/USI-Dofbot-Repo

Dofbot Reacher Reinforcement Learning Sim2Real Environment for Omniverse Isaac Gym/Sim

Primary LanguagePythonOtherNOASSERTION

Dofbot Reacher Reinforcement Learning Sim2Real Environment for Omniverse Isaac Gym/Sim

This repository adds a DofbotReacher environment based on OmniIsaacGymEnvs (commit d0eaf2e), and includes Sim2Real code to control a real-world Dofbot with the policy learned by reinforcement learning in Omniverse Isaac Gym/Sim.

We target Isaac Sim 2022.1.1 and test the RL code on Windows 10 and Ubuntu 18.04. The Sim2Real code is tested on Linux and a real Dofbot.

This repo is compatible with OmniIsaacGymEnvs-UR10Reacher.

Preview

Installation

Prerequisites:

We will use Anaconda to manage our virtual environment:

  1. Clone this repository:

    • Linux 
      cd ~
git clone https://github.com/j3soon/OmniIsaacGymEnvs-DofbotReacher.git
    • Windows 
      cd %USERPROFILE%
git clone https://github.com/j3soon/OmniIsaacGymEnvs-DofbotReacher.git

  2. Generate instanceable Dofbot assets for training:

    Launch the Script Editor in Isaac Sim. Copy the content in omniisaacgymenvs/utils/usd_utils/create_instanceable_dofbot.py and execute it inside the Script Editor window. Wait until you see the text Done!.

  3. Download and Install Anaconda.

    # For 64-bit Linux (x86_64/x64/amd64/intel64)
wget https://repo.anaconda.com/archive/Anaconda3-2022.10-Linux-x86_64.sh
bash Anaconda3-2022.10-Linux-x86_64.sh

    For Windows users, make sure to use Anaconda Prompt instead of Command Prompt or Powershell for the following commands.

  4. Patch Isaac Sim 2022.1.1

    • Linux 
      export ISAAC_SIM="$HOME/.local/share/ov/pkg/isaac_sim-2022.1.1"
cp $ISAAC_SIM/setup_python_env.sh $ISAAC_SIM/setup_python_env.sh.bak
cp ~/OmniIsaacGymEnvs-DofbotReacher/isaac_sim-2022.1.1-patch/linux/setup_python_env.sh $ISAAC_SIM/setup_python_env.sh
    • Windows 
      set ISAAC_SIM="%LOCALAPPDATA%\ov\pkg\isaac_sim-2022.1.1"
copy %USERPROFILE%\OmniIsaacGymEnvs-DofbotReacher\isaac_sim-2022.1.1-patch\windows\setup_conda_env.bat %ISAAC_SIM%\setup_conda_env.bat

  5. Set up conda environment for Isaac Sim

    • Linux 
      # conda remove --name isaac-sim --all
export ISAAC_SIM="$HOME/.local/share/ov/pkg/isaac_sim-2022.1.1"
cd $ISAAC_SIM
conda env create -f environment.yml
conda activate isaac-sim
cd ~/OmniIsaacGymEnvs-DofbotReacher
pip install -e .
    • Windows 
      # conda remove --name isaac-sim --all
set ISAAC_SIM="%LOCALAPPDATA%\ov\pkg\isaac_sim-2022.1.1"
cd %ISAAC_SIM%
conda env create -f environment.yml
conda activate isaac-sim
:: Fix incorrect importlib-metadata version (isaac-sim 2022.1.1)
pip install importlib-metadata==4.11.4
cd %USERPROFILE%\OmniIsaacGymEnvs-DofbotReacher
pip install -e .
:: Fix incorrect torch version (isaac-sim 2022.1.1)
conda install -y pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 -c pytorch

  6. Activate conda environment

    • Linux 
      export ISAAC_SIM="$HOME/.local/share/ov/pkg/isaac_sim-2022.1.1"
cd $ISAAC_SIM
conda activate isaac-sim
source setup_conda_env.sh
    • Windows 
      set ISAAC_SIM="%LOCALAPPDATA%\ov\pkg\isaac_sim-2022.1.1"
cd %ISAAC_SIM%
conda activate isaac-sim
call setup_conda_env.bat

Please note that you should execute the commands in Step 6 for every new shell.

For Windows users, replace ~ to %USERPROFILE% for all the following commands.

Dummy Policy

This is a sample to make sure you have setup the environment correctly. You should see a single Dofbot in Isaac Sim.

