IEEE Robotics & Automation Magazine - September 2015 - 151

AdultSize
TeenSize
KidSize

this change would result in less successful strikers and a reduced goal count or in an increased goal count due to more
own goals. The drop in the average number of TeenSize goals in
2013 indicates that the strikers may be less successful. However,
the drop in average goals in the KidSize subleague is only
minor, if statistically significant at all, for the respective year.
The authors expect other underlying influences to exist. With
typically more experienced teams in the TeenSize subleague,
own goals may not have played a significant role, unlike in
KidSize, where the drop in proper scoring was mostly compensated by own goals. However, no records exist to support the
explanation. The AdultSize goals do not show a similar effect,
which again can be explained by playing on a single goal in this
subleague. The Humanoid League also introduced a number of
technical improvements to robotics. The team NimbRo has
been working intensely on the stability of walking and contributed the concept of capture steps to keep robots from falling
after bumping into each other [12]. Other examples are the design of a series elastic actuator add-on to the widely used
Dynamixel servos, which was presented by a joint team from
Universidade Federal do Santa Maria in Brazil and Ostfalia
University from Germany [13]. The elastic element was intended to absorb shocks, store energy, and possibly, with an additional displacement sensor, allow for dynamic gait in the future.

United United
Kingdom States Australia
Bangladesh
The Netherlands
Thailand

Brazil
Canada

Taiwan ROC
South Korea

Chile

Singapore

China

Mexico

France

Japan
Italy
Israel

Iran

2010
2011
2012
2013
2014
2015

Germany
India
Indonesia

Figure 9. The distribution of teams by country in the RoboCup
Humanoid League.

Average Goals per Game

45
40
35
30
25
20
15
10
5
0

20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15

Number of Teams

In step with the RoboCup competition in general, the maximum number of teams that can reasonably be supported
within the current limits on infrastructure, e.g., number of
playing fields and space for the teams, has been reached. This
is especially true in the KidSize competition, where a limit of
24 fully qualified teams plus a few additional teams (usually
one or two) qualified for the technical challenges was introduced. The teams are qualified from a group of about 31 applications every year. This number has remained fairly constant
throughout the years. However, for 2015, the number of
KidSize teams will decrease slightly. Records of the geographic
origin of teams over recent years show a significant involvement from countries like China, Germany, Iran, Mexico,
Taiwan [Republic of China (ROC)], and the United States
(Figure 9). Some countries have a stable contribution, e.g.,
Germany with four to five teams every year. However, often
the individual participation appears to be subject to the host
country of RoboCup. Teams report travel costs and logistics as
increasingly relevant aspects of participation. Overall, some locations, such as The Netherlands (2013) and Istanbul (2011),
had slightly more participating teams than other locations,
such as Mexico (2012). However, the influence is rather minor,
leading to a variance of about two teams per size class.
As in regular soccer, statistics on goals in RoboCup humanoid robot soccer exist (Figure 10). The number of goals
may be considered as a suitable general performance indicator, but the Humanoid League constantly adopts the rules toward the 2050 game. One would, therefore, expect to have an
increasing average goal count that drops after the introduction of new rules. However, goal statistics show only a weak
correlation with rule changes. For example, when increasing
the field size for TeenSize in 2011, there was a drop in average goals. When doing the same change in KidSize in 2014
with, otherwise, similar conditions, the average number of
goals actually slightly increased. Then again, not observing a
similar drop in AdultSize, in 2012, when field size was increased for this subleague, can be explained by the specific
structure of the dribble-and-kick competitions with a single
robot in each team.
The consequences of the rule change of abandoning blueand yellow-colored goals in 2013 are also not reflected clearly in
the average goal count. Upon introduction, it was discussed if

9
8
7
6
5
4
3
2
1
0
2010

Kid
Teen
Adult

2011

2012

2013

2014

Year

Year
Figure 8. The number of qualified teams for RoboCup World Cup.

Figure 10. The average goals scored per game in the RoboCup
Humanoid League.

September 2015

IEEE ROBOTICS & AUTOMATION MAGAZINE

151

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2015