Scientists have designed a small drone with a “bird-like” ability to fly quickly and safely through unfamiliar environments.
The Hong Kong University team say their new micro air vehicle, dubbed SUPER, could assist first responders during disaster relief efforts or search and rescue missions into unknown areas.
“Unfamiliar environments can present many dangers to a drone, which typically don’t survive impacts with solid objects,” says David Howard, Robotic Design and Interaction Group Leader at CSIRO, who not involved in the study. He tells Cosmos, “The challenge is designing a drone that can navigate an unfamiliar environment quickly.”
While examples of rapidly manoeuvring UAVs exist, most rely on high-speed cameras, which collect high quality colour information but have restricted sensing ranges.
To address this gap, the SUPER team gave it a lightweight 3D light detection and ranging (LiDAR) sensor, which has a 70-meter sensing radius.
“LiDAR natively builds 3D maps, which is great as you get a lot of geometric – and importantly, depth – information which can be invaluable for navigation,” says Howard.
SUPER uses LiDAR to build detailed maps of its surroundings. Credit: The University of Hong Kong.
SUPER also carries an onboard computer that takes data from the LiDAR sensor to inform a novel “two-trajectory strategy” that plots a flight path in real time.
The new strategy uses the LiDAR map to calculate a “safe” route through known free spaces and a faster, “exploratory” route through unknown spaces. By switching between these two trajectories, SUPER can plan and execute high-speed, collision-free flights in real-world scenarios.
The researchers found that their flight planning framework reduces failure rates by over 35 times while halving planning time, which allows for faster flight.
SUPER can detect and avoid very thin wires. Credit: The University of Hong Kong.
In a series of tests, the researchers showed that SUPER could navigate an unfamiliar forest in both high- and low-light settings, could detect and avoid thin wires, and could even track a jogger through a densely wooded area. In some of these tests, the drone reached speeds of 20 meters per second (72 kph or 44 mph).
“SUPER represents a milestone in autonomous MAV [micro air vehicle] systems, bridging the gap from laboratory research to real-world applications,” writes first author Yunfan Ren and colleagues.
“Bird-like drones have received a good amount of research attention recently,” observes Howard. “They are attractive because they allow drones to act like birds, engaging in energy-efficient gliding, or dexterously performing aerobatics… even perching on a branch to gather sensor data or recharge battery through solar panels.
“The use of bioinspiration in the design opens up a huge range of possibilities and potential applications for these systems.”
The research is published in the journal, Science Robotics.
