Aggressive Maneuvers for Autonomous Quadrotor Flight

The simple matter is that the implications of these drones reaches far beyond simply building things, like showing one aspect of what a hammer can be used to build. The real takeaway is the radical increase in the abilities of flying drones, and the raw potential they leave to be harnessed by future engineers and entrepreneurs. Those who discount their usefulness by staying within the context of the video, and not elaborating outside of that, aren't getting it. I would refer you especially to @burgerbrain's comment and the videos he posted. Specifically: ▶ this video

in “Watch a swarm of flying robotic drones construct a tiny building” · View on HN

I wonder how well the learned policy generalizes to other environments. Places like an art gallery, outside, or a cave. Could the network have learned something fundamental about monocular vision? It would also be interesting to see if the learned policy corrects for perturbations. If we tilt the drone by hitting it, will the policy stabilize it again? While this is a really cool result, I suspect that this approach might not be the best way to control UAVs. Dragon flies are ready to fly, avoid obstacles, perch on stuff, hunt down prey right after warming up their wings for the first time. This implies that a good amount of the flight behavior is 'hard-coded.' Although I really can't wait until someone expands upon this approach. So instead of outputting left or right, the network could output 'stick vectors,' which translate to control stick commands. Maybe even have the network take in some sensor data and a 'move in this direction' vector. Add in a pinch of sufficiently fast video processing and we could probably learn how to do fly through an FPV course or do aggressive maneuvers to fly through someone's windows[0] [0] ▶ this video

in “Drone Uses AI and 11,500 Crashes to Learn How to Fly” · View on HN

I believe this is possible. Just look at these autonomous drones doing stunts: ▶ this video Delivering a packet is much easier than the stuff the are doing in this vid.

in “Amazon to deliver by drone? Don't believe the hype” · View on HN

Can't. I met the guy in person while travelling by train. He had the drone in question in his suitcase. He did tell me however he was involved with this ▶ this video (Or another such effort —I met him in France.)

in “First Victim of Amazon Drones: The Credibility of CBS and 60 Minutes” · View on HN

Projects using quadrotors to do things like fly cooperatively and do acrobatics [1,2] will often use high speed motion capture systems with multiple fixed cameras. This decouples the problem of perceiving the environment from the problem of responding to the environment. More broadly in robotics there are plenty of people working on ground-based robots that don't rely on multiple fixed high speed cameras - self-driving cars [3] would be one example. There are also people working on cooperative ground based robots that don't rely on multiple fixed high speed cameras - for example, the RoboCupSoccer Middle-Size League [4]. There are also people working on quadrotors that don't cooperate/do acrobatics and who don't rely on fixed vision systems - for example people working on autonomous quadrotor mapping [5] and using IMUs and on-robot vision [6]. [1] ▶ this video [2] [link] [3] [link] [4] [link] [5] [link] [6] [link]

in “Throwing and catching an inverted pendulum with quadrocopters” · View on HN

I think you're referring to the grasp lab at UPenn, they did stuff with quad rotors landing on Velcro. ▶ this video Georgia Tech did more with the larger Yamaha Rmax helicopter. [link]

in “Tacocopter - Flying Robots Deliver Tacos To You” · View on HN