OM1 deployment

This repository contains Docker Compose configurations for deploying the OM1 robot system.

Instructions

For a fresh Thor (JetPack 7.0) system, follow these steps to set up OM1:

Basic Setup

uv

Use curl to download and install uv:

curl -LsSf https://astral.sh/uv/install.sh | sh

Docker

The docker is pre-installed on JetPack 7.0 systems, but you need to give it proper permissions:

newgrp docker
sudo usermod -aG docker $USER
groups

You should see docker in the list of groups. If not, log out and log back in, then check again.

Docker Compose

Download and install Docker Compose with the following commands:

sudo curl -L "https://github.com/docker/compose/releases/download/v2.34.0/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose

Set the executable permissions:

sudo chmod +x /usr/local/bin/docker-compose

Verify the installation:

docker-compose --version

Poetry (Optional)

Install Poetry using the official installation script:

curl -sSL https://install.python-poetry.org | python3 -

Install poetry shell for the environment management:

poetry self add poetry-plugin-shell

Pyaudio (For microphone support)

Install the required packages:

sudo apt install portaudio19-dev python3-pyaudio

FFmpeg (For audio processing)

Install FFmpeg using the following command:

sudo apt install ffmpeg

Chrome (For web interface)

Download and install Google Chrome:

sudo snap install chromium

Hold snap updates to prevent automatic updates:

snap download snapd --revision=24724
sudo snap ack snapd_24724.assert
sudo snap install snapd_24724.snap
sudo snap refresh --hold snapd

ROS2 (Optional)

Follow the official ROS2 installation guide for Ubuntu: ROS2 Installation.

After installing ROS2, source the ROS2 setup script:

source /opt/ros/jazzy/setup.bash

You can add this line to your ~/.bashrc file to source it automatically on terminal startup.

CycloneDDS Binary (Optional)

Install CycloneDDS for ROS2 communication:

sudo apt install ros-jazzy-rmw-cyclonedds-cpp
sudo apt install ros-jazzy-rosidl-generator-dds-idl

Now, set CycloneDDS as the default RMW implementation by adding the following line to your ~/.bashrc file:

export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp

You can restart your ROS2 daemon with the following command:

ros2 daemon stop
ros2 daemon start

CycloneDDS Build from Source (Optional)

If you prefer to build CycloneDDS from source, use the following commands:

cd Documents
mkdir -p GitHub && cd GitHub
git clone https://github.com/eclipse-cyclonedds/cyclonedds -b releases/0.10.x
cd cyclonedds && mkdir build install && cd build
cmake .. -DCMAKE_INSTALL_PREFIX=../install -DBUILD_EXAMPLES=ON
cmake --build . --target install

Then you need to set the following environment variables in your ~/.bashrc file:

export CYCLONEDDS_HOME=$HOME/Documents/GitHub/cyclonedds/install

Configure Network Settings (Unitree Only)

You need to open the network settings and find the network interface that the robot connected. In IPv4 settings, set the method to Manual and add the following IP address:

192.168.123.xxx

and set the subnet mask to

255.255.255.0

CycloneDDS Configuration (Optional)

You can create a CycloneDDS configuration file to customize its behavior. Create a file named cyclonedds.xml in your home directory:

<CycloneDDS>
  <Domain>
    <General>
      <Interfaces>
        <NetworkInterface name="enP2p1s0" priority="default" multicast="default" />
      </Interfaces>
    </General>
    <Discovery>
      <EnableTopicDiscoveryEndpoints>true</EnableTopicDiscoveryEndpoints>
    </Discovery>
  </Domain>
</CycloneDDS>

Then, set the CYCLONEDDS_URI environment variable in your ~/.bashrc file:

export CYCLONEDDS_URI=file://$HOME/cyclonedds.xml

v4l2-ctl (For camera configuration)

Install v4l2-ctl using the following command:

sudo apt install v4l-utils

System Services

We assume you have bought the brain pack. If you don't have it, you can skip this section based on your needs.

Screen Animation Service

To enable the screen animation service, install unclutter first to hide the mouse cursor:

sudo apt install unclutter

Then, add the script to /usr/local/bin/start-kiosk.sh and make it executable:

#!/bin/bash

unclutter -display :0 -idle 0.1 -root &

HOST=localhost
PORT=4173

# Wait for Docker service to listen
while ! nc -z $HOST $PORT; do
  echo "Waiting for $HOST:$PORT..."
  sleep 0.1
done

# Launch with autoplay permissions
exec chromium \
  --kiosk http://$HOST:$PORT \
  --start-fullscreen \
  --disable-infobars \
  --noerrdialogs \
  --autoplay-policy=no-user-gesture-required \
  --disable-features=PreloadMediaEngagementData,MediaEngagementBypassAutoplayPolicies \
  --no-first-run \
  --disable-session-crashed-bubble \
  --disable-translate \
  --window-position=0,0

Make it executable:

sudo chmod +x /usr/local/bin/start-kiosk.sh

Add the script to /etc/systemd/system/kiosk.service to launch the kiosk mode automatically on boot.

# /etc/systemd/system/kiosk.service
[Unit]
Description=Kiosk Browser
After=docker.service
Requires=docker.service

[Service]
Environment=DISPLAY=:0
ExecStart=/usr/local/bin/start-kiosk.sh
Restart=always
User=openmind

[Install]
WantedBy=graphical.target

Enable and start the service:

sudo systemctl daemon-reload
sudo systemctl enable kiosk.service
sudo systemctl start kiosk.service

Note

To stop the kiosk service, use sudo systemctl stop kiosk.service.

AEC Service

To enable the Acoustic Echo Cancellation (AEC) service, uninstall PipWire if it's installed and install PulseAudio

sudo apt remove --purge pipewire-audio-client-libraries pipewire-pulse wireplumber

Then install PulseAudio:

sudo apt install pulseaudio pulseaudio-module-bluetooth pulseaudio-utils pavucontrol

Next, stop the PipWire daemon and start the PulseAudio daemon if it's not already running:

systemctl --user mask pipewire.service
systemctl --user mask pipewire.socket
systemctl --user mask pipewire-pulse.service
systemctl --user mask pipewire-pulse.socket
systemctl --user mask wireplumber.service
systemctl --user stop pipewire-pulse.service
systemctl --user stop pipewire.service wireplumber.service
systemctl --user disable pipewire.service wireplumber.service
systemctl --user enable --now pulseaudio.service

Next, add the script to prevent PulseAudio from going into auto-exit mode.

mkdir -p ~/.config/pulse
cat > ~/.config/pulse/client.conf << 'EOF'
autospawn = yes
daemon-binary = /usr/bin/pulseaudio
EOF

# Create daemon config to disable idle timeout
cat > ~/.config/pulse/daemon.conf << 'EOF'
exit-idle-time = -1
EOF

Now, you can restart the system to ensure PulseAudio is running properly.

sudo reboot

Note

After reboot, if the audio devices are not automatically detected, you may need to manually start PulseAudio with the command:

systemctl --user restart pulseaudio

Now, you can add the script to /usr/local/bin/set-audio-defaults.sh and make it executable:

#!/bin/bash
set -e

sleep 5

# First, set the master source volume to 200%
pactl set-source-volume "alsa_input.usb-R__DE_R__DE_VideoMic_GO_II_FEB0C614-00.mono-fallback" 131072
pactl set-source-mute "alsa_input.usb-R__DE_R__DE_VideoMic_GO_II_FEB0C614-00.mono-fallback" 0

# Unload then load AEC module
pactl unload-module module-echo-cancel || true
pactl load-module module-echo-cancel \
  use_master_format=1 \
  aec_method=webrtc \
  source_master="alsa_input.usb-R__DE_R__DE_VideoMic_GO_II_FEB0C614-00.mono-fallback" \
  sink_master="alsa_output.platform-88090b0000.hda.hdmi-stereo" \
  source_name="default_mic_aec" \
  sink_name="default_output_aec" \
  source_properties="device.description=Microphone_with_AEC" \
  sink_properties="device.description=Speaker_with_AEC"

# Wait a moment for the module to fully initialize
sleep 2

# Set defaults
pactl set-default-source default_mic_aec
pactl set-default-sink default_output_aec

# Retry volume setting until device appears and volume is set correctly
for i in {1..15}; do
  if pactl list short sources | grep -q default_mic_aec; then
    # Set volume to 200% (131072)
    pactl set-source-volume default_mic_aec 131072
    pactl set-source-mute default_mic_aec 0

    # Verify the volume was set
    CURRENT_VOL=$(pactl list sources | grep -A 7 "Name: default_mic_aec" | grep "Volume:" | awk '{print $3}')

    if [ "$CURRENT_VOL" = "131072" ]; then
      echo "Microphone volume successfully set to 200%"
      break
    else
      echo "Volume is $CURRENT_VOL, retrying... ($i/15)"
    fi
  else
    echo "Waiting for AEC source to appear... ($i/15)"
  fi
  sleep 1
done

