Tornado-resistant drone fights back against intense winds

Record ID: 1674239

  • Title: Tornado-resistant drone fights back against intense winds
  • Date: 25th May 2022
  • Summary: PASADENA, CALIFORNIA, UNITED STATES (MAY 19, 2022) (REUTERS) DRONE FLYING AGAINST WIND MACHINE WITH FIRST CO-AUTHOR, MIKE O’CONNELL HOLDING UMBRELLA SOUNDBITE (English) MIKE O’CONNELL, CO-FIRST AUTHOR, CALTECH SAYING: “We’re here at Caltech where we’re testing our drone in different wind conditions. Right now, our wind simulator is running up to 50% speed - that’s 13 miles an hour - and we’re using smoke to visualize the airflow.” DRONE HOVERING AGAINST WIND MACHINE DRONE FLYING THROUGH GATES AGAINST WIND SOUNDBITE (English) MIKE O’CONNELL, CO-FIRST AUTHOR, CALTECH SAYING: “Planes right now fly with lots of clearance between each other and these big open runways, and that's not always going to be there. So we need drones that can really precisely do what you tell it to do. So in this case, we’re really precisely trying to follow a path in different wind conditions.” DRONE FLYING THROUGH GATES AGAINST WIND O’CONNELL FLYING DRONE DRONE FLYING THROUGH GATES AGAINST WIND EXERCISE SEEN FROM POINT OF VIEW OF CAMERA MOUNTED ON DRONE AS IT FLIES THROUGH GATES SOUNDBITE (English) GUANYA SHI, CO-FIRST AUTHOR, CALTECH SAYING: “In the future, if we have, like, 3D flying cars flying in the air, it's very important to have precise control because you want to make sure the flying car can track the trajectory very accurately. Because our long term mission is that we want to generalize this technique to not only quadcopters, to flying cars, to any area robots.. such that they can accurately, safely, do complicated tasks.” VARIOUS OF DRONE FLYING THROUGH GATES AGAINST WIND, SEEN FROM ABOVE AND FROM GROUND LEVEL WIND SENSOR SOUNDBITE (English) GUANYA SHI, CO-FIRST AUTHOR, CALTECH SAYING: “We do adaptation super fast. We do it, like, 50 hertz. That means we adapt the model every 20 milliseconds.” VARIOUS OF SHI PREPARING DRONE VARIOUS OF SHI AND O’CONNELL FLYING DRONE THROUGH GATES AGAINST WINDS SOUNDBITE (English) GUANYA SHI, CO-FIRST AUTHOR, CALTECH SAYING: “The goal of our research is to have complicated algorithm on simple hardware. I think this drone is actually very standard and very easy to manufacture. For example, this computer is called the Raspberry Pi. It's super cheap, is like $50, and is very easily accessible. And the motors are also very, you know, easily accessible.” WIND MACHINE AND GATE ‘FLYING AMBULANCE’ BEING TENDED TO BY SHI AND SOON-JO CHUNG SOUNDBITE (English) SOON-JO CHUNG, PROFESSOR OF AEROSPACE AND CONTROL AND DYNAMICAL SYSTEMS, CALTECH SAYING: “So, this is called the autonomous flying ambulance. And thanks to our neural fly AI-based control method, basically we can send this flying ambulance to medevac the injured passengers from the traffic accident. But we can make sure that this drone - autonomous flying ambulance - can land under any weather conditions because of our neural flight method” ‘FLYING AMBULANCE’ BEING TENDED TO BY SHI AND SOON-JO CHUNG SOUNDBITE (English) SOON-JO CHUNG, PROFESSOR OF AEROSPACE AND CONTROL AND DYNAMICAL SYSTEMS, CALTECH SAYING: “We can probably develop the aircraft that can make the air ride more smoother and more comfortable and so on, because it can actually handle these turbulent windy conditions better.”
