Mapping Planning Actions and Partially-Ordered Plans into Execution Knowledge
Manuela Veloso, Paola Rizzo, Viale Marx · 1998
In essence, the underlying ultimate purpose of classical deliberative planning systems is to model some world task and generate plans that can then be executed. Execution systems rely on the assumption that there is a built-in library of a variety of plans and primi-tive actions for performing tasks. Although hard and time consuming, such libraries of execution plans and actions are usually handcoded. Faced with two differ-ent systems, a planner and an executor, we analyze in this work the representational map between plan-ning actions and partially-ordered plans and the exe-cution knowledge. We developed algorithms to auto-matically translate classical planning operators to ex-ecution primitive actions, and partially-ordered plans into executable tasks with partially ordered subtasks. We implemented our work using the PRODIGY planner and the RAP execution system. We provide illustrative examples of how partially-ordered plans produced by PRODIGY are translated to Reactive Action Packages suitable for execution by the RAP system.