Looks like the beginning of a hilarious series of videos, [MrL314] put us in a a quick but in-depth tour of how the AI works in Mario Kart. (Video, embedded below.) Don’t play Mario Kart much anymore? Well, take a look anyway as some very simple tricks to help Bowser get past Princess Peach without meeting her can be useful in any pre-programmed navigational situation.
Quick spoilers. The CPU players move through different regions, each with their desired speed, and the vector direction field changes the direction they should point to. Only when they run off course do they actually calculate headings for their target. This preset of the desired direction and speed greatly reduces the direct computation required.
Then you throw the other players into the mix and a very simple distance-dependent turn algorithm gets you a clean pass. However, this effect is manually tuned for the specific racetrack, because you don’t want Luigi to drive on the thin sections of Rainbow Road. Because More technical detailsyou can check [MrL314]notes of.
If anything, this video gives us a greater appreciation for clever little tricks that create seemingly complex interactions from incredibly simple rules. Remember Mario Kart when you were programming in the next multi-gigabyte neural network model, okay?