Drivers' cognitive load detection through ocular metrics classification for the development on Driver Monitoring Systems

车辆自动化可以减少因分心等人为因素造成的碰撞事件，从而使道路更加安全。然而，尽管在过去十年中观察到了技术进步，但完全不需要人类驾驶员投入的车辆尚未得到广泛采用。现有的车辆自动化(汽车工程师协会(SAE)2级和3级系统；L2/L3)仍然要求人类继续参与驾驶的总体任务，以保持车辆的安全运行或在必要时接管控制权。为了确保满足这些要求，正在开发驾驶员监控系统(DMS)，首要目标是监控驾驶员的状态。关于DMS改善道路安全的潜力，仍然存在的问题是，这些系统在L2/L3驾驶期间如何准确地监控司机的认知负荷，并检测分心等危险情况的发生。因此，本研究旨在开发更准确的DMS，能够检测驾驶员在L2/L3驾驶过程中认知负荷的变化。这项研究有三个不同的目标。目标1将采用地面真实测量来测试非侵入性眼部指标(瞳孔大小、凝视集中、眨眼频率)在检测驾驶员认知负荷方面的准确性。目标2将使用目测指标来监测驾驶员在模拟和道路L2/L3驾驶过程中认知负荷的变化。目标3将开发和测试分类模型的准确性，以区分L2/L3驾驶过程中不同水平的认知负荷(从低到高)。这个项目的影响是双重的。DMS开发。申请人与参与DMS开发的汽车供应商(德雷耶夫、佛吉亚)有现有的合作关系。本项目收集的数据将作为基准，帮助行业合作伙伴测试其DMS解决方案的可靠性。采用不同眼球指标组合的分类模型也将为开发更准确的认知负荷检测系统提供信息。L2/L3政策制定。申请人是自动驾驶汽车领域行业组织(ISO/加拿大交通部，SAE)的积极参与者，并与安大略省交通部积极合作。将利用项目结果，为安全采用L2/L3自动驾驶系统制定道路安全政策提供信息。新政策的制定和准确的DMS将为所有加拿大人带来更安全的车辆和街道。在这个研究项目中培训的HQP将拥有非常受欢迎的技能，并能够为这个快速发展的新兴行业做出贡献，无论是在学术界、政府还是行业。

Vehicle automation can make roads safer by reducing the number of collisions attributed to human factors like distraction. However, despite the technological advancements observed over the last decade, vehicles requiring absolutely no input from the human driver have not yet reached widespread adoption. Available vehicle automation (Society of Automotive Engineers (SAE) level-2 and 3 systems; L2/L3) still requires the human to stay engaged in the overall task of driving either to maintain safe vehicle operations or to take over control whenever necessary. To ensure that these requirements are met, Driver Monitoring Systems (DMS) are being developed with the overarching goal of monitoring the state of the human driver. Relating to the potential for DMS to improve road safety, there remains the question of how accurately these systems can monitor the driver's cognitive load during L2/L3 driving and detect the onset of dangerous conditions like distraction. Thus, this research aims to develop more accurate DMS that are capable of detecting changes in drivers' cognitive load during L2/L3 driving. The research is articulated in three distinct objectives. Objective 1 will adopt ground-truth measures to test the accuracy of non-intrusive ocular metrics (pupil size, gaze concentration, blink rate) in detecting driver's cognitive load. Objective 2 will use ocular metrics to monitor changes in driver's cognitive load during simulated and on-road L2/L3 driving. Objective 3 will develop and test the accuracy of classification models in distinguishing between diverse levels of cognitive load (low to high) during L2/L3 driving. The impact of this project is twofold. DMS development. The applicant has existing collaborations with automotive suppliers (Dreyev, Faurecia) which are involved in the development of DMS. Data collected in this project will be used as a benchmark to help industry partners test the reliability of their DMS solutions. Classification models adopting diverse combinations of ocular metrics will also inform the development of more accurate cognitive load detection systems. L2/L3 policy development. The applicant is an active participant in industry organizations in the area of autonomous vehicles (ISO/Transport Canada, SAE) and has an active collaboration with the Ontario Ministry of Transportation. Project findings will be leveraged to inform the development of road safety policies for the safe adoption of L2/L3 automated driving systems. The development of new policies and accurate DMS will lead to safer vehicles and streets for all Canadians. The HQP trained in this research program will have highly sought-after skills and be able to contribute to this fast-paced emerging sector whether in academia, government, or industry.