Speciation methods have been used with genetic algorithms to preserve diversity and to locate multiple solutions in an evolving population. Speciation can combat the problem of premature convergence and prevent sub-optimal soltuions from being reached. Current techniques of speciation have focused on problems having binary chromosome encodings. This paper shows how speciation can be applied to the problem domain of routing and scheduling to obtain a number of interesting solutions in a permultationbased problem space. The Open Shop Scheduling problem is used to demonstrate the technique as this problem is consideed to be a canonical routing and ordering type problem. This paper demonstrates the the use of speciation with Sharing Functions with an appropriately constructed multi-chromosomal order-based distance metric. This leads a genetic algorithm to find multiple global optima spaced well apart in the solution space in a sample open shop scheduling problem.
展开▼
