IP Library › Granted Patent US 12663767

Granted Patent B2

US 12663767 · App. 18/462,332 · Granted Jun 23, 2026

System and method for hybrid-model-based micro-assembly control with the aid of a digital computer

Inventors: Anne Plochowietz (Mountain View, CA); Anand Ramakrishnan (Worcester, MA); Warren Jackson (San Francisco, CA); Lara S. Crawford (Belmont, CA); Bradley Rupp (San Francisco, CA); Sergey Butylkov (Van Nuys, CA); Jeng Ping Lu (Fremont, CA); Eugene M. Chow (Palo Alto, CA)

Assignee: Genesee Valley Innovations, LLC

G05B13/048 G05B13/027 G05B13/042 G06N3/08 G06N7/08

View Patent ↗

Loading inventors, assignments & file history…

Monitor This Case

Get email alerts when status or documents change.

Order Certified Copies

Most orders are placed with the USPTO same day — all within 24 business hours.

Order via The Patent Place →

Pre-filled with this patent's details

Quick Facts

Patent No.: US 12663767
App. No.: 18/462,332
Granted: Jun 23, 2026
Kind: B2

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

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.

Claims (69)

1 . A method for hybrid-model-based 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 recorded data comprising input states and output states associated with previous movements of one or more of the plurality of the micro-objects due to earlier ones of the actuations;

building a hybrid model for predicting one or more positions of at least one of the micro-objects following one or more of the actuations, the hybrid model comprising a plurality of models, using the recorded data via gradient boosting, comprising:

defining a number of the plurality of models to be included into the hybrid model based on user input, wherein the plurality of models comprises a physics-based model and remaining models;

defining the physics-based model as one of the plurality of models using one or more of the parameters;

making, using the physics-based model, a prediction regarding one of the positions of one of the micro-objects using one of the input states;

using a loss function to determine a residual between the physics-based model prediction and one of the output states associated with the one input state used for the physics-based model prediction;

building one of the remaining models using the residual associated with the physics-based model; and

processing one or more of the remaining models in the hybrid model until a number of the models built equals the defined number, wherein at least some of the processed remaining models are machine-learning models, comprising:

making, using one of the remaining models, a prediction regarding one of the positions of one of the micro-objects using one of the input states;

using a further loss function to determine a residual between the remaining model's prediction and one of the output states associated with the one input state used for the remaining model's prediction; and

building another one of the remaining models using the residual for the remaining model's prediction; and

controlling the closed-loop system to perform one or more of the actuations based on the hybrid model,

wherein the steps are performed by a suitably-programmed computer.

2 . A method according to claim 1 , further comprising selecting the loss function and each of the further loss functions.

3 . A method according to claim 2 , further comprising selecting a stochastic model associated with the physics-based model and a further stochastic model associated with each of the remaining models, wherein the loss function is selected based on the stochastic model and each of the further loss functions is selected based on one of the further stochastic models.

4 . A method according to claim 3 , wherein selecting the loss function comprises comparing residual obtained using the loss function to the stochastic model and selecting each of the further loss functions comprises comparing the residual obtained using the further loss function to one of the further stochastic models.

5 . A method according to claim 4 , wherein one or more of the comparisons are performed using an autocorrelation function.

6 . A method according to claim 5 , wherein the autocorrelation function comprises a zero mean Gaussian function.

7 . A method according to claim 3 , wherein the stochastic model and the further stochastic models are Gaussian models.

8 . A method according to claim 1 , wherein the user input for the number of the models to be included in the hybrid model is based on a past performance of further hybrid models.

9 . A method according to claim 1 , further comprising:

aggregating the predictions of the models in the hybrid model into a prediction of the hybrid model, wherein the closed loop system is controlled based on the prediction of the hybrid model.

10 . A method according to claim 9 , wherein aggregation of the predictions comprises assigning a weight to each of the predictions of each of the models in the hybrid model.

11 . A system for hybrid-model-based micro-assembly control utilizing loss function selection 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 recorded data comprising input states and output states associated with previous movements of one or more of the plurality of the micro-objects due to earlier ones of the actuations;

build a hybrid model for predicting one or more positions of at least one of the micro-objects following one or more of the actuations, the hybrid model comprising a plurality of models, using the recorded data via gradient boosting, comprising:

define a number of the plurality of models to be included into the hybrid model based on user input, wherein the plurality of models comprises a physics-based model and remaining models;

build the physics-based model as one of the plurality of models using one or more of the parameters;

make, using the physics-based model, a prediction regarding one of the positions of one of the micro-objects using one of the input states;

select a loss function;

use the loss function to determine a residual between the physics-based model prediction and one of the output states associated with the one input state used for the physics-based model prediction;

build one of the remaining models using the residual associated with the physics-based model; and

process one or more of the remaining models in the hybrid model until a number of the models built equals the defined number, wherein at least some of the remaining models are machine-learning models, comprising:

make, using one of the remaining models, a prediction regarding one of the positions of one of the micro-objects using one of the input states;

select a further loss function for that remaining model;

use the further loss function to determine a residual between the remaining model's prediction and one of the output states associated with the one input state used for the remaining model's prediction; and

build another one of the remaining models using the residual for the remaining model's prediction; and

control the closed-loop system to perform one or more of the actuations based on the hybrid model.

12 . A system according to claim 11 , the at least one processor further configured to select a stochastic model associated with the physics-based model and a further stochastic model associated with each of the remaining models, wherein the loss function is selected based on the stochastic model and each of the further loss functions is selected based on one of the further stochastic models.

13 . A system according to claim 12 , wherein selecting the loss function comprises comparing residual obtained using the loss function to the stochastic model and selecting each of the further loss functions comprises comparing the residual obtained using the further loss function to one of the further stochastic models.

14 . A system according to claim 13 , wherein one or more of the comparisons are performed using an autocorrelation function.

15 . A system according to claim 14 , wherein the autocorrelation function comprises a zero mean Gaussian function.

16 . A system according to claim 12 , wherein the stochastic model and the further stochastic models are Gaussian models.

17 . A system according to claim 11 , wherein the user input for the number of the models to be included in the hybrid model is based on a past performance of further hybrid models.

18 . A system according to claim 11 , the at least one processor:

aggregate the predictions of the models in the hybrid model into a prediction of the hybrid model, wherein the closed loop system is controlled based on the prediction of the hybrid model.

19 . A system according to claim 18 , wherein aggregation of the predictions comprises assigning a weight to each of the predictions of each of the models in the hybrid model.

20 . A system for hybrid-model-based micro-assembly control utilizing stochastic-model-based loss function selection with the aid of a digital computer, comprising: