Re: Some ideeas and coments

Hello !

I am not yet in the fight club ;) but have worked
hard on my robot. So its cool for me to know the next
contest will be quite the same. 


Here is some quote from sony qrio site :

----------------------------------------------------------
http://www.sony.net/SonyInfo/QRIO/technology/index_nf.html

"QRIO determines the condition of the walking surface
using four pressure sensors in the sole of each foot
to gather data on the amount of force being received
from the walking surface."

"... an actuator was needed with the ability to produce
varying levels of torque at varying rpm speeds,
respond with quickness and agility, ..."

"If pushed by someone, QRIO will take a step in the
direction it was pushed to keep from falling over.
The control system senses that it has been pushed
through the pressure sensors in the soles of its
feet and its position sensors, and acts to maintain
stability."

"QRIO's head is equipped with two cameras. Just like
a human, through a detailed comparison of the right
and left images (stereoscopic vision) it determines
the distance to the object in view."
----------------------------------------------------------


According to me : 

1) It would be cool to get the same 4 analog pressure
   sensor on each foot of the robot

2) It would be cool to have a better control on the joint
   motor.
   There is actually no way to control the motor by speed
   instead of position. Thus, because of the PID, there is
   no way to have a fast but short motion on a joint.
   Or if is possible, I didn't succeed :(.

3) I think it can be really cool, if the speed/torque limit
   of the joint is increased (or if the weight of the robot
   decreased). Note that it was a modification done by sony
   to succeed to make the robot "run".
   And it will be really cool if the robot succeed to really
   jump. Actyally, the best I get with actual robot is a
   really small jump : about the heigth of the black part
   of the robots foot ! And I don't think it's possible
   to jump higher :(
   If a human realistic jump was possible (heigth of the
   knee ?), I think really cool moves would be implemented
   by competitors 8), and this can renew the contest a lot :)

4) Of course, self collision avoidance is an important required,
   change. Same for joint limits.

5) Concerning the availability of the contest information,
   I agree a lot. It would be like a referee counting the
   oponent when he is down :). And knowing who gets the max
   distance (or whatever measure involved in deciding who
   win the fight) is good for renew robot strategy during
   the round.

6) I also agree with the idea of more round for the top ten
   robots ! An other idea for the top 2 is to have a non
   limited round number, with a tennis "tie break" rule :
   winner is the one wo get 2 more wins over the oponent.

7) As I have not yet upload a robot on the contest,
   I don't know how it would behave on it. But it would
   be cool if we can (whatever OS or powerfull or not
   computer we use) be sure the behavior will be the
   same at home and on the contest server. Does the
   "real time" option really assure this ?
   An other point is to be sure both robot get the same
   limitation about the ressources. Not to have robot
   "stealing" processing time to the oponent by the use
   of lot of thread/cpu/memory or whatever.

8) I think robot will require more and more processing
   time to achieve all their tasks. So, an increase of it
   is probably a good idea. Note that dedicated hardware
   (and a lot of different proc) is used by sony to
   control things, probably allowing a lot more compute
   operations compared to the "real time" limitation set
   by the actual simulation.
   I tried real time balance, and behavior is really
   bad because of:
     - a lack of compute time,
     - a too slow joint reaction,
     - a detection of unbalanced done to late (analog
       pressure sensor on the foot could improve this a
       lot),
     - my balance idea is wrong and the code bad ;)

9) Concerning winner choice when none of the two robots
   was 5 second below the fall limit, I don't think fall
   count or max time below limit is a good idea, as it
   will probably prevent robot to do some original moves.
   I don't think the max distance from start point is a
   good idea. It's an easy way to evaluate the efficiency
   of the "walk" of the robot, but it's limited. The only
   way to go farther is actually to succeed to go out of
   the tatami and get back on it in less than 5 second. A
   such hard coded sequence motion is probably no more
   difficult than recover from a fall (which is done by
   most of the top robot right now). According to me,
   none of competitor is doing this right now, because they plan
   to introduce it just before the end of the contest ;)
   If no robot do it, I will be really tempted to do it ;)

   A real estimation of the distance walked by the robot
   will be a far better information to use to decide the
   winner between two robots, and its a good incitation
   to get a robot succeed to run :)

   An other idea is to integrate the distance of the robot
   from the center of the tatami, and set as winner the robot
   with the smalest value at the end of the round.
   Thus, robots will manage to keep the center position of
   the tatami, or keep the oponent far from it. I think it
   will be a good way to have the robots face each other
   instead of just staying far from the oponent, wich is
   a really boring strategy.
   I will be interested to have feedback from list members
   about this idea.
   
10) Concerning the "competition system", I find the actual
   one good. Just a modification of the rounds number for
   the top ten and/or top two seems necessary to avoid to
   much hazard when opponent "force" is closed.



I probably forgot a lot of what I wanted to write,
but this mail is long enough if I want someone to
read it until the end :)   


Lets finish with something else :

An (old) interesting article :
http://www.eetimes.com/sys/news/OEG20031219S0014

Qrio last show :
http://www.sony.net/SonyInfo/QRIO/works/20040325e_nf.html

PS : For linux user, the following command is supposed
to allow you to record the video streaming as a file
nammed qrio.ram :
mplayer -dumpstream -dumpfile qrio.ram -nocache
rtsp://cds101.bit-drive.ne.jp/shp/02-38-QR105.rm
(refer to http://linuxfr.org/~kobold/11138.html for details in french)


Bothari, not sure to be ready before the end oh this year contest ;)



 
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/