An open agent architecture: Fundamentals

Different application domains require different knowledge representation techniques. Therefore a robust agent oriented programming framework should not commit to a single KR technology. Instead, it should facilitate an easy integration of heterogeneous knowledge representation techniques in a single agent system. Similarly, different situations an agent might happen to be in, require different schemes of behaviours, such as reactive vs. sequential. Mixing these can lead to a more robust agent system. Finally, a complete agent-oriented programming system should be complemented by a logic for reasoning about resulting implemented agent programs and ideally serving as a methodological framework as well. In the first technical note of the series on open agent architecture we introduce theoretical fundaments of a novel agent programming framework of Behavioural State Machines. The presented framework draws a strict distinction between a knowledge representational and a behavioural level of an agent program. It 1) supports a high level of modularity w.r.t. employed KR technologies, 2) allows implementation of hybrid agents, i.e. mixing of both reactive, as well as sequential behaviours, and 3) provides a clear and concise semantics based on a well-studied computational model of Gurevich’s Abstract State Machines.