Feedforward Insulin Control via Semi-Coupled Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic insulin delivery systems are inadequate in controlling blood glucose levels due to the complexity of disturbances from factors like meals, activity, and stress, as they fail to accurately determine the insulin infusion rate to eliminate the effects of modeled disturbances.
Innovation Solution
A semi-coupled modeling network using a feedforward control method that incorporates input dynamics, unmeasured pseudo-blood insulin, and blood glucose dynamics, along with sensors monitoring variables such as body position, movement, and galvanic skin response, to proactively adjust insulin infusion rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If feedforward control is used to proactively cancel disturbance effects on blood glucose, then control effectiveness is improved, but model accuracy is insufficient to determine accurate insulin infusion rates
Solution Approach 1:
The system segments the complex disturbance modeling into multiple independent components: carbohydrate intake modeling, insulin sensitivity modeling, and activity/stress modeling. Each component is modeled separately with its own parameters and dynamics, allowing accurate representation of each disturbance source while maintaining overall model tractability for feedforward control calculation.
Solution Approach 2:
The system transitions from traditional single-dimensional glucose response modeling to multi-dimensional modeling by incorporating parallel pathways: carbohydrate absorption dynamics, insulin pharmacokinetics/pharmacodynamics, and activity-induced glucose utilization. This dimensional expansion captures the complex inter-relationships of multiple disturbances simultaneously, enabling accurate feedforward insulin rate determination.
2Measurement precision
If multiple disturbances (meals, activity, stress) are modeled to improve glucose control, then control accuracy is improved, but system complexity increases
Solution Approach 1:
The system divides complex disturbance modeling into separable modules: meal/carbohydrate modeling, insulin sensitivity modeling, and activity modeling. Each module processes specific input variables independently, reducing the computational burden of modeling all disturbances simultaneously while maintaining comprehensive coverage of glucose-affecting factors.
Solution Approach 2:
The system utilizes readily available sensor data (continuous glucose monitoring, activity sensors) and user inputs (meal logs) to automatically generate disturbance models without requiring additional complex measurements or manual intervention. The model self-adjusts parameters based on observed glucose responses to various disturbances, reducing system complexity while improving accuracy.
Data Source
AI summary
A system and method for automatically administering insulin based on a feedforward control. The automatic insulin delivery system may include a computing device comprising a processor, machine readable non-transitory media which stores the coupled-model of the invention, and a monitoring system. The machine readable transitory media may be configured to receive one or more inputs and the coupled-model may utilize a feedforward control to parameterize the inputs and generate an output that can be translated to a type and amount of insulin to be administered by the system. The input(s) received by the machine readable non-transitory media may be created by a user or received from one or more sensors. The method may include providing a model that is stored on a machine readable non-transitory media and administering insulin based on the output provided by the model.