Distilling Deep Reinforcement Learning into Interpretable Fuzzy Rules: An Explainable AI Framework

Researchers developed a Hierarchical Takagi-Sugeno-Kang Fuzzy Classifier System that converts opaque deep reinforcement learning agents into human-readable IF-THEN rules, achieving 81.48% fidelity in tests. The framework addresses the critical explainability problem in AI systems used for safety-critical applications by providing interpretable rules that humans can verify and understand.

• →New explainable AI framework converts deep reinforcement learning policies into interpretable fuzzy rules using clustering and regression techniques.
• →The system achieved 81.48% fidelity and outperformed decision trees by 21 percentage points in continuous control tasks.
• →Three new metrics were introduced to quantify explanation quality: FRAD, FSC, and ASG for comprehensive interpretability assessment.
• →Generated rules like 'IF lander drifting left at high altitude THEN apply upward thrust with rightward correction' enable human verification.
• →The framework addresses a critical barrier to deploying AI in safety-critical domains by making black-box systems transparent.