Advancements in reinforcement learning for robotic control: A comprehensive review

Reinforcement Learning (RL) has emerged as a transformative paradigm for training autonomous agents to achieve effective control through trial-and-error interactions. In robotics, RL has enabled the development of adaptive and intelligent control strategies that operate reliably in dynamic and uncertain environments. This review focuses exclusively on advanced RL approaches applied to robotic control over the past five years, offering a comprehensive synthesis of state-of-the-art innovations. The analysis spans key domains including robot manipulators, mobile robots, aerial robots, marine robots, and human–robot interaction, with particular emphasis on critical control tasks and their role in enabling robust robotic autonomy. By integrating methodological advances, unresolved challenges, and emerging trends, this review equips researchers and practitioners with deeper insights into the potential and limitations of RL for robotic control. To complement this analysis, we introduce an interactive web application that facilitates hands-on exploration of RL control concepts, enhancing accessibility and knowledge transfer.