Skip to main content

Appendix A: Hardware Setup Guide

Complete hardware specifications and setup instructions for Physical AI development.

Overview​

This appendix provides detailed hardware recommendations for each module, purchase links, and setup instructions.

Minimum Requirements (Modules 1-2)​

Development Workstation​

  • CPU: Intel i5-12400 / AMD Ryzen 5 5600 (6 cores, 12 threads)
  • RAM: 16GB DDR4-3200
  • Storage: 256GB NVMe SSD (for OS + ROS 2 + Gazebo)
  • GPU: Integrated graphics (sufficient for Gazebo)
  • OS: Ubuntu 22.04 LTS (native installation, not VM)

Estimated Cost: $600-800

High-Performance Workstation​

  • CPU: Intel i7-13700K / AMD Ryzen 7 7700X (8+ cores)
  • RAM: 32GB DDR5-5200
  • Storage: 1TB NVMe Gen4 SSD
  • GPU: NVIDIA RTX 4070 Ti (12GB VRAM) or RTX 4080 (16GB VRAM)
  • PSU: 750W 80+ Gold
  • Cooling: Adequate for GPU thermal management

Estimated Cost: $2,000-2,500

GPU Comparison​

GPU ModelVRAMCUDA CoresTOPS (INT8)PriceUse Case
RTX 4060 Ti8GB4,352~22$400Entry-level (Module 3 only)
RTX 407012GB5,888~29$600Good for learning
RTX 4070 Ti12GB7,680~40$800Recommended
RTX 408016GB9,728~49$1,200High-end development
RTX 409024GB16,384~82$1,600Professional / research

Recommendation: RTX 4070 Ti offers best price/performance for this course.

Edge Deployment Hardware (Optional)​

NVIDIA Jetson Orin Series​

Jetson Orin Nano (Entry-level)

  • Performance: 40 TOPS
  • RAM: 8GB
  • Power: 7-15W
  • Price: ~$499
  • Use Case: Lightweight perception, testing

Jetson Orin NX (Mid-range)

  • Performance: 100 TOPS
  • RAM: 16GB
  • Power: 10-25W
  • Price: ~$899
  • Use Case: Autonomous navigation, moderate perception

Jetson Orin AGX (High-end)

  • Performance: 275 TOPS
  • RAM: 64GB
  • Power: 15-60W
  • Price: ~$1,999
  • Use Case: Complex VLA systems, production deployment

Sensors (Optional for Hardware Testing)​

Intel RealSense Depth Cameras​

RealSense D435 (General-purpose)

  • Depth Range: 0.3m - 3m
  • RGB: 1920x1080 @ 30 FPS
  • Depth: 1280x720 @ 30 FPS
  • Price: ~$200
  • Use Case: Indoor VSLAM, object detection

RealSense D455 (Longer range)

  • Depth Range: 0.4m - 6m
  • RGB: 1920x1080 @ 30 FPS
  • Depth: 1280x720 @ 30 FPS
  • Price: ~$350
  • Use Case: Outdoor navigation, larger spaces

LiDAR (Advanced)​

SLAMTEC RPLIDAR A1 (2D)

  • Range: 12m
  • Scan Rate: 5.5 Hz
  • Price: ~$99
  • Use Case: 2D mapping, obstacle avoidance

Ouster OS0 (3D, professional)

  • Range: 50m
  • Resolution: 64/128 beams
  • Price: ~$5,000+
  • Use Case: Autonomous vehicles, outdoor mapping

Robot Platforms (Optional)​

Mobile Robots​

TurtleBot 4 (Recommended for learning)

  • Base: iRobot Create 3
  • Sensors: RealSense D435, IMU, RPLIDAR
  • Price: ~$1,500
  • Use Case: ROS 2 development, navigation

NVIDIA Carter (Advanced)

  • Base: Segway
  • Sensors: Stereo cameras, LiDAR
  • Compute: Jetson AGX
  • Price: ~$10,000+
  • Use Case: Isaac ROS development

Humanoid Robots​

Unitree G1 (Research-grade)

  • DOF: 23 joints
  • Height: 1.3m
  • Weight: 35 kg
  • Price: ~$16,000
  • Use Case: Humanoid manipulation research

ROBOTIS OP3 (Educational)

  • DOF: 20 joints
  • Height: 0.5m
  • Price: ~$10,000
  • Use Case: Education, competition

Setup Instructions​

Desktop Workstation Setup​

  1. Install Ubuntu 22.04 LTS

  2. Install NVIDIA Drivers

    sudo ubuntu-drivers autoinstall
    sudo reboot

  3. Verify GPU

    nvidia-smi

    Should show GPU model and driver version.

  4. Install CUDA 12.x

    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
    sudo dpkg -i cuda-keyring_1.1-1_all.deb
    sudo apt update
    sudo apt install cuda-toolkit-12-4

  5. Install cuDNN

  6. Continue to Appendix B for ROS 2 installation

Jetson Setup​

Flashing Jetson Orin​

  1. Download JetPack 6.x from NVIDIA SDK Manager
  2. Connect Jetson via USB in recovery mode
  3. Flash with SDK Manager (Ubuntu host required)
  4. Follow on-screen instructions

Detailed guide: https://developer.nvidia.com/embedded/jetpack

Purchasing Recommendations​

Budget Tiers​

Tier 1: Minimum ($800-1,000)

  • Desktop CPU + 16GB RAM + 256GB SSD
  • Integrated GPU (Modules 1-2 only)
  • No sensors or robots

Tier 2: Recommended ($2,500-3,000)

  • High-performance desktop with RTX 4070 Ti
  • Complete Modules 1-4 with GPU acceleration
  • Optional: RealSense D435 for testing

Tier 3: Professional ($5,000-10,000)

  • Workstation with RTX 4080/4090
  • Jetson Orin AGX for edge deployment
  • TurtleBot 4 for physical testing
  • Multiple sensors (RealSense, LiDAR)

Cloud Alternatives​

If you don't have local GPU:

  • AWS EC2 G5 instances: RTX A10G (24GB VRAM), ~$1-2/hour
  • Google Cloud GPU VMs: A100 (40GB VRAM), ~$2-4/hour
  • Paperspace: RTX 4000/5000 series, ~$0.50-1.50/hour

Cost estimate: $100-200 for entire course if using cloud GPUs.

Troubleshooting Hardware​

GPU Not Detected​

  1. Check NVIDIA drivers: nvidia-smi
  2. Verify CUDA: nvcc --version
  3. Reinstall drivers if needed

Insufficient VRAM​

  • Reduce batch size in DNN inference
  • Use INT8 quantization
  • Consider cloud GPU upgrade

Overheating​

  • Check thermal paste on CPU/GPU
  • Improve case airflow
  • Consider aftermarket cooling

See Also:

  • Appendix B: Software Installation Guide
  • Appendix E: Cloud vs. On-Premise GPU Deployment