��������

The construction industry is currently experiencing a shortage of workers at a time when there is $1.2 trillion worth of work needed to renew infrastructure systems across U.S. That dynamic has developers in the public and private sectors looking for ways to improve efficiency.

Research at CU Boulder could provide an answer or at least a road map through this problem by understanding how construction plans are read on job sites and then tailoring the information to the individual. Increasing efficiency, reducing costs and – potentially – reducing the risk of an accident.

The work is being led by Professor Paul Goodrum in the Civil, Environmental and Architectural Engineering Department and centers on the way construction plans are created, distributed and understood. 

“On most projects there is a 3D model, but by the time information is provided to a job site to those who build the project the plumbing plans are in a 2D format. My group has been using eye tracking technology to explore how sharing that information in a 3D format or tweaking the 2D representation can improve performance,” said Goodrum. “How we can better tailor that information to an individual’s abilities? The answer to that question would be big for the industry and for improving the curriculum we use to teach engineers going forward.”

Eye tracking research from Goodrum’s group shows that craft workers have difference levels of spatial cognition. That is, the ability to understand where they are in an environment or how something like a pipe is laid out, varies based on innate ability and experience gained over time. The research shows that workers with lower spatial cognition working off of 2D plans perform worse on time to completion and have a higher error rate. Both are big factors in the overall efficiency and cost of a construction project.

“By looking at the time, sequence and number of eye fixations, we can visualize how an individual looks at and processes the information. What we find is that those with lower spatial cognition leverage the 3D information when given to them and see improvements. It levels the playing field,” said Goodrum, who is part of the Imaging Science Interdisciplinary Research Theme in the college.

One way Goodrum’s group is visualizing the data is by creating convex hull areas – essentially an elastic band stretched around a set of 100 eye fixation points on a diagram. The resulting “heat map” shows how much information is being taken in at a given time and how the reader is moving through the diagram.

“A small hull means my fixations are concentrated in one area; a large hull means they are looking at it broadly. Are they looking at the details or the big picture?” Goodrum said. “That is interesting to compare to the spatial cognition scores and find correlations and we see that those with larger hulls do have score better in spatial cognition and other metrics as well.”

This research may also lead to the development of advanced visualization with augmented reality capabilities. Future technical advancements, such as safety goggles with a built-in virtual display, could transform the future work interface for millions of construction craftwowkers.

“Augmented reality is still relatively is expensive today to be used on such a large scale in construction, but the technology is going to get cheaper in the future. What we find when talking to industry about it though is that the potential cost savings is great, but it’s also the reduction of errors they are most interested in. Reducing the chance something is installed wrong which can be dangerous in the future is really important for the safety of construction craft workers and the overall project,” he said.

Goodrum said his group will continue to work on this topic and that his PhD student, Omar Alruwaythi, would be publishing a paper on the subject in ASCE Journal of Construction Engineering and Management soon.

“One of the next steps is generating more data that tracks eye movement in the same way. That will allow for statistical comparisons,” he said.