Awesome

About this Repository

This repository serves as an extension to the OmniIsaacGymEnvs framework, enhancing its capabilities with additional robots and advanced features. The primary goal is to provide a more flexible environment for deploying various code-generated robotic and performing complex navigation tasks.

Key Features

• Extended Robot Library: Introduces new robots to the existing OmniIsaacGymEnvs framework, enabling a wider range of simulations and applications.
• Navigation Tasks: Includes a variety of navigation tasks designed to test and benchmark robotic performance in different scenarios.
• Modular Extensions: Supports features like domain randomization to improve robustness and automated curriculum learning to optimize training processes.

Robots Included

• 2D Satellite: Simulates basic satellite maneuvers in a 2D plane.
• 3D Satellite: Extends satellite control to 3D space for more complex operations.
• Heron USV: A surface vessel used for aquatic navigation tasks.
• Turtle-bots: Compact mobile robots suitable for indoor navigation.
• Husky: A rugged, all-terrain robot for outdoor navigation.

Navigation Tasks Available

This library provides a set of predefined navigation tasks for robotic control and reinforcement learning. It allows for easy extensions to add new tasks or modify existing ones to suit different requirements.

Tasks list

• GoToPosition
• GoToPose
• Track Linear Velocity
• Track Linear & Angular Velocity
• Track Linear Velocity & Heading
• GoThroughPosition / Sequence
• GoThroughPose / Sequence
• GoThroughGate / Sequence

Installation

Native

Follow the Isaac Sim documentation to install the latest Isaac Sim release.

Examples in this repository rely on features from the most recent Isaac Sim release. Please make sure to update any existing Isaac Sim build to the latest release version, 2023.1.1, to ensure examples work as expected.

[!IMPORTANT] Make sure Isaac sim was installed localy. Locate it's python executable path, it can usually be found here: ~/.local/share/ov/pkg/isaac-sim-2023.1.1/python.sh or here: ~/.local/share/ov/pkg/isaac-sim-2023.1.1/python.sh

Clone this repository:

git clone https://github.com/elharirymatteo/RANS.git
cd RANS

[!CAUTION] In the following we refer to python.sh as the full path to the python executable of isaac sim. Make sure you use the full path.

python.sh -m pip install -e .

We use a modified version of the rl-games library to train our agents.

To install it clone this repository INSIDE RANS:

git clone https://github.com/AntoineRichard/rl_games

Then install the module:

cd rl_games
python.sh -m pip install --upgrade pip
pythons.sh -m pip install -e .

With these steps done you should be all set! Refer to the Getting Started section to learn how to launch your first trainings.

Docker

Before we install the simulation, please follow the procedure here to install all the required components to install IsaacSim in a docker container.

[!TIP] You will need an nvcr.io account.

Once you're all set, clone this repository as well as our own version of RL_games.

git clone https://github.com/elharirymatteo/RANS.git
cd RANS
git clone https://github.com/AntoineRichard/rl_games

[!IMPORTANT] RL games must be cloned inside the RANS repository.

One this is done, the docker image can be built by calling the build script.

./docker/build_docker.sh

With this out of the way you should be off to the races! Check the Getting Started section to start your first training.

Getting Started

[!IMPORTANT] If you are using docker you can start a new docker container by running the docker/run_docker.sh or docker/run_docker_viewer.sh. The viewer version allows the user to visualize the environments.

[!Note] All commands should be executed from RANS/omniisaacgymenvs.* python.sh referes to the python.sh inside Isaac's sim folder. In docker this will be at /isaac-sim/python.sh.

Training new agents

To train your first policy, (example for the USV robot) run:

python.sh scripts/rlgames_train_RANS.py task=ASV/GoToPose train=RANS/PPOcontinuous_MLP headless=True num_envs=1024

Modify num_envs appropriately to scale with your current machine capabilities. Turn headless to False if you want to visualize the envs while training occurs.

You should see an Isaac Sim window pop up. Once Isaac Sim initialization completes, the scene for the selected robot will be constructed and simulation will start running automatically. The process will terminate once training finishes.

Here's another example - GoToPose for the Satellite robot (MFP - modular floating platform) - using the multi-threaded training script:

python.sh scripts/rlgames_train_RANS.py task=MFP2D/GoToPose train=RANS/PPOmulti_discrete_MLP

Note that by default, we show a Viewport window with rendering, which slows down training. You can choose to close the Viewport window during training for better performance. The Viewport window can be re-enabled by selecting Window > Viewport from the top menu bar.

