Autonomous insulin delivery algorithms must prevent severe hypoglycemia during unannounced meals and strenuous exercise. High-fidelity in silico glucose simulators provide an essential safety sandbox.
Simulating subcutaneous insulin absorption, hepatic glucose production, and muscle uptake allows automated testing of control algorithms across 24-hour feeding cycles.
Testing the artificial pancreas against skipped meals and delayed carbohydrate spikes proved that rate-limiting safety clamps prevent dangerous insulin stacking.