Automatically Generating Plans for Manufacturing

Abstract

While machine planning has attracted great interest among researchers, it has seldom been used outside research labs. One impediment to wide-spread use is that existing planners are often difficult to integrate with other parts of a manufacturing system. We address this problem by showing how assembly trees (constructs often used by factories identifying how to construct an object) can easily be converted into HTN operators for our machine planner. We also demonstrate that our plans can be easily converted to a Petri-Net or matrix representation which ordinary discrete-event controllers can manipulate. We view our planner as one portion of a complete control system. We also demonstrate how our system can combine multiple alternatives into a single representation. Finally, we show that the combined representation can be converted back into a more conventional plan representation, allowing machine planners to compactly reason about alternate courses of action. keywords: Machine planning, HTN planning, Intelligent control, discrete-event control, manufacturing * This research was supported in part by the National Science Foundation, under grant GER-9355110. † wharris@cse.uta.edu, Department of Computer Science and Engineering, University of Texas at Arlington, Arlington, TX 76019 ‡ cook@cse.uta.edu, Department of Computer Science and Engineering, University of Texas at Arlington, Arlington, TX 76019 § flewis@arri.uta.edu, Automation & Robotics Research Institute, 7300 Jack Newell Blvd. South, Fort Worth, TX 76118