DPRG
DPRG List  



[DPRG] tiny-ARM from New Micros

Subject: [DPRG] tiny-ARM from New Micros
From: DeltaGraph at aol.com DeltaGraph at aol.com
Date: Mon Feb 21 19:50:43 CST 2005

In a message dated 2/21/05 5:49:49 PM Central Standard Time, k

>Speaking of which, is there a preferred technique for balancing while 
moving? 
>(Aside from the obvious "rubber side down" requirement.)

I prefer the technique of getting the robot to balance before even 
considering making it move. In fact an even easier problem is to hold the wheels in 
place so all you have to do is worry about tilt angle and tilt angular velocity. 
Once you get that working then you can move to moving then worry about telling 
it how to move.

I use the technique of lying to the robot about where "up" is, but then you 
must limit that input to avoid losing balance. There are probably some optimal 
ways to calculate the best way to fall out of balance then recover obtaining 
some target change in velocity, but I have not gone there.

You should be able to balance using 1 IR distance sensor and not two.
I like the feeler method that David Anderson suggested to me using a stick 
attached to a pot (that drags the ground). I would suggest using that first. A 
pot has some noise to deal with, but no phase error that you might find in a 
poorly/incorrectly implemented Kalman filter like I had initially.

As for computing power for balancing -- I was using a Mega 32 at 8 MHz on my 
first balancer. My second is running at 16 MHz, but I have not retuned my 
original program.

I think having something like an ARM could do nothing, but help out.
I am still using some fixed point routines on my balancer which I would love 
to trash in favor of floating point.

In fact, I think that having floating point arithmetic that cranks at 100K 
Flops or better (or sum such figure) will make a lot of tasks much easier.  I 
think of how much time I have wasted trying to use really limited precision 
arithmetic on 8 and 16 bit micros.

I remember less than two years ago writing a whole bunch of fixed point 
(24.8) bit Macros for the Tiny26 to do PID (well PD for me) TOTAL WASTE OF TIME.

Nothing like having to waste time too trying to do trig look up tables when 
all you need is a few sines and cosines...

Yes, I am looking forward to microcontrollers that can do real math at some 
reasonable pace. I should time the Mega parts (they are not too bad), but still 
the slower and less memory anything has, the more constraints to bite you -- 
IMHO

Ron
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.dprg.org/pipermail/dprglist/attachments/20050221/ddc7fc5d/attachment.html

More information about the DPRG mailing list