Image: Public Domain
MIT Researchers Develop Algorithm to Prevent Mid-Air Collision of Drones
DRONELIFE Staff Writer Ian M. Crosby
In 2020, MIT researchers unveiled MADDER, a system designed to prevent collisions between drones occupying the same airspace.
Continue reading below or listen to:
A multi-agent trajectory planner allows groups of drones to form collision-avoiding trajectories. Each drone broadcasts its trajectory and considers the trajectories of fellow drones when creating its course.
MADER is an asynchronous distributed multi-agent trajectory planner. Each drone generates its own trajectory. Each agent must agree to all new trajectories, but not at the same time. This approach makes MADER more scalable than alternative solutions. This is because it is very difficult to match a large number of drones to their trajectories at the same time.
However, when we tested the system on real drones, we found that drones that lacked up-to-date information about their partners’ trajectories could cause collisions. This prompted the researcher to develop his updated Robust MADER. This is a multi-agent trajectory planner that formulates collision-free trajectories even if communication between drones is delayed.
“MADER worked well in simulations, but was not tested in hardware. So we built a lot of drones and started flying them,” says Kondo, a graduate student in aerospace engineering. Houta said. “Drones have to communicate with each other to share their trajectory, but once they start flying, there is always a delay in communication, and you quickly realize that something is going wrong.”
The algorithm of this new system introduces a delay check step where the drone waits a specified amount of time before following a new trajectory. If additional orbit information is received during the delay period, it may abandon the planned orbit and start over if necessary. According to Kondo, the length of the delay check period depends on the distance between the drones and environmental factors that can interfere with communication.
Robust MADER achieved a 100% success rate in creating collision-free trajectories in both simulations and real drones. This new system slowed down travel times slightly, but was the only way to guarantee safety.
“If you want to fly safer, you need to be careful, so if you don’t want to hit an obstacle, it’s only natural that it will take some time to reach your destination. If you hit something, no matter how fast you run, It’s not a big deal because you can’t reach your destination,” says Kondo.
Written by Kondo paper With postdoc Jesus Tordesillas. Graduate student Parker C. Rusk. Reinaldo Figueroa, Juan Rached, Joseph Merkel, MIT undergraduate. Senior author Jonathan P. Howe, Richard C. McLaughlin Professor of Aerospace Engineering, Principal Investigator at the Information Decision Systems Laboratory (LIDS), Member of the MIT-IBM Watson AI Lab. Their work will be presented at an international conference on robotics and automation.
To test this new solution, the research team ran hundreds of simulations that artificially introduced communication delays. Robust MADER was 100% successful in formulating collision-free trajectories in all these simulations, but conversely, all tests completed using its predecessor produced collisions.
The researchers also built six drones and two aerial obstacles to test Robust MADER in a multi-agent flight environment. As a result of these tests, in this environment he could have had a total of 7 collisions using the original version of MADER, whereas in Robust MADER, any hardware experiment had him crashing only once. did not occur.
“Until you actually fly the hardware, you never know what could cause problems. , It was very rewarding to be able to actually confirm it,” said Kondo.
Drones powered by Robust MADER were able to fly at 3.4 meters per second, albeit with slightly longer average travel times than some baselines. However, Robust MADER was the only method that was completely collision-free throughout all the experiments.
Going forward, Kondo and his collaborators plan to test Robust MADER in outdoor environments where various obstacles and noise sources can disrupt communication. The researchers also hope to equip the drone with visual sensors to detect other agents and obstacles, predict their movements, and incorporate that information into trajectory optimization.
read more:
Ian attended Dominican University in California and graduated with a BA in English in 2019. With his lifelong passion for writing and storytelling and keen interest in technology, he now contributes to his DroneLife as a staff writer.
Miriam McNabb, editor-in-chief of DRONELIFE and CEO of professional drone services marketplace JobForDrones, is a fascinating observer of the emerging drone industry and drone regulatory environment. With her 3,000+ articles focused on the commercial drone space, Miriam is an international speaker and recognized figure in the industry. Miriam has a degree from the University of Chicago and high tech she has over 20 years of experience in sales and marketing new technologies.
For drone industry consulting or writing, please email Miriam.
twitter:@spaldingbarker
Subscribe to Drone Life here.
