An application of genetic and tabu searches to the freight railroad operating plan problem

This paper addresses the joint train-scheduling and demand-flow problem for a major US freight railroad. No efficient optimization techniques are known to solve the NP-hard combinatorial optimization problem. Genetic search is used to find acceptable solutions; however, its performance is found to deteriorate as the problem size grows. A "tabu-enhanced" genetic search algorithm is proposed to improve the genetic search performance. The searches are applied to test problems with known optima to gauge them for solution speed and nearness to optimality. The tabu-enhanced genetic search is found to take on average only 6% of the iterations required by genetic search, consistently achieves better approximations to the optimum and maintains its performance as the problem size grows. The tabu-enhanced search is then applied to the full-scale operating plan problem. Model results reveal a potential for 4% cost savings over the current railroad operating plan coupled with a 6% reduction in late service.