Adapting the Multi Agent Planning System for RoboCup Simulation League

Student: Jun Akiyama

Supervisor: Dr. Gordon Wyeth

Category: Electrical Engineering Thesis Project

This thesis will demonstrate the design and implementation of the Multi Agent Planning System (MAPS) on fully autonomous soccer players. MAPS will look at instances of the game, assess the player’s situation, and derive an action plan for a player. It was originally designed for the RoboRoos who participate in the RoboCup in the small size league robot competition. These robots are made to play soccer in a five a side format on a table tennis size field. In this project, MAPS was adapted to the simulation league competition in order to play soccer on eleven a side full size soccer field.

The game is played on a virtual soccer field that will implement the physics involved in the motion of the players and the ball. The simulator is run on a UNIX platform in a client and server system. The soccer field being the server, it sets all the rules inherent in the game, and the players log onto the server as clients.

The UQ Crocaroos were used as the testing environment designed to investigate the multiple agent cooperation, and reaction in the field of artificial intelligence. They competed for the first time in the RoboCup world simulation league soccer in Seattle. These programs were designed to play soccer on a virtual soccer field with no human input. These players do not feel gravity, feel the grass beneath their shoes, or get tired from soccer game. They are made to believe they do. An analysis of the solution demonstrated that MAPS functions well knowing the exact details of the game situations. However the need for the development of inter player communication was discovered.

Poster Presentation (PDF)

Thesis Document (PDF)

Other Related Files