Skip to main content

Hardware Requirements

Overview​

This course can be completed using physical hardware, cloud services, or a hybrid approach. Choose the option that best fits your budget and learning style.

Option 1: Physical Hardware Setup​

Total Cost: ~$800-1200 Best For: Hands-on learners who want tactile experience with sensors and actuators

Minimum Hardware​

ComponentSpecificationPurposeEst. CostVendor Links
NVIDIA Jetson Orin Nano8GB RAM, 1024 CUDA coresEdge AI inference for Isaac ROS$499NVIDIA Store
Intel RealSense D435iDepth camera + IMU, USB 3.0Visual SLAM and obstacle detection$279Intel Store
RPLiDAR A1360° laser scanner, 12m range, USB2D mapping and localization$99Slamtec Store
Servo Motors (12x)MG996R digital servos, 10kg-cm torqueHumanoid joint actuation$120Amazon
MicrocontrollerArduino Mega 2560 or Raspberry Pi PicoLow-level servo control$40Arduino Store
Power Supply12V 10A adapter + voltage regulatorPower for servos and sensors$30Generic vendors
Cabling & MountsJumper wires, breadboard, 3D-printed mountsAssembly$50Various
SD Card128GB UHS-I (for Jetson)OS and storage$20Generic vendors

Total: ~$1,137

Optional Enhancements​

| Component | Purpose | Cost | | ----------------------- | -------------------------------- | --------------------- | --- | | Additional Sensors | Ultrasonic (HC-SR04), GPS module | Redundancy | $30 | | Gripper | 2-finger parallel gripper | Object manipulation | $80 | | 3D Printer Filament | PLA for custom parts | Structural components | $25 |

Option 2: Cloud-Based Setup (Free Tier)​

Total Cost: $0/month (with free tier credits) Best For: Learners without access to physical hardware or working remotely

Cloud Services​

ServiceSpecificationPurposeFree TierUpgrade Cost
NVIDIA Omniverse CloudIsaac Sim streaming, RTX GPUSimulation and synthetic data generation1 year free trial$30/month after trial
GitHub Codespaces4-core CPU, 8GB RAM, 32GB storageROS 2 development environment60 hours/month free$0.18/hour after
Google Colab ProT4 GPU, 12GB VRAM, 100GB storageTraining vision modelsFree tier (limited GPU hours)$10/month Pro
AWS EC2 (g4dn.xlarge)4 vCPU, 16GB RAM, NVIDIA T4 GPUHeavy simulation/training750 hours/year free tier (new accounts)$0.526/hour

Total: $0 for first year with free tiers

Cloud Workflow​

  1. Develop on GitHub Codespaces (ROS 2 nodes, launch files)
  2. Simulate in Isaac Sim on Omniverse Cloud (physics, sensors)
  3. Train perception models on Google Colab (YOLOv8, segmentation)
  4. Deploy to cloud VM for final integration testing

Total Cost: ~$400 Best For: Most learners; balances hands-on experience with cost efficiency

Core Physical Hardware​

  • NVIDIA Jetson Orin Nano: $499 (essential for learning edge deployment)
  • Intel RealSense D435i: $279 (hands-on sensor experience)

Subtotal: $778

Cloud Supplements​

  • Isaac Sim (Omniverse Cloud): Simulate complex environments without GPU
  • Gazebo Classic: Run locally (no GPU required for simple scenes)

Total: ~$778 (one-time hardware cost)

Compromise Strategy​

  • Local development: ROS 2 nodes, URDF modeling, simple Gazebo worlds
  • Cloud simulation: Isaac Sim for photorealistic rendering and large-scale data generation
  • Physical deployment: Test final system on Jetson + RealSense

Software Requirements (All Free)​

Operating System​

  • Ubuntu 22.04 LTS (native installation or WSL2 on Windows)
  • Alternative: Ubuntu 20.04 LTS (if hardware doesn't support 22.04)

ROS 2​

  • Distribution: Humble Hawksbill (LTS until May 2027)
  • Installation: Binary packages via apt (recommended) or source build

Simulation​

  • Gazebo: Classic 11 or Gazebo Fortress (Ignition)
  • Isaac Sim: 2023.1.1+ (NVIDIA Omniverse)
  • Unity: 2022.3 LTS (optional, for high-fidelity visualization)

AI/ML Tools​

  • Python: 3.10 or 3.11
  • PyTorch: 2.0+ (with CUDA 11.8+ for GPU support)
  • OpenCV: 4.8+
  • NumPy, SciPy, Matplotlib

Development Tools​

  • Git: Version control
  • Docker: Containerization (optional but recommended)
  • VS Code: IDE with ROS extension

Architecture Summary​

System Diagram​

graph TD
    A[Sensors: RealSense, RPLiDAR] -->|Raw Data| B[Jetson Orin Nano]
    B -->|ROS 2 Topics| C[Isaac ROS Nodes]
    C -->|VSLAM/Nav2| D[Motion Planning]
    D -->|Commands| E[Arduino Mega]
    E -->|PWM Signals| F[Servo Motors]
    G[Cloud: Isaac Sim] -.->|Synthetic Data| C
    H[Development: Codespaces] -.->|Deploy| B

Data Flow​

  1. Perception: RealSense + LiDAR → Jetson → Isaac ROS (VSLAM, object detection)
  2. Planning: Nav2 (global/local planners) → Footstep planner → Joint trajectories
  3. Control: Arduino Mega receives joint commands → PWM to servos
  4. Feedback: Servo encoders → Joint states published to ROS 2

Detailed Component Specifications​

NVIDIA Jetson Orin Nano​

Why This Choice?