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/dummy_dofbot_policy.py task=DofbotReacher test=True num_envs=1

Training

You can launch the training in headless mode as follows:

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/rlgames_train.py task=DofbotReacher headless=True

The number of environments is set to 2048 by default. If your GPU has small memory, you can decrease the number of environments by changing the arguments num_envs as below:

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/rlgames_train.py task=DofbotReacher headless=True num_envs=2048

You can also skip training by downloading the pre-trained model checkpoint by:

cd ~/OmniIsaacGymEnvs-DofbotReacher
wget https://github.com/j3soon/OmniIsaacGymEnvs-DofbotReacher/releases/download/v1.1.0/runs.zip
unzip runs.zip

The learning curve of the pre-trained model:

Testing

Make sure you have model checkpoints at ~/OmniIsaacGymEnvs-DofbotReacher/runs, you can check it with the following command:

ls ~/OmniIsaacGymEnvs-DofbotReacher/runs/DofbotReacher/nn/

Please note that you may not want to use the checkpoint ./runs/DofbotReacher/nn/DofbotReacher.pth due to the randomness of the reward signal. Instead, use the latest checkpoint such as ./runs/DofbotReacher/nn/last_DofbotReacher_ep_1000_rew_XXX.pth. You can replace DofbotReacher.pth with the latest checkpoint before following the steps below, or you can simply modify the commands to use the latest checkpoint.

You can visualize the learned policy by the following command:

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/rlgames_train.py task=DofbotReacher test=True num_envs=512 checkpoint=./runs/DofbotReacher/nn/DofbotReacher.pth

Likewise, you can decrease the number of environments by modifying the parameter num_envs=512.

Sim2Real

The learned policy has a very conservative constraint on the joint limits. Therefore, the gripper would not hit the ground under such limits. However, you should still make sure there are no other obstacles within Dofbot's workspace (reachable area). That being said, if things go wrong, press Ctrl+C twice in the terminal to kill the process.

It would be possible to remove the conservative joint limit constraints by utilizing self-collision detection in Isaac Sim. We are currently investigating this feature.

For simplicity, we'll use TCP instead of ROS to control the real-world dofbot. Copy the server notebook file (omniisaacgymenvs/sim2real/dofbot-server.ipynb) to the Jetson Nano on your Dofbot. Launch a Jupyter Notebook on Jetson Nano and execute the server notebook file.

You should be able to reset the Dofbot's joints by the following script:

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/sim2real/dofbot.py

Edit omniisaacgymenvs/cfg/task/DofbotReacher.yaml. Set sim2real.enabled to True, and set sim2real.ip to the IP of your Dofbot:

sim2real:
  enabled: True
  fail_quietely: False
  verbose: False
  ip: <IP_OF_YOUR_DOFBOT>
  port: 65432

Now you can control the real-world Dofbot in real-time by the following command:

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/rlgames_train.py task=DofbotReacher test=True num_envs=1 checkpoint=./runs/DofbotReacher/nn/DofbotReacher.pth

Demo

We provide an interactable demo based on the DofbotReacher RL example. In this demo, you can click on any of the Dofbot in the scene to manually control the robot with your keyboard as follows:

  • Q/A: Control Joint 0.
  • W/S: Control Joint 1.
  • E/D: Control Joint 2.
  • R/F: Control Joint 3.
  • T/G: Control Joint 4.
  • Y/H: Control Joint 5.
  • ESC: Unselect a selected Dofbot and yields manual control

Launch this demo with the following command. Note that this demo limits the maximum number of Dofbot in the scene to 128.

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/rlgames_play.py task=DofbotReacher num_envs=64

Running in Docker

If you have a NVIDIA Enterprise subscription, you can run all services with Docker Compose.

For users without a subscription, you can pull the Isaac Docker image, but should still install Omniverse Nucleus beforehand. (only Isaac itself is dockerized)

Follow this tutorial to generate your NGC API Key, and make sure you can access Isaac with Omniverse Streaming Client, WebRTC, or WebSocket. After that, exit the Docker container.

Please note that you should generate instanceable assets beforehand as mentioned in the Installation section.

We will now set up the environment inside Docker:

