How to build NVIDIA JetBot (Jetson Nano AI education robot): the whole process
121 steps to build NVIDIA JetBot (Jetson Nano AI education robot), from buying parts, printing and assembly to software, calibration and training, with the bill of mater…
NVIDIA's docs put the parts under $250 including the Jetson Nano, or under $150 on top of a Nano; the BOM prices are undated and the Jetson Nano Developer Kit is discontinued. The Orin BOM lists the Jetson Orin Nano 8GB Developer Kit at $499. Waveshare's official store sells the JetBot AI Kit accessories pack for $105.99 (no Jetson Nano or 18650 batteries; 2026‑11‑01, tax and shipping not stated)
Bill of materials · Common parts
| Part | Qty | ANY |
|---|---|---|
| Jetson Nano · NVIDIA | 1 | USD 99 |
| Micro SD card · Amazon | 1 | USD 13.99 |
| Power supply · Adafruit | 1 | USD 7.5 |
| Motor · Adafruit, Amazon(1), Amazon(2) | 2 | USD 5.9 |
| Motor Driver · Adafruit, Amazon | 1 | USD 19.95 |
| Caster ball · Amazon | 1 | USD 6.3 |
| Battery · Amazon | 1 | USD 15.99 |
| USB cable pack · Amazon | 1 | USD 6.99 |
| *PiOLED* display · Adafruit, Amazon | 1 | USD 14.95 |
| *PiOLED* header · Adafruit, Amazon, Sparkfun | 1 | USD 5.95 |
| Chassis · STL file | 1 | |
| Camera Mount · STL file | 1 |
Bill of materials · Option 1 (default) - Leopard Imaging Camera
| Part | Qty | ANY |
|---|---|---|
| Camera · Leopard Imaging | 1 | USD 29 |
Buy parts
- Plan the Orin JetBot assembly
Plan the JetBot around the Jetson Orin Nano 8GB Developer Kit and use the Bill of Materials (Orin) page. Software for this Orin version is still work in progress.
- Buy the common components
Obtain the common components listed in the Common parts table for each JetBot.
- Select the Leopard Imaging camera
Use the default camera option for each JetBot and obtain the parts specified for it.
- Select the IMX219-160 camera
Select the IMX219-160 Camera option and obtain the parts specified for it.
- Select the RPi Camera V2 and lens
Select the RPi Camera V2 + Lens Attachment option and obtain the specified camera and lens attachment parts.
- Select the M2 card and antennas
This option suits a JetBot built with the original Jetson Nano 4GB and has been widely tested and used. It does not work with Jetson Nano 2GB because that board has no M.2 connector.
- Select the USB WiFi dongle
The USB Dongle option is compatible with both the original Jetson Nano 4GB and Jetson Nano 2GB.
- Buy TT shaft wheels
Each JetBot needs wheels with a "TT" motor shaft connector. Select an option and get the specified parts.
- Select and buy 60 mm parts
Select the 60 mm option and get the parts specified for it.
- Select and buy 65 mm parts
Select the 65 mm option and get the parts specified for it.
- Order assembly hardware
Get the following parts needed to build JetBot. They come in packs, so order the quantity needed for the number of JetBots being built.
- Choose Jetson Orin Nano 8GB
For a new JetBot build, the document recommends planning around the Jetson Orin Nano 8GB Developer Kit. The software for the Orin version is still work in progress.
- Buy Common JetBot Components
Purchase the common components listed in the table for each JetBot.
- Select a Camera Option
Choose one of the three camera options and purchase the parts specified for that option.
- Buy Leopard Imaging Camera
Select the default Leopard Imaging Camera option and purchase its specified parts.
- Buy IMX219-160 Camera
Select the IMX219-160 Camera option and purchase its specified parts.
- Buy RPi Camera V2 and Lens
Select the RPi Camera V2 + Lens Attachment option and purchase its two specified parts.
- Select a WiFi Option
Choose one of the two WiFi options and purchase the parts specified for that option.
- Buy M2 Card and Antennas
This option is a good choice for a JetBot built with the original Jetson Nano 4GB and has been widely tested and used. It will not work with Jetson Nano 2GB because that device has no M.2 connector.
- Buy USB WiFi Dongle
This option is compatible with both the original Jetson Nano 4GB and Jetson Nano 2GB. For other USB WiFi dongles usable with Jetson Nano, check the Supported Components List.
- Select the Orin Nano 8GB Kit
Plan a new JetBot assembly around the NVIDIA Jetson Orin Nano 8GB Developer Kit.
- Buy Common Parts for Each JetBot
Obtain the common components listed in the bill of materials for each JetBot.
