基于自适应神经模糊推理系统的跟驰模型研究

Study on Vehicle Following Model Based on Adaptive Neural-fuzzy Inference System

由于驾驶行为的不确定性,难以建立精确的车辆跟驰模型。针对这一问题,应用自适应模糊神经推理系统(ANFIS)建立跟驰模型,以跟随车与前车速度差及行车间距为输入量、跟随车的加速度为输出量,建立25条模糊推理规则,将模糊推理规则产生的数据作为车辆跟驰ANFIS模型的训练数据,并利用MATLAB编程对其进行训练。最后,设计了基于车载高精度GPS的跟驰试验,并结合试验数据分别对自适应模糊神经推理系统跟驰模型和传统跟驰模型进行仿真。结果表明,前者输出的跟驰车辆加速度值更接近于真实值。

Due to uncertainty of driving behavior, it is hard to set up precise vehicle following model. In order to solve this problem, adaptive neural-fuzzy inference system （ANFIS） is used to set up vehicle following model. 25 fuzzy inference rules are established by using＇speed difference and distance between following vehicle and front vehicle as input and acceleration of following vehicle an output. Data generated by fuzzy inference rules are used for training in vehicle-following model （ANFIS）, and MATLAB programming is also used for training. In the end, vehicle following test based on high precision vehicle GPS is designed. Stimulations with adaptive neural-fuzzy inference system vehicle following models and traditional vehicle following models are respectively made with the test data. The result shows that acceleration of following vehicle output from the former is much close to the actual value.