Shop OBEX P1 Docs P2 Docs Learn Events
Let's Make a Robot Do These Maneuvers — Parallax Forums

Let's Make a Robot Do These Maneuvers

ercoerco Posts: 20,256
edited 2012-12-04 09:12 in Robotics
Bicycle trials riders do such amazing maneuvers. Watch the first two minutes of this video. Where would you even begin to make a robot attempt anything close to these? Multiple gyros, accelerometers, IMUs, PID, machine vision, IDK? First, call Sebastian Thrun...

Duane, here's an extreme case where I'll admit, the software is harder than the hardware. There, I said it. :)

Comments

  • mindrobotsmindrobots Posts: 6,506
    edited 2012-11-28 00:23
    It could go back to being difficult hardware too. To do moves like that don't you need the many, many degrees of freedom of motion that the human body offers in that application? Just like balancing on a single contact point is very difficult until you redefine the robot to have a single ball with three motors driving that ball. Then the solution becomes much easier (still, complex hardware and software but much easier). for this problem, you take the balancing uniball drive and add the ability to "pogo stick" and your solution becomes simpler.

    The words easier and simpler seem to understate the problem....it's still a great challenge but it's all in relatives terms!
  • rjo__rjo__ Posts: 2,114
    edited 2012-12-04 09:12
    I planned my day around jury duty and then got excused...there is nothing sweeter than free time:)

    I have worked with the 2D accelerometers from Parallax... way back. And if programmed right, I think they are stable enough and fast enough to contribute useful data. I was able to get inertial data that was stable over a 3 mile road course in a car... but everything I have read about gyro's gives me pause. Never looked at them but I don't think they have good enough long term accuracy for a robot that could do this any old time... maybe in the beginning. or the robot could rest while it was regaining its gyro senses.

    All of the really good stuff about quad copters seems to use external vision to figure out the where and when of everything... at the hobby level that would be a kinect box... which is really interesting and really works nicely in the right light with the right contrast at the right distance but with somewhat less than the claimed accuracy. Hooking it up to a Prop...through a computer... using Processing.org is fairly painless and the data you get could keep you busy for years.

    And you can build yourself a graphical interface ... to plot out your moves... and it is all free. Speed is an issue, particularly if you have lots of code... it sits right there in the main event loop clogging everything up.
    Processing.org v1.5 works well with serial up to 115K baud. Setting up for bluetooth was easy peasy... so xbee shouldn't be a problem.
Sign In or Register to comment.