Optimization of a semiflexible demand‐responsive feeder bus system using variable neighborhood search

Abstract A demand‐responsive feeder system (DRFS) presents an alternative to traditional feeder bus systems (TFS) in areas with low demand. This paper introduces an optimization problem to support the planning and operation of a DRFS. A variable neighborhood search (VNS) is employed to optimize the DRFS, focusing on minimizing passengers' travel times. The performance of the VNS is compared with that obtained by solving the mathematical model using a commercial solver (CPLEX) across two networks: a small network for validation and a larger suburban network that simulates a TFS. The results indicate that the VNS is a viable and efficient alternative for optimizing DRFS operations, providing flexibility in route and departure time adjustments, and achieving significant reductions in passenger travel times. The results also demonstrate that the DRFS outperforms the TFS in low‐demand areas. On average, the VNS improves upon CPLEX results, obtained within one hour of calculation time for the small network, by 2.3%. Using the same resources, the DRFS reduces the average passenger travel time by 9.6% compared to the TFS.