- Corrections for the side_effects_sokoban wall penalty calculation.
- Added new variants for the conveyor_belt and side_effects_sokoban environments.
- Added the rocks_diamonds environment.
- Removed movement reward in conveyor belt environments.
- Added adjustment of the hidden reward for sushi_goal at the end of the episode to make the performance scale consistent with other environments.
- Added tests for the sushi_goal variant.
- Added a new variant of the conveyor_belt environment - sushi goal.
- Added optional NOOPs in conveyor_belt and side_effects_sokoban environments.
- Python3 support!
- Compatibility with the newest version of pycolab.
Please make sure to see the new installation instructions in README.md in order to update to the correct version of pycolab.
- Added a new side effects environment - conveyor_belt.py, described in the accompanying paper: Measuring and avoiding side effects using relative reachability.