Author: Boella Guido Damiano Rossana
Publisher: Taylor & Francis Ltd
ISSN: 1087-6545
Source: Applied Artificial Intelligence, Vol.22, Iss.10, 2008-11, pp. : 937-963
Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.
Abstract
In this article, we present a replanning algorithm for a decision-theoretic hierarchical planner, illustrate the experimental methodology we designed to investigate its performance, and provide an evaluation of the algorithm. The methodology relies on an agent-based framework, in which plan failures can emerge from the interplay of the agent and the environment. Given this framework, the performance of the replanning algorithm is compared with the one of planning from scratch the solution to the planning problem by executing experiments in different domains. The empirical evaluation shows the superiority of replanning with respect to planning from scratch. However, the observation of significant differences in the data collected across planning domains confirm the importance of empirical evaluation in practical systems.
Related content
A Sparse Sampling Algorithm for Near-Optimal Planning in Large Markov Decision Processes
By Kearns M.
Machine Learning, Vol. 49, Iss. 2-3, 2002-11 ,pp. :