Robotic task planning: independent of agents but dependent on time

A description is given of the AI-TASCS (agent-independent task specification and control system), which is designed to describe and generate, in a representation independent of an agent, plans that can subsequently be translated into agent-dependent instruction suitable for execution. Agents and objects are represented in an object-oriented fashion. Tasks can either be primitive actions based on movement or abstracted to higher level tasks (synonymous with plans) consisting of subtasks arranged in temporal networks (which allow temporal ambiguity in describing task orderings). Tasks and objects are combined to form an agent-independent plan. Plan translation transforms this into an agent-dependent plan by validating the combination of the plan with the properties and constraints of a specific agent, resulting in a set of instructions executable by the agent. Planning errors discovered during execution are recovered by either retranslating a certain portion of the agent-independent plan, or by translating a new agent-independent plan generated from the point of error. AI-TASCS is designed for the maintenance and repair of Space Station Freedom.<<ETX>>