[x] We need to provide a proposal for the robots to go outside
[ ] Also try to cooperate somehow with the rescue league
[ ] Maybe there can be more specialized arenas: An apartment, a hospital, supermarket, kindergarten.
[x] Longer term autonomy (bring back E-GPSR)
[ ] Go into the public (may replace RoboZoo): We can go shopping at some sort of Home Depot/Ikea accros the Leipziger Messe?
From the TC's themselves
[x] We want to keep the 3 attempts.
[ ] For more robot-robot interaction, there should be some sort of protocol with which the robots cooperate. Of course, there already is a good protocol and language for this, and it is called "English" Maybe the robots can just do GPSR with each other? Instead of speech it can be modem-like modulated sounds or for added awesomeness R2D2-like sounds.
[x] We should have a standardized Continue-rule, like showing the given command encoded into a QR-code. At least get rid of the typing on the robots. Then understanding the command by speech gives 100% of the points, a QR-code 25%, OCR on printed text eg. 50% and OCR of written text 75%. Anyways, make it standardized. Then, a command generator like the one for GPSR can show a command on a laptop and generate a QR code. If speech fails, just show the robot the laptop.
[ ] A refbox would be nice. Connect it to a TV for the audience.
[x] For going into the public: The RoboZoo can be replaced with something where the robots drive around in the public and ask the audience to bring a token to the referees. The most convincing robots in this wins!
[x] Of course, when going into the public, all robot should be equipped with wireless emergency buttons that can be operated from a few meters away.
[x] RoboZoo: All robots should speak English, not the local language?
[x] The robots should do more advanced grasping, e.g. grasp a pen, paper, phone etc. DO a top-grab.
[ ] Having a child as a volunteer operator could be very engaging and interesting. As an outreach, we can write a out a (robotics) competition in Leipziger schools for who will be the lucky kid. They should speak English of course.
[x] Let all teams record audio, images, map data. Next to a object recognition report file, let the teams also record audio and annotate it with the understood question.
Other suggestions heard at RoboCup
[ ] Maybe have the scores for manipulation and recognition separately. Separate challenges would take too much time maybe but is also nice.
We also got some info about the Leipzig RoboCUp event in 2016:
[ ] The rulebook's technical requirements should be ready by the first of december.
[ ] If would be handy to label in the Rulebook's TeX-files which sentences are about the areans requirements so we can make some index.
[ ] For example, we should maybe request a (more) silent room to test speech recognition. Then compare it to the normal RoboCup noise performance.
[x] It would be nice to make the 2 arenas really look like they are in an apartment building, with a doorbell and mailbox etc. The robots then start form the walkway of the building.
[ ] There should be a dedicated IT-department for RoboCup who know their shit about WiFi and can be on direct call :+1:
[ ] Speech interactions are now often interrupted by public announcements in the venue. This is very, very difficult for the robots, as also human cannot talk properly anymore. One way to deal with these is to have announcements only in 10 minute intervals. So, you an safely start a challenge after an announcement has finished, and you know there will not be another one until 10 minutes. If there is nothing to announce, just do some sort of beep, as a cue to us.
[ ] An internet connection over the WiFi. The latency should be so large that direct teleop is impossible. Also require to deal with the always crapppy and overloaded internet connections typical of RoboCup.
[x] Normizle the score for RoboZoo, so it is not influencing the benchamarks too much. Max score in a benchmark may be 200, and 50 for RoboZoo. If the best score in some benchmark is 40, then RoboZoo is renormalized to 10 (also a fourth still)
[x] Collect reports + benchmarking data from speech recognition (audio recordings), object recognition (2D images + 3D pointcloud), manipulation (3D It would be pointclouds), person recognition (2D images)
[x] For speech recognition, use more similar questions to make telling the difference harder
[ ] Finals: maybe be more relaxed with timing so teams are able to put up a good show with a minor delay. The MSL finals are also delayed when needed with hundreds of people waiting.
[ ] It would be interesting to see what the current robots can achieve with teleoperation.
[ ] The teams are allowed to use a custom container. Maybe a custom blanket will help a bit in the Wake-me-up make-the-bed part.
[ ] Speech recognition: mix direct and indirect speech recognition.
RoboCup Trustees want us to
From the TC's themselves
Other suggestions heard at RoboCup