Engineering news
Drones are hard to beat for exploring complex and unknown environments. They are fast, agile and small, and can carry sensors and payloads virtually anywhere. Autonomous drones struggle to navigate through unknown environments without a map, however, so human pilots are needed to realise their full potential.
A team of researchers from the University of Zurich in Switzerland set out to tackle this issue with a new study.
"To master autonomous agile flight, you need to understand the environment in a split second to fly the drone along collision-free paths,” said research team leader Davide Scaramuzza. “This is very difficult both for humans and for machines. Expert human pilots can reach this level after years of perseverance and training, but machines still struggle.”
Scaramuzza and his team trained an autonomous quadrotor to fly through previously unseen environments such as forests, buildings, ruins and trains, keeping speeds of up to 40 km/h without crashing into trees, walls or other obstacles. All this was achieved relying only on the quadrotor’s on-board cameras and computation, a research announcement said.
The drone’s neural network learned to fly by watching a 'simulated expert’ – an algorithm that flew a computer-generated drone through a simulated environment full of complex obstacles. At all times, the algorithm had complete information on the state of the quadrotor and readings from its sensors, and could rely on enough time and computational power to always find the best trajectory.
Such a ‘simulated expert’ could not be used outside of simulation, but its data was used to teach the drone’s neural network how to predict the best trajectory based only on data from the sensors.
“This is a considerable advantage over existing systems, which first use sensor data to create a map of the environment and then plan trajectories within the map – two steps that require time and make it impossible to fly at high speeds,” the announcement said.
“While humans require years to train, the AI – leveraging high-performance simulators – can reach comparable navigation abilities much faster, basically overnight,” said co-author Antonio Loquercio.
Fellow co-author Elia Kaufmann added: “Interestingly, these simulators do not need to be an exact replica of the real world. If using the right approach, even simplistic simulators are sufficient."
Potential applications are not limited to quadrotors, the researchers said. The same approach could help improve the performance of autonomous cars, or even open the door to a new way of training AI systems for operations where collecting data is difficult or impossible – on other planets, for example.
According to the researchers, the next steps will be to make the drone improve from experience, as well as to develop faster sensors that can provide more information about the environment in a smaller amount of time, allowing drones to fly safely even at speeds above 40 km/h.
The research was published in Science Robotics.
Join IMechE and Professional Engineering at THE virtual fair for early engineering careers and find your perfect job! Register for EngRec 2021 FREE today.
Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.