MLW sub $500 robot

Yea, I'm thinking of setting that up, but I've so many projects going on I'm
swamped.


The OAP project is very cool.

I took a different route. I'm in the process of moving stuff to the 2.6
kernel, and am using the 2.6 serio driver instead of the 2.4 psaux driver.
The Linux 2.6 kernel moves mouse stuff into the kernel under the "input"
driver model. It is a great idea for desktop stuff, but a little limited in
user space.

In dealing with the mouse issues, I've also been experimenting with the
mouse resolution. Most mice have a resolution setting 1:1, 2:1, 4:1, 8:1
the higher the resolution the better precision at lower speeds. At higher
speeds it becomes more likely that you will overflow the counter. I've also
been testing if all the various scale settings return real counts or if
they just accelerate them. A lot of turning motors by hand.

While I was redoing the code to interface with the mouse, I am also trying
to modularize the internals a bit more so that when I dump the Veleman
board, I won't have to deal with many suprises.


But, back on topic, any questions about the code?

None at the moment. Still building electronics for the base.
I have to build a PIC programmer and just haven't gotten that far yet.

I'm hoping to post some pictures and information soon to the wiki.

I started to go the same route as you, but wanted to leave more overhead
on my main board for higher functions. I won't be able to afford a
bleeding leading edge mother board for a while yet, so I'll be using a
233MHz Pentium board I'm salvaging from one of my previous PCs.

What I meant for you to do was post to the OAP wiki. It's open to
modifications to the base design and I really think you have some good
ideas that would be helpful to the small, but growing community there.
Also, instead of reinventing the wheel, I thought you might be able to
get something out of the OAP design.

I chose to go the OAP route, because it reminded me of the old Heathkit
Hero platform. I'm even going to be adding an arm or two to my unit
after I get a few more levels built up and am hoping to use your
parallel port option for that. I'm going to try and tie the kinematics
of the arm motions with the vision system after I do upgrade my main
board. I've seen it done on a couple of University level bots and the
results looked promising.

There is an alternative to the parallel port even. Boondog Automation
has the schematics for a motion control card that plugs into the main
board itself. It provides timing and resolution, so you don't have to
use so much code space to create the same affect and you don't have to
use up one of the precious ports on a pc main board. I like that since
my I2C bus is currently having to reside on my parallel port.

Matter of fact that was the reason I was asking a few weeks ago about
any implementations of I2C from USB or Serial ports that have Linux
drivers. Know of anything like that?

See ya,
Eljin

No thanks, silly rabbit, everyone knows PICs are for kids :)


Cool. If you want, I can add a space on linuxpcrobot.org for you.


233MHZ pentium, wow!! I haven't worked on one of those since my Metrabyte
days. If it helps, I'm only using a 800MHZ EPIA.



I looked at the OAP design, and think it is "a" way to go, but not what I'm
looking to do.

The sub $500 will not have user developed micro-controller code, period. If
there are peripherals, like video cards, keyboards, etc. then that's fine,
but the end-user programmability is at the PC level. It is a core design
goal.


Cool, I'm sure everyone wants to see that and source of course.


I'm not worried about timing or "code space" for the main motor control
system, it is trivial in both respects.


The 2.6 kernel has a whole I2C subsystem and drivers for a lot of chips. It
looks very cool.

I like it.  But I have a few questions.

You don't mention what kind of battery you are using.
It looks like there is one in the bottom of the robot
in the pictures.  What did you use?  It looks a little
like a motorcycle battery (or 3 6v?).

How long can the battery(ies) run the bot if it is moving
constantly and how long can it run the bot if the
thing is not moving and not sleeping but acting like
a Linux PC?

Does an 800Mhz motherboard require a fan? Will it be ok
if you cover the sides or the robot instead of it being
an open rack?

Best Wishes

Right now I am using a gel type lead acid deep cycle battery commonly used
by fire alarms and UPS systems. (12V)

The "best" consumer battery with regards to power density and weight is NiMH
batteries, unfortunately, they are sort of expensive. If I can find a
source of cheap NiMH I will switch.


I haven't done too much timing on the system, but I've left it on for about
10 hours not moving and not using the wireless ethernet system.

Unfortunately the 800MHZ board that I have does need a fan, though I bet I
could put a big-ass heat sink on it and get rid of the fan. I think the
newer 800MHZ boards can be fanless.

I'm using sub-c cell packs of NiMH. Did you take a look at maxamps.com?
That's the best place I found to buy quality GP NiMH cells and accessories.
I bought individual cells, so I can wire them as I need (electrically and
mechanically speaking). The only problem is that it is really tricky to
solder the batteries together, and some people even reccomend not to solder
at all, but I'll try either way.
I've researched LiPo, but that's really expensive...


You are correct on that. At least the newest ones from VIA (EPIA) at this
speed do not require a fan, but we always have to keep in mind that our
circunstances are a little bit different. If your robot is meant to work
outdoors, you will have to take in consideration environment temperature.
(what if your robot is meant to cross the desert on the summer peak?)

