[robocup-small] Ideas for 2007 Rules

[James Bruce]

Hi again,
I summarized the comments made so far, and follow each with some 
comments of my own.
   - Jim


*** Move to a partially automatic ethernet-based referee box
Krist W./Plasma-Z:
     likes idea, but must deal with "honesty" issues in reporting
Pailo Costa/5dpo:
     likes a dual system, but should use UDP rather than TCP

My vision of such a referee is that it would track the responses of each 
team, and perhaps even weight them by how much they match what the human 
ref says.  Another option is to allow one more team at a table to help 
auto-ref games which are not their own, which will help in breaking 
ties.  We can give a technical award at the end of robocup for the 
best/most accurate automatic refereeing by a team.

In terms of implementation, I would absolutely use UDP.  Multicast might 
be difficult to do in a portable way, but luckily SSL does not have 
referee commands that are as time-critical as they once were (only 
force-start qualifies now).  I think we can solve that with unicast UDP 
if we randomize which team gets the message first, or maybe we favor the 
most accurate automatic ref :)


*** Decrease timeout time to 5 minutes
Krist W./Plasma-Z:
     prefers 10 minutes
Smail Menani:
     5 minutes ok, faulty robots can be removed while play continues

Well, we could make a smaller step to 8 minutes.  The main thing I'm 
trying to avoid is the situation where both teams are having problems 
and play stops for almost 20 minutes as each team takes their timeouts 
one after the other.  That is too long in my opinion.  For comparison, 
the Aibo league has one timeout per team, for only 3 minutes.


*** Are chip kicks now too powerful for the goalie?
Yu Sheng/ex-ZJUNlict:
     how about an offsides rule to help with this.
Gordon Wyeth/ex-RoboRoos:
     likes offsides idea, adds more strategy
Nawarat/Plasma-Z:
     offsides rule interesting, but may be hard to ref
     should keep corner kick percentage around that of real soccer (50%?)
     allowing second defender would be ok
Krist W./Plasma-Z:
     likes offsides, hard without automated referee
BengKiat/LuckyStar:
     with fast+accurate kickers, offside less effective
Pailo Costa/5dpo:
     allowing a second defender seems a good option
Smail Menani:
     hard to notice offsides, split defense area may be better

Offsides does seem like a good idea to help with this, and seems to have 
a more support than my initial ideas.  Beng's observation that it won't 
help during regular play might not turn out to be a bad thing; We mainly 
want to help corner kicks, and not affect regular play too much.  To 
help in refereeing, perhaps we only enforce this at the time of a 
restart?  While the ref watches the kick, the assistant/ball handler can 
watch the defense and attackers.


*** Teams must prove kickers are legal
Pailo Costa/5dpo:
     wants clarification of how this would be enforced
     holding rules should handle currently "broken" cases

The holding rules should definitely be fixed, and that is something I 
complained about a few years ago when chip kickers first looked 
possible.  Right now when a ball flies over a robot, it is technically 
guilty of holding.  Refs do not call that, of course, but it should be 
fixed.  In theory, its ok for a robot to be capable of holding, but the 
referee should only call it when it actually happens.  In practice, refs 
take a dim view on robots that can consistently violate the rules, even 
if temporary, as with the dribbler controversy in 2005.  If we are going 
to enforce that for dribblers, we should enforce that for kickers as 
well.  So, if the "proof" says that a kicker is legal unless it is up 
against another robot, the ref should call holding when the robot kicks 
while up against another robot, and give a free-kick to the other team.

As far as how to prove legality, there are a bunch of possibilities as I 
see it, and I'm sure more exist as well.  Here are ones that come to mind:
(1) show that it is legal at all points of kicker travel and ball 
position (this is what most teams can do now)
(2) make a device that the ball would hit if the kicker were illegal 
(think of some sort of inverted foam wedge affixed to the top of the 
robot), and demonstrate that the kicker is not affected by the presence 
of the device (meaning the ball doesn't touch it, and thus stays out of 
that "illegal area").  This is hard to explain, but hopefully you get 
the idea.
(3) Some equations or a simulation detailing the dynamics, and evidence 
showing how the real kicker follows that model.
(4) A high-speed image sequence from the side showing the ball 
trajectory, along with evidence that the final kicker fielded at robocup 
follows a similar trajectory.


*** General kick speed limitation?
Krist W./Plasma-Z:
     agrees, to avoid an arms race, and for safety
Nawarat/Plasma-Z:
     good idea, measure with static test
Pailo Costa/5dpo:
     good idea for safety, hard to enforce
Tadashi N./RoboDragons:
     to accomplish, should commit to ball change for 2008
Smail Menani:
     supports; Ball speed should be 6 m/s or 3x robot speed

I think everyone agrees this would be a good idea in principle. 
However, the question really is how to enforce it during a game, and 
that's where I don't have any good ideas.  We need proposals to address 
*that* issue.  If auto-referees turn out to work well, that could be a 
solution, but that would be more of a 2008 thing IMO.

I liked the ball that Tim Laue brought to RoboCup last year.  Perhaps we 
should commit to some exposition games using that ball, and try to get 
them distributed to a few teams (Tim, is there a place where can we buy 
them?).  The only problem I foresee with that is the following: to slow 
kickers, we want a deforming (soft) ball, but then the holding rules 
become more difficult to measure.  20% is much easier to measure on a 
hard ball that isn't deforming.  Maybe that won't be such of an issue, 
but ideas are certainly welcome.


*** Increase team size to 6 robots
*** small field size increase (outer size unchanged)
Nawarat/Plasma-Z:
     wants a field size increase (how much?)
     robot number increase needs a bigger field
Krist W./Plasma-Z:
     field increase is good, but don't add robots
Pailo Costa/5dpo:
     doesn't want to add robots
     could start freekicks on the line again to effectively enlarge field
Smail Menani:
     keep five robots to avoid budget problems

Ok, not too much support for these changes right now.  Hopefully more 
people will weigh in on the issue, and Plasma can be more specific on 
how large the field increase should be.  I think a doubling is out of 
the question if we want to have 20 teams competing in Atlanta.  That 
kind of field increase really depends on a shared vision system (which 
itself depends on getting at least automatic referees over ethernet 
working).


*** No travel support for teams that have not participated in a local
Nawarat/Plasma-Z:
     prefers more concrete (and enforced) qualification procedures
Smail Menani:
     will discourage new teams too much, need to keep participation up

Ok, not too much support for this proposal.  The problem I'm trying to 
solve is the teams that show up and do not play, or who do not show up 
at all.  They are likely taking the spot of a team that would have 
played, and that's unfortunate.  It seems to me that first-year teams 
that haven't played in a local competition rarely show up and play, 
while ones who have been at a local competition end up playing.  Maybe 
someone can dig up some data to refute this.