Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A Concise Introduction to Models and Methods for Automated PlanningPlanning with Sensing: Logical Models

A Concise Introduction to Models and Methods for Automated Planning: Planning with Sensing:... [In this chapter we focus on models and methods for planning with uncertainty and sensing. This is usually called partial observable planning, planning with sensing, or contingent planning. In these models the true state of the environment is not assumed to be known or predictable, yet partial information about the state is assumed to be available from sensors. Uncertainty is represented by sets of states, referred to as beliefs. We will then consider probabilistic models where beliefs are not represented by sets of states but by probability distributions. Logical and probabilistic models however are closely related. A key difference is that, in the absence of probabilistic information, policies or plans are evaluated by their cost in the worst case rather than their expected cost. There may indeed be policies with small expected cost to the goal but infinite cost in the worst case, as when the state trajectories that fail to reach the goal in a bounded number of steps have a vanishing small probability. Still, as we will see, the policies that ensure that the goal is achieved with certainty can be fully characterized in the logical setting without probabilities at all as the policies that are strongly cyclic [Daniele et al., 1999].] http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

A Concise Introduction to Models and Methods for Automated PlanningPlanning with Sensing: Logical Models

Part of the Synthesis Lectures on Artificial Intelligence and Machine Learning Book Series
Geffner, Hector; Bonet, Blai
Springer Journals — Jan 1, 2013
Download PDF
13 pages

Loading next page...
 
/lp/springer-journals/a-concise-introduction-to-models-and-methods-for-automated-planning-R87nopZqkU
Publisher
Springer International Publishing
Copyright
© Springer Nature Switzerland AG 2013
ISBN
978-3-031-00436-0
Pages
65 –78
DOI
10.1007/978-3-031-01564-9_5
Publisher site
See Chapter on Publisher Site

Abstract

[In this chapter we focus on models and methods for planning with uncertainty and sensing. This is usually called partial observable planning, planning with sensing, or contingent planning. In these models the true state of the environment is not assumed to be known or predictable, yet partial information about the state is assumed to be available from sensors. Uncertainty is represented by sets of states, referred to as beliefs. We will then consider probabilistic models where beliefs are not represented by sets of states but by probability distributions. Logical and probabilistic models however are closely related. A key difference is that, in the absence of probabilistic information, policies or plans are evaluated by their cost in the worst case rather than their expected cost. There may indeed be policies with small expected cost to the goal but infinite cost in the worst case, as when the state trajectories that fail to reach the goal in a bounded number of steps have a vanishing small probability. Still, as we will see, the policies that ensure that the goal is achieved with certainty can be fully characterized in the logical setting without probabilities at all as the policies that are strongly cyclic [Daniele et al., 1999].]

Published: Jan 1, 2013

There are no references for this article.