![]() |
|||
![]()
|
![]() |
![]() |
![]() |
4. TESTING AND PERFORMANCE OF GUESSScheduling problems are often complicated by large numbers of constraints relating activities to each other, resources to activities and to each other, and either resources or activities to events external to the system (Morton and Pentico, 1993). For example, there may be "precedence constraints" connecting activities that specify which activities must precede other activities, and by how much of a delay, or by how much allowed overlap. Or two particular activities may interfere with each other and be unable to use the same required resource simultaneously. Or it may not be possible to use two resources simultaneously during certain parts of the day or on the same activity. Or a resource may be unavailable during specified intervals due to planned maintenance or planned use outside the system. Since these complex interrelationships can make exact or even approximate solutions of large scheduling problems very difficult, it is natural to attempt to solve simpler versions of a problem first in order to gain insight. Then one can test how sensitive the solution is to this complexity and find approximate solutions to difficult problems where the complexity proves central (Morton and Pentico, 1993). This is the approach that the GUESS team followed in testing GUESS for proper verification and validation, as well as for its generic capabilities. Morton and Pentico (1993) point out that a major need exists for the development of generic heuristic scheduling shells that would allow software houses or even sophisticated users to craft a scheduling system by inputting a description of the system structure and general parameters. Pinedo (1995) also suggests the need for a generic scheduling system. He indicates the following typical features of a generic scheduler (1995):
GUESS is currently being tested in a number of applications. For NASA applications, such as scheduling satellite experimenter requests to use the NASA supported satellites, GUESS can schedule 2551 events and over 14,000 constraints in 45 seconds on a Dell 486 computer. This performance corresponds well with other NASA expert scheduling systems that can schedule up to 6000 events in 2.5 to 3 minutes. We are also currently testing GUESS in other scheduling domains. These include: (1) scheduling for an Army strategic decision support application (i.e., force mobilization in a deployed theater); (2) scheduling Department of Computer Science courses and corresponding sections for a local college; (3) scheduling the baseball games; and (4) scheduling Army battalion training exercies. From preliminary results, we are optimistic that GUESS is generic enough to be easily used in these different scheduling applications. 5. SUMMARYIn the coming years, the major trends in intelligent scheduling systems will be:
We feel that GUESS is an effort that supplements well these future directions in intelligent scheduling. More testing for the generic qualities of GUESS will be conducted in the near term, as will the expansion of the number of scheduling techniques in the GUESS toolkit. ACKNOWLEDGMENTSThe authors appreciate the support from Colonel Wilkes and Professor Campbell at the Center for Strategic Leadership at the U.S. Army War College, George Washington University, and support from NASA Goddard Space Flight Center. REFERENCES
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |