Moving-horizon state-initializer for control applications

• US 9,984,773 B2
• Filed: 02/04/2017
• Issued: 05/29/2018
• Est. Priority Date: 08/06/2014
• Status: Active Grant

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Gondhalekar, Ravi L., Doyle, III, Francis J., Dassau, Eyal
• Primary Examiner(s): Chad, Aniss
• Application Number:
  US15/424,824
• Publication Number:
  US 20170147781A1
• Time in Patent Office: 479 Days
• CPC Class Codes: A61M 2005/14208 with a programmable infusion control system, characterised by the infusion program
• CPC Class Codes: A61M 2205/18 with alarm
• CPC Class Codes: A61M 2205/50 with microprocessors or computers
• CPC Class Codes: A61M 2230/201 Glucose concentration
• CPC Class Codes: A61M 5/1723 using feedback of body parameters, e.g. blood-sugar, pressure measurement of body parameters A61B5/00
• CPC Class Codes: G16H 20/17 delivered via infusion or injection
• CPC Class Codes: G16H 40/63 for local operation
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

A state-estimator for the estimation or initialization of the state of a discrete-time state-space dynamical model based on sensor measurements of the model output, comprising fitting a continuous-time function to acquired sensor measurement data-points of each model output, and subsequently sampling the continuous time function at exactly the sample-period of the state-space dynamic model for which the state is being estimated or initialized, in order to construct a model state via a synthesized output trajectory.

Claims

1. A method of state-estimator for the estimation or initialization of the state of a discrete-time state-space glucose dynamical model based on sensor measurements of output of the model, comprising;

(a) using moving-horizon optimization to fit a continuous-time function to acquired sensor measurement data-points of each model output,

(b) generating a fitted function,

(c) subsequently sampling the continuous time function at exactly the sample-period of the state-space dynamic model for which the state is being estimated or initialized, in order to construct a model state via a synthesized output trajectory, and

(d) outputting a state estimate of the glucose dynamical model;

wherein

sensor re-calibrations are included, by permitting the fitted function to be discontinuous in its value, but not its derivatives, at the point of re-calibration, and wherein the magnitude of the discontinuity is identified by the optimization, and wherein the sampling of the fitted function is performed ignoring the re-calibration discontinuity;

the function fit employs sensor measurement time-stamps and call-time of the state-estimator, wherein;

(i) delays between the sensor and state-estimator are at least partially mitigated by sampling the fitted function backwards in time starting at exactly the estimator call time; and

(ii) sensor data collected at irregular time-intervals, or time-intervals that are not the sample-period of the model for which the state is being estimated or initialized, can be accommodated;

the model for which the state is being estimated or initialized may have inputs, and if inputs are present historical input data is employed in construction of the state;

the state of the model is observable or reconstructible;

when the model for which the state is being estimated or initialized has one single output then the current model state is constructed to reflect the synthesized output trajectory without offset, and when the model for which the state is being estimated or initialized has multiple outputs then a trade-off strategy is employed to reconcile mismatching outputs,

wherein the method is adapted for medical device control of drug delivery.

2. The method of claim 1 further comprising:

constructing a state for a model by first fitting a function through acquired continuous glucose monitoring (CGM) data-points, then sampling the function at exactly the sample-period of the model for which the state is being estimated, wherein;

(i) CGM sensor re-calibrations are included, by permitting the fitted function to be discontinuous;

(ii) delays between the sensor and controller are at least partially mitigated;

(iii) data collected at irregular time-intervals, or a time-interval that is not the sample-period of the model for which the state is being estimated, can be accommodated; and

(iv) the model is not employed during or within the estimation process, rather the state is estimated for the model, wherein any plant-model mismatch effect on the final result is limited.

3. The method of claim 1 wherein the sensor measurement data is pretreated with a Kalman filter or other signal processing technique.

4. The method of claim 1 deployed in automated drug delivery in biomedical applications, such as for type 1 diabetes, or process control implementations.

5. The method of claim 1 wherein the sensor is a continuous glucose monitor (CGM) and the discrete-time state-space dynamical model is employed for predicting blood glucose concentrations.

6. The method of claim 1 operatively combined with a state-feedback control law to perform algorithmic insulin dosing.

7. The method of claim 1 operatively combined with a state-feedback control law that is a state-feedback model predictive control (MPC) law, to perform algorithmic insulin dosing.

8. The method of claim 1 operatively combined with a state-feedback control system to deliver insulin.

9. The method of claim 1 operatively combined with a state-feedback control system that is a state-feedback model predictive control (MPC) system, to deliver insulin.

10. The method of claim 1 operatively combined with an alarm and/or notification system, wherein the alarm and/or notification decision-making algorithm employs predictions, performed by a discrete-time state-space dynamical model that uses the state of a model in its alarm and/or notification decision-making algorithm.

11. The method of claim 1 operatively combined with an alarm and/or notification system, wherein the alarm and/or notification decision-making algorithm employs predictions, that are blood glucose predictions, performed by a discrete-time state-space dynamical model that uses the state of a model in its alarm and/or notification decision-making algorithm.

12. The method of claim 1 adapted for medical device control of drug delivery for a chronic medical condition or treatment that is type 1 diabetes mellitus.

13. The method of claim 1 operatively combined with:

(i) a Kalman filter, or other signal processing technique to pre-treat continuous monitoring data.

14. The method of claim 1 operatively combined with:

(ii) a model predictive control (MPC) control algorithm to deliver a drug.

15. The method of claim 1 operatively combined with:

(iii) an alarm system for open loop continuous monitoring.

