This forum is in permanent archive mode. Our new active community can be found here.

GeekNights Monday - Augmented Reality

edited February 2013 in GeekNights

Tonight on GeekNights, in light of the recent opening of applications for early access to Google Glass (nevermind forum discussions), we discuss the promise of Augmented Reality. But first, we discuss non-standard uses of naan (get it?), New York groceries, the rise and fall of SuperDaE, and a surprising "solution" to the problem of poor Youtube performance. Also, the GeekNights Gaming Grande Prix is still accepting applications!

Download MP3
Source Link
Post edited by Rym on
«1

Comments

  • How to make a margarita:

    Step 1: Find George
    Step 2: Get blasted

    Works every time.
  • Did you guys even watch my video?

    Watch it.

    George makes similar ones. ;^)
  • I watched your video. It reminded me of my video.
  • Re: "This is my hole! It was made for me!"

    It's The Enigma of Amigara Fault.

    Related:
  • edited February 2013
    Pretty much every quadcopter video which is popular on the internet is bullshit. They rely upon $50k+ Vicon systems in order to get full state estimation. They also run all their computations on large desktops while just transmitting control inputs via wireless. It's one step up from simulation and not at all representative of real robotic research outside of the lab. (Well it is for control people who have to idea of how to do robot perception ~_^)
    Post edited by Andrew on
  • The fact that this state estimation is possible realtime is still awesome.
  • They also run all their computations on large desktops while just transmitting control inputs via wireless.
    Hey, look at this picture I found of some desktops 5 years ago:
    image

    Imagine what we'll be able to do once they're miniaturized!
  • edited February 2013
    Not particularly. It's just motion capture running at high speeds. They aren't doing any sort of object recognition or real vision, they are just tracking IR reflective points in SE(3) via passive markers. Tracking a handful of points given extremely high resolution, pre-calibrated cameras was state of the art in the 90s.
    Imagine what we'll be able to do once they're miniaturized!
    The control algorithms they are running now (probably Linear quadratic regulators in state space systems or consensus-equation based approaches if their smart, otherwise some horrific PID controller on the backend) are not really that computationally heavy.

    The reason why they use Vicon systems is because proper state estimation on-board the robot is a lot more difficult (and happens to by my area of research). I'm the one actually trying to do what they do for real. It's just easier to get 1 million Youtube views when you cheat and start with only half the problem.

    Post edited by Andrew on
  • A novel on-board control algorithm for aerial automata.
    WaterIsPoison et al. Georgia Institute of Technology 2017
    See I was right! We figured it out!
  • edited February 2013
    Well, I did get just awarded a 1 year (plus an additional year of funding after review) grant on a proposal very similar to that topic... Mostly on real time, embedded state estimation and environmental mapping though.
    Post edited by Andrew on
  • Does anyone watching these videos think that the quad copters are doing all the work themselves and are not linked to outside help? I know that's what is happening and I still think it's super impressive.

    And I say that as an expert on this topic. Not robots, but my job is, literally, balancing things and flipping them and catching them in a balance again.

    I want a copter in my juggling show.
  • I'm all for this but all I can think of is Black Mirror and how shitty life will more than likely end up with augmented reality because of sod's law.
  • Does anyone watching these videos think that the quad copters are doing all the work themselves and are not linked to outside help? I know that's what is happening and I still think it's super impressive.

    And I say that as an expert on this topic. Not robots, but my job is, literally, balancing things and flipping them and catching them in a balance again.

    I want a copter in my juggling show.
    You could easily incorporate a copter into your juggling show with minimal, or even no, programming at all. The one you want to buy is the AR Drone. Lots of possibilities.
  • You could easily incorporate a copter into your juggling show with minimal, or even no, programming at all. The one you want to buy is the AR Drone. Lots of possibilities.
    LLOOOOOOOLOOOOOOOLOOOOOL

    Pretty much the only possibility is hover there. Maybe move around with someone controlling via iOS, it will be very inaccurate.
  • If he was willing to use the old "burn the set down with every show" angle, he could be attacked by drones mid-routine and fight them off with juggle-fu.
  • Some possibilities: Put a bowl on top of that bad boy or hang a basket from it on a string, or both. Juggle things into and out of the bowl/basket. Have someone else remote control it, or juggle with one hand and a remote in the other hand. Will not work outside unless there is no wind because precision is required.
  • ARDrone's have almost no payload capabilities. Additionally, they utilize ultrasonic range finders facing downward to estimate their altitude. Obstructing them by hanging something would cause very serious errors to the included autopilot and would likely make it crash.
  • edited February 2013
    Just brainstorming, but is it possible to juggle objects that you need to interact with, such as pushing a button or two on the object before the next toss/catch? Could juggle controllers and do something like trigger one of several control routines for a given drone. Might be dreaming here.
    Post edited by Matt on
  • Andrew: crushing Scott's dreams since 2013.
  • ARDrone's have almost no payload capabilities. Additionally, they utilize ultrasonic range finders facing downward to estimate their altitude. Obstructing them by hanging something would cause very serious errors to the included autopilot and would likely make it crash.
    I've seen them carry things, but not much. I don't see why it couldn't carry one ball.

