IEEE Robotics & Automation Magazine - June 2015 - 14

Colored blob segmentation,
database image recognition,
and landmark recognition are
widely available algorithms in
many computer-vision frameworks. Besides identifying the
objects in the task, landmarks
can help to solve the localization problem of the robot. Even
in a small environment, the
uncertainty added by the walking motion makes it impossible
to rely on open-loop motions
for the manipulation tasks, thus
imposing the need for visionbased closed-loop algorithms.

The Challenge
The challenge consisted of
three independent tests: 1)
switching off the kitchen stove,
2) making a shopping list, and
3) preparing a meal. Each test
was graded on a 20-point scale.
Grading was performed on the
day of the finals according to
the robot's behavior in the real
room and not in a simulation.
Each team had two attempts
Figure 2. Participants with their robots at the HUMABOT
per test, lasting 3 min each. The
Challenge, held 18-20 November in Madrid during the IEEE
best grade of the two attempts
Humanoids Conference.
was retained as the score. In
in Madrid and had to withdraw. In the addition, for each of the tests, a bonus
final challenge, the qualified teams was awarded based on the success and
The Participants
(Figure 2) came from Colombia, Mexi- speed of the task, and a penalty was
Initially, nine teams applied to partici- co, Spain (including three teams, two issued if the team had to intervene durpate in the competition, originating from Barcelona and one from Madrid), ing the robot's attempt.
from four different continents (Africa, and Sweden. Two Spanish teams used
During the first day of the competiSouth America, Asia, and Europe). the DARwIn-OP robot, provided by tion, all the teams had access to the
However, some of the teams were Ro-Botica company. All the rest of kitchen setup to test their algorithms
unable to participate at the competition teams used the NAO platform.
before the final scenario. Although the
kitchen model was well known and
accessible for all the participants, all
teams needed to adjust their algorithms in the real environment to hanCONSULTING | CONTRACT ENGINEERING | PROJECTS
dle the ambient lighting conditions. It
is notable that during these trial tests,
some teams received better results
than during the final competition. For
instance, as shown in Figure 3, one of
the robots was able to identify the
tomato toy, grasp it, and put it into the
cooking pan (all of the objectives in
the third test). Nevertheless, during
the competition, this robot was unable
to perform the tasks.
The first test consisted of a safety
task: one of the burners in the kitchen
was lit, and the robot had to turn it off
(Figure 4). The lit burner was selected
at random, and two switches allowed
the robot to control each burner independently. The relationship between
the burners and switches was known in
advance, as were the relative positions
of the elements. Markers were allowed
Advanced robotics research is our specialty and we work with
on the furniture for simplifying the
clients world wide. We design custom robot hardware. We provide
localization problem. There were two
engineering support for ongoing projects. Come view our portfolio.
rounds in this test, and only two teams
were able to switch off the burner during the competition. The other robots
suffered from poor precision in the
www.robot.engineer
localization of the robots with respect

Experienced
engineers,
exceptional
robots
14

IEEE ROBOTICS & AUTOMATION MAGAZINE

june 2015

http://www.robot.engineer

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