  1. Launch an Isaac Container: 
    docker run --name isaac-sim --entrypoint bash -it --gpus all -e "ACCEPT_EULA=Y" --rm --network=host \
-v ~/docker/isaac-sim/cache/ov:/root/.cache/ov:rw \
-v ~/docker/isaac-sim/cache/pip:/root/.cache/pip:rw \
-v ~/docker/isaac-sim/cache/glcache:/root/.cache/nvidia/GLCache:rw \
-v ~/docker/isaac-sim/cache/computecache:/root/.nv/ComputeCache:rw \
-v ~/docker/isaac-sim/logs:/root/.nvidia-omniverse/logs:rw \
-v ~/docker/isaac-sim/config:/root/.nvidia-omniverse/config:rw \
-v ~/docker/isaac-sim/data:/root/.local/share/ov/data:rw \
-v ~/docker/isaac-sim/documents:/root/Documents:rw \
nvcr.io/nvidia/isaac-sim:2022.1.1
  2. Install common tools: 
    apt update && apt install -y git wget vim
  3. Clone this repository: 
    cd ~
git clone https://github.com/j3soon/OmniIsaacGymEnvs-DofbotReacher.git
  4. Download and Install Anaconda. 
    # For 64-bit (x86_64/x64/amd64/intel64)
wget https://repo.anaconda.com/archive/Anaconda3-2022.10-Linux-x86_64.sh
bash Anaconda3-2022.10-Linux-x86_64.sh -b -p $HOME/anaconda3
  5. Patch Isaac Sim 2022.1.1 
    export ISAAC_SIM="/isaac-sim"
cp $ISAAC_SIM/setup_python_env.sh $ISAAC_SIM/setup_python_env.sh.bak
cp ~/OmniIsaacGymEnvs-DofbotReacher/isaac_sim-2022.1.1-patch/setup_python_env.sh $ISAAC_SIM/setup_python_env.sh
  6. Set up conda environment for Isaac Sim 
    source ~/anaconda3/etc/profile.d/conda.sh
# conda remove --name isaac-sim --all
export ISAAC_SIM="/isaac-sim"
cd $ISAAC_SIM
conda env create -f environment.yml
conda activate isaac-sim
cd ~/OmniIsaacGymEnvs-DofbotReacher
pip install -e .
  7. Activate conda environment 
    source ~/anaconda3/etc/profile.d/conda.sh
export ISAAC_SIM="/isaac-sim"
cd $ISAAC_SIM
conda activate isaac-sim
source setup_conda_env.sh
./vulkan_check.sh

We can now train a RL policy in this container:

cd ~/OmniIsaacGymEnvs-DofbotReacher
python omniisaacgymenvs/scripts/rlgames_train.py task=DofbotReacher headless=True num_envs=2048

Make sure to copy the learned weights to a mounted volume before exiting the container, otherwise it will be deleted:

# In container
cp -r ~/OmniIsaacGymEnvs-DofbotReacher/runs ~/Documents/runs
# In host
ls ~/docker/isaac-sim/documents/

You can watch the training progress with:

docker ps # Observe Container ID
docker exec -it <CONTAINER_ID> /bin/bash
conda activate isaac-sim
cd ~/OmniIsaacGymEnvs-DofbotReacher
tensorboard --logdir=./runs

Currently we do not support running commands that requires visualization (Testing, Sim2Real, etc.) in Docker. Since I haven't figured out how to make Vulkan render a Isaac window inside a container yet. Alternatively, it may be possible to add headless=True and view them in Omniverse Streaming Client, WebRTC, or WebSocket, but I haven't tested this by myself.

Note: below are the original README of OmniIsaacGymEnvs.

Omniverse Isaac Gym Reinforcement Learning Environments for Isaac Sim

About this repository

This repository contains Reinforcement Learning examples that can be run with the latest release of Isaac Sim. RL examples are trained using PPO from rl_games library and examples are built on top of Isaac Sim's omni.isaac.core and omni.isaac.gym frameworks.

 

 

 

 

 

Installation

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, 2022.1.1, to ensure examples work as expected.

Once installed, this repository can be used as a python module, omniisaacgymenvs, with the python executable provided in Isaac Sim.

To install omniisaacgymenvs, first clone this repository:

git clone https://github.com/NVIDIA-Omniverse/OmniIsaacGymEnvs.git

Once cloned, locate the python executable in Isaac Sim. By default, this should be python.sh. We will refer to this path as PYTHON_PATH.

To set a PYTHON_PATH variable in the terminal that links to the python executable, we can run a command that resembles the following. Make sure to update the paths to your local path.

For Linux: alias PYTHON_PATH=~/.local/share/ov/pkg/isaac_sim-*/python.sh
For Windows: doskey PYTHON_PATH=C:\Users\user\AppData\Local\ov\pkg\isaac_sim-*\python.bat $*

Install omniisaacgymenvs as a python module for PYTHON_PATH:

PYTHON_PATH -m pip install -e .

Running the examples

Note: All commands should be executed from omniisaacgymenvs/omniisaacgymenvs.

