Code for Rapid Locomotion via Reinforcement Learning

This repository provides an implementation of the paper:

Rapid Locomotion via Reinforcement Learning
Gabriel B. Margolis*, Ge Yang*, Kartik Paigwar, Tao Chen, and Pulkit Agrawal
Robotics: Science and Systems, 2022
paper / bibtex / project page

This environment builds upon the legged gym environment by Nikita Rudin, Robotic Systems Lab, ETH Zurich (Paper: https://arxiv.org/abs/2109.11978) and the Isaac Gym simulator from NVIDIA (Paper: https://arxiv.org/abs/2108.10470). Training code builds upon the rsl_rl repository, also by Nikita Rudin, Robotic Systems Lab, ETH Zurich. All redistributed code retains its original license.

Our initial release provides the following features:

Support for the MIT Mini Cheetah and Unitree Go1 robots.
Implementation of the Grid Adaptive Curriculum strategy from RLvRL.
Implementation of the teacher-student training approach from RLvRL, which is based on Rapid Motor Adaptation.
Support for scaling experiment management with ml_logger and jaynes.

Quick Start

CODE STRUCTURE The main environment for simulating a legged robot is in legged_robot.py. The default configuration parameters including reward weightings are defined in legged_robot_config.py::Cfg.

There are three scripts in the scripts directory:

scripts
├── __init__.py
├── play.py
├── test.py
└── train.py

You can run the test.py script to verify your environment setup. If it runs then you have installed the gym environments correctly. To train an agent, run train.py. To evaluate a pretrained agent, run play.py. We provie a pretrained agent checkpoint in the [./runs/](.

Option A: Using Docker

The recommended way to run the code is to use the docker image. We provide a Dockerfile. To build the docker image, use a Ubuntu 18.04 or Ubuntu 20.04 machine, and follow these steps:

Clone this repository
Obtain IsaacGym_Preview_3_Package.tar.gz from the NVIDIA website (https://developer.nvidia.com/isaac-gym). You'll have to create a free NVIDIA account. After downloading the file, place it in this repo at: docker/rsc/IsaacGym_Preview_3_Package.tar.gz.
Build the docker image: cd docker && make build
Launch and enter the docker container: cd docker && make run
[To enable GUI windows] In a separate terminal window, on the host machine, run bash docker/visualize_access.bash

Option B: Native Installation

If you'd prefer to run our code in your own python environment, you can follow the instructions below:

Install pytorch 1.10 with cuda-11.3:

pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

Install Isaac Gym

Download and install Isaac Gym Preview 3 from https://developer.nvidia.com/isaac-gym

unzip the file via:

tar -xf IsaacGym_Preview_3_Package.tar.gz

now install the python package

cd isaacgym_lib/python && pip install -e .

Verify the installation by try running an example
```
python examples/1080_balls_of_solitude.py
```
For troubleshooting check docs isaacgym/docs/index.html

Install the `mini_gym` package

In this repository, run pip install -e .

Verifying the Installation

If everything is installed correctly, you should be able to run the test script with:

python scripts/test.py

The script should print Simulating step {i}. The GUI is off by default. To turn it on, set headless=False in test.py's main function call.

Training a Model

To train the mini-cheetah robot to run and spin fast, run:

python scripts/train.py

After initializing the simulator, the script will print out a list of metrics every ten training iterations.

Training with the default configuration requires about 12GB of GPU memory. If you have less memory available, you can still train by reducing the number of parallel environments used in simulation (the default is Cfg.env.num_envs = 4000).

To visualize training progress, first start the ml_dash frontend app:

python -m ml_dash.app

then start the ml_dash backend server by running this command in the parent directory of the runs folder:

python -m ml_dash.server .

Finally, use a web browser to go to the app IP (defaults to localhost:3001) and create a new profile with the credentials:

Username: runs API: [server IP] (defaults to localhost:8081) Access Token: [blank]

Now, clicking on the profile should yield a

Evaluating the Model

To evaluate the most recently trained model, run:

python scripts/play.py

The robot is commanded to run forward at 5m/s for 5 seconds. After completing the simulation, the script plots the robot's velocity and joint angles. To modify the commanded velocity, you can edit line 109 of the script.

The GUI is on by default. If it does not appear, and you're working in docker, make sure you haven't forgotten to run bash docker/visualize_access.bash.

Support

For questions about the code, please create an issue in the repository.

Bibtex

@inproceedings{margolisyang2022rapid,
  title={Rapid Locomotion via Reinforcement Learning},
  author={Margolis, Gabriel and Yang, Ge and Paigwar, 
          Kartik and Chen, Tao and Agrawal, Pulkit},
  booktitle={Robotics: Science and Systems},
  year={2022}
}

Simonforyou/rapid-locomotion-rl