3 ms·

Wow, it seemed straight forward for the first 10 minutes or so, when he introduced the biggest challenge: gyroscopic motion due to the high speed rotation. thi

Wow, it seemed straight forward for the first 10 minutes or so, when he introduced the biggest challenge: gyroscopic motion due to the high speed rotation. this is a gripping video.

stickfigure 8y ago
I was expecting some sort of dynamic feedback system based on a gyro sensor, but apparently the compensation is just hard-coded? I'm amazed that can work.
- Animats 8y ago
  Ducted-fan VTOL aircraft are surprisingly stable. Here's the Hiller Flying Platform of the 1950s.[1] It's very stable, even though top-heavy. It's mostly steered by leaning, but there are thrust-vectoring vanes to control yaw. Many variations on this theme were tried in the 1950s, and many of them flew. But it never caught on. There's been recent interest in larger ducted fan drones. This apparently scales better than quadrotors. But nobody seems to really need large ducted fan drones. If you like early small VTOL craft, many of them ended up in the Hiller Aviation Museum in San Carlos, CA. [1] https://www.youtube.com/watch?v=W3FS3D1rCos https://www.youtube.com/watch?v=W3FS3D1rCos
- rtkwe 8y ago
  You have to pre-program in the compensation because it basically rotates your controls 90 degrees which isn't something a simple feedback loop can deal with. No matter how fancy the feedback system is if you swap 2 controls its not going to be stable, it'll spend the whole time chasing the problem.
- blt 8y ago
  The gyro effect is pretty predictable. Anything where you're pretty confident in your model, it's better to compensate in the feedforward part of the controller than to push it into the feedback part. Think about gravity compensation - adding a constant amount of thrust to cancel out your weight. It's not going to match the model perfectly due to error in your attitude estimate, but it's still better than treating gravity force as an unmodeled disturbance. Gyro compensation is similar.
RachelF 8y ago
Yes, it is impressive that he got it going with PID controllers plus whatever else is in the controller firmware. This sort of inverted pendulum type system would typically require at least 1/s^2 control. It also has the additional complexity of gyroscopic torques, center of thrust not through center of gravity and servo control lag. This is similar to controlling a rocket, something the Germans took a long time to perfect with the WW2 V2 missile.

Wow, it seemed straight forward for the first 10 minutes or so, when he introduced the biggest challenge: gyroscopic motion due to the high speed rotation. this is a gripping video.

stickfigure 8y ago

I was expecting some sort of dynamic feedback system based on a gyro sensor, but apparently the compensation is just hard-coded? I'm amazed that can work.

Animats 8y ago
Ducted-fan VTOL aircraft are surprisingly stable. Here's the Hiller Flying Platform of the 1950s.[1] It's very stable, even though top-heavy. It's mostly steered by leaning, but there are thrust-vectoring vanes to control yaw. Many variations on this theme were tried in the 1950s, and many of them flew. But it never caught on. There's been recent interest in larger ducted fan drones. This apparently scales better than quadrotors. But nobody seems to really need large ducted fan drones. If you like early small VTOL craft, many of them ended up in the Hiller Aviation Museum in San Carlos, CA. [1] https://www.youtube.com/watch?v=W3FS3D1rCos https://www.youtube.com/watch?v=W3FS3D1rCos
rtkwe 8y ago
You have to pre-program in the compensation because it basically rotates your controls 90 degrees which isn't something a simple feedback loop can deal with. No matter how fancy the feedback system is if you swap 2 controls its not going to be stable, it'll spend the whole time chasing the problem.
blt 8y ago
The gyro effect is pretty predictable. Anything where you're pretty confident in your model, it's better to compensate in the feedforward part of the controller than to push it into the feedback part. Think about gravity compensation - adding a constant amount of thrust to cancel out your weight. It's not going to match the model perfectly due to error in your attitude estimate, but it's still better than treating gravity force as an unmodeled disturbance. Gyro compensation is similar.

RachelF 8y ago

Yes, it is impressive that he got it going with PID controllers plus whatever else is in the controller firmware. This sort of inverted pendulum type system would typically require at least 1/s^2 control. It also has the additional complexity of gyroscopic torques, center of thrust not through center of gravity and servo control lag. This is similar to controlling a rocket, something the Germans took a long time to perfect with the WW2 V2 missile.