非常规交叉口设计是缓解城市道路交通拥堵的新对策，但由于其通行规则的复杂性和通行者的不适应，导致通行能力存在波动性，从而影响非常规交叉口交通功能的发挥。本项目针对确定性优化方法在处理交通运行波动性方面的缺陷，进行随机优化研究，以提高非常规交叉口运行的稳定性。首先，基于实测数据，结合非常规交叉口的运行特征，对潜在不确定因素进行统计描述和方差分析，从而揭示通行能力波动性的关键影响因素及其作用机理。在此基础上，建立随机优化模型，将车辆在冲突区内滞留、违章行驶、车道利用不均等间接不确定因素，与饱和流率、损失时间等直接影响因素相关联，进而对交叉口平面空间布置、预信号停车线位置和信号控制进行同步优化，实现通行效率的稳定。最后，分别采用蒙特卡洛模拟和灵敏度分析，对模型优化效益和非常规交叉口适用性进行评价。本项目从随机优化的角度提供了非常规交叉口优化的新思路，为制定其高效稳定的优化方案提供了必要的理论基础。

The unconventional intersection design is a new method to mitigate the traffic congestion of urban streets. However, the fluctuation of the capacity, due to the complexity of the moving regulations and the confusion of the drivers, causes a significant negative impact on the unconventional intersection operation. Aiming the deficiency of the deterministic optimization method in handling the traffic fluctuation problem, this project will develop a stochastic optimization method to improve the operation stability of unconventional intersections. Firstly, the statistic description and analysis of variance for potential uncertainty factors will be conducted based on the analysis of operation characteristic using field data. Then, the key factors and their effect mechanism on capacity fluctuation will be revealed. Secondly, the stochastic optimization model combining the design of the space layout, location of the pre-signal stop line, and signal timings will be developed to stabilize the operational efficiency. The impact of indirect uncertainty factors, such as the possibility of vehicles trapped in the conflict area, the violation of traffic regulations, and the inequality of lane utilization, are embodied in the model by formulating their correlation with the direct uncertainty factors, such as saturation flow rate and lost time. Finally, the effectiveness of the stochastic model and the applicability of unconventional intersections will be evaluated by Monte Carlo Simulation and sensitivity analysis, respectively. This project proposes a novel idea in unconventional intersection optimization, and provide a requisite theoretical basis to produce efficient and stable schemes.