- Purchase the Power Source
Select and purchase the power source listed in the bill of materials.
- Select and Purchase the Camera
Select a camera option and obtain the specified parts.
- Buy the Default IMX219 Wide-FOV Camera
Use a camera module with a 22pin cable for the Jetson Orin Nano 8GB Developer Kit carrier board and the appropriate wide field of view for each JetBot.
- Choose a JetBot Vehicle
First choose the JetBot hardware vehicle to build or use, comparing the original NVIDIA reference design with third-party hardware.
- Source Parts and Set Up JetBot
After choosing a vehicle, obtain the required parts and follow the applicable setup instructions before moving on to the JetBot examples.
- Get the DIY Kit Parts
Obtain all the parts needed for the DIY open-source JetBot using the bill of materials.
Print and make
- Print the Provided STL Parts
Some parts require 3D printing. Use the STL files provided by the project and follow the printing tips on the 3d printing page.
- Print the 3D-printed JetBot parts
Print the required parts using the provided STL files and consult the tips on the 3d printing page.
Assemble
- Remove support material from motor sockets
Remove the support material from the motor sockets using pliers.
- Clean surfaces, openings, and screw holes
Use pliers, tweezers, and/or fingernails to remove support material from all surfaces, openings, and screw holes.
- File openings that are too tight
File away material from openings if they are too tight.
- Cut and remove motor zip ties
Cut and remove the zip tie from each motor.
- Tuck motor wires beneath the transparent band
Tuck the wires beneath the transparent band to prevent stress on the solder joints.
- Insert motors into the printed chassis
Insert each motor into the 3D printed chassis.
- Secure motors with M3x25mm screws
Secure each motor with 2x M3x25mm screws.
- Separate male headers from the link
Break away the 3-pin and 2-pin male headers from the header link included with the motor driver package.
- Join screw terminal blocks
Join two screw terminal blocks to create a 4-pin screw terminal.
- Solder screw terminals to motor driver
Solder the 2-pin and 4-pin screw terminals to the motor driver.
- Solder male headers to motor driver
Solder the 2-pin and 3-pin male headers to the motor driver. The 3-pin header is offset by 1 pin from the corner, so pay careful attention to its position.
- Select the correct micro USB cable
Select the depicted micro USB cable from the pack. The other micro USB cable powers the Jetson Nano, and using the wrong power cable for Nano will block its Ethernet port.
- Cut and strip the cable
Cut the depicted cable roughly 20 cm from the Type-A terminal, then strip about 3 cm of rubber coating from the cut.
- Remove extra wires and shielding
Remove the green and white wires and the metal shielding.
- Strip the red and black wires
Strip the red and black wires about 5 mm from the cut.
- Twist and solder exposed strands
Optionally twist the exposed wire strands and coat them with solder.
- Route cable through chassis
Route the spliced cable through the chassis.
- Connect motor driver power
Secure the exposed red (+) and black (-) wires to the motor driver power terminals. Check the driver markings carefully because incorrect wiring could damage it.
- Route cable beneath driver
Route the spliced cable under the motor driver to organize it and prevent stress on the screw terminal.
- Mount the motor driver
Secure the motor driver to the chassis using self-tapping screws.
- Connect both motor wires
Secure the left and right motor wires to the motor screw terminals. The red and black wires for the two motors are flipped in the image; this will not damage the motors but will make them spin backward, so flip them to the correct orientation.
- Place the caster ball inside the shroud
Place the caster ball inside the shroud.
- Insert the caster base
Insert the caster base into the shroud and secure the ball. If the ball socket enclosure fits too tightly, use a file to remove excess material.
- Attach the caster to the chassis
Attach the assembled caster to the chassis with self-tapping screws.
- Secure the 6-pin right-angle header
Secure the 6-pin right-angle header to the PiOLED display by some method.
- Solder the header pins
Solder the depicted 3.3V, SDA, SCL, and GND pins. The pins are close to the display screen, so avoid touching it with the soldering iron tip.
- Select a four-wire jumper segment
Select a four-wire segment of female-female jumper cables. The document uses brown, red, orange, and yellow wires, and the colors can be mapped accordingly.
- Attach wires to the motor driver
As shown, attach 3.3V (red), GND (brown), SDA (orange), and SCL (yellow) to the motor driver.
- Route the jumper cable through the chassis
Route the four-wire jumper cable segment through the chassis.
- Attach wires to the right-angle header
As depicted, attach the red, brown, orange, and yellow wires to the right-angle header. Reversing the red and brown wires could damage the system.
