Robots for Football
summary
Tech United Eindhoven builds wheeled middle‑size league robots with onboard omni vision camera and a front Kinect sensor, runs CNNs offline for team‑colour ID, and experiments with running models on Jetson/TPU for real‑time use. Four years of match logs feed temporal occupancy grid maps and a convolutional model that predicts own‑robot actions with about 77% accuracy using ten frames. The group of ~15 builds more than six soccer robots, demos ball handling and kicking hardware that mostly works but still fails in live demos.
AI-written summary · about
chapters (24)
- 0:00Autonomous robot soccer 2050
- 1:45RoboCup leagues and goals
- 3:53Autonomy and team dynamics
- 6:02Tech United's championship history
- 6:48Robot soccer at Portuguese Open
- 8:38Emotional highs and lows finals
- 9:43Meet Tech United Turtles
- 10:24Ball handling and dribbling
- 11:52FIFA rules for robot soccer
- 12:38Autonomous shooting capabilities
- 13:47Sensors and vision systems
- 16:45Robot detection color challenges
- 17:48AI for opponent team ID
- 19:56CNN for robot detection
- 22:29Opponent action prediction
- 24:19Robot actions temporal maps
- 27:16Neural networks accuracy
- 29:24Why build soccer robots
- 30:51Tech United's Amigo and Sergio
- 32:14Caretaking robot challenges
- 34:21Soccer and caretaking robots
- 35:28When robotics goes wrong
- 36:52Future of robot soccer and AI
- 37:57Q&A
related talks
46mRobotics in Our Everyday Lives: A Product Designer's Perspective
23mAI.js: Robots with Brains!
43mBuilding Qrio, a Bot That Plays Videos for My Toddler
39mChoreorobotics: Teaching Robots How To Dance
47mRobolove
12mNodebots Are Back
35mDancing with Robots
37mHands on learning experience for non-technical people