[robocup-small] Rule changes for 2005

[Dietmar Walter Schreiner]

hallo everyone,
here are the final comments and ideas to planned rule changes of the vienne
cubes vision and ai divisions.

we think robocups main goal is the development of a system which allows robots
to play "real" (=fifa) soccer. in its current state, robocup is more like ice
hockey than soccer and some of the rulechanges and ideas discussed here on the
mailing list would bring robo soccer even farther away from soccer than now.

so our ideas deal with "getting a real soccer match" and no workarounds to
support weak ais.


* ad change 1.)

we agree with that. further clearification will be nescessary how to define
passive dribbling (distance ball-bot and so on...) and how to mark dribbling
devices so that dribbler-activity could clearly be spotted.

another problem will be the referees "vision" (was the dribbling device active
and for what distance), so in the long term we would like to see an autometed
referee system (more about that below).

* ad change 2.)

we can't see any benefit here. the planned 10cm rule is one step away from fifa
soccer. if any bots would need more room to navigate the out-of-bounds area of
the field could be enlarged but the ball should be placed on the lines.

* ad change 3.)

if this rule is ment to avoid passive gameplay (while you are in lead) we will
agree on that. also consider to not simply restart but to put a penalty on
inactivity.

* ad change 4.)

we fully agree on that (as we got a very good explanation on the mailing list).
we think that no used technology should be intrusive.

another thought could be to limit the number of used bandwith/channels per team
as bandwith consumption will surely raise in the future.

* ad change 5.)

see 3.)

* ad change 6.)

why would you need that ? as lighting conditions may vary, the vision system has
to be able to handle color calibration. a good vision system will also be able
to handle any pattern on top of a bot.

the only things that will make sense here are, that the rules specify a period
of for example 15 minutes befor the match, where the calibrations and
adjustmends with the opponends top-labels for the vision system could be done
and that neighbouring colors to the balls color in color space are prohibitted
for bot markers.

* ad change 7.)

that is our favorit one ;) we would like to have 11 bots per team but...

with given field size the proportions of the bots, the field and the goal are
totaly distorted. with the current field dimensions another bot would be too
much, we would overcrowd it.

again in the long term we would like to see the field enlarged so more bots
could come in play. this would also solve some other problems:

an enlarged field will reduce the problem of "high-speed"- and chip-kicks as the
ball will have to travel a much longer way. as the bots also will have to travel
long distances, passing will bekome a more important move and the relation goal
to goalkeeper-diameter will shrink forcing the ai to do more intellegent
defencive work. and finaly the energy consumption of the bots wil raise makeing
high level planning an option.

* ad change 8.)

we agree on that.

* ad change 9.)

we also think that a rule change has to be done on that topic, as the small
field size allows chip kicks to the goal from anywhere on the field.

a simple solution could be, that on any indirect kick two bots of the kicking
team have to touch the ball to score a goal (if only one bot touches the ball a
goal is not counted). this will allow to use the chip-kicker for indirect free
kicks but will deny a direct goal.

* ad sloped field

we don't like the sloped field as it brings us away from real soccer. if the
playing ai is unable to hold the ball in game this should lead to better ai
development, not to bending rules.

nevertheless we do see the problem of very short play-cycles. these cycles
result in the bad proportion of movement-speeds (bots and ball) and field size.

* ad "energy accounting"

we would like to see that as it will bring new strategies into gameplay and make
bot soccer more human-like.

each bot will get a fixed amount of "movement" energy, which decreases on moving
around. if the energy level drops down to 0 the bot can't move anymore. a bot
not moving regains energy over time. if the energy level exceeds a given
threshold the bot may move again.

to enable anything like that, we will need a automated referee with a global
vision. the referee tracks the movements, accounts energy consumption and can
"stun" a bot exceeding its personal energy reserves.

as that rule brings very komplex parameters into strategic planning it should be
introduced in a long term.



so long from vienna
dietmar schreiner