Re: Extended deadline, new version of Webots & new contest

Hi everyone,

Thanks to Olivier for another great contest!  This contest is much more
exciting than the last one.

Changes for This Contest:
1)  I agree with Xavier, Yann-Ael and Manuel that more sets should be played
in the top 10 positions.  Maybe the best 9 out of 11 instead of the current
best 3 out of 5.

2)  I think Olivier means fixing a few problems with servo limits instead of
changing physics.  A commanded position greater than the servo limits will
not cause the servos to exceed the limit, however the servo can move beyond
the limit due to inertia.  A good example of this is the beginning of the
third round of the match between Astalavista and FreeTime on March 23.
FreeTime's back_2 servo moved beyond the -0.5 limit due to inertia and the
physics engine caused the servo to move back into range by traveling
from -0.5 to +0.5 through +/-3.14. (The back rotated through the legs).   A
fix for this problem is welcome before the end of the contest, but any other
changes to physics should wait until the next contest.

Ideas for Next Contest:
1) I agree with Xavier and Yann-Ael that there should be no limit to the
number of times a robot may fall and recover.

2) The Binocular vision suggested by Yann-Ael is a good idea.  I would also
keep the distance sensor for competitors just starting.

3) Let the robot controllers run in Synchronous mode.  I have tried PID
control for balance but due to the varing time step I have not achieved good
results.  When running with 1 controller in the Judoka_1.wbt world I can get
a time step equal to the requested value and the Z-transform PID controller
will balance the robot very well, but when running with 2 controllers or on
the contest server the time step is not always equal to the requested value
and the PID controller over corrects and causes the robot to fall.

4) I like Yann-Ael's idea for scoring based on the time above the Fall
threshold instead of the distance away from start point (radial distance
traveled).

5) A better Tilt sensor based on quaternions instead of a rotation matrix.
This would give X and Z angles that are always valid.  Currently the Z angle
is not valid (locked at +/-Pi) when the X angle goes beyond +/-Pi/2.
Although the current tilt sensor is usable for the contest, a better tilt
sensor would give the controller accurate tilt angles when the robot is
laying on the ground.

Good Luck to everyone and have fun,
Mike Jost (FreeTime)








 
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/