[robocup-small] RoboCup 2018 Changes

[Christian Lobmeier]

Hey,

Thank you all for the feedback so far. It is greatly appreciated! There
were many questions about different aspects of the intended changes. I will
try to cover as many as possible with this mail.

Topic: Why will the league be split in the first place?

The RoboCup SSL has a very large spectrum of teams in terms of capability.
Our goal as TCs, OCs and ECs is to challenge the experienced teams while
providing a helpful learning environment for new teams. If we just increase
the difficulty of the SSL (by increasing the number of robots for example)
we risk discouraging potential newcomers. To be able to fit the needs of
both entry-level teams and established ones, we came to the conclusion that
splitting the SSL might be the best choice overall. It allows us to
implement big changes at a high rate in division A, while only making few
incremental adjustments in division B.

Topic: There will be some teams that don't end up in the division they
applied for.

Each team has to apply for either division A or B and add a short
rationale. If the choice is reasonable then this team gets to play in its
preferred division, especially if it is a mid-tier team. We just wanted to
avoid something like a top-seeded team choosing to play in division B or a
entry-level team applying for division A (That does not mean that new teams
are generally excluded from playing in division A. The qualification
material is the deciding factor). If the number of teams in the divisions
end up unbalanced, then we have to get by. It would be great if we get an
even split per division, but we won't move a team into a specific division
for no apparent reason. To get an estimate of how many teams would like to
participate in either division, it would be nice if each team could fill in
its preference in the following non-binding survey:
https://doodle.com/poll/92qbii8nmuvyn6gy

Topic: Dividing the teams into two divisions is unfair, because the best
team in division B will get a prize while most of the teams in division A
will not get one.

The divisions are designed to focus on different things: The goal is to
create a fair and competitive environment with high-quality games for teams
in division A while division B focuses more on learning and gaining
experience by playing a lot against other teams. Therefore, there will be
no trophy for the winner of division B. Instead, we will give out a
certificate (like the best ETDP award or the most improved team award).

Topic: The division system splits up equally strong mid-tier teams into
different tournaments.

I think that the mid-tier teams are the ones that will have the most
trouble with this new format. There are a couple of teams that historically
placed third or fourth in their group and don't really fit into either
division. On top of that, a lot of suitable opponents will end up in the
other division. When splitting up a league, there will always be some teams
who are right on the edge. Therefore we let the teams choose the division.
So if a mid-tier team is willing to spend the extra resources (more robots,
more coding, etc), it is free to participate in division A but in the same
time, if they feel like division B is better suited for them, they can
apply for division B.

Topic: We should somehow inform potential new teams and/or universities,
that the entry barrier for division B will be lower.

Yes, we should. However, we don't know yet how far we want to lower it. It
does not make sense to have games in which the robots of both teams don't
even move or touch the ball.

Topic: Will the field be increased as soon as 2018?

For the division A, yes. The size of the current field will not be
sufficient for 16 robots, at least if we want to provide enough space for
concepts like zone defense or passes in the run. We are aware of the fact
that barely any team will be able to afford a full-size field in their lab.
In fact, we may even decide to further increase the field in the future. A
lot of teams already use a partial field since the size got increased back
in 2014. It’s part of the challenge to develop a system using only a
combination of simulation and partial field testing.

Topic: Regarding the rule change that states that only linear goal shots
will be counted, does that mean that curved goal shots (like Op-Amp did)
won’t be scored?

No. The term “linear” was intended to mean that the ball has to stay close
to the ground, in contrast to chip kicks. It does not mean that the ball
has to be kicked in a perfectly straight line.

Topic: Will robots that brush each other get the same penalty as robots who
crash head-on?

We will most likely implement the following criterion: “At the moment of
collision, the difference of the speed vector is taken and projected onto
the line that is defined by the position of both robots. If the length of
this projection is greater than 1.5m/s, the faster robot committed a foul.”

Topic: What do you mean with “Joint Teams”?

We want to make it possible to cooperate and use multiple systems within
one team. This means that a joint team is explicitly allowed to use
multiple forms of strategies and robots within one single team. The team
will be treated as one, which means that it will only hand in one TDP and
can only get one trophy. The limitations of the Mixed Team Challenge do not
apply here.

