Development of an Autonomous Agent based on Reinforcement Learning for a Digital Fighting Game
Joao Ribeiro Bezerra, Luís F. W. Góes, Alysson Ribeiro da Silva · 2020
In this work, an autonomous agent based on reinforcement learning is implemented in a digital fighting game. The implemented agent uses Fusion Architecture for Learning, COgnition, and Navigation (FALCON) and Associative Resonance Map (ARAM) neural networks. The experimental results show that the autonomous agent is able to develop game strategies using the experience acquired in the matches, and achieves a winning rate of up to 90% against an agent with fixed behavior.