CONTROLLING A MANUFACTURING PROCESS USING CAUSAL MODELS

1 . A method comprising:

selecting treatment assignments for operation of a system in an environment based on a causal model mapped to a probability distribution over possible treatment assignments, wherein the causal model measures causal relationships between the treatment assignments and a performance metric associated with the environment;

mapping causal effects between the treatment assignments and measures of uncertainty associated with the environment in the causal model to corresponding probabilities using probability matching;

determining the performance metric using the treatment assignments and corresponding dependent variables obtained from the environment in response to the treatment assignments;

adjusting, based on the performance metric, the causal model; and

re-computing the adjusted causal model by computing overall causal effects based on the measures of uncertainty around the overall causal effects.

2 . The method of claim 1 , wherein the treatment assignments are selected for blocked groups according to a blocking scheme that comprises at least one of a double-blind assignment, a pair-wise assignment, a Latin-square assignment, or a propensity matching assignment.

3 . The method of claim 1 , further comprising forming the treatment assignments by assigning treatments based on the probabilities.

4 . The method of claim 1 , wherein the measures of uncertainty around the overall causal effects are represented by confidence intervals around means of d-scores, and wherein each respective d-score represents a difference between an effect when a corresponding treatment is activated and when the corresponding treatment is deactivated.

5 . The method of claim 1 , wherein selecting the treatment assignments is further based on a set of internal control parameters that define how to adjust the causal model.

6 . The method of claim 5 , further comprising adjusting the internal control parameters based on the performance metric using the treatment assignments and the corresponding dependent variables obtained from the environment in response to the treatment assignments.

7 . The method of claim 5 , further comprising adjusting the internal control parameters based on the causal model.

8 . The method of claim 1 , wherein the performance metric comprises one or more of:

a measure of success of a process associated with the environment;

one or more online metrics of a performance of the process;

one or more offline metrics of the performance of the process;

one or more measures related to an output of the process associated with the environment;

one or more measures related to defective outputs of the process associated with the environment;

a capacity of the process associated with the environment;

a stability of the process associated with the environment;

an efficiency of the process associated with the environment; or

one or more measures received from sensor data that covaries with the treatment assignments.

9 . The method of claim 1 , further comprising:

receiving one or more external inputs;

setting, based on the received external inputs, initial possible values for controllable elements of the environment; and

identifying, from the received external inputs, one or more environment responses from the environment to track during the operation of the system in the environment.

10 . The method of claim 9 , wherein the one or more environment responses include the performance metric.

11 . The method of claim 9 , wherein the initial possible values include initial values associated with the probability distribution.

12 . The method of claim 9 , wherein the initial possible values include one or more baseline values associated with one or more corresponding internal parameters that define a selection of one or more control settings associated with the causal model.

13 . The method of claim 9 , wherein the external inputs are generated by a machine learning model based on a previous operation of the system or previous operation in the environment.

14 . A system comprising:

a memory configured to store data associated with an environment; and

a processor in communication with the memory, the processor being configured to:

select treatment assignments for operation of a system in the environment based on a causal model mapped to a probability distribution over possible treatment assignments, wherein the causal model measures causal relationships between the treatment assignments and a performance metric associated with the environment;

map causal effects between the treatment assignments and measures of uncertainty associated with the environment in the causal model to corresponding probabilities using probability matching;

determine the performance metric using the treatment assignments and corresponding dependent variables obtained from the environment in response to the treatment assignments;

adjust, based on the performance metric, the causal model; and

re-compute the adjusted causal model by computing overall causal effects based on the measures of uncertainty around the overall causal effects.

15 . The system of claim 14 , wherein the treatment assignments are selected for blocked groups according to a blocking scheme by selecting control settings for controllable elements for entities in the environment.

16 . The system of claim 14 , wherein the processor is further configured to:

select a spatial extent for each respective entity of a plurality of entities associated with the environment; and

select a temporal extent for each respective procedural instance of procedural instances of the environment;

generate the procedural instances based on the selected spatial extents and the selected temporal extents.

17 . The system of claim 16 , wherein the processor is further configured to:

cluster the procedural instances based on one or more characteristics of the environment to form two or more clusters;

identify each procedural instance of each respective cluster as one of a hybrid instance or a baseline instance for a controllable element associated with the environment;

select first control settings for the identified baseline instances;

select second control settings for the identified hybrid instances; and

select the treatment assignments for blocked groups according to a blocking scheme;

form the blocked groups based on a first blocked group including the baseline instances and a second blocked group including the hybrid instances.

