Deep Q-Learning for Intelligent Non-Playable Characters in Combat Games

Abstract

In computer games, especially video games, non-playable characters (NPCs) are those characters who the player does not control but interacts with on a regular basis. They are often made utilizing algorithms that cause the NPCs to act dependent on the player’s activities. A non-playable character’s design does not necessarily need to be the result of actual artificial intelligence. To make intelligent agents function better, they may be created utilizing learning approaches. The method of constructing Mob characters, NPC characters meant to oppose players using the neural network-based model-free learning methodology known as Deep Q-learning for a game centered on hand-to-hand combat (fighting game genre), is discussed in this article.

Keywords: Intelligent agent, Close combat games, Mobs, NPC, Q-learning, Reinforcement learning