This looks super fun, but there's a bit of a backlog in my build pipeline to explore this just yet . The project described in the video doesn't look like it's had an update for 2 years. I fear the reality of dealing with the LIDAR and navigation could be a decent challenge.

If you follow the bouncing ball on the NVIDIA series, you end up at a pretty comprehensive set of instructions for both hardware (based on a TT02) and software with some working examples:
https://github.com/NVIDIA-AI-IOT/jetracer

There are probably lots of other avenues to explore. The best performing autonomous setups use multiple input types - computational power and complexity rise pretty quickly.