Dear Robocup SPL Community,

in accordance to the terms and conditions of the BHuman code release license we - the team R-ZWEI KICKERS, UAS Kaiserslautern, Germany - like to announce the use of part of the B-Human 2019 code in the upcoming competitions 2021.

Currently, our code base is a modified version of the B-Human Code Drop 2019. We modified:

- team behaviour control and the SimRobot environment to allow for automatic test procedures.

- We developed completely custom features including the teach-in of robot behaviour (using PS4-controller, mapping key presses to skills and recording them for playback), and

- a data mining analysis which clusters the robot poses to identify trigger points for our teach-in cards, and

- coupled the data mining classifier to a reinforcement learning (Q-learning) module that learns the most-effective combination of rule-based and teach-in cards while playing simulated games.

- Additionally, we developed our own image recognition and a multi-object detection model, based on Keras/Yolo

- Currently, we develop our own kick motions and dribbling skills, and we start applying

- data augmentation to increase the robustness of our object detection and

- expand the automatic referee to apply the passing-challenge rules, so we can use Q-learning here as well.

This email is sent respecting the code usage of the B-Human team available on their github page. R-ZWEI KICKERS would like to thank B-Human for developing this great platform!

You can reach us at http://smart-machines.hs-kl.de/projekte/hs-kl-robocup/ If you are interested in more technical details please mail me for a copy of our white paper.

Mit freundlichen Grüßen / Best regards

Prof. Adrian Müller, PSM-1, CSM http://www.hs-kl.de/~amueller Projektleitung AK Smart-Machines http://smart-machines.hs-kl.de/

tel: +49 631/3724-5329 fax: -5313 office: H243 ------------------------------------------------------------------------------------------------------------------------- Hochschule Kaiserslautern University of Applied Sciences Fachbereich I/MST Department of Computer Sciences and Microsystem Technology Amerikastr. 1 D 66482 Zweibruecken -------------------------------------------------------------------------------------------------------------------------