基于神经网络的交通平衡求解算法

研究目标在于采用神经网络方法替代传统的交通分配方法。回顾了交通网络平衡条件下实现交通网络配流的解析方法，然后提出了采用人工神经网络模拟处于交通平衡状态的运输网络，进而替代传统的配流方法的思想。定义了交通流的平衡函数和交通流平衡态的概念，证明了用神经网络来识别交通流的平衡状态的可行性。提出采用反向传播神经网络来模拟交通平衡状态的原则和方法，并建立了训练神经网络的判据，提出了判断神经网络模型效果的评价方法。最后，通过实例证明了良好训练的神经网络可以替代传统的交通分配，而且，容易把交通管理策略考虑到分配的过程中．并且能够政善配流速度。

Solution of Traffic Equilibrium Status with Neural Network

It is aimed to construct neural network model to assign flow links under user equilibrium of transportation network. Previous analytical method of traffic assignment under traffic equilibrium status are reviewed, and a method using neural network to simulate traffic equilibrium status is put forward. So the new methods can substitute for traditional method, and quickly finish traffic assignment. It is defined the concept of traffic equilibriums function and traffic equilibriums status, and strict mathematical proof indicates that we can recognize the equilibrium status of traffic flows by neutral network. The principle and method on how to construct, train and evaluate a neural network are also presented. At last, with an example the author explains how to use the algorithm, and to evaluate accuracy and CPU time consuming of the algorithm .