Title: Using Deep Reinforcement Learning for Autonomous Sorting in Robotic Systems
Authors: Haya Albhaisi, Samy S. Abu-Naser
Volume: 10
Issue: 7
Pages: 30-36
Publication Date: 2026/07/28
Abstract:
The increasing adoption of industrial automation has accelerated the demand for intelligent robotic systems capable of performing complex tasks with minimal human intervention. Autonomous sorting is a critical robotic application in manufacturing, logistics, warehousing, and recycling industries, where robots must identify, classify, and organize objects in dynamic and uncertain environments. Traditional rule-based robotic systems often exhibit limited adaptability when confronted with changing object characteristics or environmental conditions. Deep Reinforcement Learning (DRL), which integrates reinforcement learning with deep neural networks, has emerged as a powerful approach for enabling robots to learn optimal decision-making policies through continuous interaction with their environment. This paper presents a comprehensive review of the application of DRL in autonomous robotic sorting systems. The study examines key DRL algorithms, including Deep Q-Network (DQN), Deep Deterministic Policy Gradient (DDPG), Proximal Policy Optimization (PPO), and Actor-Critic methods, and evaluates their effectiveness in robotic manipulation, grasping, and sorting tasks. A comparative analysis of recent research is conducted to assess algorithm performance, adaptability, computational requirements, and applicability to real-world industrial environments. The review also discusses current challenges, including sample inefficiency, computational complexity, safety concerns, and the simulation-to-reality gap. The findings indicate that DRL-based robotic sorting systems outperform traditional rule-based approaches in terms of adaptability, autonomous decision-making, and operational flexibility. Furthermore, recent advances in simulation environments, transfer learning, and policy optimization have improved the feasibility of deploying DRL solutions in industrial applications. The paper concludes by identifying future research directions aimed at enhancing training efficiency, real-world reliability, and large-scale deployment of intelligent robotic sorting systems.