Timelines with temporal uncertainty

Abstract

Timelines are a formalism to model planning domains where the temporal aspects are predominant, and have been used in many real-world applications. Despite their practical success, a major limitation is the inability to model temporal uncertainty, i.e. the fact that the plan executor cannot decide the actual duration of some activities. In this paper we make two key contributions. First, we propose a comprehensive, semantically well founded framework that (conservatively) extends with temporal uncertainty the state of the art timeline approach. Second, we focus on the problem of producing time-triggered plans that are robust with respect to temporal uncertainty, under a bounded horizon. In this setting, we present the first complete algorithm, and we show how it can be made practical by leveraging the power of Satisfiability Modulo Theories. Copyright © 2013, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.