Recommended Setup (Jetson + Hardware)
Overview
This chapter describes the recommended hardware setup for students who want to:
- Run perception and control algorithms on edge hardware
- Interface with real sensors (cameras, microphones)
- Prepare for optional physical robot deployment
- Develop VLA systems with multimodal input
This setup enables both simulation-based learning AND edge hardware development, providing a bridge between simulation and physical robots.
Bill of Materials (Recommended Kit)
Core Components
| Item | Model | Qty | Cost (USD) | Purpose |
|---|---|---|---|---|
| Compute | NVIDIA Jetson Orin Nano | 1 | $199 | Edge AI accelerator for perception |
| Compute Power | Official Jetson Power Adapter (25W) | 1 | $35 | Stable power supply |
| Storage | Samsung 128GB NVMe SSD | 1 | $25 | OS + code storage |
| Camera | Intel RealSense D435i | 1 | $165 | RGB-D depth sensing |
| Microphone Array | ReSpeaker Mic Array v2.0 | 1 | $70 | Voice input, multi-channel audio |
| Networking | WiFi USB adapter (optional) | 1 | $20 | Wireless connectivity if needed |
| Cooling | Heatsink + fan for Jetson | 1 | $15 | Thermal management |
| Accessories | USB cables, brackets | - | $30 | Assembly & connectivity |
Total Estimated Cost: $559 USD
Optional Additions
| Item | Cost | Reason |
|---|---|---|
| 4G LTE modem | $40-60 | Remote robot connectivity |
| GPIO breakout board | $10 | PWM servo control |
| Robot platform (e.g., ROS Robot Base) | $300-500 | Physical mobility (optional) |
| Second RealSense camera | $165 | Stereo vision or full surround |
Hardware Component Details
1. NVIDIA Jetson Orin Nano Developer Kit
What it is: A single-board computer with NVIDIA's Orin Nano GPU, designed for edge AI inference and real-time processing.
Specifications:
- CPU: 6-core ARM Cortex-A78AE @ 2.0 GHz
- GPU: 1024-CUDA-core NVIDIA Orin Nano
- RAM: 8 GB LPDDR5 (shared)
- Storage: microSD or NVMe SSD via USB
- Power: 5-25W (passive to active cooling)
- Interfaces: USB 3.0, USB Type-C, Ethernet, HDMI, GPIO pins
- AI Performance: 8-40 TFLOPS (FP32/INT8) for inference
Why Choose Jetson Orin Nano:
- ✅ ROS 2 Humble officially supported
- ✅ Pre-built Isaac ROS packages
- ✅ GPU-accelerated SLAM and perception
- ✅ Low power consumption (fits on mobile robots)
- ✅ Cost-effective for educational use
Installation:
# Flash JetPack OS on microSD
# 1. Download JetPack 5.1.2 (supports Humble)
# 2. Use NVIDIA SDK Manager or balena Etcher
# 3. Boot Jetson from microSD
# 4. Complete onboard setup wizard
# Verify after booting
cat /etc/nv_tegra_release
# Should show: R36 (release) or later
2. Intel RealSense D435i Depth Camera
What it is: A compact RGB-D camera combining color (1280×720) and depth (1280×720) sensors.
Specifications:
- RGB Sensor: 1280×720 @ 30 FPS
- Depth Sensor: 1280×720 @ 30 FPS (5m range)
- Depth Technology: Structured light (active) + stereo (passive fallback)
- IMU: 6-axis (accelerometer + gyroscope)
- Interface: USB 3.0
- Power: 380 mA @ 5V (powered via USB)
Why RealSense D435i:
- ✅ Depth + color in one device
- ✅ High accuracy for SLAM (5-10cm over 5m)
- ✅ ROS 2 drivers readily available
- ✅ IMU useful for odometry
- ✅ Well-supported in Isaac ROS perception
ROS 2 Integration:
# Install realsense2 package
sudo apt install ros-humble-realsense2-camera
# Launch camera
ros2 launch realsense2_camera rs_launch.py depth_module.profile:=1280x720x30 rgb_camera.profile:=1280x720x30
# Verify topics
ros2 topic list | grep camera
# /camera/color/image_raw
# /camera/depth/image_rect_raw
# /camera/imu
3. ReSpeaker Mic Array v2.0
What it is: A 6-microphone circular array optimized for voice recognition in noisy environments.
