NSF Award
2302819
In high-risk workplaces such as construction sites, inattention to workplace hazards is a common factor in serious injuries and fatalities. Evidence strongly suggests that conventional, lecture-based safety training methods in classroom settings rarely capture workers’ interest and do not decrease accidents. Further, adults often learn best in the context of their own work environments and real-life situations. As a result, making safety training more closely associated with trainees’ work environments may be more effective than current methods. To this end, this project will construct a personalized augmented reality (AR) training system that uses pictures taken by trainees on their own worksites to recognize site-specific hazards and create replicas of their worksites that show simulated accidents caused by those hazards. This work will require advances in AR environment design, computer vision algorithms, and workplace training pedagogy, and will lead to better occupational safety training practices in high-risk industries and new educational materials for high schools, universities, and industry around occupational safety. It will also serve as a case study of operationalizing andragogy (adult-targeted teaching) in a learning technology setting.<br/><br/>The project will be organized into two phases. First, the project will create learner-centered AR safety training environments wherein trainees organize safety training content by themselves and experience potential accidents in the AR environment. This first phase will include (1) developing AI algorithms that detect workplace hazards from the trainees’ worksites; (2) creating an accident scenario pool, which can be adapted to various types, shapes, and scales of real-world hazards; and (3) simulating potential accidents by accurately and seamlessly overlaying AR avatars on detected real-world hazards. In collaboration with industry partners, during the second phase the team will examine the effect of the developed training on workers’ attention to real-world hazards by (1) assessing individual trainees’ hazard recognition abilities, (2) monitoring workers’ physical behaviors near workplace hazards, and (3) comparing the performance of individuals trained using the proposed methods to those trained using conventional methods as well as non-personalized and virtual reality environments.<br/><br/>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.
