Immersive virtual learning for worker-robot teamwork on construction sites

建筑业是美国最大的行业之一，雇用了数百万工人，但它面临着生产率低、工人短缺和安全问题的挑战。随着最近技术的进步，该行业有一个巨大的机会来提高其生产率和安全性。随着建筑工地自动化水平的提高，工人们需要学习如何使用这些新技术，并获得新的知识和工程技能。该项目的目标是创建和测试通过虚拟现实(网络学习)提供的培训计划，以教育和再教育工人获得新的工作经验，这些工作经验需要在建筑工地上与机器人合作。该项目旨在改善建筑工地上工人和机器人之间的团队合作，专注于不止一项特定技能，并提供全面的学习体验。该项目意义重大，因为它是一项开创性的努力，为语言熟练程度和受教育程度不同的工人提供学习机会，为他们在人类技术前沿工作做好准备。该项目涉及的基本问题包括：1)与传统培训方法相比，网络学习如何增加工人对自动化的知识、安全行为和信任？2)个体差异如何影响工人的网络学习？3)对自动化的信任如何根据施工任务发生变化？4)网络学习如何影响建筑工地的生产率和安全？5)通过网络学习获得的知识和信任水平是否会传递到实际建筑工地？6)网络学习如何影响下一代建筑自动化的发展？通过对这些研究问题的探索，本项目为建筑行业职业劳动力培训项目的实效性提供了证据。在任务1中开发的网络学习平台中包含了许多学习场景和类型的建筑机器人。通过使用离散事件模拟和沉浸式虚拟环境来模拟建筑工地，建筑工人在给定的任务中与机器人进行交互。为了摆脱今天以实验室为基础的动手培训，网络学习环境纳入了动态的建筑工作环境，例如通过模拟和使用建筑工地的准确模型，多项任务和人员共享相同的空间和条件，例如不平坦的地形、灰尘和雨水。这项工作在两个层面上推进了关于网络学习对工人-机器人团队合作的影响的知识：用户层面和现场层面。对于任务2中的用户级调查，该工作探索：1)网络学习在多大程度上增加了员工对自动化的知识、安全行为和信任，2)个人差异如何调节员工的学习，以及3)对自动化的信任如何根据任务类型发生变化。对于任务3中的现场调查，这项工作探索：1)与现场培训相比，网络学习对工作效率、安全性和自动化信任的影响程度如何；2)如何改善网络学习环境和建筑机器人，以便在建筑工地部署。最后，通过任务4中基于模拟的研究，这项工作探索了人与机器人的交互如何影响施工任务和工作流程、安全程序和生产率。网络学习计划旨在最大限度地提高工人对建筑机器人的接受和信任程度，同时释放人力资源，提高生产率和安全性，这两点对工人个人和整个社会都很重要。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The construction industry is one of the largest industries in the United States, employing millions of workers, however it is challenged by low productivity rates, worker shortages, and safety concerns. The industry has a tremendous opportunity to improve its productivity and safety with the recent advancements in technology. With the increased level of automation on construction sites, workers need to learn how to work with these new technologies and gain new knowledge and engineering skills. The goal of this project is to create and test a training program delivered through virtual reality (cyberlearning) to educate and re-educate workers for new work experiences that require collaboration with robots on construction sites. The project seeks to improve the teamwork among workers and robots on construction sites, focusing on more than one specific skill set and providing a comprehensive learning experience. The project is significant because it is a pioneering effort in providing learning opportunities to workers with varying levels of language proficiency and education, preparing them for work at the human-technology frontier. Fundamental questions addressed by the project include: 1) How does cyberlearning increase workers' knowledge, safety behavior and trust in automation compared to the traditional training methods? 2) How do individual differences impact workers' cyberlearning? 3) How does trust in automation change based on the construction task? 4) How does cyberlearning affect productivity and safety on construction sites? 5) Does the knowledge and level of trust gained through cyberlearning carry over to actual construction sites? 6) How does cyberlearning influence the development of next generation of construction automation? Through the exploration of these research questions, this project provides evidence for the utility of cost-effective training programs for vocational workforce of the construction industry. Numerous learning scenarios and types of construction robots are included in the cyberlearning platform, which is developed in Task 1. A construction site is simulated via the use of discrete event simulations and immersive virtual environments where construction workers interact with robots on a given task. To move away from today's lab-based hands-on training, the cyberlearning environment incorporates dynamic construction work environments, such as multiple tasks and crews sharing the same space and conditions, for example, uneven terrain, dust, rain through simulations and by using accurate models of construction sites. The work advances knowledge on the impact of cyberlearning for worker-robot teamwork at two levels: user-level and site-level. For the user-level investigations in Task 2, the work explores: 1) the extent to which cyberlearning increases workers' knowledge, safety behavior, and trust in automation, 2) how individual differences moderate workers' learning and 3) how trust-in-automation changes based on task type. For the site-level investigations in Task 3, the work explores: 1) the extent to which cyberlearning impacts productivity, safety, and trust-in-automation compared to in-person training across all workers on construction sites, and 2) how to both improve the cyberlearning environment and the construction robots for deployment on construction sites. Finally, through simulation-based studies in Task 4, the work explores how human-robot interactions impact construction tasks and workflows, safety procedures and productivity. The cyberlearning program is designed to maximize the extent to which the workers accept and trust construction robots, while at the same time freeing up human resources, improving productivity and safety, both of which are important to both individual workers and society at large.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.