Carnegie Mellon’s AI Drones Can Build Mid-Air Structures With 90 Percent Success Rate

Carnegie Mellon University engineers have integrated drones, 3D printing, and artificial intelligence to create flying robots, which can construct buildings in remote or dangerous locations. The drones that are controlled through AI are designed to use magnetized blocks rather than extruded plastic so that they can be assembled in a sturdy manner even during flight. A huge language model (LLM) interprets straightforward instructions into construction plans and corrects them in real time, and succeeds in constructing them 90 percent of the time.

Drone-Based 3D Printing System

According to the paper, in lab tests, small quadcopter drones carried and placed tiny magnetized blocks to form structures in mid-air. This “aerial additive manufacturing” sidesteps the stability problems of extrusion 3D printing. The blocks snap together like LEGO bricks, so the drone does not have to hold a steady position. An onboard AI (an LLM) acts like an architect, interpreting commands (for example, “build a bridge”) into step-by-step assembly instructions. A camera observes progress; if a block is misplaced, the AI recalculates a new plan on the fly rather than restarting. This closed-loop approach achieved 90% success in building test shapes.

Applications and Future Outlook

Researchers envision numerous applications other than lab tests. In one instance, Farimani proposes that someday the potholes will be filled with drones, spacecraft in orbit will be repaired, and that the pipes would be built in remote mountains where heavy machinery is unable to access. The team will take their system outside and test it against the real-life conditions and more intricate 3D designs and new materials. A combination of AI planning and aerial assembly would be able to put swarms of smart drones to work on infrastructure repairs and create emergency shelters in disaster areas.