Yea, they are really pricy.
Take a look here:
http://www.all-battery.com/index.asp?PageAction=VIEWCATS&Category 

20 3300mAH SUBC batteries for $50, thats $2.50 each.

http://www.batteryspace.com/index.asp?PageAction=VIEWPROD&ProdID@&HS=1

8 5000mAH NiMH C batteries, 21.95, that's about $2.75

I had seen those sites before, I was very tempted to buy those cheaper ones,
but when speaking about batteries, brand quality really makes a difference.
From what I know, Sanyo and GP are the ones that make real quality
batteries. If you are planning to use it a few times, I don't think there
will be any difference, but after a couple recharges the cheaper ones won't
hold a charge as they used to when new.

It is amazing how everything electronics evolved really fast, and batteries
evolved much slower compared to everything else. (ok, very vague, but you
got the idea)

I am skeptical of quality claims from manufacturors, there is typically no
real difference between a named brand battery and a generic one. (Often
times they are made in the same fctory) The chemical process is the same
for similar technologies, so I'm not sure how one can be better than
another without being defective.


IMHO batteries are all wrong. Electro-chemical power storage is ancient
technology. When we need is a super insulator to make long term electrical
storage an electrostatic technology.

Fact is when I put a pack of no-name batteries on my RC car, after a couple
of months the run time decreases significantly (let's say from 15 to 6
minutes) while when I use a GP pack, it decreases to only 12 or 13 minutes.
I need to perform more scientific experiments, but I have the clear
impression that GP's give more punch, given the same mAh ratings. (I charge
them at the same rate)

If you browse through some RC forums, you will see people telling the same
stories.


They may use the same process, but maybe using cheaper components, and being
"at the edge" of being deffective...

I work for a telephony equipment company, and we constantly have to quote on
different parts from different suppliers. Usually a part has a standard
spec, which all suppliers meet, but some with more or less quality, and
quality does affect some aspects of your final product. Sometimes you don't
care though.

Without clear measurments and process, this data is not trustworthy. If
people buy "generic" batteries, they may "feel" that they are not as good
and thus have a worse "impression."

Without some solid statistics, it is impossible to judge. Also, if there are
statistics that show one battery is better than another, I'd like to
understand why, i.e. cut apart a "good" one and a "bad" one and see what 's
the story.

More likely simply outside of one or more specs defined for the brand.
Manufacturers don't like to throw out the "rejects" if they are
perfectly fine otherwise, and to support their main brand they may not
want to put out their seconds with their name on the product. So they
create a generic brand that allows them to sell through the seconds,
while maintaining the value of the main brand..

Qualitative differences certainly do exist among products produced by
the same factory. This has been the case with every manufacturer I've
worked with. It's even true in the service sector. My folks ran a
successful carpet cleaning business for years where we had both a "brand
name" and higher prices for the affluents in the area, and a "no name"
bargain basement company where there was less attention to detail. We
just switched out the magnetic signs on the vans, and enjoyed the
pricing spectrum of the business.

-- Gordon

PS: Sorry to have missed you this Saturday.

"Gordon McComb"

No, I am sorry I couldn't go. I got a note from my advisor saying that this
upcoming Friday we'll get a visit from someone from SPAWAR, and a few things
must be ready.... so I had to use any free time to make progress. I'll try
to go to the next one.

I'm too lazy to make my own battery packs, but The Robot Marketplace
makes them: http://www.robotmarketplace.com/

These are meant for battle-bots and their ilk, though.


I've *looked* at them a bit. I'm still thinking of them
for my next bot, but I'm talking $750 for the batteries alone...

I'm *really* hoping that the price will come down quickly.

I'm thinking of a CrustCrawler Nomad for my next bot.
--
D. Jay Newman
http://enerd.ws/robots/

"D. Jay Newman"


In my case, because I need long run times, LiPos would have costed $2500
(not counting the special charger required).

Hmmm. I'm running a Pentium M (2 GHz) and sensors for 5 hours on
24V at 7 amp-hours.

I could get a 6 amp-hour LiPol 30V at 6 amp-hours for about
$500. A 7.2V 6 amp-hour for my servos is around $150.

And this is at Robot Marketplace prices.
--
D. Jay Newman
http://enerd.ws/robots/

My problem is not running the computer or sensors, it is running the dc
motors. What do you use in that case?
I need 20 amp hour just for mobility. (bear in mind it is an outdoors rover
supposed to go over rough terrain at relatively high speeds)

NiMHs or SLAs are probably best for that unless you have
government funding.  :)
--
D. Jay Newman
http://enerd.ws/robots/

Lithium Polymer are hard to beat when it comes to both those factors
and beat NiMH hands down.  Also, they can typically be discharged at 6
to 10 C (6 to 8 times their capacity rating) without damage.

They do have some down sides - cost is one.  The other is that they
can be very dangerous if they get damaged or are improperly charged or
discharged.

-Brian
--
Brian Dean
BDMICRO - ATmega128 Based MAVRIC Controllers
http://www.bdmicro.com/