but also brought the skill level of these areas improved.Among them, the path planning technology more and more attention by researchers. Emerged in a series of new path planning method.In order to guarantee the tightness of these methods, a variety of planning methods convergence proof of the growing importance attached by the researchers.This article aims to inspire potential field model of imitation of the current algorithm research.The model was combined with reinforcement learning and artificial potential field method and the application of imitation put forward flow algorithm. Simulation results show that the algorithm can overcome the potential field of the local minimum point of problem, shows that the method is correct and effective.In this paper, convergence of the algorithm on a math proof.This article first research background and practical significance for a brief introduction, Of the algorithm, especially robot navigation algorithm with the convergence of the relevant points are described. And machine learning at home and abroad on the problem of convergence studies summarized introduction.Second on the artificial potential field method and the theory of reinforcement learning are described. And then focus on how to stimulate learning model conversion potential field model of adult workers,Namely, the use of reinforcement learning and artificial potential field of the merits of the application to build a fake water law with memory function study incentive potential field model. On this basis, given the current algorithm from the imitation of the mathematical model derived.Finally, the use of iterative gradient theory and optimization theory methods to prove the convergence of the algorithm have been studied and proved.