Optimal demand-responsive connector design: Comparing fully-flexible routing and semi-flexible routing strategies

Demand-responsive connector (DRC) services are increasingly recognized for their convenience, comfort, and efficiency, offering seamless integrations between travelers' origins/destinations and major transportation hubs such as rail stations. Past analytical models for DRC optimization often failed to distinguish between two commonly used DRC operating strategies: (i) the "fully-flexible routing" strategy, where a vehicle serves only the requests received before its dispatch through an optimal tour, and (ii) the "semi-flexible routing" strategy, where a vehicle follows a predefined path through a swath to serve requests received en route. Additionally, these models often adopted oversimplified approaches for estimating local tour lengths and capturing the stochastic nature of demand. This paper distinctly identifies and analyzes the two DRC operating strategies, developing analytical models for each that accurately incorporate the second-order effects of stochastic demand and utilize refined local tour length formulas. Numerical experiments demonstrate that our models reduce cost estimation errors to within 2% for fully-flexible routing and to 0.25% for semi-flexible routing, a significant improvement over the previous errors of 8–12% and 6.3%, respectively. These enhanced models allow for more precise determination of critical demand densities for selecting between the two DRC strategies and the fixed-route feeder service. Our extensive numerical analysis offers many insights, particularly highlighting the transition from fully-flexible to semi-flexible routing as demand and region size increase, before ultimately shifting to fixed-route service. Additionally, zoning is identified as pivotal in DRC service design, with fully-flexible routing favoring square-shaped zones and semi-flexible routing preferring elongated rectangular zones.