基于ANFIS的高速公路车辆跟驰模型与仿真

Freeway Car-following Model and Simulation Based on Adaptive Neuro-fuzzy Inference System

摘要:

为了更好地描述高速公路上驾驶员在车辆跟驰过程中表现出来的模糊、不确定性的行为特征，采用自适应模糊神经网络ANFIS来建立车辆跟驰模型.首先，通过小波分析方法，对采集到的跟车数据进行降噪，消除外界因素的干扰，从而恢复数据的原始信息；根据信号处理方法，利用相关函数计算出驾驶员在跟驰过程中的反应时间.然后，建立以两车速度差、车头间距和后车速度作为输入，以及后车加速度作为单输出的自适应模糊神经网络跟车模型.最终，对该模型仿真训练，自适应生成驾驶员跟驰行为规则，并与传统的GM跟车模型对比分析.结果表明，该网络模型能较客观地反映高速公路上的驾驶员跟驰行为.

Abstract:

In order to better describe the fuzzy and uncertain characteristics of drivers when following a leading car on the freeway,the adaptive neuro-fuzzy inference system(ANFIS) was applied to the car-following model.Firstly,the real-time car-following data collected by the five-wheel system were denoised by the wavelet tool in order to eliminate the disturbance from the surrounding and recover the initial information of data.Meanwhile,the driver’s reaction time was caculated by the correlation function in term of signal processing method.Then,the car-following model based on ANFIS was developed with the relative speed,distance headway between leading vehicle and following vehicle and speed of following vehicle as inputs and car-following acceleration as the output.Finally,the model that generated the adaptive rule of drivers’ car-following behavior was simulated and trained to compare with General Motors based car-following models.The car-following model based on ANFIS proves to be able to reflect the driving behavior of real freeway situaton.