Installation and Activation

Follow the steps below to install Robotic Suite/Physical AI.

Prerequisites

Training may be executed on either CPU or GPU, and the system may therefore be installed on an NVIDIA platform, an NVIDIA Thor platform, or an Intel platform. A GPU is not a requirement; it is needed only where training is to be accelerated on an NVIDIA platform. The prerequisite components differ by platform:

Prerequisite	Intel platform	NVIDIA platform	NVIDIA Thor platform
Docker	Manual installation	Manual installation	Preinstalled
NVIDIA GPU driver	Not required	Manual installation	Preinstalled (with JetPack)
NVIDIA Container Toolkit	Not required	Manual installation	Preinstalled

On the NVIDIA Thor platform all three components are supplied with the system image; proceed directly to Pre-installation verification to confirm their status, and where verification does not pass, install the component concerned according to the corresponding section.

1. Install Docker

Docker must be installed manually on both Intel and NVIDIA platforms. The procedure below installs it from the official Docker repository:

# Remove potentially conflicting packages
for pkg in docker.io docker-doc docker-compose docker-compose-v2 podman-docker containerd runc; do
  sudo apt remove -y $pkg
done

# Add the official Docker GPG key and repository
sudo apt update
sudo apt install -y ca-certificates curl
sudo install -m 0755 -d /etc/apt/keyrings
sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc
sudo chmod a+r /etc/apt/keyrings/docker.asc
echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu $(. /etc/os-release && echo "$VERSION_CODENAME") stable" \
  | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null

# Install Docker
sudo apt update
sudo apt install -y docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin

# Add the current user to the docker group
sudo usermod -aG docker $USER

caution

Group change requires re-login The effect of usermod takes hold only after logging out and back in. Executing subsequent commands without re-logging in produces a permission denied error; this is expected behaviour and is resolved by logging out and back in.

2. Install the NVIDIA GPU driver

This step applies to NVIDIA platforms only. The installation procedure differs by platform; select the method corresponding to the model in use:

Platform	Installation method
NVIDIA platform (x86_64)	Installed from the Ubuntu package repository; see the NVIDIA driver installation guide
NVIDIA Thor platform (aarch64)	The driver is supplied with the JetPack system image and requires no separate installation

On x86_64 platforms, the available driver versions may first be queried as follows:

ubuntu-drivers devices

A restart is required following installation before the driver is loaded.

3. Install the NVIDIA Container Toolkit

This step applies to NVIDIA platforms only. The NVIDIA Container Toolkit grants containers access to the host GPU. Where it is not installed, Docker itself continues to operate normally, but the GPU is unavailable within the container and training falls back to CPU execution.

Its presence may first be confirmed as follows:

nvidia-ctk --version

Where the command cannot be executed, install it according to the following steps:

# Add the NVIDIA repository
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey \
  | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list \
  | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' \
  | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

# Install
sudo apt update
sudo apt install -y nvidia-container-toolkit

# Configure Docker to use the NVIDIA runtime and restart
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

For the complete description, see the official NVIDIA Container Toolkit installation guide.

4. Pre-installation verification

The status of the prerequisite components should be confirmed prior to executing the installation flow. On Intel platforms only the first item is required; on NVIDIA and Thor platforms all three must pass:

# 1. Docker executes correctly
docker run --rm hello-world

# 2. The host recognises the GPU (NVIDIA platforms only)
nvidia-smi

# 3. The GPU is accessible from within a container (NVIDIA platforms only)
docker run --rm --gpus all ubuntu nvidia-smi

The third item constitutes the actual criterion. Where the second passes but the third fails, the driver itself is functioning but the Container Toolkit is not correctly configured. The system remains installable and operable in this state, but training will be executed on CPU and will be markedly slower. Where GPU training is intended, step 3 should be completed before proceeding.

The installation flow below may be executed once all three verifications have passed.

Installation Flow

./Robotic-Suite_v0.0.408_x64_setup.run

The installation directory will be ~/Advantech/Robotic-Suite.

Optional: building the simulation environment

The simulation environment (simulation recording and SDG synthetic data generation) is not deployed as part of the installation flow; the source must be obtained separately and the container images built. This section may be skipped where only physical recording is used.

The simulation environment renders on the GPU, so all three items of the pre-installation verification must pass. Only the AIR-420 has been verified on actual hardware; see System Requirements.

Build steps

# Obtain the source
git clone https://github.com/ADVANTECH-Corp/Robotic-Suite-Physai-Simulation.git
cd Robotic-Suite-Physai-Simulation/

# Build the simulation environment itself
docker build -t pait/simulation:latest .

# Build the viewer service
docker build -t pait/isaacsim-viewer:latest ./isaacsim-viewer

# Start the viewer as a persistent container
docker run -d --restart always -p 5173:5173 \
  --name pait_isaacsim-viewer \
  pait/isaacsim-viewer:latest

The two images are used differently:

pait/simulation — the simulation environment itself, which only needs to be built; docker run is not required. The platform detects whether this image is present on the host and, if so, uses it to start the simulator on entry to the simulation environment.
pait/isaacsim-viewer — the viewer service, which must be run as a persistent container as shown above and occupies port 5173. --restart always causes it to start automatically after a host reboot, so that command need only be issued once.

caution

A one-off but lengthy build The built pait/simulation image occupies approximately 35 GB, and large base images are downloaded during the process; the time required depends on network bandwidth. Confirm that sufficient disk space is available before starting.

Verification

# Both images have been created
docker images | grep pait

# The viewer service is Up
docker ps --filter name=pait_isaacsim-viewer

The first is the criterion by which the platform determines the availability of the simulation environment: the appearance of pait/simulation in the list means it can be detected.

Once complete, refresh the browser page; the simulation features will then appear. For subsequent operation see Simulation & SDG Generation.

Next: Hardware preparation.

Prerequisites​

1. Install Docker​

2. Install the NVIDIA GPU driver​

3. Install the NVIDIA Container Toolkit​

4. Pre-installation verification​

Installation Flow​

Optional: building the simulation environment​

Build steps​

Verification​