Hierarchical Reinforcement Learning (HRL)

A layered reinforcement learning technique where higher-level agents define goals for lower-level agents, allowing complex tasks to be broken down into manageable parts. AI learning is structured across multiple levels of abstraction or decision-making. Useful in complex multi-agent tasks. For example, in robotics, a top-level agent might set “”clean the room,”” while lower agents handle “”move arm”” or “”avoid obstacle.””

• Speeds up learning by reusing sub-policies
• Supports multi-agent coordination and abstraction
• Powerful in planning, robotics, and autonomous systems