Highly accessible platform technologies for vision‑guided, closed‑loop robotic assembly of unitized enclosure systems

AbstractEquipping robotics with the capacity to make decisions based on real-time information about their physical environment is crucial to the success of in situ robotics and offers many process advantages in prefabrication scenarios as well. However, the perceived technical challenges of producing environmentally-aware closed-loop robotics have limited their use in construction and design applications. To address this challenge, a low-cost and largely open-source computer vision-guided closed-loop robotic control paradigm is developed. The system is used here to identify construction materials in the workspace and calculate their position in space and determine their place in the facade panel assembly. The industrial robot arm is equipped with an RGB-depth camera in an eye-in-hand configuration to give control over the positioning of the camera for greatest accuracy. The control system relies on a simple TCP client/server connection between the robot and a central control computer to pass information and instructions from the computer vision system to the robot and vice versa. This setup delivers process flexibility, enabling pick-and-place procedures of the material positioned randomly within the workspace. In this work, the technologies are deployed in a factory-type setting but would also be necessary for any on-site robotic construction system, building towards an on-site robotics future. The final product of this research is a unitized spandrel panel wherein the vision-guided robot finds and places the insulation, cement board, and masonry cladding materials.