Artificial Pancreas MMPC for Variable Glucose Dynamics Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current artificial pancreas systems lack effective control methods to manage abrupt and slow variations in glucose dynamics, relying on limited control algorithms that struggle with non-adaptive and slow-reacting responses, and often do not incorporate a comprehensive model of insulin dynamics in the human body.
Innovation Solution
A closed-loop system utilizing a Multi-Model Predictive Controller (MMPC) that integrates multiple state vectors and models with a model predictive control algorithm, propagating and filtering state vectors using a Kalman filter to determine optimal insulin delivery, including basal and meal bolus calculations, to maintain physiological glucose levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional control algorithms are used in artificial pancreas systems, then the system structure is simpler, but the system cannot effectively handle abrupt and slow variations in glucose dynamics
Solution Approach 1:
The patent implements a dynamic model selection mechanism that adapts the insulin kinetics model based on real-time glucose data. The system transitions between different model states (fast and slow insulin kinetics) using a forgetting factor and model selection criteria, allowing the controller to dynamically adjust to varying glucose dynamics without requiring a completely complex restructured system.
Solution Approach 2:
The system changes parameters (insulin sensitivity factor, insulin kinetics rate) based on the selected model state. When fast kinetics is detected, the system uses corresponding parameters for that state, and when slow kinetics is detected, it switches to the appropriate parameters. This parameter adaptation allows the system to handle varying glucose dynamics effectively.
2Speed
If a single fixed insulin model is used, then the control algorithm is simpler, but the system responds slowly to abrupt changes in glucose dynamics
Solution Approach 1:
The patent employs a dual-state dynamic model where the system can switch between fast insulin kinetics and slow insulin kinetics states. The model selection is based on real-time evaluation of glucose data using a forgetting factor that weights recent observations more heavily. This dynamic approach enables rapid response to abrupt glucose changes while maintaining simplicity through structured model states.
Solution Approach 2:
The insulin model is segmented into distinct kinetic states (fast and slow) rather than using a single continuous model. Each state has its own parameters and characteristics. This segmentation allows the system to quickly adapt to different glucose dynamics scenarios by selecting the appropriate segment (model state) without requiring a completely complex unified model.
3Adaptability or versatility
If non-adaptive control algorithms are used, then the system is more stable, but the system cannot handle variations in glucose dynamics
Solution Approach 1:
The system implements feedback through continuous monitoring of glucose data and using a forgetting factor to weight recent observations. The model selection and parameter updates are driven by feedback from actual glucose measurements, allowing the system to adapt to metabolic changes while maintaining stability through controlled adaptation rates and bounded parameter changes.
Solution Approach 2:
The patent introduces controlled dynamics into the control algorithm through time-varying model parameters and state transitions. The forgetting factor provides a mechanism for gradual adaptation, and the model selection process ensures that changes are based on actual glucose dynamics rather than random fluctuations. This dynamic approach maintains reliability by using structured adaptation mechanisms.
Data Source
Figure 1
Figure 1A
Figure 2
AI summary
The present disclosure relates to systems and methods for controlling physiological glucose concentrations in a patient using a closed loop artificial pancreas. The systems and methods may utilize a controller with control logic operative to execute a multi-model predictive controller algorithm to determine a medication dose to the patient.