Electric Mini Tesla navigates semi-autonomously with Logitech webcams and machine learning

Home > Industry Analysis > Content

【Summary】Austin Blake has created a semi-autonomous Mini Tesla using a Radio Flyer Tesla Model S toy car. The self-driving go-kart is equipped with three Logitech webcams and machine learning software to navigate. After adding two more webcams and fixing a bug, the go-kart successfully stays on track. The steering input is controlled by a salvaged steering servo motor, while the high-speed motor has been replaced with a brushless electric motor.

FutureCar Staff Sep 11, 2023 11:24 PM PT

Electric Mini Tesla navigates semi-autonomously with Logitech webcams and machine learning

Building safe, self-driving cars is no easy task, unless you're Elon Musk. However, if you're up for the challenge, starting small is a good way to begin. Austin Blake took on this challenge by creating an autonomous Mini Tesla.

Austin built the self-driving go-kart, called the Teskart, using a Radio Flyer Tesla Model S electric toy car. He equipped it with three Logitech C920 webcams - one on the right, one in the center, and one on the left - to feed the machine learning software. The model was trained using 48,000 image training sets of the walking path it was tested on.

This semi-autonomous Teskart initially had some close calls when it veered off the designated track with just one webcam. Austin then decided to add two more webcams and fixed a bug. This improvement allowed the go-kart, with the Radio Flyer Tesla frame, to navigate the makeshift track without any issues.

For the steering input, Austin created a bolt-on frame to hold the steering servo motor salvaged from an electric wheelchair in place. As for the high-speed motor capable of reaching speeds up to 6 mph, he replaced it with the new Motenergy ME0907 brushless electric motor. This upgrade allows the tiny car to reach speeds of 45mph with 48V power. Austin plans to increase the top speed even further with a 72V version in the near future.

The existing motor can produce 80 continuous amps and a peak of 220 amps for one minute, spinning the rear axle loaded on a chain. It draws power from a DIY battery pack made from 224 individual 18650 lithium-ion cells, separated into 16 modules with their own battery management system.

However, it's important to note that this DIY project is not a fully self-driving build. Currently, only the steering angle is controlled by the system, while the throttle control is still manual. Perhaps Austin intends to use the data from the behavioral cloning to add another Arduino-controlled component for throttle control as well. Only then can we truly call it a self-driving mini Tesla!