Journal of Logic and Computation, Volume 11, Issue 2, pp. 337-354: Abstract.
Minimizing Agent Specifications using a Logic of Knowledge and Actions
Department of Computer Science, William Paterson University of New Jersey, Wayne, NJ 07470, USA. E-mail: firstname.lastname@example.org
Given the specification of an agent's behaviour in terms of what it can do at every task-domain situation, we propose a mechanism to derive an equivalent specification but with fewer states that is based on Milner's `observation equivalence'. This is done by identifying internal states of an agent that are `observation equivalent'. We also introduce the notion of `pertinent fact' about an agent's behaviour at an internal state as a fact that is true at all internal states of the agent that are `observation equivalent' to this particular state, and also define a new logic, Propositional Logic of Knowledge and Actions (PLKA), to specify and reason about the behaviour of an agent in terms of pertinent facts. A complete axiomatization of this logic, which stands as a conservative extension of both the logic S5 of knowledge and Propositional Dynamic Logic (PDL), and a characterization of an agent's behaviour in terms of `pertinent facts' are also provided.
Keywords: Logic of knowledge, logic of knowledge and actions, object system, observation equivalence, procedural knowledge, propositional dynamic logic, rational behaviour, reactive systems