Game artificial intelligence model for non-playable characters

Abstract

Artificial intelligence in video games plays a crucial role in shaping gameplay, and its creation has always been a complex and challenging task. Developing effective game artificial intelligence requires understanding and knowledge of many aspects, including algorithms, rules, and strategies that a created artificial intelligence will follow. Applying machine learning methods to create an artificial intelligence model opens up many opportunities to accelerate and optimize game development processes and improve the behavior of non-playable characters, and consequently, the overall player experience. One type of machine learning is reinforcement learning, which involves receiving rewards for correct behavior or penalties for mistakes. In this process, a non-playable character gains experience interacting with the external environment. The aim of the work is to explore the possibilities of using the reinforcement learning method to develop a game artificial intelligence model for a non-playable character capable of self-learning and pursuing other game characters. Training non-playable characters is a key task for creating exciting and realistic game environments. The primary goal of NPC training was to ensure the ability to effectively pursue a moving target. The experimental results confirmed that the developed model successfully learns and makes optimal decisions in a game environment. The study proved the effectiveness of using machine learning methods in creating game artificial intelligence. In particular, reinforcement learning opens up new horizons for developing їхadaptive and unpredictable scenarios in games. Promising areas for further research include integrating machine learning methods to provide non-playable characters with additional capabilities that will contribute to a deeper player immersion in the gameplay.