为克服基于定数冲击波理论的车辆排队长度分析模型中线性冲击波假设对排队长度的求解局限性，提出匹配车辆实时到达规律的非定数排队长度分析模型。首先，基于交通流冲击波理论，将排队车辆队尾的精确定位抽象为反映车流实际行驶特征的非线性累积排队冲击波与放行冲击波的时空位置关系分析模型。然后，选取独立周期作为计算单元，根据检测数据设计了冲击波时空折线线形回溯反推算法，将剩余车辆数作为过渡变量，建立了实测数据与计算结果之间的连续校核机制。最后，以实际路段为基础验证方法的有效性。结果显示，根据车辆实测数据，非定数冲击波模型所得的冲击波线形符合车辆到达规律，排队长度绝对误差平均值为3.8 m，比两组布置位置不同的检测器下的定数冲击波模型计算结果分别少5.3 m和4.8 m。当在VISSIM仿真软件中改变路段长度并加载两种不同交通量组合时，非定数冲击波模型计算的排队长度与实测结果的绝对误差平均值分别仅为2.71 m和7.63 m。交通量组1绝对误差大于5 m、交通量组2绝对误差大于10 m的计算结果在总数中的占比分别不超过7.7%和17.8%。在交通量组1相同到达规律下，与基于定数冲击波理论所得排队长度平均值相比，非定数冲击波模型将队尾定位的精确度平均提升约3个车长。研究结果表明，所建模型能结合车辆实际到达规律复现车辆动态排队过程，为交通信控方案评估及车辆通行效能优化提供更精确的参数。

In order to overcome the limitation of the assumption of linear shock wave in solving queue length in the vehicle queue length analysis model based on constant shock wave theory, a non-constant queue length analysis model matching the real-time arrival law of vehicles was proposed. Firstly, based on the shock wave theory of traffic flow, the accurate positioning of the tail of queuing vehicles was abstracted as a spatio-temporal position analysis model concerning nonlinear arrival shock wave and release shock wave, which reflects the actual driving characteristics of traffic flow. Then, an independent period was selected as the calculation unit, and a backtracking algorithm of the spatio-temporal line shape of shock wave was designed based on detection data. A continuous checking mechanism between measured data and calculated results was established with the number of remaining vehicles used as an intermediate variable. Finally, the validity of the method was verified based on actual road sections. The results showed that according to the measured data of vehicles, the shock wave shape calculated by the non-constant shock wave model were consistent with the arrival law of vehicles. The average absolute error of the queue length obtained by the proposed model was 3.8 m, which was 5.3 m and 4.8 m less than the results of the constant shock wave model with two different detector positions of the road sections, respectively. As the length of the road section changed and two sets of different traffic volume combination were loaded in the VISSIM simulation software, the average absolute error between the queue length calculated by the non-constant shock wave model and the measured results was only 2.71 m and 7.63 m, respectively. The proportion of calculation results for traffic volume combination Ⅰ with an absolute error greater than 5 m and traffic volume combination Ⅱ with an absolute error greater than 10 m in the total number did not exceed 7.7% and 17.8%, respectively. Under the same arrival law in traffic volume combination Ⅰ, compared with the average queue length obtained based on constant shock wave theory, the model based on the non-constant shock wave theory improved the accuracy of the queue tail positioning by an average length of 3 vehicles. The research results indicate that the proposed model can reproduce the dynamic queuing process of vehicles based on the actual arrival laws of vehicles, which provides more refined parameters for the evaluation of traffic control schemes and the optimization of traffic efficiency.