Controller and method for selecting evaluation points for a Bayesian optimization method

1. A method for selecting evaluation points for a Bayesian optimization method for optimizing a physical or chemical process, which is modeled by a statistical model, comprising the following steps:

ascertaining a posterior model of the statistical model corresponding to results of one or multiple evaluations at previous evaluation points;

selecting a next evaluation point by optimizing an acquisition function over a search space, which is given by a specified limit for a predictive variance of points in the search space given by the posterior model, wherein a range, in which the next evaluation point is sought, is limited to a range that contains points for which a predictive variance of the posterior model is smaller than a specified limit, wherein each selected evaluation point corresponds to a respective control parameter; and

performing the physical or chemical process in accordance with each respective control parameter.

2. The method as recited in claim 1 , wherein the statistical model is a Gaussian process and the specified limit is a specified fraction of a signal variance of a kernel of the Gaussian process.

3. The method as recited in claim 1 , wherein a set of alternatives for the next evaluation point is ascertained by sampling from an evaluation point range and one of the alternatives is selected as the next evaluation point, for which the acquisition function assumes an optimum in a set of the alternatives.

4. The method as recited in claim 3 , wherein the evaluation range is a subset of a parameter range and the set of alternatives is ascertained by sampling using a rejection method from the parameter range, samples being rejected that are not in the search space.

5. The method as recited in claim 4 , wherein the sampling in the parameter range includes the sampling in subspaces of the parameter range, which respectively contain one or multiple of the previous evaluation points.

6. A controller configured to select evaluation points for a Bayesian optimization method for optimizing a physical or chemical process, which is modeled by a statistical model, the controller configured to:

ascertain a posterior model of the statistical model corresponding to results of one or multiple evaluations at previous evaluation points; and

select a next evaluation point by optimizing an acquisition function over a search space, which is given by a specified limit for a predictive variance of points in the search space given by the posterior model, wherein a range, in which the next evaluation point is sought, is limited to a range that contains points for which a predictive variance of the posterior model is smaller than a specified limit, wherein each selected evaluation point corresponds to a respective control parameter; and

perform the physical or chemical process in accordance with each respective control parameter.

7. A production system, comprising:

a controller configured to select evaluation points for a Bayesian optimization method for optimizing a physical or chemical process, which is modeled by a statistical model, the controller configured to:

ascertain a posterior model of the statistical model corresponding to results of one or multiple evaluations at previous evaluation points, and

select a next evaluation point by optimizing an acquisition function over a search space, which is given by a specified limit for a predictive variance of points in the search space given by the posterior model wherein a range, in which the next evaluation point is sought, is limited to a range that contains points for which a predictive variance of the posterior model is smaller than a specified limit;

a production facility, which configured to manufacture a product using the physical or chemical process; and

at least one sensor that is configured to provide the controller with output parameter values of the physical or chemical process associated with control parameter values;

wherein the controller is configured to:

ascertain one or multiple optimal control parameter values of the physical or chemical process using the Bayesian optimization method, in that it evaluates control parameter values by using the output parameter values, and

control the production facility according to the one or multiple ascertained control parameter values.

8. A processing system, comprising

a controller configured to select evaluation points for a Bayesian optimization method for optimizing a physical or chemical process, which is modeled by a statistical model, the controller configured to:

ascertain a posterior model of the statistical model corresponding to results of one or multiple evaluations at previous evaluation points, and

select a next evaluation point by optimizing an acquisition function over a search space, which is given by a specified limit for a predictive variance of points in the search space given by the posterior model wherein a range, in which the next evaluation point is sought, is limited to a range that contains points for which a predictive variance of the posterior model is smaller than a specified limit;

a processing facility, which is configured to process a workpiece using the physical or chemical process; and

at least one sensor that is configured to provide the controller with output parameter values of the physical or chemical process associated with control parameter values;

wherein the controller is configured to:

ascertain one or multiple optimal control parameter values of the physical or chemical process using the Bayesian optimization method, in that it evaluates control parameter values by using the output parameter values; and

control the processing facility according to the one or multiple ascertained control parameter values.

9. A robotic system, comprising:

a controller configured to select evaluation points for a Bayesian optimization method for optimizing a physical or chemical process, which is modeled by a statistical model, the controller configured to:

ascertain a posterior model of the statistical model corresponding to results of one or multiple evaluations at previous evaluation points, and

select a next evaluation point by optimizing an acquisition function over a search space, which is given by a specified limit for a predictive variance of points in the search space given by the posterior model wherein a range, in which the next evaluation point is sought, is limited to a range that contains points for which a predictive variance of the posterior model is smaller than a specified limit;

a robotic element that is configured to perform a movement using the physical or chemical process; and

at least one sensor that is configured to provide the controller with output parameter values of the physical or chemical process associated with control parameter values;

wherein the controller is configured to:

ascertain one or multiple optimal control parameter values of the physical or chemical process using the Bayesian optimization method, in that it evaluates control parameter values by using the output parameter values; and

control the robotic element according to the one or multiple ascertained control parameter values.

10. A non-transitory computer-readable storage medium on which are stored program instructions for selecting evaluation points for a Bayesian optimization method for optimizing a physical or chemical process, which is modeled by a statistical model, the program instructions, when executed by one or more processors, causing the one or more processors to perform the following steps:

ascertaining a posterior model of the statistical model corresponding to results of one or multiple evaluations at previous evaluation points; and

selecting a next evaluation point by optimizing an acquisition function over a search space, which is given by a specified limit for a predictive variance of points in the search space given by the posterior model, wherein a range, in which the next evaluation point is sought, is limited to a range that contains points for which a predictive variance of the posterior model is smaller than a specified limit, wherein each selected evaluation point corresponds to a respective control parameter; and

performing the physical or chemical process in accordance with each respective control parameter.