Title

Beating The Defense: Using Plan Recognition To Inform Learning Agents

Abstract

In this paper, we investigate the hypothesis that plan recognition can significantly improve the performance of a case-based reinforcement learner in an adversarial action selection task. Our environment is a simplification of an Amen-can football game. The performance task is to control the behavior of a quarterback in a pass play, where the goal is to maximize yardage gained. Plan recognition focuses on pre dicting the play of the defensive team. We modeled plan recognition as an unsupervised learning task, and conducted a lesion study. We found that plan recognition was accurate, and that it significantly improved performance. More gener ally, our studies show that plan recognition reduced the di mensionality of the state space, which allowed learning to be conducted more effectively. We describe the algorithms, explain the reasons for performance improvement, and also de scribe a further empirical comparison that highlights the utility of plan recognition for this task. Copyright © 2009, Assocation for the Advancement of ArtdicaI Intelligence (www.aaai.org). All rights reserved.

Publication Date

11-4-2009

Publication Title

Proceedings of the 22nd International Florida Artificial Intelligence Research Society Conference, FLAIRS-22

Number of Pages

337-342

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

Socpus ID

70350500065 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/70350500065

This document is currently not available here.

Share

COinS