leeurban07transfer - Transfer Learning of Hierarchical...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Transfer Learning of Hierarchical Task-Network Planning Methods in a Real-Time Strategy Game Stephen Lee-Urban H´ector Mu˜noz-Avila Department of Computer Science Lehigh University Lehigh, PA { sml3,hem4 } @lehigh.edu Austin Parker Ugur Kuter Dana Nau Department of Computer Science and Institute for Systems Research University of Maryland College Park, Maryland, 20742 { austinjp,ukuter,nau } @cs.umd.edu Abstract We describe a new integrated and automated AI planning and learning architecture, called Learn2SHOP . Learn2SHOP departs significantly from the previous works on AI planning and learning in that its modular architecture integrates Hier- archical Task Network (HTN) planning, concept learning, and computer simulations. Using simulations during the planning and learning process enables the system to get information about the outcomes of the actions. We have implemented Learn2SHOP and tested it on a transfer-learning task. The objective of transfer learning is transferring knowledge and skills learned from a wide variety of previous situations to the current, and likely different, previously unencountered prob- lems(s). The experiments with Learn2SHOP have demon- strated the advantages of integrating planning, learning, and simulation in a real-time strategy game engine. Introduction Learning in the context of automated planning has been a frequently studied research topic (Zimmerman & Kamb- hampati 2003), where the objective is to develop and use automated techniques to learn some knowledge that is used to improve the performance of a planner. Many different techniques have been developed with this objective, includ- ing learning macro-operators , e.g., (Mooney 1988; Botea, M¨uller, & Schaeffer 2005), learning search control knowl- edge (Mitchell 1977; Minton 1988; Fern, Yoon, & Givan 2004), learning of task hierarchies (Choi & Langley 2005; Reddy & Tadepalli 1997; Ruby & Kibler 1991) and learn- ing plan abstraction (Knoblock 1993; Bergmann & Wilke 1996). This paper focuses on how to take knowledge that was acquired under one model and harness it in the learning within another model (e.g., taking lessons that were learned in one planning scenario and using them in other (similar) planning scenarios). We address this issue with a mod- ular framework called Learn2SHOP based on Hierarchi- cal Task Network (hereafter HTN) planning. Unlike other systems, Learn2SHOP learns using simulation in an ac- tual gaming environment to validate the expectations of a provided model. Further, Learn2SHOP uses a general Copyright c 2007, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. HTN framework which contains methods describing stan- dard techniques. Since these techniques can be applicable to large classes of games, Learn2SHOP will be able to handle changing environments and be able to transfer knowledge from one game to the other....
View Full Document

This note was uploaded on 08/25/2011 for the course EGN 3060c taught by Professor Sukthankar,g during the Fall '08 term at University of Central Florida.

Page1 / 6

leeurban07transfer - Transfer Learning of Hierarchical...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online