Design and experimental validation of a tile-laying system based on dual-arm coordination and visual closed-loop control
This study presents and experimentally validates a dual-arm robotic system that integrates coordinated tile placement with real-time visual inspection and closed-loop control to ensure joint-width accuracy within a predefined tolerance.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Laying floor tiles by hand is a job that demands patience, a steady eye, and a body willing to kneel and bend for hours. It is work that has changed little for centuries, relying on the skill of a human to judge the tiny gaps between ceramic squares and to press them into place with just the right amount of force. When a gap is too wide or too narrow, the result is a floor that looks uneven, and fixing it often means starting over. This physical strain and the high standard for visual perfection have driven engineers to ask a simple question: can a machine do this work better than a person? The answer lies in a field where robots learn to see and move with the precision of a master craftsperson, using cameras to measure the world and arms to manipulate it. The goal is not just to automate a task, but to create a system that can check its own work as it goes, stopping immediately if a mistake is made rather than piling error upon error.
In a recent study, researchers from Chongqing University and Zhejiang University of Technology built a system to test this idea using two robotic arms working together. They did not try to build a robot that could tile an entire house at once; instead, they created a controlled laboratory setup to prove that the core logic of the system works. The setup involved two six-degree-of-freedom robotic arms, which is a technical way of saying they have joints that allow them to move and rotate in all the directions a human arm can. One arm was equipped with a suction cup gripper to pick up and place the tiles. The second arm carried a high-resolution industrial camera, acting as the system's eyes. The researchers designed a workflow where the first arm places a tile, and the second arm immediately moves in to inspect the gap between that new tile and the ones already laid down.
The process is a continuous loop of action and verification. The robot with the suction cup picks up a small ceramic tile, moves it over the work area, and lowers it into place. As soon as the tile is set, the second robot arm swings the camera into position to take a picture of the seam, or joint, between the tiles. A computer program analyzes this image to measure the width of that gap. The system has a strict rule: the gap must be exactly 4.0 millimeters wide, with a tolerance of half a millimeter in either direction. If the camera sees a gap that is 3.5 millimeters or 4.5 millimeters, the system considers the job done correctly and signals the first arm to pick up the next tile. However, if the gap is too small, like 2.5 millimeters, or too large, like 4.8 millimeters, the computer flags the error. It sends a signal to stop the entire process, preventing the robot from laying another tile on top of a mistake. This "closed-loop" control means the machine does not just blindly repeat a motion; it constantly checks the result and decides whether to continue or pause for human help.
To test this, the researchers ran the system through three separate trials, laying out nine tiles in a specific rectangular pattern. They used tiles that were 50 millimeters by 50 millimeters. In every single trial, the system successfully placed the tiles and measured the gaps. Across all three runs, the 18 gaps they measured between adjacent tiles all fell within the acceptable range of 3.5 to 4.5 millimeters. The average width of the gaps was very close to the target, hovering around 3.87 millimeters. The system was also tested on its ability to catch errors. The researchers deliberately placed tiles so that the gaps were either too narrow or too wide. In these cases, the camera correctly identified the problem, displayed a warning, and successfully stopped the robot from proceeding to the next tile. The entire cycle of placing a tile, taking a picture, measuring the gap, and making a decision took about 50 seconds per tile in their setup.
The study demonstrates that it is possible to combine the physical act of laying a tile with an immediate, automated check of the quality. By separating the tasks between two arms—one for moving and one for seeing—the system avoids the delays that would occur if a single robot had to swap tools between placing a tile and inspecting it. The results show that the machine can maintain a consistent gap width that meets the strict standards of professional tiling. While this prototype was tested in a quiet lab with small tiles and a perfectly flat surface, the findings suggest a path forward for construction robotics. The system proved it could handle the logic of the job: pick, place, measure, decide, and repeat. It showed that a robot can not only do the heavy lifting but also possess the judgment to know when a job is done right and when it needs to stop. This dual-arm approach offers a glimpse of a future where construction robots do not just follow instructions but actively ensure the quality of the work they produce, potentially reducing the physical burden on human workers while maintaining the high precision that good construction requires.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.