学术文献库

Capturing latent demand has a pivotal role in designing public transit services: omitting these riders can lead to poor quality of service and/or additional costs. This paper explores this topic in the design of OnDemand Multimodal Transit Systems with Rider Adoptions (ODMTS-DA). Prior work proposed a bilevel optimization model between the transit agency and riders with choice of adoption, and an exact algorithm to solve the resulting ODMTS-DA design problem. However, due to the complexity and combinatorial nature of the ODMTS-DA, the exact algorithm exhibits difficulties on large-scale instances. This paper aims at addressing this challenge in order to find high-quality ODMTS-DA designs in reasonable time. It proposes five heuristic algorithms whose designs are driven by fundamental properties of optimal solutions. The performance of the heuristic algorithms are demonstrated on two test cases leveraging real data: a medium size case study for the Ann Arbor and Ypsilanti region in the state of Michigan and a large-scale case study conducted in the Atlanta metropolitan region in the state of Georgia. To evaluate the results, besides directly comparing computational times and optimality gaps with the exact algorithm, this paper introduces two additional metrics that leverage the characteristics of optimal solutions with respect to customer adoption. Computational results demonstrate that the heuristic algorithms find optimal solutions for medium-size problem in short running times, and discover high-quality solutions to the large-case study that improve upon the best solution found by the exact algorithm in considerably less time. The ODMTS designs obtained by these algorithms provide substantial benefits in terms of convenience, operating cost, and carbon emissions.