Adapting the RoboCup Simulation for Autonomous Vehicle Team Information Fusion and Decision Making Experimentation

摘要

The RoboCup Simulator is a popular experimentation platform for research in artificial intelligence, from areas such as computer vision, machine learning and multi-agent systems. The primary purpose of the simulator is to provide a software analogue to the physical robot RoboCup leagues, providing an easy method of entry into RoboCup without the need for expensive hardware. This simulator provides a common framework for researchers to compare their approaches and many graduate students and professors utilize the simulator for their experiments. We wanted a platform that would encourage researchers outside the government to produce the kinds of algorithms and software that would easily transition into solutions for teams of autonomous vehicles for military scenarios. To accomplish this, we began modifying the RoboCup soccer game step-by-step to incorporate rules that simulate these scenarios. These changes affect the field of play, agent-agent communication rules, the number of balls, what constitutes a foul, and the number of teams present in a game. An initial version of the software that incorporates some of these changes is now available with the final version scheduled for completion during the summer of 2010. We anticipate publishing the final version as open source software.