摘要
城市路网交通流系统具有很强的随机性和时变性,单一固定的交通流模型难以准确地描述城市路网的实际运行情况,在考虑交通流稳态和动态特性的基础上,提出了一种含有未知时变多参数的非线性宏观交通流模型,并针对交通流固有的重复性特征,设计了一种时变多参数的自适应迭代学习辨识策略。在有限时间区间内,利用迭代学习辨识策略将参数辨识问题转化为最优跟踪控制问题,使交叉口各进口道的排队车辆数均趋于真实值,利用去伪算法的实时自适应能力调整迭代学习辨识策略的学习律增益,提高辨识策略的抗干扰能力。通过严格的数学理论推导证明了该算法的收敛性,最后采用基于模型的控制方法进行仿真实验,进一步验证了该方法的有效性。
The traffic flow system of urban road network has strong randomness and time-varying nature, and it is difficult for a single fixed traffic flow model to accurately describe the actual operation of urban road network. In order to describe the actual operation of traffic flow in urban road networks more accurately, a nonlinear macroscopic traffic flow model was proposed with unknown time-varying multi-parameter by taking into account the steady-state and dynamic characteristics of traffic flow, and a time-varying multi-parameter iterative learning identification strategy was designed by using the inherent repetitive characteristics of traffic flow. In the finite time interval, the iterative learning identification strategy is used to transform the parameter identification problem into an optimal tracking control problem, so that the number of queued vehicles at the entrance of each intersection converges on the true value, the real-time adaptive ability of unfalsified control algorithm is used to adjust the learning law gain of the iterative learning identification strategy, which improves the anti-interference ability of the identification strategy. The convergence of the algorithm is proved by a rigorous mathematical theoretical derivation, and finally the effectiveness of the method was further verified by simulation experiments using the model-based control method.
作者
仇江辰
闫飞
田建艳
QIU Jiangchen;YAN Fei;TIAN Jianyan(College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan 030024,China)
出处
《太原理工大学学报》
CAS
北大核心
2023年第1期211-224,共14页
Journal of Taiyuan University of Technology
基金
国家自然科学基金资助项目(61703300)
中国博士后科学基金资助项目(2019M651082)
山西省应用基础研究资助项目(201801D221191)。
关键词
城市路网
迭代学习
参数辨识
非线性模型
宏观交通流
urban road networks
iterative learning
parameter identification
nonlinear models
macroscopic traffic flow