基于支持向量机的车辆换道决策模型

Decision Model for Vehicle Lane Changing Based on Support Vector Machine

针对车辆换道行为受交通环境影响较大而难以识别和预测的问题,提出了一种基于支持向量机的学习模型用以仿真驾驶员在高速路上关于车辆换道的行为决策.通过分析车辆在换道阶段的特征与规律,选择适合的物理量作为模型的输入参数.以NGSIM数据库为基础用适当的方法进行样本提取,并对样本数据进行差分滤波、卡尔曼滤波、归一化预处理.在构建SVM模型过程中,运用不同的算法搜索最优参数.为了验证模型的泛用性,使用不同的数据样本对模型进行训练和测试,最终取得到了较好的预测结果与拟合度.

The vehicle lane changing behavior is difficult to identify and predict due to the influence of traffic environment then,a learning model based on support vector machine was proposed to simulate the behavior decision of the driver on a freeway.By analyzing the characteristics and laws of the vehicle in the phase of lane changing,the appropriate physical quantity was chosen as input parameter of model.Based on the NGSIM database,aproper method was used to extract the samples,and the sample data was processed by differential filtering,Kalman filtering and data normalization.In the process of building SVM model,different algorithms were used to search for the optimal parameters.In order to verify the generalization of model,different data samples were used to train and test the model.Finally,the prediction results and degree of fitting were obtained.