Mobility Allowance Shuttle Transit with Dynamic Checkpoint: A Clustering and Routing Strategy with Pricing Model

Abstract

This study explores a dynamic checkpoint strategy for an on-demand flexible transit service called Mobility Allowance Shuttle Transit with Dynamic Checkpoint (MAST-DC) with a static pricing model. The original Mobility Allowance Shuttle Transit (MAST) system operates shuttles allowing deviations from their fixed route to serve more passengers within the service area at their desired locations. The system is highly suitable for operating in a low-demand area or during the nighttime. However, we started to consider whether we could operate the system to serve higher demands and make it more profitable. Our research stems from solving this question as the following. First, as did the original MAST system, binding time constraints are entangled with fixed checkpoints. To serve more passengers given a limited amount of time for deviations, we introduced a concept of a dynamic checkpoint. A dynamic checkpoint is not a preassigned pickup and drop-off location, but the location defined by the system in each vehicle running period to gather some of the passengers and reduce deviations. Creating dynamic checkpoints may cause disutility for passengers due to extra walking distance. Therefore, we propose a sequential iterative two-phase heuristic model that first clusters to minimize passengers’ walking distance and routes second to minimize vehicle distance traveled between checkpoints. The clustering is performed using the memetic differential evolution (MDE)-based algorithm, and the routing solves a traveling salesman problem. An analytical model is developed to provide upper-bound estimates for the number of dynamic checkpoints and the passengers’ average walking distance. Second, we propose a MAST-DC operational framework with a static pricing model by combining the sequential iterative two-phase heuristic model with a discrete choice model. The system offers different fares for the two service options: fixed checkpoints and dynamic checkpoints. The fare for using dynamic checkpoints is supposedly higher than the fixed checkpoints to reflect the convenience of those who use dynamic in terms of walking distance. Here, given different at-tributes of the two service options, including fare, in-vehicle travel time, and walking distance, we observe passengers’ preferences in choosing which checkpoint to use, including rejecting all the options and leaving the system. The simulation experiments demonstrated that the proposed heuristic model utilizes approximately 90 percent of the scheduled running time regardless of the scheduled time window. The heuristic model recommends operational parameters of the MAST-DC system, including segment lengths, the maximum vehicle running times, and demand rates corresponding to the passenger walking threshold of 0.2 miles. Through the pricing model, the MAST-DC system shows potential profitability, especially when the fares of the dynamic checkpoint option are $2.5 and $3. As we collect more data by running the pricing model with different passenger configurations, the model will be able to provide the fare for the dynamic checkpoint users adaptive to the changing operational environments.