Thresholded-Rewards Decision Problems: Acting Effectively in Timed Domains

Abstract

In timed, zero-sum games, winning against the opponent is more important than the final score. A team that is losing near the end of the game may choose to play aggressively to try to even the score before time runs out. In this thesis, we consider the problem of finding optimal policies in stochastic domains with limited time, some notion of score, and in complex environments, such as domains including opponents. This problem is relevant to many intelligent decision making tasks, not just games, as nearly every decision made in the real world depends on time. The work presented in this thesis has broad applications to domains possessing the key features of control under uncertainty, limited time, and some notion of score.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 02, 2009
Accession Number
ADA507016

Entities

People

  • Colin D McMillen

Organizations

  • Carnegie Mellon University

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Algorithms
  • Computational Science
  • Computer Programming
  • Computer Programs
  • Computer Science
  • Computers
  • Control Systems
  • Distribution Functions
  • Hidden Markov Models
  • Language
  • Network Protocols
  • Probability
  • Probability Distributions
  • Reinforcement Learning
  • Simulators
  • Teamwork
  • Zero-Sum Games

Readers

  • Game Theory.
  • Systems Analysis and Design
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.