[robocup-small] Re: Looking ahead to RoboCup 2006

Oliver Purwin opurwin at gmx.de
Tue Sep 6 14:24:42 EDT 2005


Hi all,

as we already announced at the final, we're not going to compete next year.
However, I'd still like comment on the current state of the game/rules since
I feel that the small-size league has a lot of potential that is currently
not used in the most efficient way.

1) Stoppages in the game
Currently the ball goes out of bounds way too often. With the running clock,
this cuts down seriously on real game time. Also, it puts too much emphasis
on set plays as compared to playing the ball out of game situations. I
strongly advise to bring back the walls. A 90 degree wall with some padding
(to slow down balls that are bouncing off) might be a good solution. Or the
sloped walls from 2003 to discourage hard shots aimed at the wall.

2) Passing vs. dribbling
As far as I understand it, one of the main goals of the small-size is to
focus on team play, cooperation etc., which definitely involves passing.
Currently there isn't too much passing in the games, although it is getting
better. There has been an ongoing discussion about what has to be done to
make teams pass the ball more. I'd like to distinguish the suggestions into
"hacks" and ways to solve the fundamental problems.

You don't want to give the robots too much ball control, so that they can
move upfield at top speed, unhindered by the ball. Getting rid of the side
dribblers was definitely the right thing to do. However, the robots need a
certain amount of ball control to receive passes quickly, before an opponent
can get in the way. In my opinion the dribbler as it is implemented right
now (one horizontal bar) lends the robots just the right amount of ball
control to do something interesting with the ball. The robots are capable of
receiving balls and do a little bit of maneuvering, but it's not enough
control so the robot would be able to run around his opponents and score by
himself. It seems that the 50 cm dribbling rule (which is very hard to
enforce anyway) is not required. Reducing the ball control even more would
fall into the "hacks" category, since the robots will have no choice but to
repeatedly kick the ball upfield. The games would resemble some kind of
Brownian motion.

The fundamental problem is that from a tactical perspective passing only
makes sense if the ball can travel significantly faster than a robot can
move. Even though that is currently the case, teams can't really take
advantage of that fact since the field is too small: The hard kicks go
easily out of bounds without the receiving robot having any chance of
catching the ball while slow passes are just as easily intercepted by the
(fast moving) opponents. It seems there are only two possible solutions to
this:
  a) make the field larger
  b) make the robots slower

Clearly, a) is the more interesting (and conceptually simpler) solution.
However, given the constraints of everybody's lab space it seems that b) is
preferable. Options for slowing down the robots could be the introduction of
an allowed top speed (monitored by both teams or the referee?) and/or an
energy budget, as has been suggested before. An alternative would be to
limit the allowed battery power for all actuators on board.

The games last year involved more passing than in any previous year.
However, it seems that this was mainly caused by the fact that the ball went
out of bounds so often and it is much easier to pass on set plays. This is
clearly a side effect of too many stoppages in the game.

3) Chip kicks
Chip kicks are definitely a great crowd pleaser (nothing wrong with that,
per se). The Fu-Fighters showed last year that it is possible to design a
legal and highly effective chip shot, so that we can regard this engineering
problem as solved. However, from a strategic point of view chip shots don't
have a lot to offer: if you execute them right (and the Fu-Fighters did!) 
the other team has basically no way of defending against them, especially on
free kicks since the opponents are not allowed into the 50 cm circle. In
that respect they are similar to our side-dribblers from 2003: they were a
basically unstoppable gimmick. Talking about game theory and arms races: if
chip kicks are allowed in the future, every team has to have them or they
will inevitably lose against a team with chip kicks.

4) Merge small-size and mid-size
It seems that during the last couple of years the small-size has reached a
plateau. Robots are getting a little faster every year, but not much has
changed fundamentally. It would be a very cool idea to merge the small-size
and mid-size:
 - make robots a little bit larger than small-size (to fit additional
electronics, camera etc.)
 - require local vision
 - distribute AI computation
 - play on a mid-size field

That would incorporate the best of both worlds: you get the distributed
nature of the mid-size in combination with the agility and easy-to-build
(and much cheaper!) small-size.

Oliver Purwin
Cornell Big Red

-- 
5 GB Mailbox, 50 FreeSMS http://www.gmx.net/de/go/promail
+++ GMX - die erste Adresse für Mail, Message, More +++



More information about the robocup-small mailing list