一种基于驾驶员学习过程的在线估计算法(英文)

An on-line estimation algorithm of driver's learning process

为模拟驾驶员在跟踪期望路径过程中的驾驶决策过程,论文基于目标位置的思想对驾驶员的学习过程进行了建模。在建模过程中,利用三次B样条曲线灵活性的特点生成车辆对于目标位置的换道路径,同时,借助模糊神经系统模拟驾驶员跟踪目标位置过程的决策机制。并且提出了一种新的学习算法,即通过将模糊论域划分为若干局部线性网络以此拟合非线性函数。利用该算法能模仿驾驶员驾驶过程中的自学习能力。最后,对车辆躲避障碍物进行了仿真,结果表明该模型能较好的模拟驾驶员在换道过程中的驾驶行为及其自学习过程。

To simulate human decision-making in steering to follow a planned path, we modeled the driver＇s learning process by using an idea of target position. The proposed model dealt with the driver＇ s decision-making mechanisms in tracking a target position by means of a Neuro-Fuzzy system. Here, a cubic B-spline curve was used to generate more flexible paths of lane-change to a target position. Furthermore, a learning algorithm was proposed to emulate a driver＇s self-learning skill from his driving experience by splitting the fuzzy area into a number of linear systems to approximate the nonlinear function locally and online. Results from the simu-lation of the obstacle avoidance showed the capability of the proposed model to simulate a driver＇s steering behavior with emphasis on learned skill.