Research on Autonomous Mobile Robot Navigation Technology Based on Deep Reinforcement Learning

Abstract

The development of autonomous mobile robots (AMRs) is crucial for advancing automation across various sectors, including industrial, logistics, and service industries. These robots have the potential to revolutionize how tasks are performed, offering increased efficiency and reduced human intervention. However, one of the primary challenges in this field is achieving efficient and reliable navigation in complex and dynamic environments. Traditional navigation techniques, which often rely on predefined paths and static maps, fall short in such settings. This limitation necessitates the adoption of more sophisticated approaches that can adapt to real-time changes and uncertainties in the environment. Deep Reinforcement Learning (DRL), particularly algorithms like Deep Q-Learning and Proximal Policy Optimization (PPO), has emerged as a promising solution to these challenges. This review explores recent advancements in DRL-based navigation technologies, highlighting key methodologies, simulation results, practical applications, and future research directions. By analyzing various studies, this paper demonstrates how DRL can significantly enhance AMR navigation capabilities, offering marked improvements in path planning, obstacle avoidance, and overall adaptability to dynamic environments. These advancements suggest a promising future for AMR deployment in increasingly complex settings.