CAREER: Co-Adaptation and Trust in Worker-Robot Interactions: Scalable Adoption of Collaborative Robots in Construction

建筑工作目前面临着高伤害率、停滞不前的生产力、劳动力短缺和使用过时的工作流程的挑战。在未来，机器人技术可能会提供前所未有的机会来解决这些问题，但是机器人技术还没有完全解决由非结构化环境产生的一系列技术挑战，比如混乱的建筑工地和工人之间缺乏信任，他们可能不接受机器人作为合作伙伴。这个教师早期职业发展（Career）项目将推进美国国家科学基金会的使命，通过推进对人类活动、意图识别和机器人学习算法的自适应技术的基本理解，促进科学进步，促进国家健康、繁荣和福利，这些技术可以提高建筑材料处理场景中工人的安全。人类活动和意图识别将通过将来自可穿戴和环境传感器的实时信号引导到先进的机器学习和神经网络算法来执行。使用机器人运动规划器，在一定程度上优化了从机器人到人类的材料转移的人体工程学，将促进工人在材料处理过程中的安全。此外，项目团队将开发一个建立信任的模型和一个在工人-机器人团队中进行信任校准的框架，以确保建筑工人准确评估在工作现场对机器人伙伴的信任程度。该项目还包括一个教育和外联部分，为包括高中生及其教师、本科生和研究生在内的不同群体建立STEM教育能力。这个实用灵感的CAREER项目有助于未来协作机器人（co-robots）向建筑工人学习并帮助他们，从而减少体力工作量，同时促进人体工程学安全。目前的机器人算法和应用不能适应建筑工地的非结构化复杂性，也不能完全解决行业中工作、工人和工作场所的技术和行为挑战。该项目侧重于建筑材料处理，这是一项常见且费力的活动，可以使用协作机器人来促进。该项目将：(1)通过协同自适应机器人学习系统创建安全的机器人辅助材料处理工作流程，该系统响应可穿戴传感器收集的工人运动学和肌肉活动；(2)构建信任构建模型和信任校准框架，促进工人-机器人团队在建筑施工中的应用。该项目有望推进适应性和可重用建筑工人活动和意图识别方面的基础知识，并将产生新的数据集和模型。它将提高机器人辅助材料处理任务的安全性，使用优化人体工程学安全性的运动规划器。智能的工人-机器人团队将通过信任建立和信任校准模型来培养，这些模型可以用来指导工人-机器人的共同适应。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。

Construction work is currently challenged by high injury rates, stagnant productivity, labor shortages, and use of outdated workflows. In the future, robotics may offer unprecedented opportunities to address these issues, but robotics has yet to fully address the range of technical challenges that arise from unstructured environments such as cluttered construction job sites and impoverished trust among workers who may not accept robots as a collaborative partners. This Faculty Early Career Development (CAREER) project will advance the NSF mission to promote the progress of science and to advance national health, prosperity, and welfare by advancing a fundamental understanding of adaptive technology for human activity and intent recognition and robot learning algorithms that can promote worker safety during construction materials handling scenarios. Human activity and intent recognition will be performed by directing real-time signals from wearable and environmental sensors to advanced machine learning and neural network algorithms. Worker safety during materials handling will be promoted using a robot motion planner that optimizes, in part, ergonomics of the material transfer from the robot to the human. Additionally, the project team will develop a model for trust-building and a framework for trust-calibration within worker-robot teams to ensure that construction workers accurately assess how much to trust their robotic partners on the job site. The project also includes an education and outreach component that builds STEM education capacity for a diverse group of individuals including high school students and their teachers, as well as undergraduate and graduate students.This use-inspired CAREER project contributes to a future in which collaborative robots (co-robots) learn from and assist construction workers, thereby decreasing physical workload while promoting ergonomic safety. Current robotics algorithms and applications fail to adapt to the unstructured complexity of construction job sites, and do not fully address the technical and behavioral challenges of the work, workers, and workplaces in the industry. This project focuses on construction material handling, a common and strenuous activity that can be facilitated using co-robots. This project will: (1) create safe robot-assisted material handling workflows through a co-adaptive robot learning system that responds to worker kinematics and muscle activities collected by wearable sensors; and (2) develop a model for trust-building and a framework for trust-calibration that aims to promote adoption of worker-robot teaming in construction. The project promises to advance fundamental knowledge in adaptive and reusable construction worker activity and intent recognition and will generate novel datasets and models. It will promote safety in robot-assisted materials handling tasks using a motion planner that optimizes ergonomic safety. Intelligent worker-robot teaming will be fostered by models of trust-building and trust calibration that can be used to guide worker-robot co-adaptation.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.