Drone enthusiasts may be disappointed to learn a new algorithm has been shown to beat human pilots in finding the quickest trajectory to guide a quadrotor — a drone with four propellers — through a circuit.
“Our drone beat the fastest lap of two world-class human pilots on an experimental race track,” said professor David Scaramuzza, head of the robotics and perception group at the of University of Zurich and the Rescue Robotics Grand Challenge of the NCCR Robotics, which funded the research.
It’s more than just a title up for grabs – using the most time-efficient path is vital for drones used in critical situations such as search and rescue, building and site inspections, and package deliveries to maximise flight time.
“Autonomous drones have proven to be valuable tools for many applications, from delivering medications or medical supplies (like Zipline in Rwanda and Ghana, and the Swiss Post in Switzerland), to search and rescue support (like Sensefly and Werobotics in Latin America, and Fotokite in the USA),” Scaramuzza told Information Age.
Human vs algorithm
The researchers had the algorithm and two human pilots fly the same quadrotor through a race circuit, using external cameras to precisely capture the motion of the drones and – in the case of the autonomous drone – to give real-time information to the algorithm on where the drone was at any moment.
In the face off between algorithm and human pilots, the humans were given the opportunity to train on the circuit before the race to ensure fairness.
But the algorithm won: all its laps were faster than the human ones, and the performance was more consistent.
This is not surprising, because once the algorithm has found the best trajectory, it can reproduce it faithfully many times, unlike human pilots who cannot replicate the identical flight path.
The algorithm uses a series of waypoints, such as windows, rooms or specific locations to inspect, adopting the best trajectory and the right acceleration or deceleration at each segment.
“This is the first algorithm that computes the fastest trajectory by taking all quadrotor limitations into account. Previous attempts relied on simplifications of the flight trajectory or of the quadrotor model,” Scaramuzza explained.
It can even help in drone racing, which consists of passing through a sequence of gates in the least possible time.
“The path is even faster than that planned by world-class drone racing pilots, although the applications are way beyond drone racing,” he said.
Overcoming battery limitations
The algorithm was needed not just to beat humans, but because of limitations in battery technology.
“There is a key limitation to widespread drone applications today, which is battery life, typically limited to 20 to 30 minutes for drones like quadrotors,” Scaramuzza said.
Battery life is unlikely to improve much in the next decade as current batteries rely on lithium-ion polymers, while higher energy density batteries based on lithium-sulfur or even fuel cells are still not commercially deployable.
To get around this problem, one solution is to fly faster and so Scaramuzza and the research team set about addressing the problem from another way through the algorithm to improve how fast drones can complete their flight missions.
The next step is commercial application but before this can get the green light, the algorithm will need to become less computationally demanding, as the researchers admit that as it is now, it takes up to an hour for the computer to calculate the most time-optimal trajectory for the drone.
“We are making the code available to companies interested in using it,” Scaramuzza explained to Information Age, noting that the time needed to compute the fastest route is unlikely to hold it back in commercial operations.
“All flight paths must be planned beforehand since they need an authorisation by the civil aviation authority, which could take between a few hours and several days, depending on the country’s regulations,” he said.