Enhancement of the Ground Processing Scheduling System (gpss)

A scheduling process that maintains, repairs and configures a space shuttle orbiter must satisfy temporal, resource and configuration constraints. The GPSS, which stands for ground processing scheduling system, has been successfully used to schedule tasks that prepare an orbiter for its next launch. The objective of this project is to enhance and to improve the GPSS. The GPSS uses an iterative repair method that starts with a rough schedule, which satisfies only the temporal constraints. Then it repairs the resource and configuration constraint violations one by one until no more violations exist or no further improvements can be made. The conflict resolution, which repairs all the resource and the configuration constraint violations, forms the bulk of the cognitive and computational complexity of the system. We started with a study of the solution space of the GPSS which will help us to develop algorithms and strategies to improve the efficiency of the deconfliction process. To our knowledge there has been no previous effort to obtain this solution space. We probed the existing GPSS and constructed its solution spaces. The total number of iterations during the deconfliction process depends on the quality of the initial schedule better the quality the smaller the number of iterations. The quality of the initial schedule can be improved by applying a genetic algorithm on an abstract problem generated from the original problem. We have designed and developed a genetic algorithm and interfaced with the GPSS to generate a feasible schedule that satisfies temporal and resource constraints. Further, the computational efficiency of the deconfliction portion of the GPSS can be improved by using better heuristic strategies and better cost estimators instead of actually computing the cost for the heuristic table. The heuristic table used in the GPSS is very static, and the normalization process is somewhat arbitrary. We have designed and developed an evolutionary learning technique and interfaced it with the GPSS so as to learn the appropriate ways of combining the heuristic parameters and thereby reduce conflicts faster.