As we hit Toronto on our way back to Germany we took our chance to visit the active
and attentive vision lab at York university in Toronto. The special thing about this group is
that they attempt to solve computer vision problems like humans do. Meaning they
focus on research on monocular or stereo camera solutions instead of RGB-D cameras.
Markus Solbach, an old study friend of Raphael is currently pursuing his PhD. there and spontaneously organized a lab visit
including talks of the group members and integrated us as well. So we gave a short
talk about what we do for RoboCup and what our research focusses on. After the talks
we were invited to join a panel discussion about robotics. What an amazing day!
This morning we were in a sports bar and saw the soccer match.
Currently we are waiting at the Toronto airport to fly bag. Our luggage pieces with
Lisa inside are already checked in. When you want to see Lisa in action head to
the Visual Computing day on Friday. There we show Lisa and TIAGo, if he arrives in
Homer! After being very close to the Pumas before the finals it was
obvious to get a very tough match between Pumas from Mexico and us. And in the
end we made it with a demo that turned out to run just like expected. We showed that Lisa can observe a human while fulfilling a task
and imitate the action that was just observed, further the observed action was transferred
to TIAGo which then also was able to execute the action by speech with exchanged parameters.
We plan to prepare a video of the finals once we are back home. Now we are going to
celebrate with the other teams during the banquet.
We want to thank PAL Robotics for the sponsoring of a TIAGo robot and the organization of the
transport which supported us massively. And we want to thank our supporters namely the University of Koblenz-Landau,
the EINST e.V., Stiftung Zukunft Sparkasse as well as SHD.
So Stage 2 results are given, we are still leading followed closely by Pumas.
Meaning there battle for the first price between both of our teams. Mexico may
won the soccer match last time at the worlcup against Germany but we are putting
much effort that this won’t happen in the RoboCup finals.
The finals will take part later so we don’t have time to write long posts … and
instead are going back to testing.
The first competition day ended. Good news: We are leading currently, followed by Pumas from Mexico
and the SocRob team from Portugal. After the initial troubles that mostly were influenced
externally we are running well. First we scored best in the Speech and Person Recognition task.
Lisa recognized by a high margin the most speech commands and questions. In the Speech And
Person task the speech recognition and understanding capabilities are evaluated.
First the robot analyzes a crowd consisting of a mixture of females and males,
waving, pointing, lying, sitting and standing people. The robot analyzes the crowd
by it’s gender and pose and summarizes. People, like the audience standing outside
the arena need to be filtered out. After tests the knowledge of the robot by speech.
A huge variety of questions, related to trivia, the environment and even the analyzed
crowd could be asked. Lisa did well. She understood most questions and answered them properly and
even estimated the direction from which the question was given to answer it to the
Next up was the help me carry task. In this task, the robot has to support a person
in carrying in the groceries which are stored at a bag in a car. So first the robot
needs to follow a person from inside the apartment to the wild, where somewhere a car is
parked. The car location is previously unknown as well as the environment outside the
arena. This means the robot creates a map during the following stage. Once the car
is reached the person hands over a bag, which the robot should bring inside on it’s own, with the newly
generated knowledge about the outside of the arena. Once the robot is back inside
, a person that want’s to be guided outside is waiting. On the way back there will also be hard
to see objects that need to be avoided like some groceries that dropped down or
cloth hanger that magically will appear in the robot’s path or a door that will be shut and needs to be opened then.
Lisa wasn’t fit for that after the Speech and
Person recognition task. In two from three runs she lost the operator but in one
she managed to follow the operator up to the car.
As the last task for today there was the Storing Groceries task, where the
robot is supposed to put random groceries from a table into the right location
of a cupboard. Right means i.e. that drinks from the table should be stored
to drinks in the cupboard. And to make it even more complex, there will also
be unknown objects of each class that should be stored correctly. And of course the cupboard
door is closed initially and the table position is just known coarsely. TIAGo did
well in opening the door all the time but faced some problem during the gripping.
So after all we are quite happy about our achievements from today. And are looking
for tomorrow’s tasks, namely the General Purpose task and the Open Challenge.