Building a JetBot AI, Part 1 — Hardware

I recently discovered a new-found passion for DIY projects around IoT devices and set out to build an AI-enabled robot — Nvidia’s JetBot AI. In this first part of the series, I will share my experiences from building the JetBot. In Part 2, I will cover the software side of the robot.

JetBot is powered by Nvidia’s Jetson Nano — a Single Board Computer (SBC), similar to a RaspberryPi, however with one important advantage over it — it sports a powerful 128-core GPU onboard. Nvidia advertises the chip as an easy to use platform capable of embedded IoT applications like home robots, image classification or speech recognition.

Jetson Nano enjoys an active community contributing open-source project ideas based on the Nano and JetBot is its flagship project, including STL files for 3D printing its frame, a detailed Bill of Materials outlining the necessary hardware components, all needed software packages and a set of engaging example programs showcasing Jetbot’s capabilities.

There are two ways you can build your JetBot — either by printing the body elements and purchasing the hardware components individually or selecting one of the kits from an ever-growing list of partners. This being my first project of this kind, I went for an easy option and purchased a kit from SparkFun.

The kit comes in two versions — a Jetson-included ($250) package or sans Jetson, in case you already have your own ($150). The SparkFun kit contains all the components you will need, like a battery pack to power the Jetson on the move, camera, wifi module to all the body elements, and wheels. The SSD card is pre-loaded with the OS that will boot up the JetBot and also contains software examples.

SparkFun provides detailed step-by-step instructions on building the robot and it’s mostly a breeze. It roughly takes 3 hours to complete the assembly at an easy pace.

Most of the steps require just a Philips screwdriver to assembly.

The acrylic plates were a bit tight to combine and I ended up cracking one of them… Luckily it was not a show stopper.

The Serial Controlled Motor Driver (SCMD) assembly requires the soldering of a few jumper contacts and some PTH headers.

Included breadboard allows the SCMD to be connected with a Micro USB breakout.

The SCMD module eventually gets connected to the DC motors.

The kit comes with a 64x48 OLED breakout which we will find helpful later on.

Finally, we mount the Jetson Nano board on top of the body.

And voila! The JetBot is complete!

In the next part of the series, we will look at the Software driving the JetBot and play with the examples provided by Nvidia.