Flexible Plans for Adaptation by End-Users
2011.
Cyrille Martin, Humbert Fiorino, Gaëlle Calvary
International Workshop on Generalized Planning, Conference , AAAI 2011, August 8th, 2011, San Francisco
Résumé
Ubiquitous computing promotes flexibility for the enduser.
This means that some design choices have to be shifted from design-time to run-time in order to involve the end-user into the decision process.
In this paper, we study flexible plans, i.e. plans that let the end-user arrange tasks planned by an automated process seeking to achieve his needs.
More precisely, we present an algorithm -graphplan that lets the end-user to decide the order of specific treatments (loop body) execution to a set of objects (loop variants).
-graphplan is based on graph planning structure.
Its strength is that it does not require any problem dependent knowledge to compute flexible plans.
By relaxing mutex constraints in the planning graph, -graphplan discovers the loop variants and builds the
macro-actions that constitute the loop bodies.
We show that -graphplan is performant with “iterative” as well as with “linear” domains.