Planning by search through simulations

Abstract

This research is directed towards the creation of automatic planning systems capable of directing the actions of robots in solving realistic tasks. Realistic domains often involve constraints on quantitative issues such as time and cost. In order to produce successful plans, interactions and conflicts between plan steps that involve these quantities must be detected and resolved. The central assumption of this work is that the ordering of tasks, and the sequencing of the plan steps created for their solution, are major and necessary parts of the planning process. The order in which a plan's actions are to be carried out can greatly affect the efficiency and outcome of executing the plan. The actual design of the plan can and should be affected by the ordering of its steps. Keywords: Heuristic search; Scheduling algorithm; Scheduling loops.