16. The method of claim 1 operatively combined with:

(i) a Kalman filter, or other signal processing technique to pre-treat continuous monitoring data;

(ii) a model predictive control (MPC) control algorithm to deliver a drug that is insulin; and

(iii) an alarm system for open loop continuous glucose monitoring (CGM).

17. A controller programmed to implement a method of claim 1, and optionally further comprising a status monitoring system, a drug pump or metering system, and/or a drug to be delivered.

18. The method of claim 1 wherein the method further comprises synthesizing the synthesized output trajectory as follows:

(a) applying a continuous prototype function of time “

f(time,theta)”

to be employed in a data-fitting procedure to collected sensor measurements, wherein the continuous prototype function is parameterized by a set of parameters “

theta”

, and wherein the values within the parameter set “

theta”

are determined during a subsequent data-fitting procedure, wherein the continuous prototype function “

f(time,theta)”

incorporates a discontinuity in the prototype function at historical time-instants at which a sensor re-calibration occurred, wherein the size of the discontinuity is a parameter “

delta”

, resulting in a prototype function “

f(time,theta,delta)”

, and the value of the size parameter “

delta”

is determined during the subsequent data-fitting procedure;

(b) performing a data-fitting procedure of the prototype function “

f(time,theta,delta)”

to simultaneously characterize both the optimal function parameter set “

theta” and

optimal discontinuity size parameter “

delta”

;

(c) obtaining a new, continuous function of time “

F(time)”

by considering the prototype function “

f(time,theta,delta)”

with optimal parameter set “

theta”

determined above, but setting to zero the discontinuity size parameter “

delta”

, i.e., employing “

delta=0”

, where segments preceding any historical time-instant a sensor re-calibration occurred are shifted by an amount equaling the size of the discontinuity parameter identified during the first data-fitting operation, wherein the resulting function definition is continuous and equals the fitted function at time-instants following the most recent sensor re-calibration;

(d) sampling the function definition “

F(time)”

obtained during the preceding step at the current time (“

F(now)”

), and at integer multiples of the model'"'"'s sample-period “

tau”

backwards in time from the current time (“

F(now-tau)”

, “

F(now-2tau)”

, “

F(now-3tau)”

, . . . ), yielding a sequence of synthetic, synchronous, historical outputs that contain no remnants, or artifacts, of the sensor recalibration; and

(e) treating the synthesized sequence of synchronous outputs generated in the preceding step as an actual, historical measurement sequence, and constructing the model state based on this sequence, in conjunction with the historical model input trajectory if the model has inputs, thereby achieving a state-estimate or state-initialization.

19. A method of state-estimator for the estimation or initialization of the state of a discrete-time state-space glucose dynamical model based on sensor measurements of output of the model, the method comprising:

(a) using moving-horizon optimization to fit a continuous-time function to continuous glucose monitoring (CGM) data,

(b) generating a fitted function, and

(c) outputting a state estimate of the glucose dynamical model, wherein;

sensor recalibrations are accommodated by including discontinuity in the glucose output value, but not its derivatives, within the function'"'"'s definition, and wherein the magnitude of the discontinuity is identified by the optimization;

data fitting employs the CGM time-stamps and controller call time, thus asynchronous data sampling is handled naturally;

after optimization the fitted function is sampled at the controller model'"'"'s sample-period T, ignoring the recalibration discontinuity, to synthesize an output trajectory; and

in combination with historical input data, and assuming observability, the current model state is constructed to reflect the fitted output trajectory without offset,

wherein the method is adapted for medical device control of drug delivery.

20. The method of claim 19 wherein the method further comprises synthesizing the synthesized output trajectory as follows:

(a) applying a continuous prototype function of time “

f(time,theta)”

to be employed in a data-fitting procedure to collected sensor measurements, wherein the continuous prototype function is parameterized by a set of parameters “

theta”

, and wherein the values within the parameter set “

theta”

are determined during a subsequent data-fitting procedure, wherein the continuous prototype function “

f(time,theta)”

incorporates a discontinuity in the prototype function at historical time-instants at which a sensor re-calibration occurred, wherein the size of the discontinuity is a parameter “

delta”

, resulting in a prototype function “

f(time,theta,delta)”

, and the value of the size parameter “

delta”

is determined during the subsequent data-fitting procedure;

(b) performing a data-fitting procedure of the prototype function “

f(time,theta,delta)”

to simultaneously characterize both the optimal function parameter set “

theta” and

optimal discontinuity size parameter “

delta”

;

(c) obtaining a new, continuous function of time “

F(time)”

by considering the prototype function “

f(time,theta,delta)”

with optimal parameter set “

theta”

determined above, but setting to zero the discontinuity size parameter “

delta”

, i.e., employing “

delta=0”

, where segments preceding any historical time-instant a sensor re-calibration occurred are shifted by an amount equaling the size of the discontinuity parameter identified during the first data-fitting operation, wherein the resulting function definition is continuous and equals the fitted function at time-instants following the most recent sensor re-calibration;

(d) sampling the function definition “

F(time)”

obtained during the preceding step at the current time (“

F(now)”

), and at integer multiples of the model'"'"'s sample-period “

tau”

backwards in time from the current time (“

F(now-tau)”

, “

F(now-2tau)”

, “

F(now-3tau)”

, . . . ), yielding a sequence of synthetic, synchronous, historical outputs that contain no remnants, or artifacts, of the sensor recalibration; and

(e) treating the synthesized sequence of synchronous outputs generated in the preceding step as an actual, historical measurement sequence, and constructing the model state based on this sequence, in conjunction with the historical model input trajectory if the model has inputs, thereby achieving a state-estimate or state-initialization.