[robocup-small] [rc-ssl-tc] Chip kick interception challenge

Andre Ryll andre at ryll.cc
Mon Jul 24 06:41:57 EDT 2017


I recommend to use three different positions for the chip kick. As I 
have implemented our chip kick detection I am well aware of the 
difficulties of detecting such kicks. I suggest the following positions:
- From a corner to the center of the field. This is most difficult to 
detect as the ball goes through the camera axis and does not show a 
parabola in the images.
- From a corner along the long side of the field. Here, a single camera 
is available as well but the ball shows a significant parabola. The ball 
may vanish from the images at its high points.
- From the center to the goal or vice versa. This is most simple as the 
ball is visible on multiple cameras, basically providing you stereo vision.

We used these exact position at IranOpen. Furthermore, we used distances 
from 2 - 3.5m. Basically, we had 2m, 2.5m, 3m, and 3.5m in the bucket. 
Then we drew them in random order. Thereby, each team gets the same 
chips, but cannot predict the order they were executed. To speed up the 
challenge, this can certainly be reduced to two distances. The technical 
difficulties of the different positions might be used to weight scores.

Cheers,

Andre Ryll
Project Leader & Founder of
TIGERs Mannheim


Am 24.07.2017 um 19:26 schrieb Danny Zhu:
> That sounds reasonable to me, though I think we'll need to see it in 
> practice. If the range needs to be wider, we can always add variation 
> in the commanded distance as well. The exact distribution within the 
> range shouldn't matter much, as long as the same thing is done for all 
> competitors, right?
>
> Danny
>
> On 07/24/2017 07:18 PM, Alexander Danzer wrote:
> > That's a good point.
> >
> > However, when I investigated the possibilities of chip kick estimation,
> > I found that only considering a robot's orientation does not provide 
> enough
> > precision for determining the kick's direction precise enough
> > for a distance of more than 3 meters.
> >
> > But we certainly should have some variety in the distance. When I wrote
> > 3.5 meters, I had in mind that the kick was performed with an ER-Force
> > robot, which has significant variations in the actual kick distance.
> > I would estimate deviations of around 20 centimeters for said distance,
> > but I have to check our current hardware tomorrow.
> >
> > So, what do you think of a distance in the range of 3.0 to 3.5 meters?
> > (with an ER-Force robot, this is quite hard to guarantee however with
> > an equal probability distribution).
> >
> >
> > On 2017-07-24 18:55, Joydeep Biswas wrote:
> >>> On Jul 24, 2017, at 6:22 PM, Alexander Danzer <alex at dnzr.name> wrote:
> >>>
> >>> The ball touches ground after 3.5 meters in field plane.
> >>
> >> If the chip kick distance is fixed, then the intercepting robot only
> >> needs to know the kicking robots orientation when the ball is kicked -
> >> it does not really need to do chip kick estimation.
> >>
> >> -Joydeep
> > _______________________________________________
> > rc-ssl-tc mailing list
> > rc-ssl-tc at lists.robocup.org
> > http://lists.robocup.org/cgi-bin/mailman/listinfo/rc-ssl-tc
> _______________________________________________
> Management mailing list
> Management at tigers-mannheim.de
> http://tigers-mannheim.de/cgi-bin/mailman/listinfo/management

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-small/attachments/20170724/3f36ca84/attachment.html>


More information about the robocup-small mailing list