Most complex problem solving tasks are composed of several subproblems that require specialized techniques to the kind of problems to be solved. However, building integrated architectures with specialized components, such that each one can handle a specific kind of subproblems, is a difficult task. As an example, recently the fields of AI planning and scheduling have witnessed a big interest on the integration of techniques from both areas in order to solve complex problems. These problems require the reasoning on which actions to be performed as well as their precedence constrains (planning) in combination with the reasoning with respect to the time at which those actions should be executed and the resources they use. In this paper we describe IPSS (Integrated Planning and Scheduling System) a domain independent reasoner that integrates AI Planning and Constraint Satisfaction (CS) by separating both reasoning tasks. Thanks to the use of these techniques, IPSS is able to reason about precedence constraints, time (deadline, time windows, etc) and resource usage/consumption. Experimental results show that the separation of the planning and scheduling reasoning can enable it to outperform state-of-the-art planners with time and resources reasoning capabilities.
IPSS: A Hybrid Reasoner for Planning and Scheduling
Contributo in atti di convegno
16th European Conference on Artificial Intelligence (ECAI 2004), pp. 1065–1066, Valencia, Spain, August 23-27 2004