Designing an agent system for controlling a robotic soccer team

Abstract

Robotic soccer is a way of putting different developments in intelligent agents into practice, including not only problems such as multi-agent planning and coordination, but also physical problems related to vision and communication subsystems. Because these problems cannot be all taken into account beforehand, the system must be designed to be robust enough to recover from any eventualities. In this work, we present the design used as the basis for an agents system implemented for the control of a team of robots for the E-League competition in RoboCup 2004. The implementation of the system was carried out following a layered design, with the objective of having a set of Service Layers, each of which is associated with a different level of abstraction. This layered design allows to construct a functional system with basic services that can be tested and refined progressively. The layers that are proposed as a basis for the arquitecture of a robotic soccer team offer a modular design, allowing the possibility of reuse in other robotic soccer leagues. Finally, the agents are implemented using the prolog language; the three uppermost layers in the hierarchy offer interfaces designed explicitly for this language.