# Refinement vs Debugging Assignment.docx - GROUP 12...

• 6

This preview shows page 1 - 3 out of 6 pages.

GROUP 12 Refinement in Search Plans Planner A planner is a constructive strategy to achieve a goal. Thus, planning is a problem-solving technique that involves deciding upon a course of action(plan) before acting to accomplish a goal. Planning can be viewed as a search of either space of world states(state-space) or space consists of partial plans (plan space). state-space planning, a program searches through a space of world states, seeking to find a path or paths that will take it from its initial state to a goal state. State-space planning is too inflexible, because: it creates plans that are total orderings of a set of steps, and it assembles these plans in exactly the same order. Plan space- each node is a partially completed plan. Transitions in this space are accomplished by adding actions to plans. In plan-space planning, a program searches through a space of plans, seeking a plan that will take it from its initial state to a goal state. An operator creates a new plan from an old plan. REFINEMENT VS DEBUGGING Refinement Refinement is the process of gradually adding actions and constraints on a plan. Plan refinement operators modify the current plan by appending new actions to the beginning of the sequence. Refinement is one way of constructing non-hierarchical linear planners by searching a space of plans rather than the original state space. The search process that is carried out by an AI planning system corresponds to taking an AND/OR graph and generating from it an equivalent state-space graph, one OR-branch at a time. This process is known as serializing the AND/OR graph. As a result, a refinement planner’s search space is an AND/OR graph, and the planner serializes this graph by mapping it into an equivalent state-space graph. Different plan refinement strategies produce different serializations of the AND/OR graph. The sizes of different serializations of the AND/OR graph can differ by an exponential amount. Specifically, a planner whose refinement strategy produces a small serialization is likely to be more efficient than a planner whose refinement strategy produces a large serialization.
GROUP 12 Refinement Choice In a partially developed plan, there may be several elements of the plan that need to be refined in one way or another (fig. 1). These may include both unachieved goals or tasks