COOL-MC: Verifying and Explaining RL Policies for Platelet Inventory Management

Researchers developed COOL-MC, a tool that combines reinforcement learning with model checking to verify and explain AI policies for platelet inventory management in blood banks. The system achieved a 2.9% stockout probability while providing transparent decision-making explanations for safety-critical healthcare applications.

• →COOL-MC successfully verified an RL policy for platelet inventory management with 2.9% stockout and 1.1% wastage probabilities.
• →The AI policy primarily focuses on age distribution of inventory rather than other factors like day of week or pending orders.
• →Action reachability analysis revealed the policy uses diverse replenishment strategies with most order quantities reached quickly.
• →Counterfactual analysis showed that replacing medium-large orders with smaller ones had minimal impact on safety probabilities.
• →This represents the first formal verification and explanation of an RL policy for healthcare supply chain management.