  • GPU acceleration for Isaac ROS (30 FPS VSLAM vs. 3 FPS on CPU)
  • 1024 CUDA cores enable real-time perception
  • Power-efficient (15W typical, up to 25W max)
  • Compact form factor (70mm x 45mm)

Setup:

  1. Flash JetPack 5.1+ (includes Ubuntu 20.04 + CUDA 11.4)
  2. Install ROS 2 Humble from source (binary packages not available for ARM64)
  3. Install Isaac ROS packages (pre-built for Jetson)

Alternatives:

  • Jetson Xavier NX ($399, discontinued but available used)
  • Jetson AGX Orin ($1,999, overkill for this course)
  • Raspberry Pi 4 (8GB) ($75, CPU-only, slower but budget-friendly)

Intel RealSense D435i​

Why This Choice?

  • Stereo depth: 1280x720 @ 30 FPS, range 0.1m-10m
  • RGB camera: 1920x1080 @ 30 FPS
  • IMU: BMI055 6-axis (accel + gyro) at 200 Hz
  • ROS 2 support: Official realsense2_camera package

Calibration:

# Install RealSense SDK
sudo apt install ros-humble-realsense2-camera

# Launch camera
ros2 launch realsense2_camera rs_launch.py

# Verify topics
ros2 topic list | grep camera
# /camera/depth/image_rect_raw
# /camera/color/image_raw
# /camera/imu

Alternatives:

  • OAK-D ($149, stereo + AI accelerator, smaller depth range)
  • Azure Kinect ($399, higher quality but discontinued)
  • ZED 2i ($449, better outdoor performance, larger FOV)

RPLiDAR A1​

Why This Choice?

  • 360° coverage in 12m range (indoor)
  • Sample rate: 8000 samples/second
  • ROS 2 driver: rplidar_ros2 package available
  • Affordable: Best cost/performance ratio

Setup:

# Install driver
sudo apt install ros-humble-rplidar-ros

# Grant USB permissions
sudo chmod 666 /dev/ttyUSB0

# Launch
ros2 launch rplidar_ros rplidar_a1_launch.py

# Visualize in RViz2
ros2 run rviz2 rviz2
# Add LaserScan display → Topic: /scan

Alternatives:

  • RPLIDAR A2 ($269, 16m range, faster)
  • YDLIDAR X4 ($99, similar specs to A1)
  • Simulated LiDAR: Use Gazebo ray sensor (free)

Budget Breakdown by Learning Goal​

Goal 1: Learn ROS 2 Only ($0)​

  • Use Gazebo simulation exclusively
  • No physical hardware needed

Goal 2: Learn Simulation + Deployment ($778)​

  • Jetson Orin Nano + RealSense D435i
  • Deploy Isaac ROS VSLAM and Nav2

Goal 3: Build Full Humanoid ($1,137+)​

  • Add LiDAR, servos, microcontroller
  • Construct physical robot chassis

Week 1-2​

  • None: Use simulation only

Week 5 (Start Module 2)​

  • Optional: Order Jetson Orin Nano (long lead time)

Week 8 (Start Module 3)​

  • Recommended: Have Jetson + RealSense for Isaac ROS labs

Week 11 (Start Module 4)​

  • Optional: Add microphone for voice input ($15-30)

Troubleshooting​

Issue: Jetson USB Ports Not Recognizing Devices​

Solution:

# Check USB devices
lsusb

# Reinstall USB rules
sudo udevadm control --reload-rules
sudo udevadm trigger

Issue: RealSense Driver Installation Fails​

Solution:

# Install from source
git clone https://github.com/IntelRealSense/librealsense.git
cd librealsense
mkdir build && cd build
cmake ../ -DCMAKE_BUILD_TYPE=Release
make -j4
sudo make install

Issue: Servo Motors Not Responding​

Solution:

  • Check power supply (servos need 6-12V, 2A+ current)
  • Verify PWM signal (use oscilloscope or logic analyzer)
  • Test servos individually with Arduino ServoTest sketch

Safety Guidelines​

⚠️ Electrical Safety:

  • Always disconnect power before wiring changes
  • Use fuses to protect circuits from overcurrent
  • Keep liquids away from electronics

⚠️ Mechanical Safety:

  • Servo motors can exert 10+ kg-cm torque (pinch hazard)
  • Secure robot to prevent falling
  • Test in open area away from fragile objects

⚠️ Software Safety:

  • Implement emergency stop button (physical + software)
  • Set joint limits in URDF to prevent self-collision
  • Always test in simulation before physical deployment

View weekly breakdown →

View all assessments →

Return to Homepage →