Universal Planning: An (Almost) Universally Bad Idea

To present a sharp criticism of the approach known as universal planning, I begin by giving a precise definition of it. The key idea in this work is that an agent is working to achieve some goal and that to determine what to do next in the pursuit of this goal, the agent finds its current situation in a large table that prescribes the correct action to take. Of course, the action suggested by the table might simply be, “Think about your current situation and decide what to do next.” This method is, in many ways, representative of the conventional approach to planning; however, what distinguishes universal plans from conventional plans is that the action suggested by a universal plan is always a primitive one that the agent can execute immediately (Agre and Chapman 1987; Drummond 1988; Kaelbling 1988; Nilsson 1989; Rosenschein and Kaelbling 1986; Schoppers 1987). Definition 1.1. A universal plan is an arbitrary function from the set of possible situations S into the set of primitive actions A. I assume throughout this article that some fixed universal plan u describes the intended behavior of the agent being constructed; the question that I want to investigate is whether it is practical to provide the agent with an explicit description of u in its entirety as opposed to some method (such as “invoke the planner”) by which u(s) can be computed if the agent finds itself in situation s. Given Definition 1.1, I can draw distinctions between the intentions of the various authors working with universal plans. Schoppers, who coined the term universal plan, seems to feel that it will be possible to equip an agent with a completely accurate description of a universal plan for acting in its particular environment. However, other authors (Agre and Chapman 1987; Drummond 1988; Nilsson 1989) concede that only some approximation to the universal plan u will be available. Rosenschein and Kaelbling (Rosenschein and Kaelbling 1986; Kaelbling 1988) seem to agree with Schoppers, although I argue that this apparent agreement misrepresents their actual beliefs. Another choice needs to be made if an approximate universal plan is used. Agre and Articles