Construction Automation, Robotics & Visualization Laboratory
Navigating labor shortages, safety issues and stagnant productivity, the Construction Automation, Robotics and Visualization Lab strives to solve some of the most challenging problems in the construction industry. Our research emphasizes the use of automation and artificial intelligence, including robotics, computer vision and machine learning. By collaborating with industry partners and various departments at Auburn University, the CARV Lab is uniquely positioned to lead the development of innovative technical solutions for our complex industry.
Researchers: Eric Wetzel, Busra Yucel, Muhammad Umer
Can a robot answer the questions “where is it,” “what is it” and “how many” on a dynamic construction site? CARV is combining machine-learned object detection and recursive localization in an attempt to do just that.
The Core Temperature Visualization System (COTVIS) uses real-time biometric data to provide visible alerts, enabling construction crews rather than individuals to recognize and intervene when workers exhibit signs of heat stress. The goal of the COTVIS system is to help prevent severe heat-related illnesses with broader implementation pending further field testing and refinement. The device was provided a provisional patent.
Researchers: Muhammad Umer, Caleb Powell, Justin Patton (RFID) and Eric Wetzel
In partnership with the AU RFID Lab, CARV is deploying a Bayesian algorithm and RFID payload on a Boston Dynamics Spot to automate on-site material inventory management through material count and localization.
Researchers: Jian Zhang (Kennesaw State) and Eric Wetzel
In a research partnership with Kennesaw State’s Department of Information Technology, CARV is working to enhance medium and large-scale vision models’ ability to read and interpret construction documents for robot tasking.
Researchers: Eric Wetzel, Anthony Matthews (MITRE), and Dean Conte (MITRE)
In a research partnership with the MITRE Corporation, CARV is developing a dynamic ReSLAM approach to ensure safe navigation, enhance worker-robot interactions and maintain efficiency on job sites.
Researchers: Eric Wetzel, Chase Frazelle (MITRE), and Alex Rudin (MITRE)
In a research partnership with the MITRE Corporation, this project aims to achieve “context-aware robotics” capable of adapting to the dynamic nature of construction sites, enhancing worker safety, alleviating repetitive tasks and establishing robust protocols for human-robot collaboration.
Segmentation using RGB cameras is a common use case of object detection; however, visually similar objects are often incorrectly predicted or given low confidence. This research trains a spectral model on visually similar construction materials and deploys a hyperspectral payload on a robot to improve autonomous percent complete capture on construction projects.
Researchers: Jian Zhang (KSU), Liang Zhao (KSU), and Eric Wetzel
In a research partnership with Kennesaw State’s Edge Intelligence Research Lab, this project aims to establish mutual trust between non-technical users and robots using visual hand signals in voice-incapable environments such as construction and industrial sites.
Fine-tuning a large language model for course-specific information can create a specialized GPT for students in a construction estimating class.
Researchers: Eric Wetzel, Muhammad Usama, and Mahir Somalwar (HP)
In partnership with Hensel Phelps, CARV is developing a real-time object detection model specifically for three common safety glasses to be deployed on site.
