Release Notes

Version 0.8.5

Release Highlights

  • RL/GT: Improvements and fixes of training - especially stagnation detection

  • RL/GT: Introduction of new environment Double Pendulum

  • RL/GT: Improvement of installation procedure of UR5 environment

  • Numerous corrections and improvements

Fixed Issues

  • BF: Loading/storing of objects uses Dill instead of Pickle

  • RL/GT: minor corrections on MultiCartPole environment

  • RL/GT: performance optimization of SB3 wrapper (off-policy)

  • RL/GT: Wrappers OpenAI Gym and PettingZoo: removed redundant cycle counting

Documentation Changes

  • Class diagrams resized

  • Structure of environment documentation assigned to example documentation

  • Appendix 1 (examples) restructured

  • Disclaimer added

Version 0.8.1

Release Highlights

RL/GT: New environment Multi Geometry Robot

Version 0.8.0

Release Highlights

1 Sub-Package Reinforcement Learning

  • Support of model-based agents (MBRL)

  • Support of multi-agents (MARL)

  • Training and hyperparameter tuning with progress detection

  • Ready to use sample environments

  • Full 3rd party support (OpenAI Gym, PettingZoo, Stable Baselines 3)

2 Sub-Package Game Theory

  • Models for cooperative game theoretical approaches

3 Numerous ready to run examples

4 API Reference