This is a custom Reinforcement Learning environment for training a Roomba style robot to do point-to-point navigation, including with obstacle avoidance.
It can also upload a learned hugging face RL algorithm trained on the custom environment to the Hugging face repository
- Need to change is_atari to use unregistered environments
- Make sure the action and observation space are accurate
- Make sure
reset()returns ONLY obs!
- Add preview from links in readme
- Make sure new utils works
- Performance breakdown of model submission
- Fix model preview issue (model output not used correctly)
- Somehow, the particles' movements aren't that random (i.e. change of direction is constant)