摘要
汽车驾驶行为是影响燃油消耗和安全驾驶的重要因素,驾驶行为识别是对汽车安全驾驶和节能进行优化的基础。该文针对汽车转向和换道行为,通过加装汽车转向盘转角传感器,结合车载总线通信技术获取汽车行驶状态信息,基于汽车转向运动学推导车辆行驶状态与汽车行驶轨迹之间的映射关系,进一步建立汽车行驶方向向量模型,提出以车身轴线转角和最大转向盘转角为特征量的支持向量机线性分类器,并运用Lagrange数乘法和二次规划算法求解该最优分类问题。通过设计实车实验验证了该方法的有效性。实验结果表明:该方法识别汽车的转向与换道驾驶行为的准确度达98%以上。该技术可用于汽车行驶安全预警与控制系统,提升行驶安全。
Driving behavior plays an important role in fuel consumption and safe driving. Thus, driving behavior recognition can im- prove driving safety and optimize energy use. This study presents a steering and lane-changing behavior recognition system based on the vehicle status obtained from a steering wheel angle sensor. A support vector machine linear classifier is then used to analyze the vehi- cle body transfer angle and maximum steering angle given by a moving direction vector model. Lagrange number multiplication and quadratic programming are used in an optimal classifier for recogniz ing steering and lane-changing behavior. Real vehicle tests show that this methodology has 98% accuracy for steering and lane changing behavior recognition. This system can be integrated into a warning and control system to improve driving safety.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2015年第10期1093-1097,共5页
Journal of Tsinghua University(Science and Technology)
基金
国家"八六三"高技术项目(2012AA111901)
国家留学基金项目(201406215015)
清华大学汽车安全与节能国家重点实验室开放基金项目(KF14142)
关键词
驾驶行为
行驶方向向量
支持向量机
最优分类
driving behavior
moving direction vector
support vector machine
optimal classification
