Building a 1/16 Scale Racecar with Jetson Nano

Introduction

After looking at the RACECAR project from MIT and UPenn, I have always wanted to build something more compact and cheaper than the RACECAR itself. Arrival of Jetson Nano has made it possible to build one. I will be writing a series of blog posts on building this RACECAR. This post will be updated regularly. This racecar will be used as a project to implement various computer vision and deep learning techniques going ahead.

Bill of materials

Traxxas 1/16 slash 4x4 is being used as the chassis for the RACECAR. The car comes with the battery included. I upgraded slash with shock springs in both front and back to support the weight that will be added to the car. I am planning to use the ESC that came with the slash. The bill of materials is without sensors is costing around $420.

The above circuit is the one used to operate Servo and ESC using Arduino. And Arduino will be installed as an ROS node to Jetson Nano. Arduino will receive 5V power over USB and Servo will receive power from ESC header which will be powered from the car battery.

I have replaced the board of jetson nano with leopard imaging connector board for Jetson Nano. This will enable the project to use 4 cameras using CSI2 connectors which will be pursued later in the project. This project focuses on building on entire camera pipeline from scratch leveraging all the hardware accelerators available on Jetson Nano. Future scope also includes a bigger variant with more compute and more hardware accelerators. Current plan is to use multiple Jetson Nano in parallel to create a computer cluster.

Experimentation includes multiple sensors including Camera(RGB and TOF), Lidar, ultrasonics and IMU. To construct the body to install sensors, I have used Acrylic sheets. Cutting them is pretty easy. Design of the sheet is up to individual. Mine is inspired from MIT Racecar. I will introduce the path I followed in building the chassis and platform in coming posts.