As an emerging mobility service with a wide range of benefits, flexible buses balance monetary costs, travel time, and comfort, and have evolved into a powerful supplement to traditional travel modes. This study investigates the dynamic operations of an integrated mobility service system of flexible electric buses and fixed-route transits, such as traditional buses and metros. The proposed integrated framework owns the potential of making the maximum utilization of the flexibility of mobility-on-demand transports and the capacity of fixed-route transits, but it is very challenging to efficiently operate. We establish mathematical models to dynamically optimize the detailed flexible bus routing and timetabling plans, and assign travel routes for passengers while considering the transfer behavior from/to the conventional fixed-route transits in a network. To resolve the curse of dimensionality and facilitate the solution procedure, we embed a learning procedure into an optimization model and further decompose the subproblem in each period into two layers. Numerical studies verify the superior performance of the proposed learning-and-optimization framework in both stochastic dynamic and deterministic settings.

作为一种具有广泛优势的新兴出行服务，灵活公交平衡了货币成本、出行时间和舒适度，并已发展成为传统出行方式的有力补充。本研究调查了灵活电动公交车和固定路线公交（如传统公交车和地铁）的综合移动服务系统的动态运行。拟议的综合框架具有最大限度地利用按需移动交通的灵活性和固定路线过境能力的潜力，但有效运作非常具有挑战性。我们建立数学模型来动态优化详细的灵活公交路线和时间表计划，并为乘客分配旅行路线，同时考虑网络中传统固定路线公交的换乘行为。为了解决维数灾难并促进求解过程，我们将学习过程嵌入到优化模型中，并将每个时期的子问题进一步分解为两层。数值研究验证了所提出的学习和优化框架在随机动态和确定性设置中的卓越性能。