    Another possibility: Juggle the drone itself. Throw it up in the air while it is off. It turns on and flies about. Perhaps you juggle other objects around it while it doges and weaves. It turns off, you catch it.
  • edited February 2013
    Another possibility: Juggle the drone itself. Throw it up in the air while it is off. It turns on and flies about.
    Problem: When you turn on the ARDrone it initializes it's filter for the IMU and estimates the biases for the accelerometer drift rates via calibration. Turning it on mid flight would likely cause mis-calibration, again resulting in a crash immediately. Furthermore, said ultrasonic sensors would not be able to get a valid ground/altitude reading and would just fail to initialize the autopilot anyways.
    Perhaps you juggle other objects around it while it doges and weaves. It turns off, you catch it.
    Apart from someone else controlling it via a better method than iOS, it will be VERY imprecise, likely to the point where it's unreliable. Unless you have a Vicon system, the commands will be so basic that they will be benign in any performance show setting. Oh, did I also mention they are really, really loud?
    Post edited by Andrew on
  • Maybe you could calibrate it then just leave it calibrated, but stop the rotors. Then throw it while it is on, and have it spin up mid-air? It's not easy, but problems like this are solvable.

    They are REALLY loud. Not much you can do about that.
  • What's the old saying?

    Helicopters fly be being so fucking loud that the Earth rejects them.
  • Fun fact: Pope Benedict has a helicopter pilot's license (as in for real, full-sized choppers), but has no driver's license and never learned how to drive a car.
  • Oh
    Maybe you could calibrate it then just leave it calibrated, but stop the rotors. Then throw it while it is on, and have it spin up mid-air? It's not easy, but problems like this are solvable.

    They are REALLY loud. Not much you can do about that.
    They are solvable, but require lots of hardware and software hacking, something outside of a regular consumers capabilities at this point.
  • Thank you all for your suggestions. What I really want is a flying prop holder that can drop balls, clubs or rings into my juggling pattern. Check out how Viktor gets new balls from above in this video... I want balls to drop in like that!

  • You would need to make something to do that which would not be trivial, and a copter would not be the right answer, although it would be a great show if it worked.

    I think a better idea would be to have some sort of remotely triggered catapult, trebuchet, air cannon, etc. off the side of the stage. Maybe even a high powered t-shirt cannon all the way behind the audience. It could launch confetti and a ball simultaneously.

    You don't have to worry about getting something up in the air, so the trick will work in any venue, indoors or outdoors. You would just have to re calibrate the launchers on a per-venue basis depending on the distance from the stage.
  • From this Wikipedia page
    In February 2013, an unrelated song named "Harlem Shake" originally uploaded to YouTube on May 10, 2012, went viral and became an Internet meme (see Harlem Shake meme).
    From this other Wikipedia page
    Being a meme, the video was replicated using the same concept by many people, which eventually led to it becoming viral in early February 2013,[1] with thousands of "Harlem Shake" videos being made and uploaded to YouTube every day.[2]
    The art form of the meme was established in a video uploaded on February 2 by five teenagers from Queensland, Australia known on YouTube as The Sunny Coast Skate.[2][3] The video started a viral trend of people uploading to YouTube their own "Harlem Shake" videos.[4] The teenagers' video, in its turn, was a follow-up to a video by a YouTube comedy vlogger named Filthy Frank[3][5][6] featuring a part where several costumed people danced to the song "Harlem Shake" by Baauer.[7][8]
    Even Wikipedia doesn't know about the harlem shake as a meme

  • You would need to make something to do that which would not be trivial, and a copter would not be the right answer, although it would be a great show if it worked.

    I think a better idea would be to have some sort of remotely triggered catapult, trebuchet, air cannon, etc. off the side of the stage. Maybe even a high powered t-shirt cannon all the way behind the audience. It could launch confetti and a ball simultaneously.

    You don't have to worry about getting something up in the air, so the trick will work in any venue, indoors or outdoors. You would just have to re calibrate the launchers on a per-venue basis depending on the distance from the stage.
    But I want a quad copter.
Sign In or Register to comment.