Synthesizing plans for multiple domains

Abstract

Intelligent agents acting in real world environments need to synthesize their course of action based on multiple sources of knowledge. They also need to generate plans that smoothly integrate actions from different domains. In this paper we present a generic approach to synthesize plans for solving planning problems involving multiple domains. The proposed approach performs search hierarchically by starting planning in one domain and considering subgoals related to the other domains as abstract tasks to be planned for later when their respective domains are considered. To plan in each domain, a domain-dependent planner can be used, making it possible to integrate different planners, possibly with different specializations. We outline the algorithm, and the assumptions underlying its functionality. We also demonstrate through a detailed example, how the proposed framework compares to planning in one global domain. © Springer-Verlag Berlin Heidelberg 2005.