To achieve maximum performance, launch training in headless mode as follows:

python.sh scripts/rlgames_train_RANS.py task=MFP2D/GoToPose train=PPOmulti_discrete_MLP headless=True

[!NOTE] Some of the examples could take a few minutes to load because the startup time scales based on the number of environments. The startup time will continually be optimized in future releases.

Loading trained models (or checkpoints)

Checkpoints are saved in the folder runs/EXPERIMENT_NAME/nn where EXPERIMENT_NAME defaults to the task name, but can also be overridden via the experiment argument.

To load a trained checkpoint and continue training, use the checkpoint argument:

python.sh scripts/rlgames_train_RANS.py task=MFP2D/GoToPose train=RANS/PPOmulti_discrete_MLP checkpoint=runs/MFP2D_GoToPose/nn/MFP2D_GoToPose.pth

To load a trained checkpoint and only perform inference (no training), pass test=True as an argument, along with the checkpoint name. To avoid rendering overhead, you may also want to run with fewer environments using num_envs=64:

python.sh scripts/rlgames_train_RANS.py task=MFP2D/GoToPose train=RANS/PPOmulti_discrete_MLP checkpoint=runs/MFP2D_GoToPose/nn/MFP2D_GoToPose.pth test=True num_envs=64

Note that if there are special characters such as [ or = in the checkpoint names, you will need to escape them and put quotes around the string. For example, checkpoint="runs/Ant/nn/last_Antep\=501rew\[5981.31\].pth"

Going further

All scripts provided in omniisaacgymenvs/scripts can be launched directly with PYTHON_PATH.

python.sh scripts/random_policy.py task=MFP2D/GoToPose

This script will sample random actions from the action space and apply these actions to your task without running any RL policies. Simulation should start automatically after launching the script, and will run indefinitely until terminated.

python.sh scripts/rlgames_train_RANS.py task=MFP2D/GoToPosition

This script creates an instance of the PPO runner in rl_games and automatically launches training and simulation. Once training completes (the total number of iterations have been reached), the script will exit. If running inference with test=True checkpoint=<path/to/checkpoint>, the script will run indefinitely until terminated. Note that this script will have limitations on interaction with the UI.

We use Hydra to manage the config.

Common arguments for the training scripts are:

• task=TASK - Selects which task to use. Examples include MFP2D/GoToPosition, MFP2D/GoToPose, MFP2D/TrackLinearVelocity, MFP2D/TrackLinearAngularVelocity, MFP3D/GoToPosition, MFP3D/GoToPose. These correspond to the config for each environment in the folder omniisaacgymenvs/cfg/task/ + MFP2D or any robot description (eg. ASV for the boat or AGV for the turtle-bots).
• train=TRAIN - Selects which training config to use. Will automatically default to the correct config for the environment file inside the train/RANS folder (e.g., PPOcontinuous_MLP or PPOmulti_discrete_MLP).
• num_envs=NUM_ENVS - Selects the number of environments to use (overriding the default number of environments set in the task config).
• seed=SEED - Sets a seed value for randomization, overriding the default seed in the task config.
• pipeline=PIPELINE - Which API pipeline to use. Defaults to gpu, can also be set to cpu. When using the gpu pipeline, all data stays on the GPU. When using the cpu pipeline, simulation can run on either CPU or GPU, depending on the sim_device setting, but a copy of the data is always made on the CPU at every step.
• sim_device=SIM_DEVICE - Device used for physics simulation. Set to gpu (default) to use GPU and to cpu for CPU.
• device_id=DEVICE_ID - Device ID for GPU to use for simulation and task. Defaults to 0. This parameter will only be used if simulation runs on GPU.
• rl_device=RL_DEVICE - Which device / ID to use for the RL algorithm. Defaults to cuda:0, and follows PyTorch-like device syntax.
• test=TEST - If set to True, only runs inference on the policy and does not do any training.
• checkpoint=CHECKPOINT_PATH - Path to the checkpoint to load for training or testing.
• headless=HEADLESS - Whether to run in headless mode.
• experiment=EXPERIMENT - Sets the name of the experiment.
• max_iterations=MAX_ITERATIONS - Sets how many iterations to run for. Reasonable defaults are provided for the provided environments.
• warp=WARP - If set to True, launch the task implemented with Warp backend (Note: not all tasks have a Warp implementation).
• kit_app=KIT_APP - Specifies the absolute path to the kit app file to be used.

