Synergistic Integration of Q-Learning and Genetic Algorithm for Enhanced Gameplay Performance in the Snake Game

Reinforcement learning (RL) presents a powerful framework for training intelligent agents in dynamic environments, with applications ranging from robotics to gaming. This study explores the integration of two prominent techniques, Q-learning and Genetic Algorithm (GA), in the context of training an agent to play the classic Snake game. The study investigates the individual performance of standalone Q-learning and GA methodologies, as well as their combined approach termed the “convergent approach.” Through comparative analysis, including statistical assessments and tabular representations of gameplay performance, the study highlights the strengths and limitations of each approach. The convergent approach demonstrates an average performance gain of approximately 28.70% over Q-Learning and 34.95% over the Genetic Algorithm in terms of the time taken to achieve score milestones in the Snake game.

I. Introduction

Artificial Intelligence (AI) continues to redefine the boundaries of computational systems. Within this context, the amalgamation of neural networks, Q-learning, and genetic algorithms stands as a compelling avenue for research and innovation. This study delves into the interplay of these computational paradigms within the familiar domain of the classic Snake game, exploring their potential applications and implications.