Proceedings Joint IEEE/INRIA Conference on Emerging Technologies for Factory Automation, Paris, France, September, 1995

A Constraint-Posting Framework for Scheduling Under Complex Constraints

Cheng-Chung Cheng and Stephen F. Smith

The Robotics Institute
Carnegie Mellon University
Pittsburgh, PA 15213, USA

Abstract

Scheduling in many practical industrial domains is complicated by the need to account for diverse and idiosyncratic constraints. Quite often these requirements are at odds with the techniques and results produced by the scheduling research community, which has focused in large part on solutions to more idealized, canonical problems. Recent research in temporal reasoning and constraint satisfaction has produced problem solving models that operate with respect to much more general representational assumptions. These frameworks offer possibilities for developing scheduling technologies that naturally extend to accommodate the peculiarities of various application domains. One critical issue, of course, is whether such generality can be obtained without sacrificing scheduling performance.

In this paper, we investigate this issue through application of a previously developed constraint satisfaction problem solving (CSP) model for deadline scheduling to a complicated, multi-product hoist scheduling problem encountered in printed circuit board (PCB) manufacturing. The goal is to maximize throughput of an automated PCB electroplating facility while ensuring feasibility with respect to process, capacity and material movement constraints. Building from a heuristic procedure generically referred to as PCP (precedence constraint posting), which relies on a temporal constraint graph representation of the problem, we straightforwardly define an extended solution procedure for minimizing makespan. In a series of comparative experiments, our procedure is found to significantly outperform previously published procedures for solving this hoist problem across a broad range of input assumptions.

The research reported in this paper has been sponsored in part by the National Aeronautics and Space Administration, under contract NCC 2-531, by the Advanced Research Projects Agency under contract F30602-90-C-0119 and the CMU Robotics Institute. The authors can be reached through email at sfs@cs.cmu.edu.
Full paper in Postscript