Probabilistic behavioural state machines

  • Authors:
  • Peter Novák

  • Affiliations:
  • Department of Informatics, Clausthal University of Technology, Clausthal-Zellerfeld, Germany

  • Venue:
  • ProMAS'09 Proceedings of the 7th international conference on Programming multi-agent systems
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Development of embodied cognitive agents in agent oriented programming languages naturally leads to writing underspecified programs. The semantics of BDI inspired rule based agent programming languages leaves room for various alternatives as to how to implement the action selection mechanism of an agent (paraphrased from [5]). To facilitate encoding of heuristics for the non-deterministic action selection mechanism, I introduce a probabilistic extension of the framework of Behavioural State Machines and its associated programming language interpreter Jazzyk. The language rules coupling a triggering condition and an applicable behaviour are extended with labels, thus allowing finer grained control of the behaviour selection mechanism of the underlying interpreter. In consequence, the agent program not only prescribes a set of mental state transitions enabled in a given context, but also specifies a probability distribution over them.