Topic: Why should people even want to form joint teams?

The idea of allowing joint teams is the result of a lengthy discussion
about team collaboration in general. There is interest from some teams in
focusing on the AI and multi-agent coordination, rather than building and
maintaining a large team of robots. The strategy of forming joint teams
will thus allow such teams to compete together in games with up to 11
robots per team. Joint teams may choose to adopt any of a variety of
coordination strategies, ranging from the standard command protocol, to
mixed AI teams, to a hybrid high-level strategy from one team with
low-level control from the other. This is again intended to alleviate the
high barrier to entry while simultaneously pushing the state of the art in
the league.

If some points are still unclear or you have some general questions, feel
free to ask.


Cheers,

Christian Lobmeier

On Wed, Nov 15, 2017 at 7:12 PM, Christian Lobmeier <pardinensis at gmail.com>
wrote:

> Hey everyone,
>
> The SSL is in a challenging state right now. Especially in comparison to
> other leagues, the SSL suffers from a distinct lack of innovation and
> progress. The technical challenges are mostly ignored, the mixed team
> challenge has been an absolute catastrophe and even the competition games
> looked similar to 5 years ago.
>
> If we want to be able to contribute something to the oft-cited goal of
> 2050, we need changes, big changes. And we need to start to make these
> changes now. So without further ado, lets get right down to business.
>
>
>
>    1.
>
>    Divisions
>
>
> Starting in 2018, the Small Size League will be divided into two divisions
> with separate tournaments: Division A and division B. Division A is aimed
> at advanced teams whereas new and/or less competitive teams can play in
> division B. Each team will only play in one of those two divisions. When
> submitting the qualification material, the team also chooses a preferred
> division including a short rationale. The OC will have the final word. We
> are aiming at around 12 teams per division.
>
> There are multiple benefits of having two divisions.
>
>    -
>
>    The SSL has a rather high entry barrier. Our goal is to further
>    advance the league while keeping the entry barrier at a reasonable level.
>    With splitting the league, we can challenge seasoned teams while still
>    being attractive for newcomers. The qualification criteria for the division
>    B can even be lowered a bit without hurting the competition.
>    -
>
>    Feedback from this year’s RoboCup has shown that it is very important
>    for teams to face off against opponents of similar skill. Dividing the
>    league into two divisions rids a lot of the one sided games in exchange for
>    more interesting match-ups.
>
>
>
>
>    1.
>
>    Increased number of robots
>
>
> This change applies to division A only. The intermediate-term goal of the
> SSL is to be the first league to play 11v11 with actual robots (discounting
> the simulation leagues). To achieve this goal, division A will play with 8
> robots per team by 2018 and with 11 robots by 2019 or 2020.
>
>
>
>    1.
>
>    Increased field size
>
>
> This change applies to division A only. To provide enough space for 8 (and
> later on 11) robots, the field size will be increased. There are no final
> decisions regarding the concrete dimensions, mostly because we don’t know
> yet how much space the SSL is allowed to allocate. The current plan is to
> double the size of the field (to 9x12m) while increasing the width of the
> goal by 20 percent (to 1.20m). The defense area will also be increased by
> 20 percent in each dimension compared to the smaller field. Division B will
> still use a 6x9m field.
>
>
>
>    1.
>
>    Rectangular defense area
>
>
> This change applies to both divisions. Speaking of the defense area, it is
> time to introduce some changes there. Every single team uses some sort of
> defenders that always move along the defense area. It is obviously a very
> effective defense strategy and rather easy to implement. To make the
> defense more interesting, the shape of the defense area will be changed to
> a rectangle (2.40x1.20m for the large field, 2.00x1.00m for the small
> field). The rules regarding the defense area will stay the same. This means
> that attackers and defenders are still allowed to move through the defense
> area as long as they don’t touch the ball.
>
>
>
>    1.
>
>    Reduced ball speed
>
>
> This change applies to both divisions. Even though the field increases
> quite a bit for division A, the maximum allowed ball speed will be reduced
> from 8m/s to 6.5m/s.
>
> Currently, it is very common to score goals from large distance, even
