Micro air vehicles could one day make a significant contribution to search-and-rescue operations after disasters such as earthquakes or tsunamis. It’s easy to imagine quadcopters assessing buildings, entering through cracked walls, and flying through collapsed spaces to find people who are trapped.
But if these vehicles are ever to manage this task, they will need to navigate autonomously through narrow gaps at speed and at a wide variety of angular accelerations, twisting and turning as they fly to squeeze through the space available.
That’s easier said than done. Indeed, no drone has been able to do this without significant external processing power to help (see “Daredevil Drone Files through the Trees Like an Ace”).
Today that changes thanks to the work of Davide Falanga and pals at the University of Zurich in Switzerland. These guys have developed an autonomous drone that can fly quickly through narrow gaps using little more than the data from a forward-facing camera and some clever on board processing.
The team created a rectangle marked with a thick black edge to ensure that the drone can see it. They then suspend this rectangle in the middle of a room and direct the drone to fly through it under its own steam.
The drone is fitted with a forward-facing fisheye camera, which it uses to sense the gap. To simplify the task, the drone knows the size of the rectangle and needs only to calculate the required trajectory.
This is still a challenging task. The on-board processor performs the trajectory calculation in two stages. It first calculates how the drone should fly through the gap and the particular twist, yaw, or roll it has to perform to traverse the gap. It does his by maximizing the drone’s distance from the edges of the rectangle to avoid a collision.
Having decided on this traverse trajectory, the onboard processor then calculates an approach that brings the drone to the point where it can initiate the traverse trajectory.
The approach trajectory has some additional constraints. For example, this trajectory must keep the rectangle within the field of view of the camera at all times. The drone needs to see the gap so that it can determine its location.
And the processor must continually recalculate the trajectory while ensuring that any required adjustments are within the drone’s aerodynamic capabilities. The processor is capable of designing and testing 40,000 trajectories a second.
One reason the trajectory must be treated in two parts is that the drone cannot see the rectangle during the traverse. So it must perform this maneuver blind, something that is possible because this part of the flight is so short. “The trajectory is generated so as to minimize the risk of collision and, due to its short duration, does not require any visual feedback, which is not available during the traverse,” say Falanga and co.
After passing through the gap, the quadcopter must recover its attitude and hover. For this, it is fitted with a distance sensor and a downward-facing camera which it uses only for this task.
The team tested this approach using a quadrotor with dimensions of 55 by 12 centimeters and a weight of 830 grams. The quadcopter is adapted so that the motors are tilted by 15 degrees. This provides three times more yaw control but loses only 3 percent of the collective thrust.
The rectangular gap measured 80 by 28 centimeters, and the team flew 35 missions through it at speeds of up to three meters per second, requiring a roll angle of up to 45 degrees and a pitch angle of up to 30 degrees.
The results make for impressive reading and can be seen here. The team considers a flight a success if the quadcopter passes through the gap without a collision and then brings itself to a hover afterward. “We achieved a remarkable success rate of 80 percent,” they say. “To the best of our knowledge, this is the first work that addresses and successfully reports aggressive flight through narrow gaps.”
Ref: arxiv.org/abs/1612.00291: Aggressive Quadrotor Flight through Narrow Gaps with Onboard Sensing and Computing
Meta has built a massive new language AI—and it’s giving it away for free
Facebook’s parent company is inviting researchers to pore over and pick apart the flaws in its version of GPT-3
The gene-edited pig heart given to a dying patient was infected with a pig virus
The first transplant of a genetically-modified pig heart into a human may have ended prematurely because of a well-known—and avoidable—risk.
Saudi Arabia plans to spend $1 billion a year discovering treatments to slow aging
The oil kingdom fears that its population is aging at an accelerated rate and hopes to test drugs to reverse the problem. First up might be the diabetes drug metformin.
Yann LeCun has a bold new vision for the future of AI
One of the godfathers of deep learning pulls together old ideas to sketch out a fresh path for AI, but raises as many questions as he answers.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.