  • Embargoed: 8th June 2022 09:39
  • Keywords: ARTIFICIAL INTELLIGENCE CALTECH DRONE FLIGHT INNOVATION MEDEVAC SCIENCE TECHNOLOGY WIND
  • Location: PASADENA, CALIFORNIA, USA
  • City: PASADENA, CALIFORNIA, USA
  • Country: USA
  • Topics: Science,United States
  • Reuters ID: LVA001933020052022RP1
  • Aspect Ratio: 16:9
  • Story Text: Researchers at the California Institute of Technology - known as Caltech - have shown off their drone that can fight back against powerful wind gusts which would render most currently-sold equivalents unusable, powered by machine learning, AI and complex algorithms.
    Drones that could potentially fly themselves unaided in tornado-strength winds have the capability to revolutionize everything from air travel to medical rescues, the researchers said.
    “Planes right now fly with lots of clearance between each other and these big open runways, and that's not always going to be there. So we need drones that can can really precisely do what you tell it to do," said Mike O’Connell, co-first author and researcher.
    “Our method is general, right? It can be applied to any robot and it works well for flight or flying vehicles where we have to have really fast response times. It has to be able to, you know.. a sudden gust of wind comes.. we have to respond quickly to it, maintain control," he said.
    The drone only weighs around five pounds (2.3 kg), but intelligently predicts how to react to winds in real time using a deep-learning method developed by Caltech’s engineers called Neural-Fly, recalculating conditions around it five times per second, and adapting its path.
    Put up against the Institute’s Real Weather Wind Tunnel, a custom-made simulator that uses 1,200 fans to create almost gale-force winds, the drone managed to stay upright with little effort and even performed figure of eight movements between tight obstacles without veering off course - as scientists standing by saw their umbrellas turned inside out by the gusts.
    Limitations of the simulator make it hard to power winds higher than 26 miles per hour, but based on the drones reliability at 100% of the RWWT’s power, the team said it is confident the drone will hold its own against some of the harshest conditions out there.
    “In the future, if we have 3D flying cars flying in the air, it's very important to have precise control because you want to make sure the flying car can track the trajectory very accurately. Because our long term mission is that we want to generalize this technique to not only quadcopters, to flying cars, to any area robots, such that they can accurately, safely, could do complicated tasks, we do adaptation super fast. We do it, like, 50 hertz. That means we adapt the model every 20 milliseconds” said co-first author and researcher, Guanya Shi. It took more than two years for the team to develop the system which was delayed by the COVID pandemic, Shi said.
    The drone isn’t custom made and is an adapted quadcopter of the kind commonly used for TV filming. The team removed the cameras, but the components - including the on-board computer powering the machine learning - are off-the-shelf items that can be purchased from many retailers, Shi said.
    Soon-Jo Chung, Professor of Aerospace and Control and Dynamical Systems at Caltech showed off another drone which would be able to run the same software, but looks more like a plane with four rotors under the wings rather than engines that would give it the vertical lift of a helicopter, on the larger body of an aircraft:
    “This is called the autonomous flying ambulance. And thanks to our neural fly AI-based control method, basically we can send this flying ambulance to medevac the injured passengers from the traffic accident. But we can make sure that this drone - autonomous flying ambulance - can land under any weather conditions because of our neural flight method”, he said.
    Nervous flyers could also benefit along the way, with the technology applied to commercial aircraft to help alleviate bumpy rides.
    “We can probably develop the aircraft that can make the air ride more smoother and more comfortable and so on, because it can actually handle these turbulent windy conditions better” Soon-Jo Chung said.
    The technology is still in development but the team is confident they are about halfway to their final goal. It could even be in practical use within a year, Soon-Jo Chung said.
    (Production: Sandra Stojanovic, Phil Lavelle)
  • Copyright Holder: REUTERS
  • Copyright Notice: (c) Copyright Thomson Reuters 2022. Open For Restrictions - http://about.reuters.com/fulllegal.asp
  • Usage Terms/Restrictions: None
Register