基于RBF神经网络的高速公路车辆换道行为决策模型研究

A RBF Neural Network Based Intelligent Vehicle Lane-changing Decision Model on Highways

自由换道是一种在高速公路中常见的驾驶行为,将智能车辆的换道行为决策过程拟人化,不仅可以提高车辆的乘坐舒适度与安全性,同时也能提高交通流的通行效率,作者提出了一种基于RBF神经网络的驾驶员换道意图产生模型,模型的输入为本车与当前车道前车的距离及相对速度、前车类型、左右两侧相邻车道车流平均速度,并将该模型与基于最小安全距离的换道可行性判断模型结合,构建了一个换道行为决策模型。文中使用NGSIM(Next Generation Simulation)中的数据对RBF进行训练,并使用未参与训练的部分数据对换道行为决策模型进行数值仿真,结果证明文中提出的模型模拟准确率较高,可以较好地模拟人类驾驶员在高速公路上的换道决策行为。

prove not only a better level of driving comfort but also a higher level of transportation efficiency.In order to make lanechanging behavior of intelligent vehicles personification,a RBF neural network model was put forward»which has five inputs,including the distance and the relative speed between the subject vehicle,the preceding vehicle in the current lane.the category of the preceding vehicle*and the average speed of the traffic flows in the adjacent lanes.Then we built into a lanechanging decision-making process.The model was trained by NGSIM(Next Generation Simulation)data.To verify the proposed decision-making model,using real vehicle trajectory data in NGSIM,numerical simulation tests were conducted.The experimental results show that the proposed model could accurately resemble the drivers true lane-changing decision,and initiate free lane-changing behaviors similarly to human drivers in real world highway driving.