Abstract for the October 10, 1997 Robotics Institute Seminar

Warning:
This page is provided for historical and archival purposes only. While the seminar dates are correct, we offer no guarantee of informational accuracy or link validity. Contact information for the speakers, hosts and seminar committee are certainly out of date.

Mixed-Initiative Planning and Scheduling Systems

Stephen Smith
Robotics Institute
Carnegie Mellon University

4:00pm, Adamson Wing, Baker Hall

In practice, planning and scheduling is rarely (if ever) a one-shot generative task aimed at satisfying pre-specified constraints and objectives. It is an iterative (and ongoing) process, concerned concurrently with (1) building understanding of the evolving problem and execution state, (2) determining what the constraints and objectives are (or should be), and (3) generating and maintaining an acceptable solution. Despite this fact, most current planning and scheduling systems are based rigidly on a "specify and solve" model of user-system interaction, where the user specifies problem inputs and constraints up-front, and the back-end problem solver is then invoked to generate a solution. This model of interaction does not match the requirements of the larger planning and scheduling process. There is no persistence of decisions over time, no ability to control how solutions change in response to changed inputs or constraints, and no ability to converge to and maintain an acceptable solution in an incremental, controlled manner.

A better framework for user-system interaction follows from a view of planning and scheduling as an incremental change process. This is the approach taken in the design of Ozone, a configurable framework for mixed-initiative scheduling and planning under continuing development at CMU. Within Ozone, incremental constraint-based problem solving techniques are combined with graphical solution visualization and manipulation techniques to provide a flexible, "spreadsheet-like" planning and scheduling model. The user analyzes solution elements and formulates change directives from aggregate task-oriented perspectives; the system manages the details of implementing change directives in accordance with user goals and expectations. The Ozone framework has been applied to develop application systems in a range of complex planning and scheduling domains, most notably the Ditops transportation scheduler. In this talk, I will summarize the technical basis of this mixed-initiative problem solving framework, demonstrate current capabilities, and describe current research, application and technology transition efforts.

Biographical Sketch:
Stephen Smith is a Senior Research Scientist in the Robotics Institute at Carnegie Mellon University, where he is Director of the Intelligent Coordination and Logistics Laboratory. He holds a B.S. in Mathematics from Westminster College and M.S. and Ph.D. degrees in Computer Science from the University of Pittsburgh. For the past several years, his research has focused generally on development of knowledge-based techniques and tools for complex planning, scheduling and coordination problems. His current research interests include constraint-based reasoning, mixed-initiative problem solving, reactive planning and scheduling, (re)configurable and self-organizing systems, and agent-based modeling and analysis of supply chain dynamics.

For appointments, please contact the speaker: sfs@ri.cmu.edu

Last Modified on: Mon Oct 6, 1997

This appears on the World Wide Web at http://www.frc.ri.cmu.edu/~mcm/seminar.html

Martin C. Martin, <mm+@cmu.edu>