网联环境下混合交通流偶发拥堵演化机理研究

针对混合交通流中智能网联车辆(Connected and Autonomous Vehicles, CAVs)和人工驾驶车辆的交织干涉问题，本文在传统交通流统计理论模型和一阶连续介质模型的基础上，通过引入智能驾驶员跟驰模型(Intelligent driver model, IDM)和协同自适应巡航控制模型(Cooperative Adaptive Cruise Control, CACC)，构建人工驾驶车辆和CAVs的混合交通流偶发拥堵演化模型，探索CAVs混入和诱导干涉措施对混合交通流偶发性拥堵传播规律的影响。实验选取重庆市华陶立交至巴南立交路段为路网原型，对CAVs不同渗透率( Pc )下的路段拥堵演化情况进行仿真。实验结果表明：CAVs渗透率越高，混合流流量、占有率和速度的改善情况越显著，但只有当 Pc ≥ 0.2 时，网联车辆对拥堵消散的改善效果才较为明显；Pc ≤ 0.8 时，干涉措施下，拥堵消散状态的持续时间约为不采用干涉措施的 50%；当 Pc = 1.0 时，网联车辆的通行能力是纯人工驾驶交通流的2.34倍；分别在非干涉措施和干涉措施下计算拥堵评价指标，与仿真结果进行对比，最大相对误差在5.38%之内，验证了模型的准确性。研究成果对疏散交通拥堵具有重要意义。

Mechanism of Non-recurring Congestion Evolution Under Mixed Traffic Flow with Connected and Autonomous Vehicles

This paper investigates the interference between connected and autonomous vehicles (CAVs) and the traditional vehicles in mixed traffic flow, and proposed a model to describe the evolution of the non-recurring congestions in the mixed traffic flow. The model is developed based on the traditional traffic flow statistical theory model and the first-order continuous medium model, and introduced the intelligent driver model (IDM) and cooperative adaptive cruise control (CACC). The study uses the roadway section from Huatao Interchange to Banan Interchange in Chongqing City as a case study to examine the congestion evolution under different permeability ( Pc ) of CAVs in the traffic flow. The results show that higher penetration rate of CAVs relates to more significant improvement of the flow, occupancy, and speed of the mixed traffic flow. However, the improvement of congestion dissipation by CAVs is more obvious when Pc ≥ 0.2 . When Pc ≤ 0.8 , the duration of the congestion dissipation state with interference measure is approximately 50% of that without interference measure. When Pc = 1.0 , the traffic capacity of CAVs is 2.34 times of that in traffic flow with only traditional vehicles. The traffic congestion evaluation indexes are calculated under noninterference and interference measures and compared with the simulation results. The maximum relative error is within 5.38 %, which verified the model's accuracy. The research results provide important references for traffic congestion evaluations.