驾驶员操纵能力动态演变规律及其对智能汽车人车控制权分配影响研究

In this project, we will focus on the new trend of human-vehicle interactive design in intelligent vehicle system around the world and the research demand of human-vehicle interactive testing and evaluating system in Chinese OEM. The driver’s instant psychological and physiological state and their potential take-over ability when the controlling authority between human and vehicle interacts under the human-vehicle cooperation circumstance will be analyzed. A novel relationship, the characteristically expression and the dynamic progress law of driver’s operation work load and ability based on the machine learning and the pattern recognition of multi-bioelectricity information, will be established. The advantages of the operation work load expression (electrocardiogram, electroencephalogram, and electromyogram), operation ability (road cognition, traffic trend prediction and musculoskeletal motion control) and the intelligent vehicle control will be combined. A new theory, the mixing distribution and interactive learning mechanism of human-vehicle control authority based on multi-object decision and reinforcement learning method, will be established while the synthesized performance of safety, comfort, energy saving, low conflict and high efficiency is considered as the optimal goal. An integral human-vehicle interaction testing platform and a subjective and objective assessing method based on the driver’s dynamic open-loop simulation and the psychological physical model in experimental psychology will also be built. These will provide theoretical foundations and key technologies for the humanized and individualized design of human-vehicle interaction in intelligent vehicles. Furthermore, these will provide local research and test requirements and basic data support, which are adapted to the Chinese driver and the Chinese road condition, for the driving assistant system design and the human-machine system design in China.

人车深度融合与协同控制是智能汽车人车交互的前沿与热点。以往相关研究往往集中于驾驶员认知负荷的评测以及控制权切换过程驾驶员反应时间等方面，缺乏对人车控制权分配机理以及该过程的驾驶员自身操纵能力的适应性研究。本项目瞄准国际智能汽车人车交互的相互适应性前沿动态和我国整车厂家人车交互测试评价体系研发需求，分析人车控制权交互过程驾驶员即时心生理状态与潜在的接管控制能力，建立基于多模态生物电信息的驾驶员操纵负荷特征表达及其动态演变规律，构建驾驶员操纵负荷-操纵能力的数据-机理动态演变模型；以此为基础，设计基于多目标决策与增强学习的人车控制权混合型分配与交互学习方法，搭建智能汽车人车交互性能一体化测试平台与基于心理物理学的主客观评价方法，为智能汽车人车交互人性化、个性化设计提供理论基础和关键技术，也为我国先进驾驶辅助系统开发提供适配于中国驾驶员、中国路况的本地化研发实验条件和基础数据支撑。

人车深度融合与协同控制是智能汽车人车交互的前沿与热点。以往相关研究往往集中于驾驶员认知负荷的评测以及控制权切换过程驾驶员反应时间等方面，缺乏对人车控制权分配机理以及该过程的驾驶员自身操纵能力的适应性研究。.本项目在驾驶员正常态-临界态-异常态研究中，在多个工况下探究驾驶员操纵能力变化，并采用驾驶模拟器方法，在实验场景中设置危急工况，探究驾驶员在紧急情况下的决策机理以及驾驶行为。基于多导生物电信号的机器学习和模式识别，建立正常态-临界态-异常态、长时-短时、主次任务共存下驾驶员操纵负荷与操纵能力动态演变数据机理模型，定量化揭示驾驶员在不同任务不同驾驶负荷下具备的潜在操纵能力。基于模糊隶属度函数方法，利用实时获取的驾驶员心生理状态操纵负荷与操纵能力的特征值、智能汽车运动学动力学特性及道路交通环境态势信息，建立反映人车路协同共驾的安全、舒适、节能、环保、低冲突、高效通行的性能指标的定量化描述模型。建立基于深度学习的驾驶员与智能汽车控制性能的交互学习机理，实现人车控制权的相互适应性分配，推进智能汽车人车交互性能的人性化。项目期间发表学术论文17篇，其中SCI检索11篇、EI检索6篇，获得发明专利7项，参与国际国内重要学术论坛9次。“汽车智能主被动安全一体化设计关键技术及产业化”获得中国汽车工业科学技术进步奖一等奖，本人排序第一位。本人获中国汽车工业优秀科技人才奖。.本项目为智能汽车人车交互人性化、个性化设计提供理论基础和关键技术，也为我国先进驾驶辅助系统开发提供适配于中国驾驶员、中国路况的本地化研发实验条件和基础数据支撑。