Extrinsic Motivation: And You Thought Inverted Pendulums Were Hard

An extremely common project for a control systems class is the inverted pendulum. Basically, it’s a robot mounted on a linear rail, a hinge, and a pendulum sticking straight up in the air. Get your algorithms right, and you have a pendulum that seemingly resists the inexorable pull of gravity and a great understanding of how Segways, balancing robots, and quadcopters work.

[zakowy] is taking this to the next level with his entry to The Hackaday Prize. It’s an inverted pendulum with two counter-rotating propellers in a gimballed fan, and the most unstable UAV design we’ve ever seen.

The mechanics of the build consist of a carbon and epoxy frame, with a motor mount that can move in the X and Y axes. This mount holds two brushless motors and is actuated with rather large pitch and roll servos. The electronics consist of the usual suite of sensors found in a quadcopter – gyros, accelerometers, magnetometers, and a barometric altimeter. Everything is controlled by an Arduino Due, getting commands from an RC receiver and sending telemetry back to a computer

[zakowy]’s project didn’t make the cut for the quarterfinalist selection, but he is undeterred. He’s building this strange contraption because he can, not because we’re dangling some great prizes in front of his nose. Right now, [zakowy] is working on a testing rig. This thing will fly, make no mistakes about that.

Videos available below.


SpaceWrencherThis project is an official entry to The Hackaday Prize that sadly didn’t make the quarterfinal selection. It’s still a great project, and worthy of a Hackaday post on its own.

24 thoughts on “Extrinsic Motivation: And You Thought Inverted Pendulums Were Hard

    1. Me too! The gimbal actuation tests I ran last month with the motors spinning surprised me–while I didn’t try to quantify servo response rate, visually it didn’t seem to change at all.

    2. Gyroscopic effects arise from changing angular momentum, which is a vector quantity.
      Since the propellers are counter-rotating, the vectors will be in opposite directions and so the over-all angular moment will be very small (or zero if the mass and angular speed of the propellers is the same).
      This means when you change the direction of the axes of the motors, there is little to no change in angular momentum and hence no gyroscopic effects.

    1. Hahaha thanks! Spec creep has hit this project pretty hard, but I think the ducting will help me get the thrust:weight ratio above 1:1. I certainly won’t give up; winter is coming in MN (it’s going to hit the high 30s tonight) so I need to get my house projects buttoned up, but I’ll continue pushing for tethered hover tests this fall. Stay tuned!

  1. This is really ambitious, they gyroscopic effect is probably going to interfere with stabilization instead of helping it, if he gets it to fly he should win, this is far more complex than anyone realizes.

    1. I appreciate the vote of confidence. I’ve really only started to gather instrument data, but I haven’t tried to model and correct for gyroscopic effects. Early on, I forced myself to focus on the basics–get the thrust:weight ratio high enough to get off the ground and sustain uncontrolled tethered flight. I haven’t started to deal with other basic measurement concerns, like the massive amount of vibration transmitted through the frame by the motors and servos. Should be an interesting journey!

      1. on the macro scale they will, but he has to deal with the micro scale on the orientation of the motor, THAT is going to need to be programmed around, the blades look to be at least 3 inches apart, that is going to make a lot of wobble when it swivels.

  2. your probably going to have to spin that lower motor much faster to offset the desire of the system to twist. The down-wash from the first motor with make the second encounter less force and therefore need more rpm to account for it. pretty awesome, good luck!

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.