Imagine entering a construction site in the near future and seeing a group of robots grading and laying out the site, constructing a masonry wall, or even putting together scaffolding trusses. This might sound far-fetched, like something out of a sci-fi movie or a Ray Bradbury or Isaac Asimov short story, but developments in construction robotic technology are rapidly turning this into reality.

You may already be using robots at the construction site, depending on the sort of work you do. There are now widely available robotic heat welders and remote-controlled demolition robots. There are many ways that robots could be used to speed up, simplify, and lessen the labor-intensive nature of construction work, though it's unlikely that they will ever fully replace the need for human labor in the sector.

In this article, we will take a look at some cutting-edge AI developments in the construction vertical and the data annotation that is needed to train this high-end technology.

Robot Assistants

Imagine a construction worker who is working on top of a scaffolding, but then he realizes that he forgot to bring some tools needed to complete his job. Instead of having to climb down, pick up the tools, and then climb back up, a robot could fetch the tools for him. An AI-powered robot named Atlas uses a claw gripper to pick up a wooden plank and create a bridge for itself onto the scaffold. The robot then grabs a toolbag, climbs up the scaffold, spins around, and tosses the toolbag up to the worker. The wooden box is then propelled off the platform by Atlas, who then flips and twists his way to the ground.

These new capabilities reflect a logical progression in the capabilities of humanoid robots, especially in the areas of perception, manipulation, and autonomy. Atlas' enhanced locomotion and sensing abilities allow it to pick up and move items of various sizes, compositions, and weights while maintaining balance.

Bricklaying Robots

Laying brings a very labor-intensive job, which makes it a good task to offload to a robot. A bricklaying robot was developed to work with a mason in order to reduce costs, increase productivity and increase the quality of work. By lifting the brick, applying mortar, and placing each brick in place, the system removes the laborious work. The mason is in charge of making sure the bricks are placed precisely, clearing away extra mortar, and making sure the job is finished correctly.

An interesting AI bricklaying robot that is still in the development stages but certainly shows a lot of promise, called Hadrian X, can lay 1,000 bricks in an hour and can build an entire house in two days, complete with pathways for electrical and plumbing and spaces for doors and windows.

Monitoring Job Progress

In addition to performing manual labor, robots can help humans monitor the progress of job sites, thereby providing real-time, actionable data. The technology uses autonomous drones and rovers equipped with computer vision cameras to photograph and scan the construction site each day with pinpoint accuracy. Their AI then uses those scans to compare against your BIM models, 3D drawing, schedule, and estimates to inspect the quality of the work performed and to determine how much progress has been made each day.

What Types of Data Annotation are Necessary to Train AI Construction Robots?

If we take a look at the first robot, called Atlas, it needs to identify various objects and navigate in its surroundings. Therefore, data annotation ranging from tagging and 2D/3D bounding boxes to more advanced annotation types, like semantic segmentation, will be necessary. The same is true for the bricklaying robot since it needs to identify where the bricks are located and the proper place they need to go. Since the bricks are rectangular in shape, they will fit neatly into 3D bounding boxes.

The job progress robot requires polygon annotation since the structure will be of irregular shape. A lot of detailed tagging will also need to be done to allow the system to compare the progress with the BIM files.