大规模场景下网约车与城市交通拥堵交互影响仿真研究

Large-scale simulation for the interaction effect of ride-sourcing and urban congestion

网约车为居民提供了便捷的出行服务,同时也给城市交通性能带来影响,受影响的城市交通环境又将反作用于网约车运营.为了在大规模场景中探究这种交互影响机制,本文首先基于宏观基本图理论和元胞传输模型建立了同时考虑交通需求时变性,状态相关性和区域边界各项异性的交通动力学模型,将其集成到时间和事件混合驱动的模块化仿真组件中.其次,将考虑接驾半径约束的二部图匹配模型集成到出行模块中.最后,基于仿真结果,通过引入交互项的回归模型分析了出行延误率、网约车车队规模及其交互作用对网约车运营指标的影响机制与边际效应.结果表明,回归模型中各项指标的拟合优度均达到0.97以上.对于道路拥堵,网约车每增加1000辆,道路出行延误率全天平均增加1.59%,高峰期平均增加2.04%.对于网约车,不同的拥堵条件下都存在使网约车达到利润峰值的最佳车队规模,多项指标表明大型车队不利于网约车的运营.此外,在精度方面,仿真结果与真实数据差距较小.在效率方面,与微观仿真相比,仿真计算时间至少节约83.3%.

The emergence of online ride-sourcing service dramatically facilitates the travel of residents.At the same time,it is bound to impact urban traffic,which will also affect the ride-sourcing operation.This paper explored the interaction effect mechanism in large-scale scenarios.Firstly,it establishes a traffic dynamics model based on macroscopic fundamental diagram theory and cell transmission model,which considers the time variability of traffic demand,state correlation,and regional boundary anisotropy.The Dynamics model is integrated into a simulation module driven by a hybrid drive mechanism.Secondly,the bipartite graph matching model considering the constraints of pickup radius is integrated into the travel module.Finally,based on the simulation results,a multiple nonlinear regression model analyzes the influence mechanism and marginal effect of road congestion rate,fleet size,and their interaction effect on ride-sourcing service operation indicators.The results show that the goodness of fit of each operation index is above 0.97.As for road congestion,every 1,000 additional ride-sourcing taxis will increase the road travel delay rate by 1.59%on average throughout the day and 2.04%during peak hours.For ride-sourcing services,the optimal fleet size for the taxi to reach peak profits exists under different congestion conditions.Several indicators show that large fleets are not conducive to the operation of ride-sourcing services.Meanwhile,the accuracy of the simulation is proved by comparing it with actual data.The simulation calculation time can save at least 83.3%compared with micro-simulation.