# Final verification
pactl list sources | grep -A 7 "Name: default_mic_aec" | grep -E "Name:|Volume:"

Use the following command to get the list of audio sources and sinks:

pactl list short

Note

Replace alsa_output.platform-88090b0000.hda.hdmi-stereo with your speaker source and alsa_input.usb-R__DE_R__DE_VideoMic_GO_II_FEB0C614-00.mono-fallback with mic source

Make it executable:

sudo chmod +x /usr/local/bin/set-audio-defaults.sh

Create a systemd user service to run the script on login:

mkdir -p ~/.config/systemd/user
sudo vim ~/.config/systemd/user/audio-defaults.service

Add the following content:

[Unit]
Description=Set Default Audio Devices
After=pulseaudio.service
Wants=pulseaudio.service

[Service]
Type=oneshot
RemainAfterExit=yes
ExecStart=/usr/local/bin/set-audio-defaults.sh

[Install]
WantedBy=default.target

Enable and start the service:

systemctl --user daemon-reload
systemctl --user enable audio-defaults.service
systemctl --user start audio-defaults.service

Now, you need to export USER ID as an environment variable in your ~/.bashrc file:

export HOST_USER_ID=$(id -u)

to allow the docker containers to access the PulseAudio server properly. Then, reload your Bash profile to apply the changes:

source ~/.bashrc

Enable Cloud Docker Management Service

The cloud docker management service allows remote management of Docker containers via a web interface. To enable this service, follow these steps:

Sign up for an account on OpenMind Portal.
Create your OpenMind API key from the Dashboard page.
Set the API key as an environment variable in your Bash profile:

vim ~/.bashrc

export OM_API_KEY="your_api_key_here"

Get the API Key ID from the Dashboard page. The API Key ID is a 16-digit character string, such as om1_live_<16 characters>. Now, export the API Key ID as an environment variable:

vim ~/.bashrc

export OM_API_KEY_ID="your_api_key_id_here"

Set the robot type that you are using

vim ~/.bashrc

export ROBOT_TYPE="go2"  # or "go1", "tron"

Now, reload your Bash profile to apply the changes:

source ~/.bashrc

Setup OTA Update Services

To enable the Over-The-Air (OTA) update service for Docker containers, you need to set up two docker services: ota_agent and ota_updater. These services will allow you to manage and update your Docker containers remotely via the OpenMind Portal.

To create a ota_upater.yml file, follow these steps:

cd ~

vim ota_updater.yml

Add the following content from this ota_upater.yml to the ota_updater.yml file.

Note

You can use the stable version as well. The file example provided on the top is the latest version.

Save and close the file. Now, you can start the OTA updater service using Docker Compose:

docker-compose -f ota_updater.yml up -d

A .ota directory will be created in your home directory to store the OTA configuration files.

Now, you can set up the ota_agent service. Create an ota_agent.yml file:

cd .ota

vim ota_agent.yml

Add the following content from this ota_agent.yml to the ota_agent.yml file.

Note

You can use the stable version as well. The file example provided on the top is the latest version.

Save and close the file. Now, you can start the OTA agent service using Docker Compose:

docker-compose -f ota_agent.yml up -d

Now, both the OTA updater and agent services should be running. You can verify their status using the following commands:

docker ps | grep ota_updater
docker ps | grep ota_agent

You can now manage and update your Docker containers remotely via the OpenMind Portal.

Model Downloads

Riva Models

Riva models are encrypted and require authentication to download. To download Riva models, you need to set up the NVIDIA NGC CLI tool.

Install NGC CLI

Warning

Please run the following command in your root directory. Otherwise, the docker-compose file we provide for Riva services may not work properly.

To generate your own NGC api key, check this video.

wget --content-disposition https://ngc.nvidia.com/downloads/ngccli_arm64.zip && unzip ngccli_arm64.zip && chmod u+x ngc-cli/ngc
find ngc-cli/ -type f -exec md5sum {} + | LC_ALL=C sort | md5sum -c ngc-cli.md5
echo export PATH=\"\$PATH:$(pwd)/ngc-cli\" >> ~/.bash_profile
source ~/.bash_profile
ngc config set

This will ask several questions during the install. Choose these values:

Enter API key [no-apikey]. Choices: [<VALID_APIKEY>, 'no-apikey']: <YOUR_API_KEY>
Enter CLI output format type [ascii]. Choices: ['ascii', 'csv', 'json']: ascii
Enter org [no-org]. Choices: ['<YOUR_ORG>']: <YOUR_ORG>
Enter team [no-team]. Choices: ['<YOUR_TEAM>', 'no-team']: <YOUR_TEAM>
Enter ace [no-ace]. Choices: ['no-ace']: no-ace

Warning

ngc cli will create a .bash_profile file if it does not exist. If you already have a .bashrc file, please make sure to merge the two files properly. Otherwise, your bash environment may not work as expected.

