Integrating artificial intelligence techniques to generate ground station schedules

Scheduling of contacts between space vehicles (SVs) and ground stations is of extreme significance since it is absolutely essential for data receipt from and transmission to satellites, vehicle maintenance, and orbit tracking and maintenance. We looked at the problem of scheduling contacts between SVs and the Air Force's Satellite Control Network (SCN). Our work integrates case-based reasoning, rule based systems, and generate-and-test techniques, all adopted from Artificial Intelligence. Our system creates a preliminary, daily SCN schedule with between approximately 500 to 1500 contact requests and other tasks (such as station down times). The goal is to create a schedule with as few conflicting contact requests as possible, which is then finalized by expert schedule planners. We evaluated our system looking at its performance using only one scheduling algorithm and also using the integration of the algorithms. The system was tested on real SCN schedules and removed on average 46.2% of conflicts, generating schedules which were on average 75.3% clear of conflicts. We also tested the system on schedules created by experts and which contained scheduling conflicts that the experts could not resolve; in these tests our system managed to resolve on average 44.4% of these conflicts, showing performance better than human expert schedulers.