Specifications:
- Microphones: 6× MEMS omnidirectional mics
- Beamforming: 4 directional pickup patterns
- Audio Format: 16-bit, 16 kHz mono output
- Interface: USB
- Power: 100 mA @ 5V
- LED Ring: 12-LED RGB indicator for feedback
Why ReSpeaker:
- ✅ Beamforming reduces background noise
- ✅ Compact and easy to integrate
- ✅ ROS 2 drivers available
- ✅ Supports multiple voice commands simultaneously
- ✅ LED feedback for visual status
ROS 2 Integration:
# Install ReSpeaker driver
pip install respeaker
# Or use raw audio via pulseaudio
pactl list sources | grep respeaker
# ROS 2 audio bridge
ros2 run audio_common_nodes audio_node --ros-args -p device:=/dev/snd/by-id/usb-Respeaker_Mic_Array_v2.0*
# Verify
ros2 topic list | grep audio
# /audio (audio_common_msgs/AudioData)
Complete Hardware Setup Guide
Physical Assembly
Step 1: Jetson Preparation
-
Attach heatsink + fan to Jetson:
- Apply thermal paste to GPU
- Screw heatsink firmly
- Connect fan header
-
Insert NVMe SSD:
- Open M.2 slot cover
- Insert SSD at 45° angle
- Press down and screw
-
Optional: Mount in case with cooling
Step 2: Camera & Microphone Mounting
-
RealSense D435i:
- Mount on robot front
- Connect USB 3.0 to Jetson
- Test initial alignment
-
ReSpeaker Mic Array:
- Mount on robot top or body
- Connect USB to Jetson
- Position for omnidirectional input
Step 3: Power Delivery
- Official Jetson power adapter → USB-C on Jetson
- RealSense powered via USB 3.0 port
- ReSpeaker powered via USB port
- Total power draw: ~15W nominal; budget for 25W peaks
Step 4: Network Connectivity
# If using Ethernet
sudo apt install network-manager
nmtui # Configure network
# If using WiFi
sudo apt install wpasupplicant wireless-tools
# ... configure via NetworkManager
Software Installation on Jetson
Step 1: Flash JetPack OS
- Download NVIDIA SDK Manager on host PC
- Select Jetson Orin Nano + JetPack 5.1.2
- Follow guided flashing process
- Complete onboard setup on Jetson
Step 2: Install ROS 2 Humble
# Add ROS 2 repository (same as Ubuntu)
sudo apt install curl gnupg lsb-release ubuntu-keyring
curl -sSL https://raw.githubusercontent.com/ros/rosdistro/master/ros.key | sudo apt-key add -
sudo sh -c 'echo "deb [arch=$(dpkg --print-architecture)] http://packages.ros.org/ros2/ubuntu $(lsb_release -cs) main" > /etc/apt/sources.list.d/ros2-latest.list'
sudo apt update
sudo apt install ros-humble-desktop
Step 3: Install Isaac ROS (Perception Accelerated)
# Clone Isaac ROS repository
mkdir -p ~/isaac_ros_ws/src
cd ~/isaac_ros_ws/src
git clone https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_common.git
git clone https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_visual_slam.git
# ... other Isaac ROS packages
# Build
cd ~/isaac_ros_ws
colcon build
Step 4: Install Sensor Drivers
# RealSense
sudo apt install ros-humble-realsense2-camera librealsense2 librealsense2-dev
# ReSpeaker (via pip in Python environment)
pip install respeaker
# Audio common
sudo apt install ros-humble-audio-common
Step 5: Verify Hardware Integration
# Test RealSense
ros2 launch realsense2_camera rs_launch.py &
ros2 topic echo /camera/color/image_raw | head -10
# Test ReSpeaker
ros2 run audio_common_nodes audio_node &
ros2 topic echo /audio | head -10
# Test Isaac ROS VSLAM (if installed)
ros2 launch isaac_ros_visual_slam isaac_ros_visual_slam_jetson.launch.py
Performance Benchmarks
Jetson Orin Nano Performance
| Task | Performance | Notes |
|---|---|---|
| Gazebo simulation | 20-30 FPS | Single 640×480 world |
| Isaac ROS VSLAM | 30 FPS @ 640×480 | Real-time visual odometry |
| YOLO v8 inference | 10-15 FPS | On 1280×720 RGB stream |
| ROS 2 node spin rate | 100-1000 Hz | Depending on workload |
| Whisper ASR (small) | ~2 sec per 30 sec audio | GPU accelerated |
Power Consumption
| Workload | Power (W) | Duration |
|---|---|---|
| Idle (desktop) | 5 | Continuous |
| ROS 2 idle spin | 7 | Continuous |
| Gazebo simulation | 12-15 | During simulation |
| Camera + SLAM | 15-18 | Real-time perception |
| Peak (all sensors + SLAM + inference) | 20-25 | During heavy compute |
Cost-Performance Tradeoffs
Why This Specific Configuration?