Hydra also allows setting variables inside config files directly as command line arguments. For example, to set the minibatch size for an rl_games training run, you can use train.params.config.minibatch_size=64. Similarly, variables in task configs can also be set, such as task.env.episodeLength=100.

Hydra Notes

Default values for each of these are found in the omniisaacgymenvs/cfg/config.yaml file.

The way that the task and train portions of the config works are through the use of config groups. You can learn more about how these work here The actual configs for task are in omniisaacgymenvs/cfg/task/<TASK>.yaml and for train in omniisaacgymenvs/cfg/train/<TASK>PPO.yaml.

In some places in the config you will find other variables referenced (for example, num_actors: ${....task.env.numEnvs}). Each . represents going one level up in the config hierarchy. This is documented fully here.

Tensorboard

Tensorboard can be launched during training via the following command:

python.sh -m tensorboard.main --logdir runs/EXPERIMENT_NAME/summaries

WandB support

You can run (WandB)[https://wandb.ai/] with OmniIsaacGymEnvs by setting wandb_activate=True flag from the command line. You can set the group, name, entity, and project for the run by setting the wandb_group, wandb_name, wandb_entity and wandb_project arguments. Make sure you have WandB installed in the Isaac Sim Python executable with PYTHON_PATH -m pip install wandb before activating.

Citation

If you use the current repository in your work, we suggest citing the following papers:

@article{el2023drift,
  title={DRIFT: Deep Reinforcement Learning for Intelligent Floating Platforms Trajectories},
  author={El-Hariry, Matteo and Richard, Antoine and Muralidharan, Vivek and Yalcin, Baris Can and Geist, Matthieu and Olivares-Mendez, Miguel},
  journal={arXiv preprint arXiv:2310.04266},
  year={2023}
}

@article{el2023rans,
  title={RANS: Highly-Parallelised Simulator for Reinforcement Learning based Autonomous Navigating Spacecrafts},
  author={El-Hariry, Matteo and Richard, Antoine and Olivares-Mendez, Miguel},
  journal={arXiv preprint arXiv:2310.07393},
  year={2023}
}

Directory Structure (most relevant folders shown)

.
├── cfg                                 # Configuration files
│   ├── controller                      # Controller configurations
│   └──  hl_task                         # High-level task configurations convergence task
│   └── train                           # Training configurations
│       └── MFP                         # Training configurations for Modular Floating Platform
├── demos                               # Demonstration files (e.g., gifs, videos)
├── doc                                 # Documentation files
│   ├── curriculum.md                   # Documentation for curriculum
│   ├── domain_randomization.md         # Documentation for domain randomization
│   ├── figures                         # Figures used in documentation
│   │   └── ...                         # Other figure files
│   └── penalties.md                    # Documentation for penalties
├── envs                                # Environment scripts
│   ├── vec_env_rlgames_mfp.py          # Vectorized environment for rlgames with MFP
│   ├── vec_env_rlgames_mt.py           # Vectorized environment for rlgames with multiple tasks
│   └── vec_env_rlgames.py              # General vectorized environment for rlgames
├── extension.py                        # Extension script
├── images                              # Image files
│   ├── 3dof_gotoxy.png                 # Image for 3DOF GoToXY task
│   └── ...                             # Other image files
├── __init__.py                         # Initialization script for the package
├── lab_tests                           # Lab test scripts and data
├── mj_runs                             # Mujoco run scripts and data
├── models                              # Model files
├── robots                              # Robot related files
│   ├── articulations                   # Articulation files for robots
│   ├── sensors                         # Sensor files for robots
│   └── usd                             # USD files for robots
├── ros                                 # ROS related files
├── scripts                             # Utility scripts
├── tasks                               # Task implementations
│   └── MFP                             # Task implementations for Modular Floating Platform
│       ├── curriculum_helpers.py       # Helper functions for curriculum
│       └── unit_tests                  # Unit tests for MFP tasks
├── utils                               # Utility functions and scripts
│   ├── aggregate_and_eval_mujoco_batch_data.py  # Script to aggregate and evaluate Mujoco batch data
│   ├── rlgames                         # RL games related utilities
│   │   ├── __pycache__                 # Compiled Python files
│   │   ├── rlgames_train_mt.py         # Training script for RL games with multiple tasks
│   │   └── rlgames_utils.py            # Utility functions for RL games
├── videos                              # Video files
└── wandb                               # Weights and Biases integration files