BLAZ3D was initially planned to run on devices such as the Nokia N95, Nokia N82 and later on iPhone 2g. So, considering the specs of the phone we were quite restricted with the hardware. Now we needed to find a sturdy solution that could give us intelligent AI that acted smart, and yet won’t eat a lot of processing. Actually considering the amount of load that bullet physics was sucking on the device, we had allocated near zero space for AI.
Now the AI was supposed to be able to pickup powerups, use shortcut ramps, and avoid obstacles. With these goals in mind we started designing the AI, but we quickly realized that we needed to keep AI constantly competitive too, as the levels had a large variety in their difficulty so the AI would quickly become too easy to defeat or at times too hard.
So the solution we came for the near zero CPU intensive AI was to use simple splines. But not using only the traditional best path splines, but instead we distributed splines. Through each level we manually placed small regions of splines, connecting options for the AI.
For example a straight line to go through a straight piece of tunnel would keep going straight until we reached powerups. After which 3 splines would cut out of the main spline, and go through the powerups, and finally merging into one. We had some complex scenarios as well when we extended splines, to give the AI logic to turn sharp through the next turn and catch the near powerup, or to turn loose and catch the powerup near the far wall of the tunnel.
Each line had an identifier in its name and a list of children in the name to tell the AI options it had to choose the next spline. But this still meant the AI was pretty dumb, so to give it a bit more intelligence we simply added one more variable to each line, for the probability of using that line. Then we assigned each AI its own probability on load. So if there were 3 AI in a race, there would be one AI how would always be tough to defeat (much like the red enemy in pacman <sad to see that wordpress spellcheck tries to correct pacman>).
So by this simple variable of probability in each spline, we managed to suddenly bring life in each character with this equation:[highlight type=”one”]finalProbability = nextLinesProbability + enemyProbability + Random value between 0 – 0.25[/highlight]
We looped through all the lines in the child list with this equation and if one of the finalProbability goes over 1, we go ahead with that spline. This simple system allowed us to give each AI a dynamic feel with just a few additional operations.