Jetson Orin Nano vs. Alternatives:
| Platform | Cost | GPU | ROS 2 Support | Best For |
|---|---|---|---|---|
| Jetson Orin Nano | $199 | 1024 CUDA | ✅ Excellent | Learning + edge deployment |
| Jetson Orin Nano Developer Kit | $249 | 1024 CUDA | ✅ Excellent | Same, with accessories |
| Raspberry Pi 4 (8GB) | $75 | None (CPU) | ✅ Good | Low power; no GPU inference |
| Laptop (Intel i7) | $800+ | Optional | ✅ Excellent | Development; not portable |
| Jetson Orin NX | $149 | 512 CUDA | ✅ Excellent | Budget; half the performance |
| Jetson AGX Orin | $999+ | 12288 CUDA | ✅ Excellent | Industry-grade; excessive for learning |
RealSense D435i vs. Alternatives:
| Camera | Cost | Depth Tech | Accuracy | ROS 2 |
|---|---|---|---|---|
| RealSense D435i | $165 | Structured light | 5-10 cm @ 5m | ✅ Official drivers |
| RealSense D455 | $199 | Stereo | 5-10 cm @ 5m | ✅ Official drivers |
| Azure Kinect DK | $299 | ToF | 5 cm @ 5m | ✅ Community drivers |
| ZED 2i | $399 | Stereo | 10 cm @ 10m | ✅ Official support |
| OAK-D | $100 | Stereo (mono+spatial) | 10-20 cm | ✅ Community |
ReSpeaker vs. Alternatives:
| Mic Array | Cost | Channels | Beamforming | ROS 2 |
|---|---|---|---|---|
| ReSpeaker v2.0 | $70 | 6 | ✅ Yes | ✅ Community |
| ReSpeaker USB | $50 | 4 | ✅ Yes | ✅ Community |
| Respeaker Core+ | $120 | 6 | ✅ Yes | ✅ Community |
| Blue Yeti Pro | $130 | 4 | ❌ No | ❌ Limited |
| Generic USB mic | $20 | 1 | ❌ No | ✅ Native audio |
Integration Checklist
- Jetson Orin Nano flashed with JetPack 5.1.2
- ROS 2 Humble installed and verified
- RealSense camera detected:
realsense-viewershows streams - ReSpeaker mic array detected:
pulseaudiolists device - All 3 devices accessible via ROS 2 topics
- Power supply rated for 25W sustained
- Wireless or Ethernet connectivity verified
Troubleshooting
Jetson Not Booting After JetPack Flash
Solution:
# Reflash using SDK Manager on host PC
# Select "Force Recovery Mode" if needed
RealSense Camera Not Detected
Solution:
# Check USB connection
lsusb | grep Intel
# Install librealsense
sudo apt install librealsense2 librealsense2-dev
# Test
realsense-viewer
ReSpeaker No Audio Input
Solution:
# List audio devices
pactl list sources | grep respeaker
# Test via PulseAudio
parecord -d <device-name> | aplay -f cd
Next Steps
Once this hardware is set up:
- Complete Modules 0–2 on desktop/laptop (faster iteration)
- Deploy to Jetson for Module 3 (Isaac ROS, SLAM acceleration)
- Integrate VLA on Jetson for Module 4 (voice + perception)
- Capstone on Jetson: Voice-controlled perception system
Estimated time to first working example on Jetson: 2-3 hours (after JetPack flashing)
Cost Breakdown
| Category | Items | Cost |
|---|---|---|
| Compute | Jetson + power + storage | $259 |
| Perception | RealSense D435i | $165 |
| Audio | ReSpeaker Mic Array | $70 |
| Accessories | Cables, cooling, mounting | $65 |
| Total | All-in-one edge AI kit | $559 |
Compare to typical robotics systems: $2,000–$10,000 for equivalent hardware-enabled platform.
Further Resources
- Jetson Orin Nano Documentation: https://developer.nvidia.com/jetson-orin-nano
- RealSense Documentation: https://github.com/IntelRealSense/librealsense
- ReSpeaker Documentation: https://wiki.seeedstudio.com/ReSpeaker-Mic-Array-v2.0/
- Isaac ROS: https://github.com/NVIDIA-ISAAC-ROS
Last Updated: 2025-12-10 Relevant For: Modules 2–4 (Simulation, Isaac, VLA) Capstone Connection: Enables real-world perception and voice input for capstone integration