基于隐马尔可夫模型的驾驶行为预测方法研究  

Research on Driving Behavior Prediction Method Based on HMM

作　　者：肖献强[1] 任春燕[1] 王其东[1] 

机构地区：[1]合肥工业大学,合肥230009

出　　处：《中国机械工程》2013年第21期2972-2976,共5页China Mechanical Engineering

基　　金：中央高校基本科研业务费专项资金资助项目(2011HGBZ1322);国家自然科学基金资助项目(51175135)

摘　　要：根据驾驶行为受驾驶意图驱使,在时空上应先产生驾驶意图后有驾驶动作执行,再有具体的驾驶行为的时空顺序,利用意图到行为实现过程中的时间差,研究建立了基于多驾驶操作动作特征观测信息和隐马尔可夫模型的驾驶行为预测方法,实现了跑偏驾驶、一般转向和紧急转向3种驾驶行为的预测。从机动车安全预警的角度进行分析,预测出驾驶员的驾驶行为等同于获得驾驶员对车辆未来运行状态的需求,有助于及时纠正或干预驾驶员正在或即将实施的危险行为。According to time and space relationship of driving behavior and driving intention, driv- ing intention of driver drived driving behavior and manoeuvre,The driver first produced driving inten- tion, and then driving manoeuvre. On the basis of the time difference between driving intention and driving manoeuvre, this paper researched driving behavior prediciton method based on multi--charac- teristics of driving manoeuvre and HMM. The results show that the method has successfully forecas- ted three type driving behaviors of lane departure, steering and urgent steering. In the vehicle safety warning term, if was understood the demand of the driver to vehicle in the future can be calculated the driving behavior of driver. The method presented herein will help to correct and intervene the hazard- ous driving behavior of driver for the vehicle safety warning.

关 键 词：驾驶行为预测 操作观测信息 隐马尔可夫模型 多指标分析法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]