Our Raspberry Pi 3 B is the core processing unit for our humanoid robotics platform [1]. This is in the context of the RoboCup competition [2]. With it we run:

* Vision - A custom CNN using YOLO [3], where we are able to process a 256x256 (input is scaled) at 10fps to detect bounding boxes for balls and goal posts

* Localization - Kalman filter (mainly currently used for tracking rotation)

* Networking - Game controller (referee) [4], team communication [5] and a debug interface [6]

* Behaviour - A hybrid state machine

* Walking - Inverse kinematic walk with a balance system [7]

Feel free to ask questions. We plan to open source everything (everything) in a month to two months.

[1] https://humanoid.science/

[2] https://www.robocup.org/

[3] https://pjreddie.com/darknet/yolo/

[4] https://github.com/RoboCup-Humanoid-TC/GameController

[5] https://github.com/RoboCup-Humanoid-TC/mitecom

[6] https://github.com/hellerf/EmbeddableWebServer

[7] https://github.com/Rhoban/IKWalk

EDIT: Bullet points on different lines