18 . The system of claim 17 , wherein to cluster the procedural instances, the processor is configured to:

compute d-scores for the first and second control settings, wherein each respective d-score represents an effect of the respective treatment assignment being active;

monitor interaction terms between the causal effects and one or more external factors associated with the environment; and

apply a factorial analysis of variance (ANOVA) to generate the two or more clusters across space and time such that each respective cluster of the two or more clusters is representative of a distinct state of the one or more external factors.

19 . The system of claim 17 , wherein to cluster the procedural instances, the processor is configured to apply a machine learning technique.

20 . The system of claim 19 , wherein the machine learning technique comprises one of a decision tree-based technique, a conditional inference tree-based technique, a deep neural network, or a k-means clustering technique.

21 . An apparatus comprising:

means for selecting treatment assignments for operation of a system in an environment based on a causal model mapped to a probability distribution over possible treatment assignments, wherein the causal model measures causal relationships between the treatment assignments and a performance metric associated with the environment;

means for mapping causal effects between the treatment assignments and measures of uncertainty associated with the environment in the causal model to corresponding probabilities using probability matching;

means for determining the performance metric using the treatment assignments and corresponding dependent variables obtained from the environment in response to the treatment assignments;

means for adjusting, based on the performance metric, the causal model; and

means for re-computing the adjusted causal model by computing overall causal effects based on the measures of uncertainty around the overall causal effects.

22 . A non-transitory computer-readable medium encoded with instructions that, when executed, cause a processor to:

select treatment assignments for operation of a system in an environment based on a causal model mapped to a probability distribution over possible treatment assignments, wherein the causal model measures causal relationships between the treatment assignments and a performance metric associated with the environment;

map causal effects between the treatment assignments and measures of uncertainty associated with the environment in the causal model to corresponding probabilities using probability matching;

determine the performance metric using the treatment assignments and corresponding dependent variables obtained from the environment in response to the treatment assignments;

adjust, based on the performance metric, the causal model; and

re-compute the adjusted causal model by computing overall causal effects based on the measures of uncertainty around the overall causal effects.

23 . The non-transitory computer-readable medium of claim 22 , wherein the causal relationships are first causal relationships, the non-transitory computer-readable medium being further encoded with instructions that, when executed, cause the processor to:

measure second causal relationships between different possible values for an internal parameter of the causal model and a figure of merit for the internal parameter;

apply probability matching to map the causal model the probability distribution based on the second causal relationships; and

sample values from the possible values based on the probability distribution.

24 . The non-transitory computer-readable medium of claim 22 , further encoded with instructions that, when executed, cause the processor to:

select one or more control settings associated with the environment;

monitor environment responses from the environment to the selected control settings;

determine, based on the monitored environment responses, a change to one or more properties of the environment; and

based on the determined change to the one or more properties, adjust one or more internal parameters associated with the causal model.

25 . The non-transitory computer-readable medium of claim 24 , wherein the instructions that cause the processor to adjust the one or more internal parameters comprise instructions that, when executed, cause the processor to shrink a data inclusion window parameter.

26 . The non-transitory computer-readable medium of claim 24 , wherein the instructions that cause the processor to adjust the one or more internal parameters comprise instructions that, when executed, cause the processor to decrease a hybrid-to-explore ratio such that the procedural instances include fewer hybrid instances than explore instances.

27 . The non-transitory computer-readable medium of claim 24 , further encoded with instructions that, when executed, cause the processor to select the treatment assignments for blocked groups according to a blocked scheme, wherein the instructions that cause the processor to adjust the one or more internal parameters comprise instructions that, when executed, cause the processor to decrease a number of the blocked groups.

28 . The non-transitory computer-readable medium of claim 22 , further encoded with instructions that, when executed, cause the processor to:

receive one or more external inputs;

set, based on the received external inputs, initial possible values for controllable elements of the environment; and

identify, from the received external inputs, one or more environment responses from the environment to track during an operation of the environment.

29 . The non-transitory computer-readable medium of claim 28 , wherein the one or more environment responses include the performance metric.

30 . The non-transitory computer-readable medium of claim 28 , wherein the external inputs are generated by a machine learning model based on a previous operation of the system or previous operation in the environment.