Using UAVs To Find Mannequins Lost In The Outback

Every other year the Australian Research Center for Aerospace Automation, the government of Queensland, Australia  and other government and research institutions hold a contest to develop technology for unmanned aerial vehicles for the wastes of central Australia. Canberra UAV – a group of autonomous drone enthusiasts from the Make, Hack, Void hackerspace – took part in this UAV challenge this year. They ended up with one of the most successful UAVs to every compete, and while they didn’t quite finish the competition they were one of the most successful entries to date.

The goal of the Search and Rescue Outback Challenge is to take off from a landing strip, search for a mannequin named Outback Joe, and deliver 500 ml of life-giving water via air drop. Out of 72 teams entered in to this year’s Outback Challenge, only 6 were allowed to take off – safety restrictions, don’t you know – and all but two hadn’t been destroyed via ‘rapid, unplanned descent’ during qualifications.

UAV Canberra was the only team able to search the entire 5 nautical mile radius search area with their cameras and find Outback Joe. Contest judges gave them permission to drop their payload, but unfortunately the bottle of water was snagged on the engine.

In the many iterations of the Search and Rescue Outback Challenge, UAV Canberra is only the second team to locate a mannequin in the outback, and the first to do it autonomously. A shame, then, that they were unable to claim a victory, but we’ll look forward to their entry in the next Search and Rescue competition.

You can check out a few videos of the Canberra UAV team’s flights after the break.

[youtube=https://www.youtube.com/watch?v=9_WTa6y-P74&w=470]

[youtube=https://www.youtube.com/watch?v=DucMZswn8rM&w=470]

40 thoughts on “Using UAVs To Find Mannequins Lost In The Outback

  1. And now think about how valuable this technology would be to humanity if it would drop a grenade instead of water. Then imagine a rogue government building a couple of thousand of these things and sending them your way.

    It is a great accomplishment from a technological standpoint, but please, ALWAYS think about what this could be used for.

    1. I’m sure most people understand the problems with using drones for military purposes, and considering most people who use them question their effectiveness, this probably is not much of a deal.

      1. The first thing I was thinking when I read the headline was: “Oh, that’s certainly for the US border police.”

        That said, I don’t believe in not developing technology that can be misused for military uses.

        I believe in having social control on your government; and developing the appropriate passive-action counter-technology that makes the military uses costly and unreliable. (as if drones weren’t already…)

    2. What? your gibberish can be summed up in car analogy – Henry Ford says “Now, I am become Death, the destroyer of worlds” while selling first car to a potential drunk driver.

    3. I could strangle someone with a coat hanger, crash plane into a city, bash someones head in with a baseball bat, drown someone in a pool, use bleech and amonia to kill people, should we stop making all of that too?
      Your argument is invalid.

    4. To clarify:
      I think it is a really good idea to create robots to help people and I exaggerated a bit on the given example.
      I do not think that we could ever stop people abusing technology, but we, as in “we that create this technology”, should always keep in the back of our heads, what the technology we create could be abused for and to which people we give this technology.
      There is a reason we don’t let drunk people drive cars and the same counts for this.

      1. I nominate you as supreme overseer of the committee to ensure only qualified individuals have access to technology. I further propose you start by screening the world’s population for responsible use of fire and the wheel.

    5. Problem with people like you is that you think about new technology not as a good thing, but as a weapon. as with the hammer you can do harm, as with the regular vehicles that are used for military purposes and civilian as well, like it or not, but everything may be used to harm somebody else. and only people with mindset like yours are stopping technological progress.

    6. My first thought was of the potential to really help people do what it was designed for. I’m a member of the State Emergency Service (voluntary group in Australia). One of the common tasks we are involved in is searching for missing people in the bush. There’s often a *huge* area that needs covered and/or difficult terrain combined with high temperatures (45°C/113°F or more). If something like this could be combined with ground searches then it could increase the chances of somebody being found in time to be of help.

