共驾型智能车模式选择方法及自主干预机制研究

Considering the complexity of the traffic environment, fully autonomous driving may still be decades away, the driving modes of manual driving, warning assistance driving, and autonomous driving would be co-existence in a long time. During this period, one principal research question is how to choose the suitable driving mode of an intelligent vehicle in the complex traffic environment. What’s more, the evaluation of driving safety is also another indispensable step. Therefore, feature selection, human factor analysis and artificial intelligence methods were used, the on-road and driving simulator experiments were conducted in this study to deal with the following research topics: (1) the definition and classification of driving mode for intelligent vehicle in different levels of driving risk status; (2) the decision-making model and feature extraction of driving mode for intelligent vehicle; (3) the autonomous intervention mechanism and control method of intelligent vehicle. We focus on to achieve a breakthrough in the area of data analyzing of high dimensional physiological indexes, the decision making model of intelligent vehicle, and the quantitative of switching condition in different driving mode. The conclusions can be applied in the intelligent vehicle; it can not only meet the demands of the safety and efficiency of intelligent vehicle, but also can provide theoretical support for designing and marketing of intelligent vehicle.

在复杂的交通环境中进行全自动驾驶将经历一段漫长的过程，多驾驶模式（人工驾驶、警示辅助、自动驾驶）协同共驾状态将长期存在。在该阶段，针对当前的交通环境及人车状态，如何选择合适的驾驶模式是研究的重点问题之一。此外，对驾驶安全性进行评价也是提高车辆行驶安全性的关键步骤。基于此，本项目综合运用属性选择、人因工程和人工智能等方法，采用模拟实验和实车实验相结合的技术手段，系统研究（1）不同险态交通情境下共驾型智能车模式分级量化标准 （2）共驾型智能车模式选择表征参数与决策模型 （3）不同交通情境下智能车自主干预机制和方法等内容。旨在突破高维异构生理指标数据的在线处理和分析方法、多模式共驾智能车驾驶模式决策建模和不同驾驶模式切换的触发条件及其量化计算方法等关键技术，将在智能车协调安全行驶方面取得创新。项目研究成果应用于自动驾驶系统，将有效的促进智能汽车智能化水平的提高，推动智能车的市场化和产业化进程。

项目研究针对人机共驾智能车辆在行驶过程中的模式接管和安全性评价问题，采用模拟驾驶实验和实车实验相结合的方式深入研究了不同险态交通情境下共驾型智能车辆的模式选择分类方法，以及人机共驾智能车辆不同模式下运行的安全接管方法。项目主要进行了三个方面的研究：（1）采用时间窗和动态K-均值聚类方法构建了智能车辆模式分级标准；（2）采用改进的马尔科夫毯（MB-NEW）和基于信息增益的属性排序方法确定了模式选择的决定性因子；（3）采用贝叶斯网络和C5.0算法构建了智能车辆行驶安全性评估模型及自主干预机制。通过对关键技术的联合攻关，采用在线分析等技术实现了高维异构生理指标数据的在线处理和分析，实现了多模式共驾智能车驾驶模式决策建模和不同驾驶模式切换的触发条件及其量化计算，从而为人机共驾智能车辆的安全性接管奠定了基础。课题在理论和实践上都能够为智能车辆在过渡阶段的发展提供新思路和新方法，加速智能车辆的应用和产业化推广。