THE FEEDER BUS ROUTE DESIGN PROBLEM (TRANSIT VEHICLE ROUTING, BUS NETWORK DESIGN)

The U.S. transit industry is in a financial crisis, caused primarily by the rising operating cost and shrinking resources. Among the deficit-reducing strategies that have been proposed, a better integrated transit system has significant advantages in that it offers the possibility of simultaneously reducing cost and increasing revenue. This research develops an optimization-based methodology for the design of an integrated feeder bus/rail transit system. The proposed methodology includes two types of models, an analytic model and a network model.; The study defines the Feeder Bus Route Design Problem (FBRDP) as that of locating bus routes and bus stops as well as determining the service characteristics of a feeder bus system, designed to access an existing rail system. The proposed analytic model provides approximate values for the optimal route density, operating headway and bus stop spacing. Within the network approach, we consider demand with a single (many-to-one) and with multiple (many-to-many) destinations. We develop a conceptual representation of the many-to-one FBRDP as a spanning tree network, which enables us to formulate the problem as a nonlinear programming model with integer variables. The model is then generalized to the many-to-many FBRDP. We show that the proposed mathematical programming models can represent both the short-run and long-run problems.; The network optimization models for the FBRDP are routing-type models which cannot be solved exactly and need to be solved heuristically. We develop a two-phase heuristic which can be viewed as a generalization of the sequential-savings approach to vehicle routing in that (1) it determines not only the route structure but also the operating frequencies, and (2) it represents the more general FBRDP objective of finding the optimal balance between user and operator costs. An analysis is performed to validate the proposed models and their capabilities as strategic planning tools for the design of a feeder bus network. The proposed models are shown to provide consistent and reasonable responses to changes. Our analysis shows that changes which increase the relative weight of user cost often result in a network with more circuitous routes operated at higher frequency with shorter wait time at bus stops.