近年来，随着科学技术的发展，市场上最先进的高级驾驶辅助系统(ADAS) 已经可以实现同时自动化的横向和纵向车辆控制。然而，受限于技术水平，这些系 统仍然要求驾驶员实时监视道路，并在遇到系统限制或者技术故障时，及时进行接 管。使用过程中，驾驶员可能因过度依赖这些系统，而投入过多的注意力来执行次 任务，给道路交通安全带来隐患。本文以“提出问题-探索行为认知机理-提出并验 证解决方案”为研究主线，探究了有限自动驾驶下，与驾驶员行为适应和注意力自 我调节相关的工效学问题。本文首先采用半结构化访谈的研究方法，基于定性行为适应模型，调查了特斯 拉车主使用有限自动驾驶过程中的行为适应、信任水平和心智模型。访谈结果表明， 驾驶员们对自动驾驶系统信任程度很高。在实际使用的过程中，驾驶员逐步建立了 相对准确的心智模型，借以帮助他们筛选出相对安全的使用情景。而驾驶员在这些 情景下，倾向于投入更多注意力来执行次任务。与此同时，一些自我调节行为，又 帮助他们将风险控制在一定范围内。为了进一步研究影响驾驶员在执行次任务时的自我调节机制，本文开发了模 拟有限自动驾驶任务形式的结构化抽象任务，并探究了三层次的自我调节在执行 次任务时的作用。结果表明，在规划层面，自我调节受到预期的接管事件率和接管 场景的时间预算的影响;在决策层面，则受短期时间期待效应的影响;在控制层面， 当预期的接管险情的时间预算较少，被试倾向于停止执行次任务来等待险情出现， 而预期的时间预算较多时，则倾向于在任务之间频繁切换，并减少一次切换到次任 务的时间。在此基础上，本文又进一步探究了注意网络模型中警觉、定向和执行控 制与注意力自我调节的关系。结果表明，在注意力调节过程中，被试在警觉任务上 停留的时间与警觉网络效率负相关，而次任务的投入程度，则与定向网络效率负相 关。根据这些认知过程，研究设计了一种基于进度条的连续反馈方式的驾驶员监视 系统，并通过模拟器实验验证了该方案对于促进驾驶员监视道路的有效性。本文的研究成果是在有限自动驾驶技术条件下，对驾驶员行为适应理论的深 入和扩展，加深了对于驾驶员自我调节中认知过程的理解，为如何促进驾驶员在即 将普及的有限自动驾驶中执行监视控制，设计了有效干预和管理的手段，并提出了 新的研究方法和思路。
In recent years, with the development of driving technologies, the state-of-the-art ADAS can achieve lateral and vertical control automatically and simultaneously. However, limited by the current technological level, these systems still require the driver to take over in response to systematical malfunctions. When the autonomous driving mode is employed, drivers may spend too many cognitive resources to secondary task engagement. This research followed the route of “raising questions, exploring the behavioral and cognitive mechanism, proposing and verifying the solution”, and explored the ergonomics problems related with behavior adaptation and attentional self-regulation in the context of partially automated driving.This research first employed semi-structured interviews to probe into the short-term behavioral adaptation of Tesla drivers. Based on qualitative behavioral adaptation model, drivers’ behavioral adaptation, trust level, and mental model were analyzed. The interview results illustrated that drivers kept an optimistic attitude towards this system, with a considerably high trust level. They developed a relatively accurate mental model mainly through practice, helping them to filter relatively safer driving situation. Then, they were widely involved in secondary tasks, while some self-regulation behavior helped them to control the risk within a specific range.To further analyze the self-regulation problems, and to explore the cognitive mechanism of attention regulation, this research developed a structural and abstract task for partially automated driving and explored the three-level self-regulation model. The results showed that drivers’ self-regulation was influenced by the taking-over scenario's event rate and urgency level in the planning level. While in the decision level, drivers’ self-regulation was influenced by the short-term expectation effect. In the control level, theses results showed that participants tended to disengage from the secondary task with the anticipation of a more urgent hazard but to continue the secondary task with frequent switching-back for a less urgent hazard. Moreover, this research explored the relationship between the attentional networks (phase alerting, orienting, and executive control) and the self-regulation. The results showed that the participants with more efficient alerting network spent less time on the vigilance task and participants with more efficient orienting networks was less engaged in the secondary task. Based on the findings, this research developed a driver monitoring system with continuous feedback using a progress bar. The driving simulation study was conducted to prove that the continuous feedback was better in keeping drivers engaged, compared with current scaling feedback.This research was a further development of drivers’ behavioral adaptation theory under partially automated driving conditions, and provided new perspectives and methods to explore how to keep drivers engaged while supervising driving automation.
