Abominations, schedule & Euler angles

Frans wrote:

>what is the match schedule in the coming 2 weeks full of holidays?
>  
>
I will be out of office from Dec 24th, 2003 to Jan 5th, 2004. However, I 
now have a script for launching the round server every week day at 2 pm 
(CET). So, unless the server crashes you can expect rounds to run on the 
every week day from now until I am back. So, a round will run today, 
tomorrow, on Wed. 24th, Thu 25th, Fri 26th, Mon 29th, Tue 30th, Wed 
31th, Thu 1st, Fri 2nd and Mon 5th when I am back.

asfsdf sdfsdfsdf wrote:

> I've been observing the contest for quite a while now and in my 
> opinion it has undergone some really sad development during this week, 
> having robots that do actually nothing but random attacks or strange 
> movements placed in the top ranks. Climax of this was Curry and 
> Sonovobic overhelming Coyote. Beating the old grandmaster, capable of 
> tracking his enemy and planning, just by blindly and continuously 
> punching (Curry) or by doing a weirdo frogstyle random bouncing 
> movement (Sonovobic) sounds so ridicoulous to me. Kp2 made a lot of 
> ranks using a dive attack that acutally rotates his legs through his 
> chest in a ugly looking manner.
>  
> As it looks right now, a silly and ugly competitor is more likely to 
> win a online contest in robotics than a reflecting and humanoid 
> looking robot. I really hope that there won't be many clones of such 
> behavior, as it was the case when everyone started adopting Coyotes 
> dive attack. If so, the Judo contest will turn into a freakshow of 
> abominations not worth watching anymore.
>  
> However, merry Xmas and a happy new year to all competitors who 
> provided their fans with funny and astonishing fights in the tatami.
>  
> A quite disappointed spectator

I do agree with you. I believe the problem comes from two facts: (1) the 
robots can stand up too easily and in a non-realistic way, it seems that 
the leg servos are too strong and (2) there are no joint limits. To fix 
that and keep the realism of this contest as high as possible I will 
change the robot models again.

The following changes will be introduced as soon as possible:

(1) Make it impossible to stand up too easily (from doing the splits 
directly to standing up). This can be achieved by 3 different ways: (a) 
increase the mass of the robot (by increasing the density from 0.1 to 
0.2 or so) or (b) decrease the maxForce parameter of each servo, or (c) 
increase again the friction coefficient with the ground. It will be 
probably a mixture of these 3 possibilities.

(2) Add joint limits for each joint. Madrag and others could not any 
more make impossible moves (leg turning around).

This will clearly break the behavior of many existing robot controllers. 
But I believe this is a necessary price to pay to keep this contest 
interesting for all.

Mike wrote:

>The DecimalFormat.format(double arg) function creates a string equal to "?"
>when the argument is equal to Double.NaN.
>The Double.NaN value occurs when the argument of the asin( ) function is
>outside of the range from {-1, 1}, such as 1.0000000001.
>I have added a check in the routine to force the argument of asin( ) be in
>the valid range so the Double.NaN does not occur.
>  
>
>>> Do you care to publish the routine on the Wiki pages? I can make the
>>> page if you don't have time.
>>>
>>    
>>
>There is a new Wiki page showing the Inclinometer Angle Calculations.
>
>Olivier wrote:
>  
>
>>>Thanks Frans for this feedback. I will add the routines of Mike into the
>>>Controller class and document them, so that anyone can use them easily.
>>>This will be part of Webots 4.0.19, which I hope to release next tuesday.
>>>
>>    
>>
>The noise error added during the contest should probably be added to the
>resultant X,Y, and Z angles and not to the OpenGL matrix elements.  If the
>noise is added to the matrix elements the actual error in the computed
>angles will vary with the angle and be a maximum of (2*NoiseOfElement) at
>angle=PI/4, and a minimum close to 0 at angle=0 and PI/2.
>angle=PI/4,    atan2(1+noise / 1+noise)
>angle=0, PI/2,    atan2(largevalue+noise / 0+noise ),  atan2(0+noise /
>largevalue+noise)
>
>Also, the argument of asin( ) should be forced into the valid range of {-1,
>1} to prevent returning an angle of Double.NaN.
>
Many thanks Mike for the Wiki page and for this continuously maintained 
and debugged inclinometer / compass code. The noise only applies to the 
matrix[12] matrix[13] and matrix[14] values and only affect the robot 
position but not it's orientation. As you use only matrix[1], matrix[5], 
matrix[8], matrix[9] and matrix[10], this should not affect your 
computation of the inclination. I will keep this function as open source 
code in the Judoka0 and Judoka1 sample classes so that anyone can still 
improve / fix it.

-Olivier



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/