Vehicle prognostics and remedial response

• US 10,553,046 B2
• Filed: 04/05/2018
• Issued: 02/04/2020
• Est. Priority Date: 04/05/2018
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations LLC (General Motors Company)
• Inventors: Thron, Esther, Montgomery, Wallace P., Nudurumati, Ravi P.
• Primary Examiner(s): Nguyen, Tan Q
• Application Number:
  US15/946,548
• Publication Number:
  US 20190311556A1
• Time in Patent Office: 670 Days
• Field of Search: 701 311
• CPC Class Codes: G07C 5/008 communicating information to a remotely located station transmission systems for measured values G08C
• CPC Class Codes: G07C 5/0808 Diagnosing performance data testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005

Abstract

A system and method of method of carrying out a remedial action in response to a vehicle prognosis, the method including: receiving vehicle feature data from a vehicle; extracting a plurality of feature combination data from the vehicle feature data, wherein each of the feature combination data pertains to a feature combination, wherein each of the feature combinations includes two or more vehicle features; for each extracted feature combination data, then: (i) evaluating the extracted feature combination data using an anomaly detection function based on a multivariate distribution mixture model; and (ii) obtaining an anomaly detection score for each extracted feature combination based on the evaluating step; determining a vehicle subsystem that comprises a portion of vehicle electronics installed on the vehicle and that is likely experiencing a problem or unusual behavior based on the anomaly detection scores; and carrying out a remedial action in response to the determining step.

Claims

1. A method of carrying out a remedial action in response to a vehicle prognosis, the method comprising:

receiving vehicle feature data from a vehicle;

extracting a plurality of feature combination data from the vehicle feature data, wherein each of the feature combination data pertains to a feature combination, wherein each of the feature combinations includes two or more vehicle features;

for each extracted feature combination data, then;

evaluating the extracted feature combination data using an anomaly detection function configured particularly for the feature combination, wherein the anomaly detection function is based on a multivariate distribution mixture model; and

obtaining an anomaly detection score for each extracted feature combination based on the evaluating step;

determining a vehicle subsystem that comprises a portion of vehicle electronics installed on the vehicle and that is likely experiencing a problem or unusual behavior based on the anomaly detection scores; and

carrying out a remedial action in response to the determining step.

2. The method of claim 1, wherein the vehicle feature data is vehicle sensor data, and wherein the vehicle feature data is obtained at the vehicle through use of a plurality of onboard vehicle sensors.

3. The method of claim 2, wherein the onboard vehicle sensors are connected to a wireless communications device via a communications bus, and wherein the wireless communications device is used to send the vehicle feature data to a remote facility.

4. The method of claim 2, further comprising the step of generating a plurality of multivariate mixture models for each possible feature combination of a particular class of vehicles, wherein the multivariate mixture model used in the evaluating step is one of the plurality of multivariate mixture models, and wherein the vehicle is included in the particular class of vehicles.

5. The method of claim 1, wherein the multivariate mixture model is a bivariate Gaussian mixture model that includes a plurality of mixture components.

6. The method of claim 1, wherein each of the anomaly detection functions are based on a different multivariate mixture model, wherein each of the different multivariate mixture models are generated for a particular feature combination.

7. The method of claim 6, wherein a first one of the plurality of feature combinations includes two vehicle features and wherein the first feature combination is associated with a bivariate Gaussian mixture model.

8. The method of claim 1, wherein the remedial action includes sending a warning message to the vehicle.

9. The method of claim 1, wherein the remedial action includes sending a vehicle command to the vehicle that causes the vehicle to automatically carry out a vehicle function pursuant to the vehicle command.

10. A method of carrying out a remedial action in response to a vehicle prognosis, the method comprising:

receiving vehicle feature data from a vehicle, wherein the vehicle feature data includes data for a plurality of vehicle features, and wherein each of the vehicle features are associated with an onboard vehicle sensor;

extracting a plurality of feature combination data from the vehicle feature data, wherein each of the feature combination data includes data pertaining to two or more vehicle features;

for each extracted feature combination data, obtaining an anomaly detection score for each extracted feature combination based on the evaluating step, wherein the anomaly detection scores are each determined by;

obtaining an anomaly detection function for a given feature combination, wherein the anomaly detection function is based on a multivariate distribution model that is specifically generated for the feature combination; and

calculating the anomaly detection score based on the anomaly detection function and the extracted feature combination data;

determining a vehicle subsystem that comprises a portion of vehicle electronics installed on the vehicle and that is likely experiencing a problem or unusual behavior based on the anomaly detection scores; and

carrying out a remedial action in response to the determining step.

11. The method of claim 10, further comprising the step of generating the anomaly detection function.

12. The method of claim 11, wherein the generating step includes modelling a set of training data for each feature combination of a particular type of vehicle, wherein the modelling includes using a multivariate Gaussian mixture model to obtain a feature combination mixture model that includes one or more mixture components.

13. The method of claim 10, wherein the anomaly detection function is a negative log likelihood function.

14. The method of claim 13, wherein the determining step is carried out based on selecting one or more feature combinations that are associated with top anomaly detection scores.

15. The method of claim 14, wherein one or more vehicle features included in the one or more selected feature combinations are analyzed to determine which vehicle subsystem is experiencing or is likely experiencing abnormal behavior or problematic behavior.

16. The method of claim 10, wherein the remedial action is particularly tailored for the vehicle subsystem.

17. A remote vehicle prognosis and remediation system, comprising:

a server that includes a processor and computer-readable memory, the computer-readable memory storing a computer program; and

a vehicle prognostics database that stores vehicle telemetry information including a plurality of anomaly detection functions;

wherein the computer program, when executed by the processor, causes the server to;

receive vehicle feature data from a vehicle;

extract a plurality of feature combination data from the vehicle feature data, wherein each of the feature combination data pertains to a feature combination, wherein each of the feature combinations includes two or more vehicle features;

for each extracted feature combination data, then;

evaluate the extracted feature combination data using an anomaly detection function configured particularly for the feature combination, wherein the anomaly detection function is based on a multivariate mixture model; and

obtain an anomaly detection score for each extracted feature combination based on the evaluating step;

determine a vehicle subsystem that comprises a portion of vehicle electronics installed on the vehicle and that is likely experiencing a problem or unusual behavior based on the anomaly detection scores; and

carry out a remedial action in response to the determining step.