irom-lab / dragen

Distributionally Robust Policy Learning via Adversarial Environment Generation

DRAGEN - Distributionally Robust Policy Learning via Adversarial Environment Generation

This repository includes codes for synthetic trainings of the two robotic tasks in the paper:

1. Swinging up a pendulum with onboard vision
2. Grasping realistic 2D/3D grasping

Although the codes for all experiments are included here, we only provide instructions for re-producing pendulum experiments here. Re-producing grasping experiments relies on several custom packages (SDF sampling, mesh processing, off-screen rendering) that requires careful installation; we are happy to provide instructions upon further request (but would be difficult to keep anonymity). We do provide the instructions for running grasping policy training in 2D setting (with PyBullet simulation).

Dependencies:

pip install with python=3.7/3.8: pybullet, Pillow, torch, torchvision, gym, psutil, python-mnist, h5py, pyyaml, perlin_noise (pendulum baseline); shapely, trimesh (grasping)
pip install -e . to install locally: rlpyt folder included in the supplementary materials, which is a custom fork of rlpyt library for SAC training. Running pendulum experiments requires a GPU.

File naming:

ae is for DRAGEN training; dr is for baselines (noise, domain randomization, etc).

Running pendulum experiments:

1. (For Digit) install and download MNIST dataset here and download USPS dataset here; (for Urban) download Apollo Synthetic dataset here.
2. Generate Landmark/Digit/Urban images by running generate_landmark.py and process_mnist/usps/apollo.py.
3. Specify dataset paths in provided config files.
4. Run python run_pen_ae.py {config filename}, or python run_pen_dr.py {config filename} for baselines.

Running grasping policy training in 2D setting:

1. Generate training and testing objects by running generate_2d_shapes.py.
2. Specify object path in train_grasp.py and run it.

dragen's Issues

Recommended way to install Python requirements

Hello, is there a recommended way to install the Python package requirements for this repo? I've tried a lot of combinations with conda but haven't been able to get all of the packages to install without incompatibilities. Specifically, is there a particular PyTorch/CUDA version that works well for this repo? Thanks!