I-Corps: Personalized AI-Driven Training for Construction Workers with Non-Intrusive Measures

Purdue University, West Lafayette IN

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is the development of a software platform in combination with wearable sensors for training construction workers. Currently, construction companies rely on training programs to improve construction workers’ skills. However, traditional classroom-based training environments may not adequately prepare workers for the current challenges in the construction industry. The proposed software platform is designed to capture performance and personalize learning in non-invasive ways that may enable rethinking of the current pedagogical approach. The proposed technology provides smart training systems that observe human metrics to strengthen workplace and academic educational practices and knowledge acquisition among diverse learners. In addition, the adaptive systems may incorporate mixed reality that provide context-dependent support from multiple sources of information and include personalized tracking of the individual worker’s capabilities, work history, goals for the task, and prior performance on the task. The project is aimed at architecture, engineering, and construction worker training; however, the proposed platform may be adapted for other labor-intensive industries. This I-Corps project is based on the development of personalized artificial intelligence (AI)-driven training for construction workers that includes non-intrusive measures. The proposed technology uses the Human-Error Detection Framework that harnesses real-time psychophysiological data collected from wearable sensors (e.g., eye tracker, electroencephalogram, electrodermal activity, and photoplethysmography). The sensors are designed to measure, track, and predict workers’ performance and capabilities using the multimodal heterogeneous sensor data. Predictive models resulting from this study may contribute to significant accident reduction as well as provide a critical validation measure to confirm the effectiveness of training programs on enhancing workers' risk-analysis skills. Validated at real construction jobsites, the algorithms, classifiers, and predictive models developed by the research reveal which physiological metrics characterize training effectiveness. Since the results of this study link training proficiency to direct measures of cognitive load and attentional demands, it lays the foundation for developing personalized training environments that provide the optimum amount of challenge for each user in dynamic and hazardous workplaces such as construction. These results challenge the passivity paradigm of construction training by creating methods to boost workers’ cognitive abilities by considering their individual differences to overcome challenges on work sites. 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.