The popularity of the racing game genre is still visible today. Factors that supporting the popularity of this genre are speed driving which provides an exciting experience, interesting track variations, stunning graphics, and unique challenges presented by artificial intelligence. An important factor to be developed and in line with the core of the racing game that provides a fun challenge is artificial intelligence. Artificial intelligence behavior that is not varied and easy to guess, or even playing badly will affect the fun of the challenge of racing games. To avoid this, artificial intelligence is needed which is able to learn mindset and imitate human decisions when playing, especially braking decision or gas and brake determination. That is the basis of the Neural Network algorithm implemented for Artificial Intelligence in the Racing Game Starter Kit. The complexity of the code on the machine is simplified by changing the decision process to some intelligent neural networks that are similar to human neurons, especially how it works. Coupled with the adaptation process in a dynamic environment makes this algorithm interesting for AI researchers. By utilizing Cross-Validation, learning this algorithm with human behavior has a similarity rate of 76 percent. In a 10-round trial, the time results showed 12% or 72 seconds faster than the kit's AI, and a stable frame rate with an average of 59 frames per second.
Copyrights © 2019