Re: Abominations, schedule & Euler angles

My viewpoint is make the physics match the actual hardware.
I agree the self collision detection should be implemented, however the
robot servo torque should not be limited only because many competitors
have found a way to stand-up unless the movements violate self-collision or
violate commercially available servos.  I have no problem with robots diving
either.
The engineer in me asked the question "Is the current model valid?".

Since this is a model of the Sony SDR-4X, it should be limited to the
capabilities of that design.
I tried to find the servo specs of the SDR-4X but, the best I could do was
to estimate the specs based on a comparison with the PINO open source
robot design.  http://www.symbio.jst.go.jp/PINO/
The PINO is a humanoid robot similar to SDR-4X.

Comparison:
The SDR-4X weighs 6.5 kg, from
http://www.sony.net/SonyInfo/News/Press/200203/02-0319E/
The PINO robot weighs 4.5 kg.
The PINO robot uses servos with a torque of 2.06 N*m for the leg, waist and
shoulder servos.
The Judoka robot weighs 0.725 kg. (Note: physics density could be changed to
900 kg/m^3 to better match the SDR-4X weight).

Scaling the servos torque, based on weight, from the PINO robot to the
Judoka and SDR-4X robot gives a torque for the servos equal to 1.03 N*m.
1.03 = (4.5 * 0.725) / 6.5

Results:
I set all the servo torques to 1.0 N*m in my Judoka robot and it was still
able to stand-up using the "splits" method of correction.
My Judoka robot failed to stand-up when the torques were set lower than 0.7
N*m.
My Judoka robot failed to walk when the torques were set lower than 0.5 N*m
(Due to collapsing of joints from inertia).

So, it would appear doing the "splits" to stand-up is a valid method for the
PINO and SDR-4X robots.  Also a few humans, ballet dancers, olympic
ice skaters, can use the "splits" method to stand-up (not me).

Recommended Changes to Webots:
1.  Physics density = 900 kg/m^3
2.  Maximum torque of servos = 9 N*m
3.  stopping self-collision

Notes:
Servo manufacturers use a torque unit of kg*cm instead of N*m as this unit
is more user friendly.  A kg*cm is the torque required for the servo
to lift a 1 kg mass on earth with a torque arm of 1 cm.  The conversion from
kg*cm to N*m is *= 9.807/100.
Webots uses N*m for torque.

Judoka mass Excel worksheet:
http://www.mikejost.com/cyberbotics/JudokaModel/JudokaMass.xls

Movies of SDR-4X:
goto bottom of this page   http://www.droidlogic.com/


Merry Christmas to all,
Mike Jost


----- Original Message ----- 
From: "Benoît Pothier" <[email protected]>
To: <[email protected]>
Sent: Wednesday, December 24, 2003 7:09 AM
Subject: Re: R: [webots-contest] Abominations, schedule & Euler angles


> I disagree, here is my point of view, simply because I am reading 2
reactions
> against Olivier proposal:
>
> I agree with the fact than the model should be changed a bit. I am not
against
> diving, but robots shouln't be able to stand up so easily, in a non
realistic
> way.  Why? I would simply give a more interesting/challenging contest.
Last
> year rules have been changed to prevent robots from loading multiple times
on
> the same feeder and that was a very good thing. Controllers based on this
> strategy had to be adapted. Olivier can't think about strategies spoiling
the
> contest interest before they are created. And so, rules have to be changed
> during the contest. Competitors should rely on sensors and not on
> trials/errors and manual servo positionning to build their robots. This
way
> they would easily adapt themselves to new rules or models.
>
> I hope than the joint limit and the friction factor ajustements will be
> suffisant to prevent fast-recovery. Because, the force of the joints is
not so
> important : it is allready difficult to jump.
>
> -- 
> Benoit
>
>
> * fransklip <[email protected]> [031223 15:29]:
> > Good reply, Caramine
> >
> > Here is 1 cent from me too.
> >
> > A spectator, who obviously has not tried to make a two-legged robot
> > walk, cannot appreciate the work and thinking and testing that has
> > gone into just making a two-legged robot move purpously. (his "name"
> > came from just hitting the middle row of keys at the left side "asdfg"
> > a couple of times, he liked to be anonymous obviously.)
> > Also this contest runs for many months and the "diving" is just one of
> > the things that came up. It will go again, because amoungst the top
> > ranking robots it is not decisive anymore.
> > I added the dive to my robot becaus I programmed it for my "sparring
> > partner" judoka to test a defence, and then thought: attack is the
> > best defence, and make it faster then Coyote's. It did work for a few
> > days, then other robots did it better.
> >
> > And about the contest rules:
> >
> > This was a new type of match / robot sport, and the rules had to
> > evolve with the development of the robots. In human sports, rules
> > change too. And I learned to forget about the human form of judo early
> > on in the contest (I have practiced judo 8 years in my teens).
> >
> > It is a wise to think about the newly proposed model changes.
> > There are only 4 months to go. And model changes are much more
> > destructive for controller programs then changes is match rules. It
> > will mean a lot of programming and testing, and a lot of time. Early
> > in the contest there ware two such changes in the model (the floor was
> > slippery like a dancing foor, and the maximum servo speed/force was
> > double the current value of 5), and even now there are some "broken"
> > robots that are not updated for these change.
> >
> > Perhaps it would be wise to leave the proposed drastic changes for the
> > next contest.
> > There is a chance that a number of contenders won't update for this
> > model change, or perhaps don't have the time for it. And the primary
> > interest of the EPFL students stops in January.
> >
> > Another point is, that perhaps some contenders could have some more
> > technical help. The Wiki pages give some, they actually were created
> > as a spin-off from technical discussions on this message board. A
> > second message board perhaps?
> >
> > Frans (Kimaita)
>
>
> To unsubscribe from this group, send an email to:
> [email protected]
>
>
>
>
> Yahoo! Groups Links
>
> To visit your group on the web, go to:
>  http://groups.yahoo.com/group/webots-contest/
>
> To unsubscribe from this group, send an email to:
>  [email protected]
>
> Your use of Yahoo! Groups is subject to:
>  http://docs.yahoo.com/info/terms/
>
>



To unsubscribe from this group, send an email to:
[email protected]

 

Yahoo! Groups Links

To visit your group on the web, go to:
 http://groups.yahoo.com/group/webots-contest/

To unsubscribe from this group, send an email to:
 [email protected]

Your use of Yahoo! Groups is subject to:
 http://docs.yahoo.com/info/terms/