GOALI: Coadaptation of Intelligent Office Desks and Human Users to Promote Worker Productivity, Health and Wellness

这个学术联络与工业（GOALI）项目的主要目标是进行基础研究，最终使GOALI团队能够开发和测试智能办公工作站（智能办公桌），通过调整姿势、温度和照明设置来优化用户的健康和生产力。智能办公桌使用可穿戴和安装在工作站上的传感器来推断人类的意图、生理状况和当前的任务。该项目推进基础研究，解决如何最好地结合传感器数据、机器学习方法和用户与工作站之间的结构化通信，使热、视觉和姿势条件更接近经过验证的最佳实践，同时提高用户满意度和使用系统的意愿。该项目将生成实验数据，人类意图和偏好的任务特定模型，以及开发机器人设备所需的自适应控制算法，该设备将与工人进行物理和直观的交互，以提高他们的身体舒适度和工作效率。该项目意义重大，因为在办公室中增加智能工作站有可能改变在工作场所促进和实现健康和福祉的方式。该项目通过促进科学探索人类工人和智能机器人系统之间的互动模式，从而促进个体工人、雇主和国家的健康、繁荣和福利，直接服务于美国国家科学基金会的使命。该项目通过旨在招募和留住代表性不足的学生参与研究以及促进创业和创新的外展活动，支持教育和促进多样性。这种智能工作站将通过持续的、双向的、自适应的感知、反馈和操纵环境参数的过程来学习工人的偏好和塑造工人的行为，这些参数有可能直接影响姿势、热和视觉舒适度，并提高工人的生产力。研究了四个任务。任务1评估感知和学习方法来推断工人的热、视觉和姿势舒适的现有状态。传感模式将包括：用于皮肤温度、皮肤电反应和心率的可穿戴传感器；用于温度、湿度、气压、光强和色温的环境传感器；结构光深度传感器（用于姿势评估）；“被动”传感器记录用户对桌椅高度、风扇/加热器速度/设定值、光线强度和颜色的改变。学习方法包括监督学习（由用户对工作站的调整驱动，以及使用Likert-like量表对用户舒适度进行真实评估）、无监督学习和半监督学习（使用有限的用户反馈来标记数据簇）。在推断用户偏好时，系统将考虑任务上下文。任务2检查用户和自主工作站如何最好地协商本地工作站环境的控制，以优化工人的生产力、健康和福祉。研究了两个子任务。第一个探索极端条件下的共享控制（机器提示的完全用户控制或手动覆盖的完全工作站控制）。第二个子任务探讨了当人类和工作站完全访问传感器数据时，工人和工作站之间自适应协商控制环境状态的方法。任务3将使用焦点小组和用户体验研究来确定最适合促进共享控制和用户/工作站交互的通信和反馈提示类型，从而将环境条件推向理想状态。任务4集成了前三个任务的结果，以合成一个工作站控制器，该控制器将促进用户/工作站的共同适应，从而促进生产力、健康和保健。然后，该团队将在为期6个月的功效研究中测试共享自主工作站，研究人机协同适应系统在多大程度上能够使热、视觉和姿势条件更接近已被证明的最佳实践，同时还能提高用户满意度和使用该系统的意愿。项目成果可能通过提高个人和社会工作场所的生产力、健康和福祉而产生长期影响。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The main objective of this Grant Opportunity for Academic Liaison with Industry (GOALI) project is to perform fundamental research that will ultimately allow the GOALI team to develop and test an intelligent office workstation (a smart desk) that will optimize the user's wellbeing and productivity through adjustment of postural, thermal, and lighting settings. The smart desk uses wearable and workstation-mounted sensors to infer human intent, physiological condition and current task. The project advances fundamental research addressing how best to combine sensor data, machine learning approaches and structured communication between the user and the workstation to bring thermal, visual and postural conditions closer to proven best practices over time, while simultaneously improving user satisfaction and willingness to use the system. The project will generate the experimental data, task-specific models of human intent and preferences, and adaptive control algorithms needed to develop a robotic device that will interact physically and intuitively with workers to enhance their physical comfort and workplace productivity. The project is significant because the addition of intelligent workstations in offices has potential to change the way health and wellbeing are promoted and achieved in the workplace. This project directly serves the NSF mission by promoting science that explores modes of interaction between human workers and intelligent robotic systems that advance the health, prosperity and welfare of individual workers, their employers, and the nation. The project supports education and promotes diversity through outreach activities aimed at recruiting and retaining under-represented students in research, as well as by promoting entrepreneurship and innovation.This intelligent workstation will learn worker preferences and shape worker behavior through an ongoing, bi-directional, adaptive process of sensing, feedback and manipulation of environmental parameters that have the potential to directly impact postural, thermal and visual comfort and to increase worker productivity. Four tasks are researched. Task 1 evaluates sensing and learning methods for inferring the worker's existing state of thermal, visual and postural comfort. Sensing modalities will include: wearable sensors for skin temperature, galvanic skin response, and heartrate; environmental sensors for temperature, humidity, air pressure, light intensity and color temperature; a structured light depth sensor (for postural assessment); and "passive" sensors to record user changes to desk/chair height, fan/heater speed/set-points, and light intensity and color. Learning approaches include supervised learning (driven by user adjustments to the workstation as well as ground truth assessments of user comfort using Likert-like scales), unsupervised learning, and semi-supervised learning (using limited user feedback to label data clusters). The system will consider task context when inferring user preferences. Task 2 examines how the user and the autonomous workstation might best negotiate control of the local workstation environment to optimize worker productivity, health and well-being. Two sub-tasks are researched. The first explores shared control under the extreme conditions (full user control with machine cueing or full workstation control with manual overrides). The second sub-task explores approaches for adaptive, negotiated control of environmental state between the worker and the workstation when the human and workstation have full access to the sensor data. Task 3 will use focus group and user experience studies to identify the kinds of communication and feedback prompts best suited to promote shared-control and user/workstation interactions that drive environmental conditions toward the ideal. Task 4 integrates the results from the first three tasks to synthesize a workstation controller that will facilitate user/workstation co-adaptations promoting productivity, health and wellness. The team will then test the shared-autonomy workstation in a 6-month efficacy study examining the extent to which the co-adaptive human/machine system can bring thermal, visual and postural conditions closer to proven best practices over time, while also improving user satisfaction and willingness to use the system. The project outcomes may have long-term impact by improving individual and societal workplace productivity, health and well-being.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.