Genetic Fuzzy Modeling of Supervisory Scheduling of Freight Rail Systems

Abstract

This chapter develops a genetic fuzzy modeling approach for train scheduling of freight rail network systems. A genetic fuzzy algorithm is suggested as a means to solve train scheduling problems. The algorithm uses fitness estimation model based on participatory learning fuzzy clustering to improve its processing speed and to keep solution quality. The approach is particularly useful in scheduling problems involving dynamic environments because in these instances fitness evaluation usually is costly. In dynamic environments such as rail network systems, decision-making demands feasible train movement plans to control traffic and operate yards, stations and terminals. The genetic fuzzy algorithm is compared against exact optimal solutions given by classic optimization and genetic algorithms. To illustrate the usefulness of the approach, a real-world freight rail system problem is solved using the genetic fuzzy approach and the classic genetic algorithm. Results suggest that the genetic fuzzy approach constitutes a promising alternative to solve scheduling problems in general, but performs particularly well to produce supervisory train schedules.