Integrating Planning and Acting in Autonomous Robots through a Unified Hierarchical Representation Model

Abstract

One way to apply AI is to create intelligent agents that break down challenging problems into multiple layers of abstraction, thereby simplifying the problem at each level. The Hierarchical Task Network (HTN) architecture enables us to perform these tasks. The task planning process for an autonomous robot involved operating under the assumption of a closed and deterministic world, (as a domestic environment), in order to accomplish the designated tasks with a level of precision. HTN planners utilize descriptive action models to account for the next states in the state transition system during the planning process. Our HTN planning algorithm is designed to work with and integrate with actors to choose optimal paths and for re-planning when execution errors arise.