# Integrating Neural Differential Forecasting with Safe Reinforcement Learning for Blood Glucose Regulation

arXiv: https://arxiv.org/abs/2511.12417 · HTML: https://mlsanity.com/paper/2511.12417/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in RL for clinical treatment and surgery, one of 15 directions in Deep reinforcement learning methods.
- Builds on 6 of its 25 references that are in the map.
- First posted on arXiv 16 Nov 2025.

## Placed in

- Deep reinforcement learning methods / [RL for clinical treatment and surgery](https://mlsanity.com/deeprl/rl-for-clinical-treatment-and-surgery/)

## Builds on (most cited first, 6 of 6)

- [Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems](https://mlsanity.com/paper/2005.01643/) (2020)
- [Design and Validation of an Open-Source Closed-Loop Testbed for Artificial Pancreas Systems](https://mlsanity.com/paper/2208.06479/) (2022)
- [Data-driven Design of Context-aware Monitors for Hazard Prediction in Artificial Pancreas Systems](https://mlsanity.com/paper/2104.02545/) (2021)
- [Towards Developing Safety Assurance Cases for Learning-Enabled Medical Cyber-Physical Systems](https://mlsanity.com/paper/2211.15413/) (2022)
- [Short: Basal-Adjust: Trend Prediction Alerts and Adjusted Basal Rates for Hyperglycemia Prevention](https://mlsanity.com/paper/2303.09913/) (2023)
- [KnowSafe: Combined Knowledge and Data Driven Hazard Mitigation in Artificial Pancreas Systems](https://mlsanity.com/paper/2311.07460/) (2023)

## Abstract (arXiv)

Automated insulin delivery for Type 1 Diabetes must balance glucose control and safety under uncertain meals and physiological variability. While reinforcement learning (RL) enables adaptive personalization, existing approaches struggle to simultaneously guarantee safety, leaving a gap in achieving both personalized and risk-aware glucose control, such as overdosing before meals or stacking corrections. To bridge this gap, we propose TSODE, a safety-aware controller that integrates Thompson Sampling RL with a Neural Ordinary Differential Equation (NeuralODE) forecaster to address this challenge. Specifically, the NeuralODE predicts short-term glucose trajectories conditioned on proposed insulin doses, while a conformal calibration layer quantifies predictive uncertainty to reject or scale risky actions. In the FDA-approved UVa/Padova simulator (adult cohort), TSODE achieved 87.9% time-in-range with less than 10% time below 70 mg/dL, outperforming relevant baselines. These results demonstrate that integrating adaptive RL with calibrated NeuralODE forecasting enables interpretable, safe, and robust glucose regulation.