System and method for machine-learning-based position estimation for use in micro-assembly control with the aid of a digital computer
Control loop latency can be accounted for in predicting positions of micro-objects being moved by using a hybrid model that includes both at least one physics-based model and machine-learning models. The models are combined using gradient boosting, with a model created during at least one of the stages being fitted based on residuals calculated during a previous stage based on comparison to training data. The loss function for each stage is selected based on the model being created. The hybrid model is evaluated with data extrapolated and interpolated from the training data to prevent overfitting and ensure the hybrid model has sufficient predictive ability. By including both physics-based and machine-learning models, the hybrid model can account for both deterministic and stochastic components involved in the movement of the micro-objects, thus increasing the accuracy and throughput of the micro-assembly.
1. A method for machine-learning enabled micro-assembly control with the aid of a digital computer, comprising steps of:
obtaining one or more parameters of a closed-loop system for positioning one or more of a plurality of micro-objects, the system comprising a plurality of programmable electrodes, the electrodes configured to induce movements of the micro-objects when the micro-objects are proximate to electrodes upon actuations of the electrodes, each of the actuations comprising a generation of one or more electric potentials by one or more of the electrodes;
obtaining training data comprising states associated with previous movements of one or more of the plurality of the micro-objects due to earlier ones of the actuations;
defining one or more physics-based models for predicting a position of at least one of the micro-objects following one of the actuations;
building a hybrid model for predicting a position of at least one of the micro-objects following one of the actuations, the hybrid model comprising a plurality of models comprising one or more of the physics-based models and one or more machine-learning models, using the training data via gradient boosting;
obtaining a data set;
dividing the data set into the training data and interpolative testing data;
extrapolating the states of previous movements to generate extrapolated data;
testing the hybrid model using the interpolative testing data and the extrapolated data;
upon the hybrid model passing the testing, controlling the closed-loop system, comprising:
predicting a position of one or more of the micro-objects following one of the actuations based on at least one sensor measurement;
receiving a further position of one or more of the micro-objects; and
performing another one of the actuations using the predicted position and the further position,
wherein the steps are performed by a suitably-programmed computer.
2. A method according to claim 1 , wherein the gradient boosting comprises a plurality of stages, wherein one of the models is built during each of the stages, further comprising:
for each of the models, using a loss function to determine residuals between a prediction of the one model based on an input portion of the training data and output data associated with the training data;
for each of the models other than the model built during a first one of the stages, defining that model using the residuals associated with the model built during the stage preceding the stage during which that model is built.
3. A method according to claim 2 , further comprising:
selecting the loss function used to determine the residuals for each of the models based on a stochastic model for each of the models based on the stage associated with that model.
4. A method according to claim 3 , wherein the loss function is a decorrelation function.
5. A method according to claim 4 , wherein a Ljung-Box loss function is used for selecting the loss function for at least one of the stages.
6. A method according to claim 4 , wherein the model built during the first stage is one of the physics-based models and the model during a last one of the stages is one of the machine-learning models.
7. A method according to claim 1 , wherein the one or more processors define a recurrent neural network.
8. A method according to claim 7 , wherein the machine-learning models comprise one or more of a linear model, a dense model, a deep model, and an ARIMA model.
9. A method according to claim 1 , wherein the computer comprises at least one of a central processing unit and a graphics processing unit.
10. A system for machine-learning enabled micro-assembly control with the aid of a digital computer, comprising:
at least one processor configured to:
obtain one or more parameters of a closed-loop system for positioning one or more of a plurality of micro-objects, the system comprising a plurality of programmable electrodes, the electrodes configured to induce movements of the micro-objects when the micro-objects are proximate to electrodes upon actuations of the electrodes, each of the actuations comprising a generation of one or more electric potentials by one or more of the electrodes;
obtain training data comprising states associated with previous movements of one or more of the plurality of the micro-objects due to earlier ones of the actuations;
define one or more physics-based models for predicting a position of at least one of the micro-objects following one of the actuations;
build a hybrid model for predicting a position of at least one of the micro-objects following one of the actuations, the hybrid model comprising a plurality of models comprising one or more of the physics-based models and one or more machine-learning models, using the training data via gradient boosting;
obtain a data set;
divide the data set into the training data and interpolative testing data;
extrapolate the states of previous movements to generate extrapolated data; and
test the hybrid model using the interpolative testing data and the extrapolated data;
upon the hybrid model passing the testing, control the closed-loop system, comprising:
predict a position of one or more of the micro-objects following one of the actuations based on at least one sensor measurement;
receive a further position of one or more of the micro-objects; and
perform another one of the actuations using the predicted position and the further position.
11. A system according to claim 10 , wherein the gradient boosting comprises a plurality of stages, wherein one of the models is built during each of the stages, the at least one processor further configured to:
for each of the models, use a loss function to determine residuals between a prediction of the one model based on an input portion of the training data and output data associated with the training data;
for each of the models other than the model built during a first one of the stages, define that model using the residuals associated with the model built during the stage preceding the stage during which that model is built.
12. A system according to claim 11 , the at least one processor further configured to:
select the loss function used to determine the residuals for each of the models based on a stochastic model for each of the models based on the stage associated with that model.
13. A system according to claim 12 , wherein the loss function is a decorrelation function.
14. A system according to claim 13 , wherein a Ljung-Box loss function is used for selecting the loss function for at least one of the stages.
15. A system according to claim 13 , wherein the model built during the first stage is one of the physics-based models and the model during a last one of the stages is one of the machine-learning models.
16. A system according to claim 10 , wherein the one or more processors define a recurrent neural network.
17. A system according to claim 16 , wherein the machine-learning models comprise one or more of a linear model, a dense model, a deep model, and an ARIMA mode.
18. A system according to claim 10 , wherein the at least one processor is at least one of a central processing unit and a graphics processing unit.
19. A system for machine-learning-enabled hybrid-model-based micro-assembly control with the aid of a digital computer, comprising:
at least one processor configured to:
obtain one or more parameters of a closed-loop system for positioning one or more of a plurality of micro-objects, the system comprising a plurality of programmable electrodes, the electrodes configured to induce movements of the micro-objects when the micro-objects are proximate to electrodes upon actuations of the electrodes, each of the actuations comprising a generation of one or more electric potentials by one or more of the electrodes;
obtain training data comprising states associated with previous movements of one or more of the plurality of the micro-objects due to earlier ones of the actuations;
define one or more physics-based models for predicting a position of at least one of the micro-objects following one of the actuations;
build a hybrid model for predicting a position of at least one of the micro-objects following one of the actuations, the hybrid model comprising a plurality of models comprising one or more of the physics-based models and one or more machine-learning models, using the training data via gradient boosting that comprises a plurality of stages, wherein one of the models is built during each of the stages, the at least one processor further configured to:
select a loss function, the loss function comprising a decorrelation function, used to determine residuals for each of the models based on a stochastic model for each of the models based on the stage associated with that model;
for each of the models, use the loss function to determine the residuals between a prediction of the one model based on an input portion of the training data and output data associated with the training data; and
for each of the models other than the model built during a first one of the stages, define that model using the residuals associated with the model built during the stage preceding the stage during which that model is built;
control the closed-loop system, comprising:
predict a position of one or more of the micro-objects following one of the actuations based on at least one sensor measurement;
receive a further position of one or more of the micro-objects; and
perform another one of the actuations using the predicted position and the further position.