Learning Hierarchical Task Networks

This project aims at developing an unified architecture for learning and reusing Hierarchical Task Networks (HTNs): learning of Hierarchical Task Network (HTNs), Applying learned HTNs for problem-solving and an educational component where secondary-school students experience knowledge engineering activities (e.g., to encode HTNs).

Activities:

• We built a prototype of a Hierarchical Task Network Learner, called HTN-MAKER. HTN-Maker is an offline and incremental algorithm for learning the structural relations between tasks in a Hierarchical Task Network (HTN). HTN-MAKER receives as input a STRIPS domain model, a collection of STRIPS plans, and a collection of task definitions, and produces an HTN domain model.
  • We are studying ways to augment partial HTNs with episodic knowledge (i.e., cases).

• We are currently enhancing experiments on RepairSHOP. RepairSHOP is a plan adaptation algorithm build on top of the HTN planner SHOP, which implements a variant of HTN planning called Ordered Task Decomposition. In this variant tasks are totally ordered and conditions are evaluated relative to the current state of the world, which is updated during planning.

• A course on Automated Planning developed under this project.

Publications:

• A paper describing HTNMaker

• A paper describing RepairSHOP

• An application of RepairSHOP for creating adaptive game AI

• An overview on plan adaptation

This material is based upon work supported by the National Science Foundation under Grant No. NSF 0642882. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF)

Last updated: Tue Sept. 25 14:49:51 EST 2007