Simulating Human Behavior in Fighting Games using Reinforcement Learning and Artificial Neural Networks

The study of intelligent agent training is of great interest to the gaming industry due to its wide application in various game genres and its capabilities of simulating a human-like behavior. In this work two machine learning techniques, namely, a reinforcement learning approach and an Artificial Neural Network (ANN), are used in a fighting game in order to allow the agent/fighter to emulate a human player. We propose a special reward function for the reinforcement learning approach that is capable of integrating specific human-like behaviors to the agent. The ANN is trained with several recorded battles of a human player. The proposed methods are compared to other two reinforcement learning methods presented in the literature. Furthermore, we present a detailed discussion of the empirical evaluations performed, regarding the training process and the main characteristics of each method used. The results obtained in the experiments indicated that the proposed methods have a good performance against human players and are also more enjoyable to play against when compared to the other existing methods.