Evolving engineering mission schedules: a machine-learning approach to scheduling

摘要

Scheduling is frequently a militarily significant procedure. In a battlefield environment, it is often important to have access to rapid scheduling techniques that produce effective and efficient schedules. Standard approaches to scheduling may be ineffective whenever the characteristics of the schedule to be generated-its size, its complexity, interactions among its components, etc.-make it difficult to generate a satisfactory schedule in a reasonable amount of time. In such cases it may be possible to produce near-optimal schedules rapidly through the use of Genetic Algorithms, a sub-symbolic machine learning technique. This approach evolves a schedule probabilistically from a population of schedules, rather than attempting to generate one deterministically. Results obtained in a project to generate mission schedules for U.S. Army engineering units demonstrate that schedules can be evolved relatively rapidly, and that the quality of these schedules is high.