Constraint-Directed Reasoning as a Basis for Real-Time Planning

Constraint-directed reasoning has emerged as a powerful problem-solving paradigm used widely in artificial intelligence applications. This paper explores the use of constraints as a foundation for real-time planning in complex domains. We describe the function of constraints as a mechanism for representing and working with domain knowledge in resource bounded agents, and examine the types and organization of constraints necessary for agents in complex domains. We also describe an architecture for real-time planning based on constraint-directed reasoning. This architecture allows the resource-bounded consideration of constraints on an agent's behaviour, and employs constraint-directed reasoning both to represent domain knowledge as well as to control the extent of the agent's reasoning.