Racing is a video game genre that is still popular today. Its development processes cannot be separated from the need to have Non-Player Character (NPC) in them. NPCs act as the opponents for the players, and thus the developers are always challenged with the problem of how to make the NPCs smarter than them. One of the problems is related with breaking decision, specifically when the NPCs decided to slow down their speed during races by using brakes. One commonly used method for this type of experiment is the Brake Zone. Although, this method also has its own shortcomings, such as the devs have to manually place the zone themselves in the designated locations for the brake test. Other solution that can be applied is Smart AI System by Racing Game Starter Kit (RGSK), but this also has its problem in which to get the best result, a proper configuration is needed. To resolve the problem, researcher proposes the method of machine learning, Naive Bayes for the braking decision. Naive Bayes use three features for the data input, and two output class in which the data will be obtained from the player. The test result showed that the braking decision from Naive Bayes was able to prevent the vehicle from crashing with the outer wall without dropping the game's FPS (Frames per Second). Time acquisition each lap from Naive Bayes was able to keep up with the player's time at an average of 52,5 seconds during 10 laps.
Copyrights © 2019