How I visualize what the AI does
🕑 Added 2025-12-03 14:00:11 +0000 UTCIn the videos, I often show hundreds of AI runs overlaid on top of each other. These ingame clips look nice and make the AI’s behaviour easy to visualize…
but they also take a while to produce. And since I’m often training the AI several times a week, with tens of thousands of runs in every session, those polished visuals aren’t very practical.
Instead, during training I regularly save some trajectories into simple files, containing things like the car’s position and rotation over time. Then, with a few small scripts, I can turn that into lightweight 2D graphs and quickly check what the AI did.
Here are a few examples I made with matplotlib (in Python) over the last days to examine a problem I’ve been stuck on recently. Good luck guessing what they represent :)
Comments
Yosh
I never heard of QBasic, I’m a bit younger, and in 1999-2000 the only language I was using was French, and just as a beginner :D About Python: it’s often considered one of the easiest languages to start with, and it has a huge community. It’s also extremely popular in machine learning / reinforcement learning, most libraries/projects/courses etc. in this field are built around it. That said, I don't have much experience with other languages either (aside from some R I used in the past for statistics), so I'm probably not the best person to ask
QotSA Fan
How hard is it to program in Python? I know when I did programming (in High School) I programmed in QBasic! Keep in mind that was 1999-2000. Me being 42 years old now and still a gamer it's funny to see how lines of code have evolved over time. All i know it looks like that one looks like the AI is taking multiple lines to a curve make a kink in the road or a small chicane.