Closed-Loop Insulin Infusion Control System Based on Reinforcement Learning and IoT: Adaptive Glucose Management Algorithm for Type 1 Diabetes Patients

View at publisher Summarise with VIRAT
Authors: Rong Zhang
Cited by: 0
References: 12

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Type 1 diabetes treatment still poses difficulties because it has complex and time-varying glucose-insulin dynamics. The proposed study aims to provide a closed-loop insulin infusion system using the combination of deep reinforcement learning and Internet-of-Things architecture for adapting glucose levels. The co-designed Proximal Policy Optimization algorithm for basal and bolus insulin infusion, using highly customized Actor networks, has also been proposed. The proposed system has been validated using the FDA-approved UVA/Padova Type 1 Diabetes Simulator for the analysis on 100 virtual participants. The obtained results clearly showed that the average time in range is 78.3 ± 6.2% for glucose levels, along with level 2 hypoglycemia < 0.3%, performing better than PID controllers, model predictive controllers, and conventional therapy. The proposed system’s edge-computing framework allows real-time processing for inference in less than 20 milliseconds, using less resources. The proposed system has notable potential for enhancing glucose levels without sacrificing safety in closed-loop infusion systems for administering insulin.

Fields of Study & Subjects

Diabetes Management and Research Diabetes Management and Education Diabetes Treatment and Management

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
1569–1578
Publication date
28 January 2026
Language
EN