Scheduling and Routing for Demand Responsive Transit Operations

Abstract

The multi-vehicle dial-a-ride problem has been proven to be intractable and NP-hard. Researchers have proposed numerous techniques for furnishing approximate solutions, but most ofthese aim at conquering a single objective, such as minimization of eet size. The chosen technique must handle multiple (different or competing) objectives that characterize the problem, so that it may lend itself as a tool in transit research. This thesis explores a new technique based on the cluster-first, route-second strategy, where the multiple competing objectives of the dial-a-ride problem are successfully addressed. The key highlights in this technique are the development of different routers for different stages of the optimization, and organism filters to guide the population of prospective solutions in the clustering phase of the optimization. Application and comparison to case study data indicate that the techniques developed outperform a commercial scheduling package in generating optimal schedules, and lend themselves to configuration and customization in order tosimulate different scenarios, which help in selecting demand-responsive transit policies.