      As to the “think of what the military could do with this” train of thought; they already *are* doing such things. Projects such as this are trying to put them into beneficial civilian use instead.

      1. @Wholostwhat,
        We’re very interested in making contact with the SES to see if we can work with them on making what we have developed more suitable for real S&R work – for example making the ground station easier to use and more compact, making the airframe suitable for takeoff/landing in very restricted space etc.
        Can you suggest the right person to talk to in the SES about this?
        Cheers, Tridge
        PS: I’m in the white pages for Canberra if you want to give me a call

  2. While this is cool, using a mannequin really takes away from the ability to develop your technology to sense biological organisms. You could use electric sensors, heat sensors, or sound sensors to find people lost in the wilderness, but not a mannequin; thus the technology isn’t perfectly faced toward humans.

    1. Dan: if you look into the Outback Challenge, you’ll see that it was designed to mimic a human signature. There’s an IR lamp in the mannequin, and the winning team spotted its IR signal.

      1. Ding ding ding! Uncontrolled flight into terrain means you lost control and crashed. Controlled flight into terrain means you flew into the ground. Both describe a very specific kind of crash and aren’t euphemisms at all.

        By the same token, rapid unplanned descent means it wasn’t a planned maneuver that went wrong (eg diving and failing to pull up), it didn’t disintegrate and rain out of the sky, and it didn’t lose power and gently drift out of control.

        All of these are examples of the precision possible with the English language, and the skill with which readers may miss the point. Though I admit “Controlled Rapid Ascension of Surface Height” owns bones. Sounds like something from the Kerbal Space Program.

  3. If less than 10% of your teams make it past the safety inspection, you didn’t do a very good job of stating the expectations or requirements, now did you?

    Either that or your interpretation is wildly different from theirs.

    1. That’s what I thought. 72 teams, all but 6 disqualified, and 4 of those crashed? Somethin’ ain’t right there.

      Also, the editors need to do another pass on this summary.

    2. While some teams did crash, most of the reduction from 73 down to 4 that flew happened months before, when teams decided to pull out at one of the 3 deliverable stages. The 3 pre-competition deliverables get increasingly difficult, with the 3rd deliverable requiring teams to provide flight logs showing 5 hours of autonomous flight. Only 9 teams got past that deliverable.
      Cheers, Tridge

    1. @rasz,
      The core image recognition code we used is here:
      https://github.com/tridge/cuav/blob/master/image/scanner.c
      additionally we had a scoring system as a post recognition filter, which is here:
      https://github.com/tridge/cuav/blob/master/lib/cuav_region.py
      It’s really quite a simple algorithm compared to state of the art computer vision systems. The motivation for making it simple was to allow it to run in real time at 7.5 fps on a dual core ARM (a pandaboard) in the plane. The pandaboard was running flat out to keep up with the images coming in.
      We did use some OpenCV library routines in various places, but the main code was written from scratch. We found that OpenCV was just too slow on the CPU we had on the plane to use it for the main recognition task.
      Cheers, Tridge

  4. That water bottle sure looks and is shaped like a missile. And the fact that it’s designed to specifically look for humans and drop it on them on a remote location, while being autonomous. I see what you did there.

    1. There are already a whole lot of drones to choose from that can find and kill people much more reliably than all this.

      I think the competition is probably just as it says.

  5. Very well done guys. +1

    This also has huge potential for fire fighting in the accurate location of firefronts.

    {We call the outback a lot of things, but not “waste” since much of it is actually used, e.g. for grazing.}

  6. Amazing job! Keep up the good work, people are silly to think we’ll attach bombs, seriously, the US government has been doing this with bigger and better drones for a while now – this just shows how an open source community can develop cheap and reliable (although slow progress) drones for a great purpose!

    Finally someone who can think outside of the “bomb” box.

Leave a Reply to Blue Footed BoobyCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.