智能汽车人车协同感知与失效特性研究

Intelligent vehicle technologies have become the trendiest research topic. However, as technique, regulation, and road environment are not developed for fully automation, most intelligent vehicles on road are now and will be in the Cooperative Human-Vehicle Driving state such as assistant driving and conditional automation. In such systems, drivers can decide whether to control the intelligent vehicle, safety reliability is relying more on drivers rather than vehicles...This study focuses on cognition reliability of Cooperative Human-Vehicle Driving. Cognition characteristics are collected from “Human in Loop” driving simulation experiments. Fuzzy logic is involved to analyse the overall cognition impact of multi-media warning information on Cooperative Human-Vehicle. Personal differences information process ability and threshold of information overload are then analysed for each participant. Furthermore, occurring mechanisms of cognition failures are modelled and analysed. Prevention strategies are then proposed to reduce cognition failures for complex driving scenarios and low-ability drivers. The study will provide bases for improving information interaction between human and machine in intelligent vehicles.

智能汽车是交通智能化发展的必然趋势，但目前，由于技术法规等方面尚未成熟，未来一段时期内，在路行驶的智能汽车多为辅助驾驶、条件自动驾驶等“人车协同驾驶”车。协同驾驶环境下，驾驶人的风险感知受其自身观察与多模态交互信息的双重影响，存在信息过载、感知失效、车辆控制权归属错位等问题，导致错误驾驶行为，严重影响行车安全。. 本课题针对多模态人车协同感知与失效问题，搭建“人在回路”式仿真平台模拟人车协同系统，采集分析感知特性数据；基于模糊逻辑构建融合多模态信息的人车协同感知模型，测算分析驾驶人个体感知特性与负荷边界信息量；基于蒙特卡洛法实现感知模型统计模拟，获取人车协同感知失效事件库，分析复杂驾驶场景下的感知失效致因及失效机制，并提出面向重点失效场景与重点驾驶人的优化方案。本课题将为预防协同感知失效、优化多模态人车协同系统、提高智能汽车安全可靠性等方面提供相关方法与实践依据。

本项目面向智能汽车人车协同驾驶系统环境中，多模态交互信息对驾驶人的感知影响尚不明晰，以及个体感知能力差异对感知失效特性与机制影响未知的问题，搭建了一种多等级多用户的人在回路协同仿真平台，实现了人车协同环境下的多模态信息感知数据采集；引入非线性附加噪声因果推理模型与有向无环图，提出了基于模糊逻辑的群体感知模型，定量解析交互信息对驾驶人感知的影响，并进一步构建个体层面的模糊感知模型，分析驾驶人感知能力的个体差异。最后考虑人车系统感知失效的随机性和多样性，识别出典型失效原因与失效路径，提出加强人车协同感知可靠性的针对性改善对策。相关研究工作整理后共形成9项研究成果，其中专著章节1章、期刊论文3篇、会议论文4篇、授权专利1项。研究成果可应用于高精度车载驾驶员监控系统开发，为提高智能汽车安全可靠性、优化智能汽车驾驶体验提供依据。