Could Quadrotor Robots Construct Buildings? (video)
What works in teams, can build any tower you design, and makes a horrifically angry buzzing sound? No, not super intelligent construction wasps. Quadrotors. UPenn's GRASP Lab has been experimenting with these autonomous aerial drones to make them perform amazing feats. We've seen them dart through windows, spy on people, and even haul lumber. Now, GRASP has gotten them to work together and build simple structures. In the video below you can watch the quadrotors use magnetically interlocking struts to autonomously build towers individually or in concert. It makes me think of bees constructing a hive. They sound like it, too. While this work is still very preliminary it alludes to ways in which aerial drones could revolutionize disaster relief, construction, and exploration.
The drones in the following video are operating without direct human control. These aren't RC vehicles. Instead, humans provide the design for the tower the quadrotors should build. Using well defined algorithms, the drones locate construction materials, grasp them in the correct orientation, and bring them to the construction site. A wireless network, and a combination of centralized and decentralized processing, allows teams of drones to communicate and build the tower as efficiently as possible. Make sure to watch at the end to see the sped-up video of several different structures being constructed. I really wish I had one of these quadrotor crews when I was in my fort-building phase.
This GRASP quadrotor project is the work of Daniel Mellinger, a PhD candidate working with Vijay Kumar. The drones themselves are purchased from Ascending Technologies, and are actually fairly standard devices. What's really exciting is Mellinger's ability to manage these complex dynamic systems. If a robot on the ground stalls or experiences a bug in its programming, no big deal, it just shuts down and waits. But a flying drone has to be controlled all the time, otherwise it will plummet to its doom. Special care has to be taken to ensure that the weight and orientation of construction pieces don't cause errors in the drone's flight. Drones also have to recognize when each piece is properly grasped and placed.
To that end, Mellinger does use a bit of a cheat. His quadrotors are tracked by an external camera system that allows them to be positioned in the frame of reference of the construction set. Red LEDS and small reflective white spheres allow the Vicon cameras to locate the drones (and construction struts) in 3D space. Besides sensors on board the drones that are used for IMU, the camera system is the main source of guidance for the drones. That's not really a practical solution if you wanted to use these drones outside the lab. As we've seen before, however, GRASP is working on quadrotors that could navigate in real world environments without external camera systems.
Good thing, too. According to comments made by Mellinger and Kumar on the Robots Podcast, the ultimate goal for many of these projects is to get quadrotors to function as helpful assistants in many dangerous situations. In disaster relief they could function as mobile eyes and ears, especially in areas that are inaccessible to workers on foot. Judging by this most recent video, quadrotors may also be used to help construct temporary structures for victims - even if aid workers cannot access them directly. In larger swarms they might be able to move heavy loads or be used in even more complex construction projects.
There's still a long way to go before we get there, however. A drone with a full complement of onboard sensors has a battery life around 8 minutes. That's a pittance compared to what would be needed in real world applications. Also, Mellinger seems to be focused on small groups of drones working together. Dreams of huge swarms of quadrotors building your home in an afternoon are completely impractical at the moment- the coordination is still beyond us.
There's no denying, however, that quadrotors are a fascinating field of robotics. These autonomous drones, fully developed, have all sorts of advantages (mainly mobility) over their terrestrial cousins. Mellinger has already won an award for his work with his hive of minions, and it should be very interesting to see what he comes up with next. Maybe the quadrotors could take up Jenga?*
*I'll leave you with Stephen Colbert's take on Mellinger's work. Skip to around 3:05 in the following video: