Drones trained to tackle obstacle course at high speeds using virtual racetrack
Image credit: MIT
Drones have been trained to find the fastest route around an obstacle course without crashing, thanks to an algorithm developed by MIT engineers.
The algorithm was developed using simulations of a drone flying through a virtual obstacle course alongside data taken from experiments of a real drone flying through the same course in a physical space.
The researchers found that a drone trained with their algorithm flew through a simple obstacle course up to 20 per cent faster than a drone trained on conventional planning algorithms.
The new algorithm did not always keep a drone ahead of its competitor throughout the course, as it sometimes opted to conserve energy by slowing down only to speed up and ultimately overtake its rival by the end.
“At high speeds, there are intricate aerodynamics that are hard to simulate, so we use experiments in the real world to fill in those black holes to find, for instance, that it might be better to slow down first to be faster later,” said researcher Ezra Tal. “It’s this holistic approach we use to see how we can make a trajectory overall as fast as possible.”
Training drones to fly around obstacles is relatively straightforward if they are meant to fly slowly, the researchers said. That’s because aerodynamics such as drag don’t generally come into play at low speeds and can therefore be left out of any modelling of a drone’s behaviour.
At high speeds, a whole host of additional physics need to be accounted for, meaning how the vehicles will handle is much harder to predict.
The researchers developed a high-speed flight-planning algorithm that combined simulations and real-world trials in a way that minimises the number of experiments required to identify fast and safe flight paths.
They started with a physics-based flight planning model, which they developed to first simulate how a drone is likely to behave while flying through a virtual obstacle course. They simulated thousands of racing scenarios, each with a different flight path and speed pattern.
They then charted whether each scenario was feasible or resulted in a crash and zeroed in on a handful of the most promising scenarios to try out in the lab.
To demonstrate the approach, a simulation of a drone flying through a simple course with five large, square-shaped obstacles arranged in a staggered configuration was created. This set-up was then replicated in a physical training space and a drone was programmed to fly through the course at speeds and trajectories that had been previously picked out from the simulations.
The researchers also ran the same course with a drone trained on a more conventional algorithm that does not incorporate experiments into its planning.
Overall, the drone trained on the new algorithm 'won' every race, completing the course in a shorter time than the conventionally trained drone.
In some scenarios, the winning drone finished the course 20 per cent faster than its competitor, even though it took a trajectory with a slower start. This kind of subtle adjustment was not taken by the conventionally trained drone, likely because its trajectories, based solely on simulations, could not entirely account for aerodynamic effects that the team’s experiments revealed in the real world.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.