A testbed for intelligent soccer strategies for the RoboCup.be team

摘要

Robocup.be is part of an international project to encourage AI and robotic research. Guided by DSP Valley and in association with other colleges and universities KaHo Sint-Lieven cooperates in a Belgian team of small, driving soccer robots to take part in international RoboCup competitions in the future. This demonstration paper will present an environment to simulate and test intelligent strategies for the RoboCup.be soccer robots. The testbed developed is the result of the first author's master thesis. A physical simulation environment is created, using a physical engine. Next this is extended with a multi-agent system to get a testbed in which intelligent multi-agent soccer strategies can be developed and tested. We focus on incorporating reinforcement learning in our RoboCup team. The ultimate goal is to hybridize the testbed by replacing the physical engine with the real-life robotic soccer game.