• Chad Green

Working with DARPA

Yesterday was a landmark day. I submitted my first proposal as the principal investigator and prime contractor to DARPA (Defense Advanced Research Projects Agency). Who's DARPA? They're the arm of the Defense Department that reaches out to private industry to solve the hardest problems in science and engineering. Yes, you can thank them for the internet (not Al Gore) and Google Maps (not Google) and Siri (not Apple). Working with DARPA is exciting...and nerve-wracking...and heart-breaking at times, but always interesting.

Remember that time in 2004 when a dozen or so nerdy engineers built robot cars to cross the desert without a driver and none of them made it? That was the DARPA Grand Challenge. The deal was, "Bring your best ideas. If you win, you get a million dollars." I have some good friends that competed in that challenge. The ink on my Bachelors Degree was still drying at the time, but out of that crazy horse race came autonomous vehicles. In fact, one of my favorite technologies, lidar, went from the thing that police use to write tickets, to the thing that lets cars "see". One guy, a loud speaker specialist (maybe my second favorite technology from way back in my car audio days) lined up 64 lasers and spun them around to give his autonomous vehicle a better view of its surroundings. That guy, David Hall, President and CEO of Velodyne LiDAR, was just awarded 2018 inventor of the year and rightly so because today his lidar sensor is used by almost everybody in the autonomous vehicle market.

I can't tell you how much fun it was working at my previous place of employment, Artis, LLC, and I'm really proud of what they've accomplished with Active Protection technologies. I made life-long friends on random test ranges, in bunkers, or in the middle of the desert while testing DARPA technologies. We blew stuff up and put it back together time and time again until we got it right. The best part is, success on DARPA projects is not scored (usually) by pass or fail. Success is scored by what you learn and whether you've pushed the state-of-the-art a little further forward. If you did, that's a win!

So what's my proposal about? Well, I love computer vision. I love remote sensing, and I love what you can do with Convolutional Neural Networks (the undergirding structural steel to all the Artificial Intelligence hype and buzz these days). DARPA wants to be able to detect and pursue non-cooperative drones in the sky--those terrifying ones that show up unexpectedly at public venues or harass you at the beach. However, they want to do it without any active sensors, i.e., using mostly standard cameras. The project is called Visual Relative Navigation and it's meant to be mounted on a drone to help it chase or stay in formation with those other drones:

I was fortunate to have the help and partnership of friends like the University of Maryland Robotics Center; and a Northern Virginia AI and aviation company, Mosaic ATM; and even some help from a smart, promising lidar startup, Blackmore (we're going to try to sneak a lidar in there. We have to. Everything is better with lasers and Blackmore has the coolest one, IMHO. "Excuse me sir, would you like instantaneous velocity with your lidar pulse? Pfft, yes!! I'll take 1.2 million per second!!"). Also, huge thanks to my partners-in-crime Colin Fischer, a flight instructor from the Air Force, and Ted Markson, a man who builds his own UAVs to deliver medicine to remote locations in the Outback.

So, what we're proposing is a really cool, airframe-agnostic, setup with a bunch of stereo cameras and lots of Nvidia Jetson TX2 GPU firepower to compute the snot out of those drone pixels. Hopefully, the work I've done building my little company up with AI in everything will count for something and we'll be able to meet or exceed human pilot performance in chasing down pesky drones. I'm excited.

If you'd like to know more about working with DARPA, please feel free to reach out!

343 views0 comments

Recent Posts

See All