- Attach the camera
Attach the camera to the camera mount using 4x self tapping screws.
- Attach the camera mount
Attach the camera mount to the chassis with 4x self tapping screws.
- Insert the ribbon cable
Insert the other end of the ribbon cable into the Jetson Nano developer kit.
- Tuck the ribbon cable
Optionally, tuck the ribbon cable against the heat sink to organize it.
- Peel off the original sensor
Peel off the original camera sensor.
- Remove excess adhesive
Remove any excess adhesive material.
- Remove the sensor adhesive film
Remove the adhesive film cover from the wide angle replacement sensor.
- Attach the wide angle sensor
Attach the wide angle replacement sensor to the camera module.
- Fasten the sensor connector
Ensure the sensor connector is fastened to the camera module.
- Attach battery adhesive tabs
Attach double-sided adhesive tabs to each end of the battery, leaving some tab exposed.
Software
- Download the pre-built JetBot SD card image
Download the pre-built JetBot SD card image from the table and select the version matching the Jetson being used, for example Jetson Nano 2GB.
- Download the original JetPack image
For JetBot SD card images based on JetPack 4.5, first download the plain JetPack 4.5 SD card image for the device; the document links images for Jetson Nano 2GB and Jetson Nano 4GB.
- Complete Jetson Nano initial setup
Boot the Jetson Nano with the plain JetPack 4.5 SD card image and complete the operating system setup. This performs a one-time configuration that enables the device to use SD card images based on JetPack 4.5.
- Use the JetBot JetPack 4.5 image
After the one-time setup is complete, use the JetPack 4.5-based JetBot SD card images listed in the document on the device.
- Insert the SD card into your desktop
Insert the SD card into the desktop running Etcher.
- Flash the JetBot image with Etcher
Use Etcher to select the image downloaded above and flash it onto the SD card.
- Remove the SD card from the desktop
After flashing, remove the SD card from the desktop machine.
- Insert the SD card into Jetson Nano
Insert the SD card into the Jetson Nano; the micro SD card slot is under the module.
- Power on Jetson Nano with the charger
Power on the Jetson Nano by connecting the micro USB charger for the 4GB model or USB-C charger for the 2GB model to the port.
- Log in to Jetson Nano
Log in with the user jetbot and password jetbot.
- Connect to a WiFi network
Run the provided nmcli command in the command line to connect to a WiFi network.
- Shut Down JetBot Using the Command Line
After the robot is connected to WiFi, shut down JetBot using the command line.
- Unplug the Jetson Nano Connections
Unplug the HDMI monitor, USB keyboard, mouse, and power supply from the Jetson Nano.
- Power JetBot with the USB Battery Pack
Plug in the micro-USB cable to power the JetBot from the USB battery pack.
- Wait for JetBot to Boot
Wait a little while for JetBot to boot.
- Check the IP Address on the piOLED Display
Check the robot's IP address on the piOLED display, then use it in place of `<jetbot_ip_address>` in the next command.
- Open JetBot in the Web Browser
Navigate to `http://<jetbot_ip_address>:8888` from your desktop's web browser. You can do this from any machine on your local network.
- Sign In with the JetBot Password
Sign in using the password `jetbot`.
- Access the Remote Programming Environment
After signing in, you can access JetBot's remote programming environment and the JetBot examples.
- Load JetBot environment variables
Enter the docker directory and source configure.sh to configure environment variables related to JetBot Docker.
- Set the L4T base image
For L4T 32.6.1, manually set JETBOT_BASE_IMAGE to the specified NVIDIA L4T PyTorch container image.
- Set the NVIDIA Docker runtime
Set the default Docker runtime to NVIDIA so the containers can use CUDA-related components.
- Set Jupyter memory limits
If needed, set the Jupyter container memory limit to 500m and the memory swap limit to 3G.
- Build the JetBot containers
Run build.sh to build the JetBot Docker containers from scratch.
- Enable all JetBot containers
Enable the Docker service to run at boot and use the specified directory as the working directory for enable.sh.
- Follow the Software Setup
Complete the JetBot software setup by following the SD card installation instructions in the Software Setup page.
- Download the object detection model
Download the object detection model. The latest engine targets Jetson Nano running JetPack 4.3, so use the JetPack 4.3 JetBot SD card image or the JetPack 4.3 base SD card image.
- Connect to the robot interface
Navigate to http://<jetbot_ip_address>:8888 in your browser to connect to the robot.
- Shut down running notebooks
Select Kernel -> Shutdown All Kernels... to shut down all other running notebooks.
- Open the object following folder
Navigate to the ~/Notebooks/object_following/ folder.
