Application of Deep Reinforcement Learning to Event-Triggered Control for Networked Artificial Pancreas Systems

2026-04-30 · Source: stat.ML updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Health & Medical Research · Depth: Expert, extended

Summary

This paper introduces a deep reinforcement learning (DRL)-based event-triggered controller, called CGM-ETPPO, for networked artificial pancreas (AP) systems. The goal is to maintain blood glucose levels within the 70-180 mg/dL target range while significantly reducing communication frequency and energy consumption. Unlike existing DRL methods that assume periodic updates or jointly learn insulin dosing and update timing, CGM-ETPPO decouples these tasks by using a rule-based event-triggering mechanism based on changes in continuous glucose monitor (CGM) values. This approach formulates the problem as a semi-Markov decision process (SMDP), extending a standard DRL algorithm to handle irregular decision intervals. Numerical experiments using the UVA/Padova T1D simulator with 10 virtual adult patients demonstrate that CGM-ETPPO achieves over 90% Action Update Reduction Rate (AURR) and comparable Time in Range (TIR) to well-tuned PID controllers, outperforming hierarchical DRL methods in training success.

Key takeaway

For AI Scientists and Machine Learning Engineers developing control systems for medical devices, adopting a decoupled, rule-based event-triggered DRL approach can significantly enhance energy efficiency without compromising control performance. Your teams should consider formulating problems with irregular decision intervals as Semi-Markov Decision Processes and extending standard DRL algorithms to manage complexity, especially when deploying on resource-constrained wearable devices like artificial pancreas systems.

Key insights

A DRL-based event-triggered controller for artificial pancreas systems improves energy efficiency by decoupling insulin dosing from update timing.

Principles

• Rule-based triggering simplifies DRL for networked control.
• Semi-Markov Decision Processes handle irregular decision intervals.
• Reduced communication frequency improves energy efficiency.

Method

The CGM-ETPPO algorithm extends PPO for SMDPs, using a rule-based event-triggering mechanism based on CGM value changes to determine insulin infusion rate updates.

In practice

• Implement rule-based event triggers to simplify DRL tasks.
• Use SMDP formulations for systems with variable action durations.
• Prioritize communication efficiency in wearable medical devices.

Topics

• Deep Reinforcement Learning
• Event-Triggered Control
• Artificial Pancreas Systems
• Networked Control Systems
• Semi-Markov Decision Process

Code references

• jxx123/simglucose

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Deep Reinforcement Learning for Flexible Job Shop Scheduling with Random Job Arrivals — DRL with PPO and MLPs effectively solves FJSP with random arrivals by learning dispatching rule selection.
• Timescale Separation Enables Deep Reinforcement Learning Control of Rotating Detonation Engine Mode Transitions — Reformulating DRL in a moving reference frame enables effective control of multi-timescale systems like RDEs.
• Discovering Lattice Reduction Strategies via Self-Play — Deep reinforcement learning can discover superior, generalizable lattice reduction strategies beyond traditional algorithms like LLL.
• DiPRL: Learning Discrete Programmatic Policies via Architecture Entropy Regularization — DiPRL learns discrete programmatic policies directly, avoiding performance drops from post-hoc continuous-to-discrete conversion.
• Why the Next Leap in AI May Depend on Learning by Trial and Error — Deep reinforcement learning offers AI a crucial mechanism for learning through action and discovery, moving beyond mere pattern recognition.
• Safe Deep Reinforcement Learning for Building Heating Control and Demand-side Flexibility — A real-time adaptive safety filter enhances DRL for building heating, ensuring demand-side flexibility compliance and efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by stat.ML updates on arXiv.org.