Document Type

Journal Article

Publication Date


Subject Area

ridership - demand, mode - bus, mode - paratransit


Waiting time, Vans, Stochastic processes, Simulation, Random processes, PUD, Private carriers, Pickup and delivery service, Paratransit services, Minibuses, Miniature buses, Flexible transit system, Dial a ride, Demand responsive transportation, Demand, Cost effectiveness, Computer simulation


Conceptual design and preliminary feasibility simulation results are presented for a flexible transit system for travel from any point to any point based on real-time personalized travel desires, which is now possible because of advances in communications and computing technologies. Although it is demand-responsive, the concept is significantly different from older demand-responsive transit systems, which were often failures. The proposed system requires high coverage, referring to the availability of a large number of transit vehicles (often minibuses or vans), which could also operate in conjunction with private transit and paratransit systems. The design strictly eliminates more than one transfer for any passenger. The system could potentially provide a transit alternative that is much more competitive with personal automobile travel than are conventional transit systems because of significantly lower waiting times. The passenger demand for such a system is uncertain, but preliminary simulations show that under a variety of acceptable demand levels, the system can operate with high cost-effectiveness. The focus is on describing the details of the concept and providing arguments in favor of the system based on simulations. The system essentially attempts to solve a stochastic real-time passenger pickup-and-delivery problem with a large number of vehicles. A strict optimization formulation and solution for such a problem are computationally prohibitive in real time. The design proposed is effectively geared toward a decomposed solution using detailed rules that achieve vehicle selection and route planning. If real-time update of probabilities is included, this scheme could be considered as a form of quasi-optimal stochastic control.