3D Robotics

Visual obstacle avoidance with Centeye and Crazyflie

Geoff Barrows at Centeye has been working for many years on his small, low-cost vision chips and they're really starting to get good. See the above video for visual obstacle avoidance on the Crazyflie nano-drone platform. Sadly you can't buy these boards (Geoff mostly does custom engineering for defense industry clients), but maybe we can convince him to make them a product? 

More details here

E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones


  • Chris, Everyone- Thank You for posting this. We are deliberately developing this technology for both civil and military use, and so I do envision this hardware finding it's way into, at least, commercial nano drones. The hardware can be made available, but the cost (based on current cost to make) is higher than what would be viable here.

    I do love the work going on in this community. I've been in the "micro drone" space since it's inception 20 years ago. Without a doubt- the recent explosion of the drone space was triggered by the collective work of thousands of hobbyists here. I think it is possible for the same thing can happen for "nano" scale drones as it did for small/micro ones. And so I do like the idea of eventually putting out a less expensive system that people here can hack and innovate with. As a company, though, we are still focused on the current hardware- what you saw in the video is just a small fraction of what we aim to achieve. (Google our contract number and you will find out a bit more.)

    Regarding nano air vehicles- I do believe that one of the next waves for drones will be "nano" scale and smaller. If you look at the trajectory of drones the past 20 years, and if you look at society's expectations, this is almost inevitable. They may have a limited payload and flight duration, but they have all sorts of advantages over larger ones, beyond what was mentioned in the video. One, they are *very* agile. Vijay Kumar discussed this at length in his TED talk, where he showed how the maneuverability of a drone can scale at up to the negative fifth power of linear dimension! This means you can get right up close to an obstacle before you have to evade it. They are also quite safe- I've picked up many a Crazyflie with still-spinning rotors with no cuts to my fingers, and when they crash, well, you just pick it up and fly it again.

    And as you get smaller, the whole world actually increases in size relative to you. So the whole nature of what you have to do and perceive, and what you do NOT have to do and perceive, changes. The problem is still one of "autonomy" and "obstacle avoidance", but it is different from that of larger drones, and certainly different from that of full-size aircraft. And when I speak about "minimalism" in design, what I am referring to is first eliminating the task of doing what you do not need to do, so that you are freer to do what you need to do. (Yes, I devoured Richard Koch's 80/20 book back in 2000!)

    If you are curious about the above, and are willing to listen to me for 35 minutes, I gave a high level talk on this to Simon Levy's class to the Venture Club at Washington and Lee University Venture Club. https://www.youtube.com/watch?v=duKz0WdfYcQ It doesn't talk about the current system, but it does talk about the lessons I learned in past work.

  • MR60

    Wow. Smaller ships still providing this high level functionality is the obvious future as it solves all regulations constraints, reduces to practically zero all safety hazards to people and property, reduces costs to bare minimum. 

  • Hi Chris,

    I've watched Centeye over the last few years with interest, tyheir concept of preprocessed image data always intrigued me.

    It permits a minimalist computer system to actually make some sense of the world around it and respond accordingly.

    Clearly it has severe limits as compared with more heavy duty systems such as 3D laser scanning and high compute power requiring conventional stereoscopic vision solutions, but there are a lot of applications where this more minimalist approach might be satisfactory and even excel.

    Tiny quadcopters and Robobugs seem an excellent vehicle for them.

    It is a matter of getting the pre-processed video to do what is appropriate for your vehicle/application and the environment it works in and what it is supposed to be doing in relation to it.

    I suspect that more specialized firmware can be adapted to specific applications / tasks.

    It reminds me of Stiquitos - If you know what those are that ages both of us.

    The were a minimalist robotic (sort of) bug made with as few parts (and as little logic) as possible, that still often seemed to do semi-intelligent (or at least interesting) things.

    If they can make this board available in a form and with firmware that will actually allow us to play around with it, I will definitely buy one.



  • For years now, on this site and others, I have been expressing my belief in micro UAS. The technology being packed into smaller, and smaller (and cheaper) packages is astounding. The one area where there seems to be room for significant improvement to be seen is the power system. The most complex of systems are being miniaturized and implemented in pocket carried nano quads used for applications as important as search and rescue, and as mundane as aerial selfies. The limit, overall, is flight time. I think a bio-battery could be a major step in the right direction. Insects, nature's very own nano-drones (male worker bees are even called drones), are not only capable of refueling themselves, but can even manufacture newer, smaller versions of themselves.
This reply was deleted.