Skip to main content

Foundational Setup

Overview​

Before starting the 13-week Physical AI course, you need to choose a hardware setup and install the required software stack. This guide helps you select the right option based on your budget, performance needs, and learning goals.

Total Setup Time: 4-6 hours (varies by hardware option)

Hardware Setup Options​

You have three hardware options for this course. Each has trade-offs in cost, performance, and use cases.

Quick Comparison Table​

FeatureDigital Twin WorkstationJetson Orin NanoCloud GPU
Upfront Cost$1,530-2,450$599$0
13-Week Cost$0 (one-time)$200-300 (cloud supplement)$200-600
Total Cost$1,530-2,450$799-899$200-600
Isaac Sim Support✅ Full support❌ Not supported✅ Full support
Gazebo Performance30-40 FPS10-15 FPS30-40 FPS
VLA Inference2-3 tokens/sec0.5-1 token/sec (quantized)2-3 tokens/sec
Offline Work✅ Yes✅ Yes❌ Requires internet
Physical Deployment❌ Not portable✅ Edge deployment❌ Cloud only
Power Consumption400-600W7-15WN/A
Best ForLong-term use, max performanceEdge deployment testingNo hardware budget

Decision Tree​

START: What's your priority?

├─ **Maximum Performance** + Budget for Hardware Investment
│  └─> Digital Twin Workstation ($1,530-2,450)
│     - Runs all modules locally
│     - Best for offline development
│     - 3-5 year hardware lifespan
│
├─ **Edge Deployment** + Portable Testing
│  └─> Jetson Orin Nano + Cloud Supplement ($799-899)
│     - Test on physical robot hardware
│     - Low power consumption
│     - Use cloud for Isaac Sim (Module 3)
│
└─ **No Upfront Cost** + Remote Learning
   └─> Cloud GPU Only ($200-600)
      - Pay-as-you-go pricing
      - Access from anywhere
      - Requires stable internet

Hardware Setup Guides​

Choose your hardware option and follow the corresponding guide:

  1. Digital Twin Workstation Setup

    • Hardware requirements (RTX 3060+, 32GB RAM, Ubuntu 22.04)
    • GPU driver and CUDA installation
    • Performance optimization tips
    • Cost: $1,530-2,450 (one-time)

  2. Jetson Orin Nano Edge Kit Setup

    • Jetson specifications and accessories
    • JetPack 6.0 installation
    • NVMe SSD setup and power configuration
    • Cost: $599 hardware + $200-300 cloud supplement

  3. Cloud GPU Development Setup

    • Provider comparison (AWS, GCP, Lambda Labs, Vast.ai)
    • Cost optimization strategies (spot instances, auto-shutdown)
    • Remote desktop configuration
    • Cost: $200-600 for 13 weeks

Software Installation​

After setting up hardware, install the software stack in this order:

Step 1: ROS 2 Humble (Required for All Modules)​

ROS 2 Humble Installation Guide

  • Install ROS 2 Humble Desktop on Ubuntu 22.04
  • Configure environment and create workspace
  • Verify installation with talker/listener demo
  • Time: 45-60 minutes

Module Coverage: Required for Modules 1-4

Step 2: Isaac Sim (Required for Module 3 Only)​

Isaac Sim Installation Guide

  • Install NVIDIA Omniverse Launcher
  • Install Isaac Sim 2023.1.1
  • Configure ROS 2 bridge
  • Verify GPU acceleration
  • Time: 2-3 hours (including download)

Module Coverage: Module 3 (Weeks 8-10)

⚠️ Hardware Requirements:

  • GPU: RTX 3060+ (12GB+ VRAM)
  • NOT supported on Jetson Orin Nano
  • Use cloud GPU if you have Jetson setup
# Install VS Code
sudo snap install code --classic

# Install Git (if not already installed)
sudo apt install git -y

# Install Python tools
pip3 install --upgrade pip
pip3 install jupyter notebook matplotlib numpy scipy

# Install Docker (optional, for containerized deployments)
sudo apt install docker.io -y
sudo usermod -aG docker $USER

Verification Checklist​

After completing setup, verify all components:

Hardware Verification​

  • Ubuntu 22.04 LTS installed and updated
  • GPU detected (if applicable): nvidia-smi
  • At least 32GB RAM (workstation) or 8GB (Jetson)
  • 100GB+ free disk space
  • Stable internet connection (>50 Mbps recommended)

Software Verification​

  • ROS 2 Humble installed: ros2 --version
  • ROS 2 environment sourced: echo $ROS_DISTRO returns humble
  • Talker/listener demo works across terminals
  • Workspace created: ~/physical_ai_ws/
  • Isaac Sim installed (if using Workstation/Cloud for Module 3)
  • Git configured: git --version
  • VS Code or preferred IDE installed

Network Configuration​

  • Firewall allows ROS 2 DDS communication (ports 7400-7500)
  • ROS_DOMAIN_ID set (default: 0)
  • SSH access configured (for remote development)

Common Issues​

Issue: ROS 2 nodes can't communicate between terminals​

Cause: Firewall blocking DDS multicast.

Solution:

# Allow DDS traffic
sudo ufw allow 7400:7500/udp
sudo ufw allow from 224.0.0.0/4

Issue: Isaac Sim won't launch​

Cause: Insufficient VRAM or driver issue.

Solution:

# Check VRAM (need 12GB+)
nvidia-smi --query-gpu=memory.total --format=csv

# Update driver if needed
sudo ubuntu-drivers autoinstall
sudo reboot

Issue: Slow performance on Jetson​

Cause: Not using maximum power mode.

Solution:

# Set to 15W mode
sudo nvpmodel -m 0
sudo jetson_clocks

Cost Summary​

Option 1: Digital Twin Workstation​

  • Hardware: $1,530-2,450 (one-time)
  • Software: $0 (all free/open-source)
  • 13-Week Total: $1,530-2,450
  • Cost per Week: $118-188
  • Amortized (3 years): $10-16/week

Option 2: Jetson + Cloud Hybrid​

  • Jetson Hardware: $599 (one-time)
  • Cloud Supplement: $200-300 (13 weeks for Isaac Sim)
  • 13-Week Total: $799-899
  • Cost per Week: $61-69

Option 3: Cloud Only​

  • Hardware: $0
  • Cloud GPU: $200-600 (13 weeks, depends on usage)
  • 13-Week Total: $200-600
  • Cost per Week: $15-46

Next Steps​

✅ Hardware and Software Setup Complete!

Start Learning:

  1. Module 1: ROS 2 Fundamentals (Weeks 3-5)

    • Nodes, topics, services, actions
    • Multi-node system architecture
    • ROS 2 package development

  2. Module 2: Digital Twin Simulation (Weeks 6-7)

    • Gazebo simulation
    • Unity ML-Agents
    • Sim-to-real transfer

  3. Module 3: NVIDIA Isaac Sim (Weeks 8-10)

    • High-fidelity perception pipeline
    • Object detection and pose estimation
    • Sensor fusion

  4. Module 4: VLA Models (Weeks 11-13)

    • Natural language control
    • Task decomposition
    • Autonomous humanoid capstone

Support and Troubleshooting​

If you encounter issues during setup:

  1. Check Troubleshooting Guide: Reference → Troubleshooting
  2. Search Glossary: Reference → Glossary
  3. Community Forums: ROS Discourse, NVIDIA Isaac Sim Forums

Setup Complete! → Start with Module 1: ROS 2 Fundamentals