Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Precise Insulin Delivery for Artificial Pancreas: A Reinforcement Learning Optimized Adaptive Fuzzy Control Approach

Omar Mameche, Abdelhadi Abedou, Taqwa Mezaache, Mohamed Tadjine

TL;DR

This paper employs a reinforcement learning agent tasked with adjusting 27 parameters of the Takagi-Sugeno fuzzy controller at each time step, ensuring real-time adaptability and significantly enhances the robustness of the controller against variations in meal size and timing, while also stabilizing glucose levels with minimal exogenous insulin.

Abstract

This paper explores the application of reinforcement learning to optimize the parameters of a Type-1 Takagi-Sugeno fuzzy controller, designed to operate as an artificial pancreas for Type 1 diabetes. The primary challenge in diabetes management is the dynamic nature of blood glucose levels, which are influenced by several factors such as meal intake and timing. Traditional controllers often struggle to adapt to these changes, leading to suboptimal insulin administration. To address this issue, we employ a reinforcement learning agent tasked with adjusting 27 parameters of the Takagi-Sugeno fuzzy controller at each time step, ensuring real-time adaptability. The study's findings demonstrate that this approach significantly enhances the robustness of the controller against variations in meal size and timing, while also stabilizing glucose levels with minimal exogenous insulin. This adaptive method holds promise for improving the quality of life and health outcomes for individuals with Type 1 diabetes by providing a more responsive and precise management tool. Simulation results are given to highlight the effectiveness of the proposed approach.

Precise Insulin Delivery for Artificial Pancreas: A Reinforcement Learning Optimized Adaptive Fuzzy Control Approach

TL;DR

This paper employs a reinforcement learning agent tasked with adjusting 27 parameters of the Takagi-Sugeno fuzzy controller at each time step, ensuring real-time adaptability and significantly enhances the robustness of the controller against variations in meal size and timing, while also stabilizing glucose levels with minimal exogenous insulin.

Abstract

This paper explores the application of reinforcement learning to optimize the parameters of a Type-1 Takagi-Sugeno fuzzy controller, designed to operate as an artificial pancreas for Type 1 diabetes. The primary challenge in diabetes management is the dynamic nature of blood glucose levels, which are influenced by several factors such as meal intake and timing. Traditional controllers often struggle to adapt to these changes, leading to suboptimal insulin administration. To address this issue, we employ a reinforcement learning agent tasked with adjusting 27 parameters of the Takagi-Sugeno fuzzy controller at each time step, ensuring real-time adaptability. The study's findings demonstrate that this approach significantly enhances the robustness of the controller against variations in meal size and timing, while also stabilizing glucose levels with minimal exogenous insulin. This adaptive method holds promise for improving the quality of life and health outcomes for individuals with Type 1 diabetes by providing a more responsive and precise management tool. Simulation results are given to highlight the effectiveness of the proposed approach.

Paper Structure

This paper contains 30 sections, 19 equations, 13 figures.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Fuzzy Logic Control
  4. Fuzzification
  5. Fuzzy Inference / Fuzzy Rule base
  6. Defuzzification
  7. Reinforcement Learning
  8. Value Functions and Bellman Equations
  9. Reinforcement Learning Algorithms
  10. Twin Delayed Deep Deterministic Policy Gradient
  11. Model and Control Strategy
  12. Model Used for Simulation
  13. Glucose Subsystem
  14. Non-Insulin-Dependent Glucose Flux and Renal Clearance
  15. Gut Absorption Rate
  16. ...and 15 more sections

Figures (13)

  • Figure 1: Glucose level (mg/dL) response of the Direct insulin control through Reinforcement Learning, along with corresponding meals intake (mg/dl) and insulin injections in units in the nominal case.
  • Figure 2: Glucose level (mg/dL) response of the Direct insulin control through Reinforcement Learning, along with corresponding meals intake (mg/dL) and insulin injections in units in the nominal case.
  • Figure 3: Glucose level (mg/dL) response of the Non-adaptive optimized fuzzy control, along with corresponding meals intake (mg/dL) and insulin injections in units in the nominal case.
  • Figure 4: Glucose level (mg/dL) response of the Non-adaptive optimized fuzzy control, along with corresponding meals intake (mg/dL) and insulin injections in units in the nominal case.
  • Figure 5: Glucose level (mg/dL) response of the adaptive FLC along with corresponding meals intake (mg/dL) and insulin injections in units in the nominal case.
  • ...and 8 more figures