Intelligent agent optimization of urban bus transit system design

The transit route network design (TRND) problem seeks a set of bus routes and schedules that is optimal in the sense that it maximizes the utility of an urban bus system for passengers while minimizing operator cost. Because of the computational intractability of the problem, finding an optimal solution for most systems is not possible. Instead, a wide variety of heuristic and meta-heuristic approaches have been applied to the problem to attempt to find near-optimal solutions. This paper presents an optimization system that synthesizes aspects of previous approaches into a scalable, flexible, intelligent agent architecture. This architecture has successfully been applied to other transportation and logistics problems in both research studies and commercial applications. This study shows that this intelligent agent system outperforms previous solutions for both a benchmark Swiss bus network system and the very large bus system in Delhi, India. Moreover, the system produces in a single run a set of Pareto equivalent solutions that allow a transit operator to evaluate the trade-offs between operator costs and passenger costs.