AutoQD: Automatic Discovery of Diverse Behaviors with Quality-Diversity Optimization

Researchers present AutoQD, a new AI method that automatically discovers diverse behavioral policies without requiring hand-crafted descriptors. The approach uses mathematical embeddings of policy occupancy measures to enable Quality-Diversity optimization algorithms to find varied high-performing solutions in reinforcement learning tasks.

• →AutoQD eliminates the need for manually designed behavioral descriptors in Quality-Diversity optimization algorithms.
• →The method uses random Fourier features to approximate Maximum Mean Discrepancy between policy occupancy measures for automatic behavior discovery.
• →Theoretical guarantees prove that embeddings converge to true behavioral distances as sample size increases.
• →Experiments demonstrate successful diverse policy discovery across multiple continuous control tasks.
• →The approach enables open-ended learning without requiring domain-specific knowledge or predefined diversity metrics.