by Dan B.
In the late 1980s, Martin Marietta had a robust artificial
intelligence (AI) research team. We built one of the world’s most sophisticated
scheduling systems, called MAESTRO. It allowed representation of very complex
activities and constraints, and built “good” schedules hundreds of times faster
than humans. We applied it to the problem of scheduling all activities on what
was the Martin Marietta design for the common modules on the proposed Space
Station Freedom. A common module power system breadboard was developed, and
interfaced with a fault isolation and detection system (another AI effort at
Martin Marietta), which was interfaced with MAESTRO.
Operators were able to schedule activities automatically, then execute the power usages for them on the breadboard. When a researcher threw a wrench on the breadboard, shorting out some power paths, the fault isolation system safed the breadboard, then communicated the power system changes to the scheduler. MAESTRO determined which activities had been interrupted, found ways to continue some using alternate resources, curtailed others and scheduled other activities that could use released resources, displaying the changes for the user. The operator was given a limited time to review the changes and then released the revised schedule. All of this took on the order of five minutes, from fault to continuing with work-arounds in place. The system won a NASA New Technology Award, and resulted in a number of research publications and invited talks at major conferences. Being part of that effort was a highlight of my career.
Operators were able to schedule activities automatically, then execute the power usages for them on the breadboard. When a researcher threw a wrench on the breadboard, shorting out some power paths, the fault isolation system safed the breadboard, then communicated the power system changes to the scheduler. MAESTRO determined which activities had been interrupted, found ways to continue some using alternate resources, curtailed others and scheduled other activities that could use released resources, displaying the changes for the user. The operator was given a limited time to review the changes and then released the revised schedule. All of this took on the order of five minutes, from fault to continuing with work-arounds in place. The system won a NASA New Technology Award, and resulted in a number of research publications and invited talks at major conferences. Being part of that effort was a highlight of my career.
