Systems and methods for hybrid automata mining from input-output traces of cyber-physical systems

Patent metadata

Field	Value
Publication number	US-11054807-B2
Application number	US-201916413018-A
Country	US
Kind code	B2
Filing date	May 15, 2019
Priority date	May 15, 2018
Publication date	Jul 6, 2021
Grant date	Jul 6, 2021

Abstract

Systems and methods for mining hybrid automata from input-output traces of cyber-physical systems are disclosed herein.

First claim

What is claimed is:

1. A system for inferring and fine tuning control parameters of a cyber-physical system, comprising:
a physical system including an actuator and one or more sensors that produces an input signal and an output signal, a relationship between the input signal and the output signal being representative of physical dynamics of the physical system; and
a controller that generates a hybrid automaton based on continuous variables measured from the one or more sensors of the physical system and computes a next controller output for the actuator of the physical system based on the hybrid automaton, wherein to generate the hybrid automaton, the controller
accesses as inputs the continuous variables including time series input and output traces of the physical system,
segments the time series input and output traces into a plurality of segments according to a sudden change in a slope of the continuous variables,
infers a controller strategy for each segment of the plurality of segments such that the controller clusters each segment into one of a plurality of equivalence classes based on their respective controller strategies, each equivalence class corresponding to a controller mode of a plurality of controller modes,
derives one or more polynomial flow equations representative of physical dynamics of the physical system for each mode, and
derives one or more guard conditions representative of a boundary that causes a controller mode change for each mode.

2. The system of claim 1 , wherein the hybrid automaton defines an output of the controller and includes the plurality of control modes, the one or more polynomial flow equations, and the one or more guard conditions the cyber-physical system.

3. The system of claim 2 , wherein the controller strategy includes a reset condition and one or more jump conditions.

4. The system of claim 1 , wherein the controller is adapted to:
compare a set of collected controller input-output traces associated with the physical system with a set of inferred controller input-output traces to validate safety and optimal performance of the physical system and the hybrid automaton,
wherein the set of inferred controller input-output traces are generated by the hybrid automaton.

5. The system of claim 4 , wherein the root mean square error is taken between the set of collected controller input-output traces and a set of inferred controller input-output traces, wherein an amount of input-output traces taken as input by the system within a predetermined period of time is iteratively increased until the root mean square error is close to zero.

6. The system of claim 1 , wherein each of a set of timestamps associated with the plurality of segments is representative of a potential controller mode change, wherein each one of the set of timestamps corresponds with a peak in the rate of change of observed controller inputs or observed controller outputs defined by the continuous variables.

7. The system of claim 6 , wherein the controller is further adapted to:
associate a linear equation with each one of the plurality of segments, wherein the linear equation maps the set of observed controller outputs to the set of observed controller inputs.

8. The system of claim 1 , wherein the physical system is an artificial pancreas such that the actuator is an infusion pump and the continuous variables include one or more glucose-meter values, the hybrid automaton generated by the controller being implemented by the controller to access the one or more glucose-meter values and to output by the controller an amount of insulin infusion rate for the infusion pump.

9. The system of claim 8 , wherein hybrid automaton implemented by the controller linearizes a plurality of non-linear differential equations associated with blood glucose levels and insulin levels of a specific patient such that the amount of insulin infusion rate for the infusion pump is computed for the specific patient.

10. A system for optimizing control parameters of a cyber-physical system, comprising:
a physical system including an actuator and one or more sensors that produces one or more input signals and one or more output signals, a relationship between the one or more input signals and the one or more output signals being representative of physical dynamics of the physical system; and
a controller that generates a hybrid automaton based on continuous variables measured from the one or more sensors of the physical system and computes a next controller output for the actuator of the physical system based on the hybrid automaton,
wherein to generate the hybrid automaton, the controller:
identifies a mode transition of the physical system by identifying a sudden change in a derivative of the one or more output signals, wherein a time interval between a mode transition is a segment of a plurality of segments,
identifies jump conditions and reset conditions relating the one or more output signals to the one or more input signals for each segment of the plurality of segments,
clusters the plurality of segments into one or more discrete modes according to their respective jump conditions and reset conditions,
determines one or more polynomial flow equations for each discrete mode, each polynomial flow equation relating the one or more input signals and the one or more output signals and being indicative of dynamics of the physical system, and
derives one or more guard conditions representative of a boundary that causes a controller mode change for each discrete mode based on the one or more input signals and the one or more output signals,

wherein the hybrid automaton includes the one or more discrete modes, the one or more polynomial flow equations and the one or more guard conditions.

11. The system of claim 10 , wherein reset conditions are representative of an initialization of each new segment and wherein reset conditions include a linear relationship between the input signal and the output signal for each segment.

12. The system of claim 10 , wherein jump conditions include values for one or more input signals and the one or more output signals associated with two neighboring segments, wherein jump conditions compare a mode transition between the two neighboring segments.

13. The system of claim 10 , wherein polynomial flow equations are determined for each mode by determining relationships between the one or more input signals and the one or more output signals associated with each mode based on an unbiased estimator.

14. The system of claim 10 , wherein guard expressions are determined for each mode transition by determining relationships between controller output values at each mode transition.

15. The system of claim 14 , wherein to determine the guard expressions for a mode transition, the controller:
expresses each of a plurality of continuous variables associated with the one or more input signals or the one or more output signals as a linear combination of each other continuous variable of a plurality of continuous variables associated with the one or more input signals or the one or more output signals; and
determines a coefficient matrix for the linear combination using Fisher Information,
wherein the linear combination expression is expressed as a conditional inequality.

16. A method for inferring and fine tuning control parameters of a cyber-physical system, comprising:
accessing, by a controller, continuous variables defining input and output traces from a cyber-physical control system, the cyber-physical system including a physical system and a control system for controlling

Assignees

• Gupta Sandeep K S
• Banerjee Ayan
• Lamrani Imane
• Univ Arizona State

Inventors

• Gupta Sandeep K S
• Banerjee Ayan
• Lamrani Imane

Classifications

• G06F7/552

  Powers or roots {, e.g. Pythagorean sums} · CPC title

• G05B2219/35261

  Use of mathematical expression, functional equation · CPC title

• G05B19/4155 Primary

  characterised by program execution, i.e. part program or machine function execution, e.g. selection of a program · CPC title

• G06F17/12

  Simultaneous equations {, e.g. systems of linear equations} · CPC title

• G06N5/04 Primary

  Inference or reasoning models · CPC title