- Go to "chrome://flags/"
- enable "Experimental Web Platform features"
- restart chrome
- Use beta.pybricks.com to edit and upload the hub program
conda create --name lego python=3.8
conda activate lego
pip install pybricksdev --pre
pip3 install torch torchvision torchaudio
pip3 install tensordict
pip3 install torchrl
pip3 install hydra-core
pip install wandb
Update your client script on the PyBricks Hub whenever you want to run a new environment
We created a PyBricks GPT which is an expert on the PyBricks documentation and can help you with coding and debugging client code for the PyBricks Hub.
Code examples from PyBricks. Might be interesting as expert policies for offline RL. GitHub Pybricks examples
-
Milestone Cleanup []
- cleanup base environment
- addons? balance?
- runaway-v0 as it currently is [X]
- runaway-v1 with independent wheel control [X]
- might play around with reward function to stand still when max dist reached
- if reward done well could be maybe applied to a labyrinth ?
- fix disconnect / reconnect error [X]
- cleanup base environment
-
Milestone CatRobot
- integrate CatRobot environment
- create CatRobot Env [X]
- create CatRobot offlline datasets []
- other ideas but walking max distance? [ ]
-
Milestone Offline Datasets [ ]
- collect dataset for wall-distance/run away [ ]
- create datasets that can be loaded for torchrl offline datasets [ ]
- do offline training for run-away robot []
- create script for offline training and evaluation [ ]
-
Milestone RoboArm
- integrate RoboArm environment [ ]
- Possible task?
- reach some point e.g. ReachingEnv
- Grab cube
- more ?
- Create offline dataset