Building a Robot Soccer Team David Cohen and Paul Vernaza University of Pennsylvania
Dec 25, 2015
Building a Robot Soccer Team
David Cohen and Paul Vernaza
University of Pennsylvania
What is Robocup? International robotics competition
and symposium Goal: By the year 2050,
to develop a team of fully autonomous humanoid robots that can win against the human world soccer champion team.
Many different leagues in competition
The Four-Legged League Team composition Rules Field Uniforms Ball Halves Challenges Robots used…
The Sony AIBO
“Robotic Pet”/research platform Construction (deconstruction) Battery, memory stick, wireless
card Sensors, effectors What kinds of sensors and
effectors are present on the AIBO?
Major Challenges
Can you guess what the major challenges are?
We will talk about four of them…
Major Challenges
Vision: making sense of the camera image
Localization: finding out where the dogs are located on the field
Locomotion: teaching a dog to walk Behavior: decision-making and
cooperation
Vision
1. Build a color table2. Convert image pixels to distinct
colors3. Connect the dots to make blobs
Images courtesy of CMU
Localization
What does it give us? Why localization? Humans don’t
have it Beacon-based localization Line-based localization
Locomotion A good walk is omnidirectional,
fast, and stable Teams differ greatly in locomotive
ability Being able to walk fast gives a
team a great advantage Walk is generated using inverse
kinematics
Behavior Consists basically of
telling the dog what to do in different circumstances
Fully autonomous Modeled with a state
machine Important issues include
collaboration, obeying rules, robustness
Robocup 2003 Penn finished 2nd out
of 24 teams Teams from around
the world participated Each team had its
own characteristic strengths and weaknesses (UNSW, Germany, U. of Washington, etc.)
Plans for the Future New robots: ERS-7 Improving walk speed Improving localization
with more natural visual cues
Using AI techniques for behavior
Passing