Study of Cooperation Strategy of Robot Based on Parallel Q-Learning Algorithm

Authors:
Shuda Wang;Feng Si;Jing Yang;Shuoning Wang;Jun Yang
Affiliations:
College of Computer and Information Engineering, Harbin University of Commerce, Harbin, China;College of Computer and Information Engineering, Harbin University of Commerce, Harbin, China;College of Computer and Information Engineering, Harbin University of Commerce, Harbin, China;College of Computer and Information Engineering, Harbin University of Commerce, Harbin, China;College of Computer and Information Engineering, Harbin University of Commerce, Harbin, China
Venue:
ICIRA '08 Proceedings of the First International Conference on Intelligent Robotics and Applications: Part I
Year:
2008

Citing 4
Cited 0

Asynchronous Stochastic Approximation and Q-Learning

Machine Learning
Between MDPs and semi-MDPs: a framework for temporal abstraction in reinforcement learning

Artificial Intelligence
Global Planning from Local Eyeshot: An Implementation of Observation-Based Plan Coordination in RoboCup Simulation Games

RoboCup 2001: Robot Soccer World Cup V
Credit assigned CMAC and its application to online learning robust controllers

IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics

Quantified Score

Hi-index	0.00

Visualization

Abstract

How to solve MR (Multi-Robots) in a dynamic environment of the study of knowledge, and to complete a task or solve a problem, the robot can have the same goal , also different goals. Therefore, to put forward two architectures, which are more suitable for MR studying, according to the architecture, to design the improved learning methods algorithm Q for MR, which solve the problems of coordination and cooperation, such as the credit distribution, distribution of resources, tasks and conflict resolution. MR may be learning in independent environment, and fusing results after learning cycle, and the final results is going to be shared by all the robots, and as the basis of reference passing into next learning cycle, increase learning chances between MR and environment. Simulation results show that the learning algorithm enables MR learning rapidly and quickly surrounded by a mobile group, complying with better effective.