Learning a hierarchy

Researchers have developed a hierarchical reinforcement learning algorithm that learns high-level actions to efficiently solve complex tasks requiring thousands of timesteps. The algorithm was successfully applied to navigation problems, where it discovered high-level actions for walking and crawling in different directions, enabling rapid mastery of new navigation tasks.

• →A new hierarchical reinforcement learning algorithm can learn high-level actions useful across multiple tasks.
• →The algorithm enables fast solving of complex tasks that require thousands of timesteps.
• →When applied to navigation problems, it automatically discovers walking and crawling behaviors in different directions.
• →The learned high-level actions allow agents to quickly master new navigation tasks.
• →This represents progress in creating AI systems that can efficiently learn and transfer skills across related problems.