Comments for Mechanically Inclined

Comment on Robot Baby Steps To First Victory by Raa

Raa — Thu, 03 Feb 2011 17:38:42 +0000

The sample time for the PID would need to be extremely fast to be able to detect
movement at the speed necessary to adjust the path.
Its likely the control loop is just too slow.
Having tried simliar control of systems with extreme input variations.
Also the input resolution from the camera capture would need to be increased greatly for the control loop to have any realtime effect.
The trouble is even if you can get a processor to actually sense the movement
and get the PID loop to function, you will find the hysterisis unmanageable.
There is a point in any proportional control known as the minimum controllable
veclocity, at which point the signal becomes an on/off function. With the
equipment you have that minimum is likely 1000 times larger than what one would need to enable real time control.
Using optical recognition only you will never get the input resolution required nor the speed needed with the current state of technology.
You need an input signal that is much more responsive to ball movement.
If you could accomplished that though, then you would find you have a new bottleneck, likely the mechanics of the servo and the minimum required movement.

If you could put a nano sized velocity sensor and processor ‘in the ball’ and get the signal with minimal error at perhaps every .00001 seoonds you might be able
to do it. Or maybe try to think of another way to sense ball movement with some type of field generated signal.. maybe something magnetic that could be external to
the ball and produce a signal with hyper resolution and speed. Even with that though you’ll need to increase the reaction time of the control loop to a thousand times a second… (depending on what you determine is the fastest the ball will ever move). You need a 10000 ghz processor which they don’t make yet, but if the current exponential trend continues as Kurzweil predicts may be available by 2020.

Comment on More Parts by swighton

swighton — Thu, 13 Jan 2011 13:49:09 +0000

Actually I used a small boring head which is a sort of high precision adjustable drill. you can see a video of one in action here The bearings actually just slide in - I drill hole smaller than the bearing then measure the precise size of it with a bore gauge in conjunction with a micrometer and adjust the boring head so that it makes the hole just slightly larger than the bearing by about .01mm. The bearing is held in naturally by the geometry of the mechanism - if you look at the assembled machine you'll see spacers between the parts that hold the bearings in. If I needed the bearings to stay in without the spacers, I would have made the hole slightly smaller than the bearing and pressed it in, if I needed it to be removable I would use a small screw from the side or if I needed greater holding strength than the screw, i could make something more complicated If you don't have a boring head you can get a reamer the size that you need which is used to enlarge a small hole to a precise size. I didn't a reamer which is why I used the boring head. Before I had access to a machine shop I would have drilled the hole smaller then filed it out by trial and error. No machine shop means lots of filing and sanding and grinding. ;)

Comment on More Parts by samiznaetekto

samiznaetekto — Thu, 13 Jan 2011 05:59:03 +0000

How do you affix the bearings inside their holders? Do you just drill a hole of slightly less diameter, then file it out by trial and error? Or you heat up the holder to make the hole slightly bigger? Or you have a screw to push the bearing tight?

Thanks You’re doing amazing stuff.

Comment on Extracing information from an image by Tony

Tony — Wed, 12 Jan 2011 16:24:51 +0000

Wow, this is confusing. But good job explaining! I think I can wrap my head around it…

Comment on Disparate parts come together to = machine by Esolitos

Esolitos — Wed, 12 Jan 2011 12:37:02 +0000

Here’s an italian iPhone blog talking about your “machine”
http://www.ispazio.net/158459/ecco-larduino-un-joystick-motorizzato-per-i-giochi-che-utilizzano-laccelerometro

Comment on Camera Holder by swighton

swighton — Wed, 12 Jan 2011 04:20:52 +0000

@Raa
I’ve considered throwing a dual axis accelerometer onto it – I can sample an accelerometer faster than I can the camera which would allow me to have more finesse in my positioning.

I completely agree about the project – it’s really a mechanical engineering, electrical engineering, and computer science all wrapped into one tasty package – the reason that I chose it!

Comment on Camera Holder by swighton

swighton — Wed, 12 Jan 2011 04:15:30 +0000

@Stefan
Yeah it is – I didn’t have the camera mount constructed at the time, but I have since played watching through the camera. It’s a bizarre experience watching through the camera you can’t even tell that it is being tilted all over the place – its all about your frame of reference.

Comment on Camera Holder by Raa

Raa — Wed, 12 Jan 2011 04:02:56 +0000

Courageous project
Your probably going to want an accelerometer or two on that rig to give
you the feedback control you may need as well as a camera. Precision on the analog signal to the actual servo position and hysterisis could be critical.
Your project is really like 3 projects…

Comment on Camera Holder by Stefan

Stefan — Tue, 11 Jan 2011 12:23:05 +0000

Wouldn´t it be more playable if you watch through the camera. The iPhone wouldn´t move then and you have a better view on the screen.

Comment on Test footage by Pixowl

Pixowl — Mon, 10 Jan 2011 21:19:55 +0000

Hi!
Very interesting!

I developed a tilt based iPhone game named Doodle Grub (free)
If you have time, try it with the joystick!

Have a good day!