【摘要】 针对间断流建模影响因素多、仿真随机性大的特点，提出了一种考虑横向阻断的障碍泛化形式交通流模型。通过将经典智能驾驶模型（IDM）拓展为静态避障形式（IDM+）并引入车辆预决策控制流程（Pre-IDM+），搭建交通完全随机加载的间断流仿真平台，以单车道行人穿越场景为例对运行轨迹、交通流特征、车辆控制状态3类交通分析指标的随机收敛性和仿真步长敏感度进行定量讨论。结果表明：200仿真样次能较好兼顾模型精度和分析效率（平均误差<2%），仿真步长改变主要对车辆停车及避障控制相关指标产生明显影响（最高误差43%）。在指标可靠度方面，平均速度、减速度均值、控制状态切换频次以及自由/跟驰状态时间占比具有较好的评估一致性，随机误差和仿真步长的影响均<1%。更多还原

【Abstract】 The intermitted traffic flow modeling needs to take many impact factors into consideration and usually suffer large randomness in traffic simulation analysis. In response, this paper presents a traffic flow model with generalized obstacle form which includes transverse traffic blocking and its interference into modeling. By extending the IDM car-following model to obstacle avoidance cases(IDM+) and simultaneously introducing a vehicle pre-decision control workflow(Pre-IDM+), an intermitted traffic flow simulation platform under totally random traffic loading is built. Based on the platform, the random convergence and simulation updated time step sensitivity of traffic indexes on driving trajectory, traffic flow characteristics and vehicle control state are discussed quantitively in a single-lane pedestrian crossing scenario. The results show that the simulation accuracy and analysis efficiency can be balanced(average error<2%) by setting 200 of simulation sample size, and the change of simulation updated time step has significant impact on parking times and obstacle avoidance(maximum error up to 43%). From the perspective of index reliability, the average speed, the average deceleration, the frequency of driving state switching, and the time proportion of free/following driving state perform a good evaluation consistency(random error and simulation updated time step impact both<1%).更多还原