To train your first policy, run:

PYTHON_PATH scripts/rlgames_train.py task=Cartpole

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

Here's another example - Ant locomotion - using the multi-threaded training script:

PYTHON_PATH scripts/rlgames_train_mt.py task=Ant

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, you can launch training in headless mode as follows:

PYTHON_PATH scripts/rlgames_train.py task=Ant headless=True

A Note on the Startup Time of the Simulation

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 // 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_PATH scripts/rlgames_train.py task=Ant checkpoint=runs/Ant/nn/Ant.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_PATH scripts/rlgames_train.py task=Ant checkpoint=runs/Ant/nn/Ant.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"

We provide pre-trained checkpoints on the Nucleus server under Assets/Isaac/2022.1/Isaac/Samples/OmniIsaacGymEnvs/Checkpoints. Run the following command to launch inference with pre-trained checkpoint:

Localhost (To set up localhost, please refer to the Isaac Sim installation guide):

PYTHON_PATH scripts/rlgames_train.py task=Ant checkpoint=omniverse://localhost/NVIDIA/Assets/Isaac/2022.1/Isaac/Samples/OmniIsaacGymEnvs/Checkpoints/ant.pth test=True num_envs=64

Production server:

PYTHON_PATH scripts/rlgames_train.py task=Ant checkpoint=http://omniverse-content-production.s3-us-west-2.amazonaws.com/Assets/Isaac/2022.1/Isaac/Samples/OmniIsaacGymEnvs/Checkpoints/ant.pth test=True num_envs=64

When running with a pre-trained checkpoint for the first time, we will automatically download the checkpoint file to omniisaacgymenvs/checkpoints. For subsequent runs, we will re-use the file that has already been downloaded, and will not overwrite existing checkpoints with the same name in the checkpoints folder.

Training Scripts

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

To test out a task without RL in the loop, run the random policy script with:

PYTHON_PATH scripts/random_policy.py task=Cartpole

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.

To run a simple form of PPO from rl_games, use the single-threaded training script:

PYTHON_PATH scripts/rlgames_train.py task=Cartpole

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.

Lastly, we provide a multi-threaded training script that executes the RL policy on a separate thread than the main thread used for simulation and rendering:

PYTHON_PATH scripts/rlgames_train_mt.py task=Cartpole

This script uses the same RL Games PPO policy as the above, but runs the RL loop on a new thread. Communication between the RL thread and the main thread happens on threaded Queues. Simulation will start automatically, but the script will not exit when training terminates, except when running in headless mode. Simulation will stop when training completes or can be stopped by clicking on the Stop button in the UI. Training can be launched again by clicking on the Play button. Similarly, if running inference with test=True checkpoint=<path/to/checkpoint>, simulation will run until the Stop button is clicked, or the script will run indefinitely until the process is terminated.

Configuration and command line arguments

We use Hydra to manage the config.

Common arguments for the training scripts are:

  • task=TASK - Selects which task to use. Any of AllegroHand, Ant, Anymal, AnymalTerrain, BallBalance, Cartpole, Crazyflie, FrankaCabinet, Humanoid, Ingenuity, Quadcopter, ShadowHand, ShadowHandOpenAI_FF, ShadowHandOpenAI_LSTM (these correspond to the config for each environment in the folder omniisaacgymenvs/cfg/task)
  • train=TRAIN - Selects which training config to use. Will automatically default to the correct config for the environment (ie. <TASK>PPO).
  • 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, and overrides the default seed in the task config
  • pipeline=PIPELINE - Which API pipeline to use. Defaults to gpu, can also 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.

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

Tasks

Source code for tasks can be found in omniisaacgymenvs/tasks.

Each task follows the frameworks provided in omni.isaac.core and omni.isaac.gym in Isaac Sim.

Refer to docs/framework.md for how to create your own tasks.

Full details on each of the tasks available can be found in the RL examples documentation.

Demo

We provide an interactable demo based on the AnymalTerrain RL example. In this demo, you can click on any of the ANYmals in the scene to go into third-person mode and manually control the robot with your keyboard as follows:

  • Up Arrow: Forward linear velocity command
  • Down Arrow: Backward linear velocity command
  • Left Arrow: Leftward linear velocity command
  • Right Arrow: Rightward linear velocity command
  • Z: Counterclockwise yaw angular velocity command
  • X: Clockwise yaw angular velocity command
  • C: Toggles camera view between third-person and scene view while maintaining manual control
  • ESC: Unselect a selected ANYmal and yields manual control

Launch this demo with the following command. Note that this demo limits the maximum number of ANYmals in the scene to 128.

PYTHON_PATH scripts/rlgames_play.py task=AnymalTerrain num_envs=64 checkpoint=omniverse://localhost/NVIDIA/Assets/Isaac/2022.1/Isaac/Samples/OmniIsaacGymEnvs/Checkpoints/anymal_terrain.pth

A note about Force Sensors

Force sensors are supported in Isaac Sim and OIGE via the ArticulationView class. Sensor readings can be retrieved using get_force_sensor_forces() API, as shown in the Ant/Humanoid Locomotion task, as well as in the Ball Balance task. Please note that there is currently a known bug regarding force sensors in Omniverse Physics. Transforms of force sensors (i.e. their local poses) are set in the actor space of the Articulation instead of the body space, which is the expected behaviour. We will be fixing this in the coming release.