基于动态规划的城市交通流预测与分配

Urban Traffic Flow Prediction and Assignment with Dynamic Programming

城市交通系统是一个复杂的大系统,针对交通流的不确定性,本文基于动态规划思想提出了交通流预测与分配的方法.考虑路段容量对交通分配的影响,建立了路网流量预测和分配模型.为了保持并充分利用传统网络模型的性质和特征,引入惩罚函数,将容量约束条件转换到目标函数中,使模型符合传统均衡网络流结构.将凸规划法作为一个子过程植入惩罚函数,得到模型的求解算法.控制策略上采取预测控制、反馈校正和滚动优化的方式.最后,通过算例分析,进一步阐述模型和算法的应用,验证算法的有效性.为交通流预测和分配提供一定的参考.

In view of the uncertainty of the complex urban traffic system, this paper presents a traffic flow prediction and assignment method with dynamic programming. The model is developed by considering the capacity limitation in each section of the road network. In this model, the newly-added restraint condition is transferred by introducing punishment function, and the fine characters of traditional assignment model are maintained. Then, the algorithm is documented, in which Frank-Wolfe algorithm is combined with the punishment function. The control strategy consists of prediction control, feedback correction, and rolling optimization. Finally, a numerical example is used to illustrate the availability of the proposed method which provides reference for traffic flow prediction and assignment.