Model predictive control for diabetes management
Disclosed herein are techniques related to model predictive control. The techniques may involve generating a desired glucose trajectory that approaches a desired steady state setpoint from a current glucose value over a prediction horizon. The techniques may involve generating a plurality of insulin delivery patterns. Each insulin delivery pattern may correspond to an amount of insulin to be delivered over a control horizon. The techniques may involve generating a plurality of predicted glucose trajectories over the control horizon. Each predicted glucose may be generated based on the current glucose value and a respective insulin delivery pattern. The techniques may involve comparing the desired glucose trajectory against each predicted glucose trajectory and selecting a predicted glucose trajectory that is more similar to the desired glucose trajectory than any other predicted glucose trajectory. The techniques may involve determining an insulin delivery pattern used to generate the selected predicted glucose trajectory.
1. A processor-implemented method comprising:
obtaining a current glucose value for a user of an insulin delivery device;
generating a desired glucose trajectory that approaches a desired steady state setpoint from the current glucose value over a prediction horizon;
generating an insulin delivery pattern corresponding to an amount of insulin to be delivered to the user over a control horizon;
generating a predicted glucose trajectory over the control horizon, the predicted glucose trajectory being generated based on the current glucose value and the insulin delivery pattern;
comparing the desired glucose trajectory against the predicted glucose trajectory;
responsive to a difference between the desired glucose trajectory and the predicted glucose trajectory being no larger than a predetermined threshold, selecting the predicted glucose trajectory, wherein the difference is determined based on a cost function;
determining the insulin delivery pattern used to generate the predicted glucose trajectory; and
causing insulin delivery by the insulin delivery device in accordance with the determined insulin delivery pattern.
2. The processor-implemented method of claim 1 , wherein generating the predicted glucose trajectory comprises using a metabolic model to replicate insulin and glucose kinetics and dynamics of the user.
3. The processor-implemented method of claim 1 , wherein the cost function comprises an error function describing a difference between the desired glucose trajectory and the predicted glucose trajectory.
4. The processor-implemented method of claim 3 , wherein the cost function further comprises a penalty function that provides a check against sudden changes in insulin delivery.
5. The processor-implemented method of claim 3 , wherein the cost function further comprises a penalty function that provides a check against large increases in a total amount of insulin delivered.
6. The processor-implemented method of claim 1 , wherein the current glucose value is obtained from a glucose sensor system configured to measure glucose values of the user at periodic intervals.
7. The processor-implemented method of claim 6 , wherein the method is repeated at regular intervals corresponding to the periodic intervals.
8. A system comprising:
one or more processors; and
one or more processor-readable storage media storing instructions which, when executed by the one or more processors, cause performance of a method comprising:
obtaining a current glucose value for a user of an insulin delivery device;
generating a desired glucose trajectory that approaches a desired steady state setpoint from the current glucose value over a prediction horizon;
generating an insulin delivery pattern corresponding to an amount of insulin to be delivered to the user over a control horizon;
generating a predicted glucose trajectory over the control horizon, the predicted glucose trajectory being generated based on the current glucose value and the insulin delivery pattern;
comparing the desired glucose trajectory against the predicted glucose trajectory;
responsive to a difference between the desired glucose trajectory and the predicted glucose trajectory being no larger than a predetermined threshold, selecting the predicted glucose trajectory, wherein the difference is determined based on a cost function;
determining the insulin delivery pattern used to generate the predicted glucose trajectory; and
causing insulin delivery by the insulin delivery device in accordance with the determined insulin delivery pattern.
9. The system of claim 8 , wherein generating the predicted glucose trajectory comprises using a metabolic model to replicate insulin and glucose kinetics and dynamics of the user.
10. The system of claim 8 , wherein the cost function comprises an error function describing a difference between the desired glucose trajectory and the predicted glucose trajectory.
11. The system of claim 10 , wherein the cost function further comprises a penalty function that provides a check against sudden changes in insulin delivery.
12. The system of claim 10 , wherein the cost function further comprises a penalty function that provides a check against large increases in a total amount of insulin delivered.
13. The system of claim 8 , wherein the current glucose value is obtained from a glucose sensor system configured to measure glucose values of the user at periodic intervals.
14. The system of claim 13 , wherein the method is repeated at regular intervals corresponding to the periodic intervals.
15. One or more non-transitory processor-readable storage media storing instructions which, when executed by one or more processors, cause performance of a method comprising:
obtaining a current glucose value for a user of an insulin delivery device;
generating a desired glucose trajectory that approaches a desired steady state setpoint from the current glucose value over a prediction horizon;
generating an insulin delivery pattern corresponding to an amount of insulin to be delivered to the user over a control horizon;
generating a predicted glucose trajectory over the control horizon, the predicted glucose trajectory being generated based on the current glucose value and the insulin delivery pattern;
comparing the desired glucose trajectory against the predicted glucose trajectory;
responsive to a difference between the desired glucose trajectory and the predicted glucose trajectory being no larger than a predetermined threshold, selecting the predicted glucose trajectory, wherein the difference is determined based on a cost function;
determining the insulin delivery pattern used to generate the predicted glucose trajectory; and
causing insulin delivery by the insulin delivery device in accordance with the determined insulin delivery pattern.
16. The one or more non-transitory processor-readable storage media of claim 15 , wherein generating the predicted glucose trajectory comprises using a metabolic model to replicate insulin and glucose kinetics and dynamics of the user.
17. The one or more non-transitory processor-readable storage media of claim 15 , wherein the cost function comprises an error function describing a difference between the desired glucose trajectory and the predicted glucose trajectory.
18. The one or more non-transitory processor-readable storage media of claim 17 , wherein the cost function further comprises a penalty function that provides a check against sudden changes in insulin delivery.
19. The one or more non-transitory processor-readable storage media of claim 17 , wherein the cost function further comprises a penalty function that provides a check against large increases in a total amount of insulin delivered.
20. The one or more non-transitory processor-readable storage media of claim 15 , wherein the current glucose value is obtained from a glucose sensor system configured to measure glucose values of the user at periodic intervals, and wherein the method is repeated at regular intervals corresponding to the periodic intervals.