Optimal design of community shuttles with an adaptive-operator-selection-based genetic algorithm

• Published in: Transportation research. Part C, Emerging technologies Vol. 126; p. 103109
• Main Authors: Xiong, Jie, Chen, Biao, He, Zhengbing, Guan, Wei, Chen, Yanyan
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.05.2021
• Subjects: Community shuttle system; Frequency setting; Genetic algorithm; Network design; Network design; Frequency setting; Genetic algorithm; Community shuttle system
• ISSN: 0968-090X; 1879-2359
• DOI: 10.1016/j.trc.2021.103109

•The optimal design problem of community shuttles provides a more generalized network layout pattern.•The trip demands are set to originate from each small zone in a large community area as a grid.•A service frequency setting method including two stages is proposed.•Customized multiple crossover operators and mutation operators are developed in the GA.•An adaptive-operator-selection scheme is embedded in GA to spread out the evolution. This paper investigates the optimal design problem of a shuttle system provided by a large-scale community with the last-mile service feeding to metro stations. A mixed integer optimization problem is formulated to jointly optimize the route network and the service frequency for each shuttle. This problem aims to minimize the total transit system cost, including user and supplier costs, subject to the constraints on route length, coverage area, vehicle capacity and total fleet size. A solution approach that consists of the following three components is then proposed. The first component is a network analysis procedure that assigns the demand of each network zone to a set of paths and determines the service frequency of each route with a fleet size adjusting heuristic. The second component is an initial route network generation procedure, ensuring all the divided zones within the coverage of at least one shuttle route with appropriate length. The third component is a genetic algorithm procedure that contains multiple crossover and mutation operators to guide the evolving process of generating feasible solutions. Synthetic and real-world case studies are conducted to test the proposed model and the solution, and sensitivity analysis on key parameters and variables are also investigated.