Feedback Predictive Controller for Insulin Delivery Time Delay
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic insulin delivery systems are inadequate in controlling blood glucose levels due to complexities arising from disturbances like meals, activities, and stress, as they rely on inaccurate predictive models and lack the capability to address unmeasured disturbances effectively.
Innovation Solution
A feedback predictive controller (FBPC) system that uses a predictive feedback error to minimize deviations in blood glucose concentration by employing a Wiener/Semi-Coupled method and pre-whitening to model serial correlation structures, allowing for accurate future predictions and reducing the need for cause-and-effect models between insulin flow rate and glucose levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If model predictive control (MPC) is used to address time delay in manipulated variable, then future prediction capability is provided, but prediction accuracy decreases as prediction horizon increases
Solution Approach 1:
The patent applies preliminary action by using pre-whitening filtering on the controlled variable measurements before feeding them to the predictor. This preprocessing step removes serial correlation structures from the data, allowing the predictor to achieve better accuracy without extending the prediction horizon. The pre-processed measurements are then used to predict future values at the minimal horizon θ, maintaining high accuracy while providing the necessary future prediction capability.
2Device complexity
If classical feedback control is used, then system simplicity is maintained, but inability to address unmeasured disturbances and complex correlations reduces control effectiveness
Solution Approach 1:
The patent implements feedback by using a predictor that continuously updates future predictions of the controlled variable based on pre-processed measurements and past prediction errors. The predictor incorporates feedback from actual measurements to correct prediction deviations, allowing the system to adapt to unmeasured disturbances and complex correlations while maintaining a relatively simple control structure.
Solution Approach 2:
The patent introduces an intermediary element - the pre-whitening filter - that processes measurements before they enter the predictor. This intermediary removes serial correlation structures and prepares the data for more accurate prediction, enabling the system to handle complex disturbances without requiring a complex control algorithm.
3Reliability
If cause-and-effect models between insulin flow rate and glucose levels are used, then physiological accuracy is improved, but model complexity and difficulty of accurate modeling increase
Solution Approach 1:
The patent uses copying by creating a predictive model that replicates the behavior of the glucose regulation system without requiring a detailed physiological cause-and-effect model. The predictor copies the essential dynamics by using pre-processed measurements and statistical relationships, achieving sufficient accuracy for control purposes while avoiding the complexity of detailed physiological modeling.
Data Source
AI summary
This invention relates to a feedback predictive controller, systems comprising and methods employing the same. Preferably the feedback predictive controller and/or systems comprising the feedback predictive controller are part of an automatic insulin delivery system. The methods described herein can be used to control blood glucose concentration in a patient with diabetes. Preferably, the insulin delivery system is an artificial pancreas.