English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

This paper proposes a solution based on reinforcement learning in the soccer dribbling tasks of RoboCup. The dribbler spares no efforts to keep possession of the ball from beginning to end while adversaries attempt to gain possession. In this paper, the dribbler uses SARSA possession algorithm and the opposite perform traditional strategies respectively. These two kinds of strategies are applied to robots in the environment of 4V3.The results show that by learning many times robots have made great progress in terms of dribbling time. Introduction RoboCup simulation game provides a fully distributed control and real-time asynchronous multi-agent environment. Through this platform, you can test various theories, algorithms as well as Agent architecture. And you can also research cooperation and confrontation problems between multi-agent in real-time asynchronous and noisy confrontation environment. The game is conducted using the Client / Server mode in a standard computer environment when teams write their own client program to simulate the real soccer players to compete. Applying reinforcement learning algorithm to robot soccer simulation in RoboCup must overcome several problems such as multi-dimensional continuous state space, the effect of noise, the multi-agent and the need for real-time action. In the past period of time, machine learning has been applied in many RoboCup sub-tasks. This paper presents the SARSA algorithm in multi-agent (4V3) dribbling task and desired results have been achieved after a period of study. Reinforcement learning The theoretical foundation of reinforcement learning Reinforcement learning, also known as stimulate learning or assessment learning, is a process during which agents constantly test and learn to make the value of the accumulated reward of system behavior to get the maximum from the environment. The basic model is shown in Figure 1. In reinforcement learning, we design algorithms to put the external environment into actions to maximize the amount of reward way. The agent has not been told what to do or what action to take, but discoveries by looking at which action been most rewarding. Agent movements affect not only immediate rewards, but also the next action and the final reward. In reinforcement learning, the environment is in a certain state s of the state set S and the agent selects an action a in action set A. The action a acting on the environment will immediately receive a reward r, when r is greater than 0 the agent will enhance the tendency of this action in the future. When the control behavior of learning system repeatedly interacts with the state and evaluation of the environment, mapping strategies from state to action have been constantly modified by the way of learning in order to optimize system performance. International Industrial Informatics and Computer Engineering Conference (IIICEC 2015) © 2015. The authors Published by Atlantis Press 458

The New solution to Cooperation Strategy in Dribbling Task of RoboCup