Online Reinforcement Learning for Decision-Making and Explicit Policy Learning in Autonomous Agents: A Systematic Literature Review

Online reinforcement learning (RL) enables autonomous agents to adapt their policies in real time, responding to dynamic and uncertain environments. This systematic literature review (SLR) examines high-impact studies from 2004 to 2026 on online RL methods for decision-making and explicit policy learning in autonomous agents, excluding approaches that do not adapt during execution. We analyze algorithmic strategies for handling non-stationarity, exploration-exploitation trade-offs, and multi-agent interactions. The review highlights emerging trends, practical implementations, and remaining challenges, offering actionable insights for designing adaptive and efficient online RL controllers in robotics, UAVs, and intelligent systems.