考虑异质性出行需求的主支线公交树网络优化设计

公交线网优化设计是指在一定的运行约束条件下，选择1组公交线路和相关频率以达到优化目标的设计过程，可以表示为一个优化问题。针对具有高异质性出行需求的主支线公交树网络，在考虑客流需求和运营约束的前提下，以用户和运营者的成本最小为目标，提出了1种多目标非线性混合整数优化模型。优化变量为候选线路服务频率。为求解这一模型，设计了1种基于改进的布谷鸟算法的高效元启发式方法。该方法包括初始候选路线集生成过程；基于MNL模型的公交分配过程；确定路线服务频率的改进布谷鸟算法过程。通过算例验证了该方法的有效性和适用性。数值分析结果表明，该算法通过对所有可能的候选路径的服务频率选择得到接近最优的公交线路网络。另一方面，通过保持高峰时的公交线路为有效备择线路，为具有异质性出行需求的网络的重新设计提供了更好的解决方案。此外，该系统在1次运行中产生了1组帕累托解，其允许公交线网设计师评估运营商成本和乘客成本并做出折中方案。通过比较3种算法的计算结果和CPU时间，证明了改进的布谷鸟算法的可靠性和有效性。另外还研究了最优公交网络设计与公交运行速度、总需求规模等关键设计输入参数之间的关系，分析结果表明，关键设计输入参数与最优公交网络具有一定的协同效应。模型与算法为实际的大规模主支线公交树网络的优化设计提供了1种有效的工具。

Optimal Design for Trunk-and-feeder Bus Transit Tree Network with Heterogeneous Demand

Optimal bus transit route network design is the process of selecting a group of bus routes and associated frequencies to achieve the optimal goal,subject to the operational constraints. This can be formulated as an optimization problem. In this paper,a multi-objective nonlinear integer optimization model for a trunk-and-feeder bus transit tree network design problem was proposed,aimed at minimizing the costs for users and suppliers by considering the high heterogeneous travel demands of passengers and operating constraints. The decision variables were the candidate route service frequencies. To solve this model,an efficient meta-heuristic method based on an improved cuckoo search( CS) algorithm was proposed. The solution method included the initial candidate route set generation process,the bus travel demand assignment process based on the MNL model,and the improved CS algorithm for determining route service frequency. The validity and applicability of this improved method was demonstrated through specific examples. The numerical analysis results showed that using this improved algorithm to select the service frequency of all potential candidate routes will yield a near-optimal bus transit route network. On the other hand,maintaining the existing efficient bus routes at peak hours yields a better solution for the redesign of the network with high heterogeneous travel demands. The system produces a set of Pareto-like solutions in one run that allow a bus transit network designer to assess operator and passenger costs and make tradeoffs. By comparing the results and the CPU times of the three algorithms, this study demonstrated that the improved CS algorithm is reliable and more efficient than GA and CS. This study also showed the relationship between optimal transit network design and critical design input parameters,such as transit operating speed,OD demand size,and value of time. As a result of the analysis,the synergistic effect of critical design input parameters and the optimal transit network were discussed. The proposed general analytic framework provides an effective tool for the optimal design of a large-scale trunk-and-feeder bus transit tree network in practice.