A new quadcopter drone can learn to pull off acrobatic maneuvers, researchers report.

While a power loop or a barrel role might not be needed in conventional drone operations, a drone capable of performing these kinds of maneuvers is likely to be much more efficient.

The maneuvers can push the drone to its physical limits, help make full use of its agility and speed, and allow it cover more distance within its battery life.

Since the dawn of flight, pilots have used acrobatic maneuvers to test the limits of their airplanes. The same goes for flying drones: Professional pilots often gauge the limits of their drones and measure their level of mastery by flying such maneuvers in competitions

Now, the researchers have developed a navigation algorithm that enables drones to autonomously perform various maneuvers—using nothing more than onboard sensor measurements.

To demonstrate the efficiency of their algorithm, the researchers flew maneuvers such as a power loop, a barrel roll, or a matty flip, during which the drone is subject to very high thrust and extreme angular acceleration.

“This navigation is another step towards integrating autonomous drones in our daily lives,” says Davide Scaramuzza, robotics professor and head of the robotics and perception group at the University of Zurich.

At the core of the novel algorithm lies an artificial neural network that combines input from the onboard camera and sensors and translates this information directly into control commands. The neural network is trained exclusively through simulated acrobatic maneuvers.

This has several advantages: Maneuvers can easily be simulated through reference trajectories and doon’t require expensive demonstrations by a human pilot. Training can scale to a large number of diverse maneuvers and does not pose any physical risk to the quadcopter.

Only a few hours of simulation training are enough and the quadcopter is ready for use, without requiring additional fine-tuning using real data. The algorithm uses abstraction of the sensory input from the simulations and transfers it to the physical world. “Our algorithm learns how to perform acrobatic maneuvers that are challenging even for the best human pilots,” says Scaramuzza.

However, the researchers acknowledge that human pilots are still better than autonomous drones. “Human pilots can quickly process unexpected situations and changes in the surroundings, and are faster to adjust,” says Scaramuzza.

Nevertheless, the robotics professor is convinced that drones used for search and rescue missions or for delivery services will benefit from being able to cover long distances quickly and efficiently.

A paper on the work will appear in Robotics: Science and Systems. Intel also contributed to the work.

Source: University of Zurich

Related

‘Dodgeball’ drone flies fast and avoids incoming stuff

Sharp-edged wings cut turbulence for little drones

Watch a roofing drone fire its nail gun

A new aerial vehicle called an “octocopter” can attach asphalt shingles to a roof without a human at the controls.

Engineers demonstrated that the drone can autonomously position a nail gun on a nailing point, place the nail, and then move on to the next point.

“For me, the biggest excitement of this work is in recognizing that autonomous, useful, physical interaction and construction tasks are possible with drones,” says Ella Atkins, a professor of aerospace engineering and robotics at the University of Michigan.

Is roofing a robot job?

Atkins says many consider tasks best suited to robotization “dull, dirty, and dangerous,” and robots like octocopter could offer a way to move the human workforce to cleaner, safer, and more interesting jobs.

Already, drones spare humans some high-stakes fall risks by inspecting bridges, wind turbines, and cell towers. The natural next step, Atkins says, is to upgrade from surveillance alone to performing physical tasks.

The problem of nailing down a shingle breaks down into several smaller problems—including telling the octocopter where the nails should go and triggering the nail gun.

Atkins’ team used a system of markers and stationary cameras to enable the octocopter to precisely locate itself in space and then to tell the octocopter where the nails should go.

To fire the nail gun, they first measured the force needed to compress the point of the nail gun, a necessary task before a nail will deploy. Then, they wrote software that to enable the octocopter to apply that force.

The off-the-shelf version of the electric nail gun requires the compressing of a trigger as well, but the team turned that into a virtual switch which activated when the octocopter was in position to place a nail.

Slow but learning

For now, the drone doesn’t move as fast as human roofers.

“Initially, we tried using faster approach speeds to minimize nailing time,” says Matthew Romano, a robotics PhD student and first author of a paper submitted to the International Conference on Robotics and Automation.

“However, for those attempts, the nail gun tip often bounced off the roof, which meant it either wouldn’t trigger or it would trigger in the wrong place.”

“A novice roofer—who’s never climbed on a roof, who’s never used a nail gun—they start out slow. That learning process, the evolution from them being a complete novice to being successful, is something that we’ll need to see in this system as well,” Atkins says.

In addition to speed, the team identified other improvements needed for a practical system. First, a tether should power it, not a battery. Because both batteries and nail guns are heavy, the system can only run for a little more than 10 minutes at a time. A tether would allow it to run indefinitely. Also an air line running alongside the power cable could make the nail gun a more effective pneumatic model.

Finally, a system of cameras and markers is more complicated than a roofing drone would actually need. Shingles are marked with a shiny adhesive strip, in addition to the color difference between the exposed surface and the portion that lies beneath the next layer of shingles.

“It would be pretty easy to have a camera system mounted on the octocopter that understands both the orientation of the shingle and its position,” Atkins says.

The National Science Foundation funded the work.

Source: University of Michigan

Related

Math paves the way for fleets of connected drones

Robot made with smaller bots has a funky way of getting around

These robot arms can fold laundry