Analysis of Reinforcement Learning Developments in Games and Warehouse Robots: A Systematic Literature Review
TL;DR
The review indicates that ReinforcementLearning has evolved from classical Q-Learning algorithms into Deep Reinforcement Learningcapable of solving high-dimensional decision problems using deep neural networks.
Abstract
Artificial Intelligence (AI) has significantly evolved through advances in machine learningtechniques capable of producing adaptive intelligent systems. One of the most prominentparadigms is Reinforcement Learning (RL), which enables intelligent agents to learn optimaldecision-making policies through repeated interactions with their environments. The success ofAlphaGo in defeating a professional Go player in 2016 demonstrated that RL can solve extremelycomplex problems with enormous search spaces. However, studies discussing RL applications indigital games and autonomous warehouse robots remain fragmented across multiple researchdomains. This study presents a Systematic Literature Review (SLR) on the development ofReinforcement Learning, covering its theoretical foundations, mathematical formulation, majoralgorithms, Deep Reinforcement Learning, game-based applications, and warehouse roboticswithin the context of Industry 4.0 and Industry 5.0. Relevant scientific publications wereidentified, evaluated, and synthesized systematically. The review indicates that ReinforcementLearning has evolved from classical Q-Learning algorithms into Deep Reinforcement Learningcapable of solving high-dimensional decision problems using deep neural networks. Furthermore,RL has demonstrated significant potential in warehouse automation by optimizing routingstrategies, reducing energy consumption, minimizing congestion, and improving operationalproductivity. Nevertheless, high computational costs, large-scale training requirements, safetyconsiderations, and ethical issues remain important challenges for real-world deployment.