AI Planning: Systems and Techniques

that can describe a set of actions. .. that can be expected to allow the system to reach a desired goal. A long-standing problem in the field of automated reasoning is designing systems that can describe a set of actions (or a plan) that can be expected to allow the system to reach a desired goal. Ideally, this set of actions is then passed to a robot, a m a n u f a c t u r i n g system, or some other form of effec-tor, which can follow the plan and produce the desired result. The design of such planners has been with AI since its earliest days, and a large number of techniques have been introduced in progressively more ambitious systems over a long period. In addition , planning research has introduced many problems to the field of AI. Some examples are the representation and the reasoning about time, causality, and intentions; physical or other constraints on suitable solutions; uncertainty in the execution of plans; sensation and perception of the real world and the holding of beliefs about it; and multiple agents who might cooperate or interfere. Planning problems, like most AI topics, have been attacked in two major ways: approaches that try to understand and solve the general problem without the use of domain-specific knowledge and approaches that directly use domain heuristics. In planning , these approaches are often referred to as domain dependent (those that use domain-specific heuristics to control the planner's operation) and domain independent (those in which the planning knowledge representation and algorithms are expected to work for a reasonably large variety of application domains). The issues involved in the design of domain-dependent planners are those generally found in applied approaches to AI: the need to justify solutions, the difficulty of knowledge acquisition, and the fact that the design principles might not map well from one application domain to another. Work in domain-independent planning has formed the bulk of AI research in planning. The long history of these efforts (figure 1) has led to the discovery of many recurring problems as well as to certain standard solutions. In addition, there have been a number of This article reviews research in the development of plan generation systems. Our goal is to familiarize the reader with some of the important problems that have arisen in the design of planning systems and to discuss some …