> against strong teams. The reason is that the ball is fast enough to make it
> through a hole in the defense without the defenders being able to react in
> time. This is not a bad thing by itself, the defense should just anticipate
> passes and future attack maneuvers. However, the offense is currently very
> one-dimensional. It all boils down to finding or forcing a hole in the
> defense and exploiting it. There is only little advantage gained by being
> able to bring the ball towards the opponent goal without losing possession
> or getting marked. By reducing the maximum ball speed, long distance shots
> are less likely to succeed resulting in a stronger focus on positional play.
>
> The signature high pace of our league will not be affected by this change.
> First, the viewer will most likely not even notice the difference in shoot
> speed and second, the high pace is not affected by the speed of the robots
> or the speed of the ball, but of the frequency of actions (e.g. passes,
> goal shots).
>
>
>
>    1.
>
>    Excellence Award
>
>
> The RoboCup unites two very different ideas: research in robotics and a
> soccer competition. While it usually synergizes quite well, there are some
> conflicts when it comes to the actual goals and motivations. The goals of
> the RoboCup is to promote research and innovation and advance the field of
> robotics as a whole. A major goal of the teams is to win the competition.
> The problem is that investing time into research does not necessarily
> improve the results in the RoboCup competition, especially if the new ideas
> don’t work out as intended. Instead, the most effective way to succeed is
> to only make small, incremental changes and use tested and trusted
> technologies. Thankfully, a lot of teams do actual research and try out new
> technologies. To honor that, we will introduce the Excellence Award.
>
> The Excellence Award is a judged award given to teams that show a lot of
> overall skill and dedication to the long-term goals of the RoboCup. The
> criteria include:
>
>    -
>
>    Performance in competition games
>    -
>
>    Performance in Technical Challenges
>    -
>
>    Sportsmanship
>    -
>
>    Research Contributions
>    -
>
>    Innovation
>
>
> This Excellence Award will be the most important award given out to SSL
> teams on the RoboCup. More information regarding awards will be released in
> spring.
>
>
>
>    1.
>
>    Qualification process
>
>
> Obviously, the (E)TDP is a central part of the qualification and is used
> to determine whether a team is able to participate or not. However, in 2017
> the (E)TDPs and their reviews have been ignored for the qualification
> results. This won’t happen next year. Each team is required to submit a
> (E)TDP that follows the given guidelines
> <http://wiki.robocup.org/Small_Size_League/RoboCup_2018/Qualification>.
> If the paper lacks quality, the team will only be conditionally qualified
> and has the chance to incorporate the feedback into the paper. If the
> second version does not improve significantly, the team will not be
> qualified for the RoboCup.
>
> Since the RoboCup is in June, the whole qualification process will start
> earlier this year:
>
>    -
>
>    TDP Submission Deadline: January 16, 2018
>
>
>    -
>
>    TDP Peer Review Deadline: February 2, 2018
>
>
>
>
>    1.
>
>    Additional changes
>
>
> The following rule changes will be made this year, the points below are a
> summary of the changes and may differ slightly from the final text in the
> rules.
>
>
>    -
>
>    Only linear shots can be used to score a goal. If the ball height
>    exceeds the robot height between the last contact of an attacker and the
>    crossing of the goal line, the goal will not be scored.
>    -
>
>    Automatic ball placement becomes mandatory for division A.
>    -
>
>    The collision rule will be reworked. If two robots collide with a
>    relative speed of more than 1.5m/s, the faster robot committed a foul.
>    -
>
>    The bar above the goal will be removed to provide better vision in the
>    goal.
>    -
>
>    If two teams want to collaborate and appear and participate as one,
>    they will now be able to register as a joint team.
>
>
>
> We are aware that these changes mean a lot of work for everyone, but we
> think that they are necessary to improve our league as a whole. If you have
> any criticism, good or bad, or if some points are not clear to you, don't
> hesitate to write us a mail. Even though these changes are the result of
> months of discussion, nothing is written in stone, yet.
>
>
> Cheers,
>
> Christian Lobmeier
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-small/attachments/20171128/16537ae3/attachment-0001.html>