Method and apparatus for supplying interpolation point data for a data-based function model calculation unit

Patent metadata

Field	Value
Publication number	US-9805313-B2
Application number	US-201414326110-A
Country	US
Kind code	B2
Filing date	Jul 8, 2014
Priority date	Jul 9, 2013
Publication date	Oct 31, 2017
Grant date	Oct 31, 2017

Abstract

A method for identifying a set of interpolation point data points from training data for a sparse Gaussian process model, encompassing the following tasks: successively selecting training data points from the set of training data for acceptance into or exclusion from a set of interpolation point data points in accordance with a selection criterion; and terminating selection when a termination criterion exists; the selection criterion depending on a divergence between a target value of the selected training data point and a function value, at the selected training data point, of the Gaussian process model based on the respectively current set of interpolation point data points.

First claim

What is claimed is:

1. A method for identifying a set of interpolation point data points from training data for implementing a sparse Gaussian process model in an engine control module having a main unit, a memory unit, and a model calculation unit, the method comprising:
successively selecting training data points from the set of training data for acceptance into or exclusion from a set of interpolation point data points in accordance with a selection criterion; and
terminating selection when a termination criterion exists;
wherein the selection criterion depends on a divergence between a target value of the training data point to be selected and a function value, at the training data point to be selected, of the Gaussian process model based on the respectively current set of interpolation point data points;
wherein the model calculation unit is hard-wired or uses a restricted, specialized instruction set, and
wherein the sparse Gaussian model enables a model value to be determined by the main unit in the engine control module,
wherein the training data points are of at least variable that describes a physical behavior of an internal combustion engine,
wherein at least one of the following is satisfied: (i) the successive selection of the training data points from a set of training data for acceptance into the set of interpolation point data points in accordance with the selection criterion provides that in each selection cycle the selection criterion identifies for acceptance into the set of interpolation point data points a particular training data point having a target value which maximally diverges from a function value of the sparse Gaussian process model defined by a previously determined interpolation point data, and (ii) the successive selection of training data points from the set of training data for exclusion from the set of interpolation point data points in accordance with the selection criterion that includes the fact that in each selection cycle the selection criterion identifies, for exclusion from the set of interpolation point data points, that training data point whose target value diverges minimally from the function value of the sparse Gaussian process model defined by the previously determined interpolation point data,
wherein hyperparameters and the interpolation point data are stored in the memory unit,
wherein the model calculation unit is configured only for carrying out a specific calculation protocol that is based on repeated calculations of a sum, a multiplication, and/or an exponential function, and
wherein the hyperparameters are for the Gaussian model and are determined based on the interpolation point data.

2. The method of claim 1 , wherein the termination criterion encompass ascertaining that a defined number of selected interpolation point data points has been reached, and/or ascertaining that the maximum divergence falls below a defined error threshold.

3. The method of claim 1 , wherein determination of the set of interpolation point data points is carried out iteratively, such that after each determination of the set of interpolation point data points, the hyperparameters based thereon are determined for the sparse Gaussian process model and the set of interpolation point data points is determined again, the iterative determination being carried out as long as an iteration criterion is met.

4. The method of claim 3 , wherein the iteration criterion is met if the average divergence of the target values of the set of interpolation point data from the function values of the determined sparse Gaussian process model falls below a defined limit value.

5. The method of claim 1 , wherein the hyperparameters for the sparse Gaussian process model are identified after determination of the set of interpolation point data.

6. The method of claim 1 , wherein the termination criterion includes ascertaining that a defined number of selected interpolation point data points has been reached, and/or ascertaining that the minimal divergence exceeds a defined error threshold value.

7. An engine control module apparatus for identifying from training data a set of interpolation point data points for implementing a sparse Gaussian process model, comprising:
a memory unit;
a model calculation unit; and
a main unit having a processor arrangement configured to perform the following:
successively selecting from the set of training data, in accordance with a selection criterion, training data points for acceptance into or exclusion from a set of interpolation point data points; and
terminating selection when a termination criterion exists;

wherein the selection criterion depends on a divergence between a target value of the training data point to be selected and a function value, at the training data point to be selected, of the sparse Gaussian process model based on the respectively current set of interpolation point data points;
wherein the model calculation unit is hard-wired or uses a restricted, specialized instruction set, and
wherein the sparse Gaussian model enables a model value to be determined by the main unit in the engine control module,
wherein the training data points are of at least variable that describes a physical behavior of an internal combustion engine,
wherein at least one of the following is satisfied: (i) the successive selection of the training data points from a set of training data for acceptance into the set of interpolation point data points in accordance with the selection criterion provides that in each selection cycle the selection criterion identifies for acceptance into the set of interpolation point data points a particular training data point having a target value which maximally diverges from a function value of the sparse Gaussian process model defined by a previously determined interpolation point data, and (ii) the successive selection of training data points from the set of training data for exclusion from the set of interpolation point data points in accordance with the selection criterion that includes the fact that in each selection cycle the selection criterion identifies, for exclusion from the set of interpolation point data points, that training data point whose target value diverges minimally from the function value of the sparse Gaussian process model defined by the previously determined interpolation point data,
wherein hyperparameters and the interpolation point data are stored in the memory unit,
wherein the model calculation unit is configured only for carrying out a specific calculation protocol that is based on repeated calculations of a sum, a multiplication, and/or an exponential function, and
wherein the hyperparameters are for the Gaussian model and are determined based on the interpolation point data.

8. A system, comprising:
an engine control module apparatus for identifying from training data a set of interpolation point data points for implementing a sparse Gaussian process model, including:
a memory unit;
a model calculation unit; and
a main unit having a processor arrangement configured to perform the following:
successively selecting from the set of training data, in accordance with a selection criterion, training data points for acceptance into or exclusion from a set of interpolation point data points; and
terminating selection when a termination criterion exists;

wherein the selection criterion depends on a divergence between a target value of the training data point to be selected and a function value, at the training data point to be selected, of the sparse Gaussian process model based on the respectively current set of interpolation point data points;
wherein the engine control module is configured to receive the interpolation point data and ca

Assignees

• Bosch Gmbh Robert

Inventors

• Nguyen-Tuong The Duy
• Markert Heiner
• Schreiter Jens
• Hanselmann Michael

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• F02D2041/1433

  using a model or simulation of the system · CPC title

• F02D2041/1423

  Identification of model or controller parameters · CPC title

• G06F30/20

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title

• F02D41/2416

  Interpolation techniques · CPC title