https://github.com/berkeleyautomation/iifl
Implicit Interactive Fleet Learning
https://github.com/berkeleyautomation/iifl
Last synced: 2 months ago
JSON representation
Implicit Interactive Fleet Learning
- Host: GitHub
- URL: https://github.com/berkeleyautomation/iifl
- Owner: BerkeleyAutomation
- License: mit
- Created: 2023-06-27T01:26:37.000Z (almost 2 years ago)
- Default Branch: master
- Last Pushed: 2023-11-01T20:55:43.000Z (over 1 year ago)
- Last Synced: 2025-01-25T21:26:37.102Z (4 months ago)
- Language: Python
- Size: 41.3 MB
- Stars: 3
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# IIFL: Implicit Interactive Fleet Learning
Code for the following paper:
G. Datta*, R. Hoque*, A. Gu, E. Solowjow, K. Goldberg. IIFL: Implicit Interactive Fleet Learning from Heterogeneous Human Supervisors. *Conference on Robot Learning (CoRL)*, 2023.
## Installation
Installation instructions are similar to the [IFL Benchmark](https://github.com/BerkeleyAutomation/ifl_benchmark) on Github. First create a Python 3.8 virtual environment and install dependencies by running `. install.sh`.
To run the IFL Benchmark you will need to install Isaac Gym. Download Isaac Gym 1.0rc3 from https://developer.nvidia.com/isaac-gym (you may need to send a request but it should be quickly approved) and read the installation instructions in the docs to pip install into the virtual environment. You will need NVIDIA driver version >= 470.
Then clone NVIDIA IsaacGymEnvs from https://github.com/NVIDIA-Omniverse/IsaacGymEnvs and pip install it into the virtual environment. **Note**: make sure to run `git checkout 347cfbfaeeb708e7e94bc3bd8e7f2ef069e24fde` for the correct version of IsaacGymEnvs (1.3.0), since IsaacGymEnvs is actively under development.
## Reproducing Results
Simply run
```. scripts/run_[env].sh```
where `env` is one of `{ant, anymal, ball_balance, franka_cube}`. This will run with default expert checkpoints and offline datasets, which you can re-generate if you wish.
## Acknowledgement
IFL implementation is based on the [IFL Benchmark](https://github.com/BerkeleyAutomation/ifl_benchmark).
IBC implementation is adapted from [Kevin Zakka's PyTorch implementation](https://github.com/kevinzakka/ibc).