基于变结构离散动态BN的最优交通路径规划

为了确保城市路网交通流平稳运行和各路段交通流量合理分配，提出了一种基于变结构动态BN的最优交通路径规划方法。该方法考虑驾驶员偏好，按时间序列建立适用于交通路径规划的变结构离散动态BN模型，采用最大似然估计算法和参数的自适应产生算法学习网络参数，用基于时间窗的动态BN近似推理算法中固定窗口宽度方法进行在线推理。结合实例对算法进行仿真，并与Dijkstra算法所得结果进行比较。实验结果表明变结构离散动态BN能利用实时采集到的信息对最优路径进行实时更新，在线决策。

Optimal Traffic Path Planning Based on Structure-variable Discrete Dynamic BN

In order to ensure the smooth operation of urban road network traffic flow and the reasonable distribution of traffic flow in each road section, an optimal traffic route planning method based on variable-structure dynamic BN was proposed. This method considers driver preferences,establishes a variable-structure discrete dynamic BN model for traffic path planning according to time series.The maximum likelihood estimation algorithm and the adaptive parameter generation algorithm are used to learn the network parameters,and the fixed window width method in the dynamic BN approximate reasoning algorithm based on time window is used for on-line reasoning.Simulation with an example and the proposed method is compared with Dijkstra algorithm. The experimental results show that Variable- Structure Discrete Dynamic BN can make use of real-time collected information to update the optimal path in real time, online decision-making.