Diploma Thesis: XABSL - A Behavior Engineering System for Autonomous Agents

In the area of agent systems, as throughout in computer science, formal methods are applied to specify complex systems, to ensure certain properties of a system, or to generally simplify the development of solutions. In traditional symbolic artificial intelligence, logic and planning theories are usually used for modelling autonomous agents. But these approaches turned out to be not applicable for agent systems in highly dynamic environments. Therefore, mainly in intelligent robotics, theories or formalisms are often not used when programming agents to perform certain tasks. In this thesis, the Extensible Agent Behavior Specification Language (XABSL) is introduced as a pragmatic approach to agent engineering. It does not follow any agent theory but in return provides a powerful set of tools for the convenient and rapid programming of agent behavior. The system was succesfully applied by the GermanTeam in the RoboCup Sony Four Legged League, resulting in the win of the RoboCup world championship 2004.