O S Sumukh , Prinson Fernandes , Merin Meleet
Department of Information Science and Engineering ,, RV College of Engineering, Bangalore, Karnataka
One of the most popular subjects being investigated in AI nowadays is game learning. Addressing such issues require proper domain specific knowledge. So one such game was developed ie., flappy bird where the agent learns itself on how to avoid the obstacles and also tries to maximize the score based on the rewards and punishments it receives. No prior knowledge was given to the agent regarding the environment. Instead of utilizing raw pixels, the agent was trained using domain-specific features such as the birds speed, the distance between pipes, and the height of the pipes, which significantly simplifies the feature space and avoids the need for deeper models to automatically extract underlying data. The agent was trained using the Neuro Evolution of Augmenting Topologies (NEAT) algorithm and is talked about in this paper.
Keywords: NEAT, feature extraction, pygame, reward,fitness, reinforcement learning
Journal Name :
EPRA International Journal of Multidisciplinary Research (IJMR)

Published on : 2022-07-28

Vol : 8
Issue : 7
Month : July
Year : 2022
Copyright © 2024 EPRA JOURNALS. All rights reserved
Developed by Peace Soft