In 2018, when SparkFun was still hosting its annual Autonomous Vehicle Competition (AVC), Tawn Kramer wanted to compete with his friends but was unable to make it. Since he wanted to be a part of the action, he asked his friends Will Roscoe and Adam Conway (who were competing with the Donkey Car) to take lots of pictures of everything they could, from the building facades and tent structures to the hay bales and track obstacles. Tawn had already been working on a simulator for the AVC, but these extra steps made it even closer to the real thing. Using Blender, he rendered everything he saw in the images his friends supplied and even got the line details of the parking lot spot on.
Here's what the course looked like!
After sending his friends the video above, they were impressed! Tawn says they all tried their hand at making an autonomous racer that could conquer the track. "It was very challenging!" he said. "I finally had some luck after using some reinforcement learning and allowing it to run for a couple of days."
The simulator project is open source and you can find it here on Tawn's Github.
It's easiest to download pre-build binaries from here.
Also, the original Donkey Car project is still developing. You can check it out here.
If you fire it up and get it running, let us know how it goes! Shoot us a tweet @sparkfun, or let us know on Instagram, Facebook or LinkedIn.
The current maintainers, Maxime Ellerbach and Jack Silberman, deserve lots of credit, too!