Analysis of Agent Programs Using Action Models

摘要

The use of action models for the analysis of control programs can be useful for two reasons. First, it promises to deliver better tools for the simulation, verification and synthesis of control programs, and second it presents challenging problems for theories of action and knowledge. In this paper we use a theory of actions and knowledge developed elsewhere to analyze control programs for navigation tasks. We model both physical and sensing actions and establish conditions under which different control programs are executable and lead the agent to the intended goal.