- Upload the pretrained detection model
Upload the pretrained ssd_mobilenet_v2_coco.engine model to this folder. The collision avoidance model is assumed to already be at ~/Notebooks/collision_avoidance/best_model.pth.
Calibrate and test
- First boot without connected peripherals
The document recommends first booting without the piOLED or motor driver connected, so you can verify that the system boots properly from the SD card image.
- Check WiFi setup help
If connecting to WiFi is difficult, check the Wi-Fi setup page for more detailed instructions.
- Configure the Jetson system
Run configure_jetson.sh to set the Jetson power mode and other parameters.
- Set Jetson to 7W Mode
With this power source, set Jetson Orin Nano to 7W mode; shutdown in 15W mode was observed when motors suddenly ramped from 0% to 100% or stalled.
Data and training
- Train a Collision Avoidance Model
Collect data and train a custom Image Classification model so JetBot can avoid collisions in varied scenarios.
- Run the live object following demo
Open and follow the live_demo.ipynb notebook. JetBot will follow common objects such as Person, Cup, and Dog while running the collision avoidance model from Example 3.
Use it
- Follow the JetBot Examples
Work through the JetBot examples to learn robotics and AI basics and prepare to explore your own applications.
- Program Basic Motion
Program JetBot from your web browser using basic motor controls.
- Teleoperate With a Gamepad
Attach a gamepad to your laptop to stream live images from JetBot and control it.
- Follow Objects With a Pretrained Model
Use a pre-trained object detection model to make JetBot follow an object such as a person, cup, or dog.
- Modify Examples and Create Projects
Modify the examples, make JetBot do something new, or create an entirely new AI project.
- Share Your JetBot Project
Share what you create with JetBot on the NVIDIA developer forum.
- Open JupyterLab for programming
From a browser on the local network, access port 8888 on the JetBot and log in with the password jetbot. The directory specified for enable.sh is mounted at /workspace in the container as the JupyterLab root; work outside that directory is lost when the container shuts down.
- Start programming in the browser
After JetBot boots, enter its IP address in a web browser and start programming.
Common problems
- Motor control fails to initialize or stop due to I2C, interface, or package errors #45 #97 #231 #473
- Prebuilt SD images and Docker containers fail to install, build, or boot #21 #385 #390 #406 #446
- Camera is undetected, fails to initialize, or stops updating #29 #35 #63 #306
- Object detection model initialization fails, including TensorRT version errors #126 #166
- Road-following and collision-avoidance logic cannot be combined reliably #267
- Wi-Fi fails after updates and computers cannot connect to the robot #50 #79
- Following and obstacle response is slow, with straight driving drifting left #68 #182 #198 #233
- Recorded images still yield zero samples, causing model training to fail #193
Parts
- Jetson Nano Developer Kit 4GB (472 GFLOPS, 128-core Maxwell GPU, quad-core Cortex‑A57, 5–10 W); the docs now recommend the Jetson Orin Nano 8GB for new builds
- 2 × TT gear motors (Adafruit 3777) with an Adafruit DC Motor + Stepper FeatherWing driver, 60 or 65 mm wheels and a 1-inch ball caster
- IMX219 wide-angle CSI camera (three options)
- Intel AC8265 M.2 card with two U.FL antennas, or a TP‑Link Archer T2U USB dongle
- 10,000 mAh power bank (two 5V/3A outputs); the Orin version uses a 20 W PD power bank and recommends 7 W mode to avoid shutdowns when the motors surge
- Adafruit PiOLED 128×32 display with a right-angle header, 64 GB microSD, M2/M3 screws and jumper wires
- 3D-printed chassis, camera mount, caster base and shroud (10–12 hours of printing, under 100 g of PLA, about $2)
Tools
- Long-nose pliers, wire strippers, Phillips and M2 hex screwdrivers, tweezers
- Soldering iron and solder (motor driver terminals and headers, PiOLED header)
- A 3D printer or a printing service (the docs tested the Monoprice Voxel and Ultimaker 2+)
Skills
- Basic soldering and wiring
- Flashing the SD card image with Etcher, basic Linux and Docker
- Python and Jupyter; collecting data and training image classification and regression networks (on the Jetson or a GPU workstation)
Software
- JetBot 0.4.3 SD card image (JetPack 4.5, Nano 4GB and 2GB), or Docker containers (Jupyter, camera, display)
- jetbot Python library and Jupyter examples: basic motion, teleoperation, collision avoidance, road following, object following (SSD MobileNet V2 COCO TensorRT engine)
- Third-party kits use their own images and tutorials (Waveshare adds ROS node code, for example)