Using genetic algorithms for complex, real-time scheduling applications

Applications that require real-time scheduling of large-scale problems in complex domains present a number of difficulties for search and optimization techniques. These difficulties include: (i) a search space whose size grows exponentially with the size of the problem, (ii) a problem that is constantly changing due to a changing environment and user interaction, and (iii) the need to trade off between a variety of different criteria measuring the relative fitness of a particular schedule. BBN has used genetic algorithms to solve a variety of real-world scheduling problems, including applications in areas such as field service scheduling, job shop scheduling, transportation scheduling, and laboratory experiment scheduling. With the recent acquisition of BBN by GTE, we expect that our scheduling technology will soon be deployed to solve hard scheduling problems in network operations management. Our genetic algorithm technology addresses the issues above and provides a domain-independent infrastructure upon which we have rapidly developed customized software solutions for various customers. Because the infrastructure is domain independent, adding network operations management to the list of domains in which our scheduling technology has been successfully applied should be straightforward