Download Riva Models

Download Riva Embedded version models for Jetson 7.0:

ngc registry resource download-version nvidia/riva/riva_quickstart_arm64:2.24.0

cd riva_quickstart_arm64_v2.24.0
sudo bash riva_init.sh

# initialize riva model locally
# this will ask the NGC api key to download the model, use <YOUR_API_KEY>
# this will take a while to download

Note

The following command is for testing.

Run Riva locally:

cd riva_quickstart_arm64_v2.24.0
bash riva_start.sh

Now, please expose these environment variables in your ~/.bashrc file to use Riva service:

export RIVA_API_KEY=<YOUR_API_KEY>
export RIVA_API_NGC_ORG=<YOUR_ORG>
export RIVA_EULA=accept

source ~/.bashrc

OpenMind Riva Docker Image for Jetson

We create a openmindagi/riva-speech-server:2.24.0-l4t-aarch64 docker image that has Riva ASR and TTS endpoints with the example code to run Riva services on Jetson devices. You can pull the image directly without downloading the models from NGC:

docker pull openmindagi/riva-speech-server:2.24.0-l4t-aarch64

The dockerfile can be found here and the docker-compose file can be found here.

Note

Once you download the models from NGC and export the environment variables, you can use OpenMind Portal to download Riva dockerfile and run Riva services.

Once you have Riva services running, you can use the following script to test the ASR and TTS endpoints:

git clone https://github.com/OpenMind/OM1-modules.git

cd OM1-modules

# Activate poetry shell
poetry shell

# Install dependencies
poetry install

# Test ASR
python3 -m om1_speech.main --remote-url=ws://localhost:6790

# Test TTS
poetry run om1_tts --tts-url=https://api-dev.openmind.org/api/core/tts --device=<optional> --rate=<optional>

Port Usage

1935: MediaMTX RTMP Server
6790: OM Riva ASR Websocket Server API
6791: OM Riva TTS HTTP Server API
8000: MediaMTX RTMP Server API
8001: MediaMTX HLS Server API
8554: MediaMTX RTSP Server API
8860: Qwen 30B Quantized API
8880: Kokoro TTS API
8888: MediaMTX Streaming Server API
50000: Riva Server API
50051: Riva NMT Remote TTS/ASR API

Name		Name	Last commit message	Last commit date
Latest commit History 55 Commits
.github/workflows		.github/workflows
latest		latest
script		script
v1.0.0		v1.0.0
v1.0.1-beta.1		v1.0.1-beta.1
.gitignore		.gitignore
README.md		README.md

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

OM1 deployment

Instructions

Basic Setup

uv

Docker

Docker Compose

Poetry (Optional)

Pyaudio (For microphone support)

FFmpeg (For audio processing)

Chrome (For web interface)

ROS2 (Optional)

CycloneDDS Binary (Optional)

CycloneDDS Build from Source (Optional)

Configure Network Settings (Unitree Only)

CycloneDDS Configuration (Optional)

v4l2-ctl (For camera configuration)

System Services

Screen Animation Service

AEC Service

Enable Cloud Docker Management Service

Setup OTA Update Services

Model Downloads

Riva Models

Install NGC CLI

Download Riva Models

OpenMind Riva Docker Image for Jetson

Port Usage

About

Uh oh!

Releases

Packages

Uh oh!

Contributors 4

Uh oh!

Languages

OpenMind/OM1-deployment

Folders and files

Latest commit

History

Repository files navigation

OM1 deployment

Instructions

Basic Setup

uv

Docker

Docker Compose

Poetry (Optional)

Pyaudio (For microphone support)

FFmpeg (For audio processing)

Chrome (For web interface)

ROS2 (Optional)

CycloneDDS Binary (Optional)

CycloneDDS Build from Source (Optional)

Configure Network Settings (Unitree Only)

CycloneDDS Configuration (Optional)

v4l2-ctl (For camera configuration)

System Services

Screen Animation Service

AEC Service

Enable Cloud Docker Management Service

Setup OTA Update Services

Model Downloads

Riva Models

Install NGC CLI

Download Riva Models

OpenMind Riva Docker Image for Jetson

Port Usage

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors 4

Uh oh!

Languages

Packages