System and methods for detecting vehicle braking events using data from fused sensors in mobile devices

• US 9,809,159 B1
• Filed: 12/28/2016
• Issued: 11/07/2017
• Est. Priority Date: 12/28/2016
• Status: Active Grant

• Current Assignee: Arity International Limited (Allstate Corporation)
• Original Assignee: Allstate Insurance Company (Allstate Corporation)
• Inventors: Snyder, Jared S., Ferguson, Dana
• Primary Examiner(s): Singh, Hirdepal
• Application Number:
  US15/391,986
• Time in Patent Office: 314 Days
• Field of Search: 340903
• CPC Class Codes: B60T 17/22 Devices for monitoring or checking brake systems; Signal devices
• CPC Class Codes: B60W 2040/0818 Inactivity or incapacity of driver
• CPC Class Codes: B60W 2040/0863 due to erroneous selection or response of the driver
• CPC Class Codes: B60W 2540/22 Psychological state; Stress level or workload
• CPC Class Codes: B60W 2540/229 Attention level, e.g. attentive to driving, reading or sleeping
• CPC Class Codes: B60W 2540/24 Drug level, e.g. alcohol
• CPC Class Codes: B60W 40/08 related to drivers or passengers
• CPC Class Codes: B60W 40/09 Driving style or behaviour
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06Q 40/08 Insurance

Abstract

One or more braking event detection computing devices and methods are disclosed herein based on fused sensor data collected during a window of time from various sensors of a mobile device found within an interior of a vehicle. The various sensors of the mobile device may include a GPS receiver, an accelerometer, a gyroscope, a microphone, a camera, and a magnetometer. Data from vehicle sensors and other external systems may also be used. The braking event detection computing devices may adjust the polling frequency of the GPS receiver of the mobile device to capture non-consecutive data points based on the speed of the vehicle, the battery status of the mobile device, traffic-related information, and weather-related information. The braking event detection computing devices may use classification machine learning algorithms on the fused sensor data to determine whether or not to classify a window of time as a braking event.

Claims

1. A braking event detection system comprising:

at least one processor; and

memory storing computer-readable instructions, that when executed by the at least one processor, cause the system to;

collect, by a sensor data collection device of the system, raw sensor data from sensors associated with a mobile device within a vehicle during a window of time using a polling frequency, wherein the sensors comprise a GPS receiver, an accelerometer, and a gyroscope, wherein the raw sensor data comprises information relating to a location, a speed, and an acceleration of the vehicle, and wherein the polling frequency is determined based on at least one of;

the speed of the vehicle, a battery status of the mobile device, traffic information within a first threshold radius of the vehicle, and weather information within a second threshold radius of the vehicle;

process, by a sensor data processing device of the system, the raw sensor data collected from the sensors associated with the mobile device to remove duplicate data points and generate processed sensor data;

apply, by a braking event classification device of the system, a classification machine learning algorithm to the raw sensor data and processed sensor data to determine that a window should be classified as a braking event, wherein the classification machine learning algorithm is stored in a braking event classification model of the system; and

generate and transmit, by a braking event notification device of the system, a notification to at least one of;

a second mobile device within a first predetermined distance of the vehicle and a second vehicle within a second predetermined distance of the vehicle relating to the braking event of the vehicle.

2. The system of claim 1 wherein processing the raw sensor data collected from the sensors associated with the mobile device to remove duplicate data points includes:

replacing a duplicate data point with an average value of a first data point and a second data point, wherein the first data point immediately precedes the duplicate data point, and wherein the second data point immediately follows the duplicate data point.

3. The system of claim 1, further including instructions that, when executed by the at least one processor, cause the system to:

perform, by a sensor data calibration device of the system, an alignment of a first axis, a second axis, and a third axis of the raw sensor data with an x-axis, a y-axis, and a z-axis of a reference frame of the vehicle based, at least in part, on a vector indicating acceleration due to gravity and a detection of acceleration of the vehicle along one of the first axis, the second axis, and the third axis.

4. The system of claim 1, further including instructions that, when executed by the at least one processor, cause the system to:

collect, by the sensor data collection device, supplemental sensor data from sensors associated with the vehicle during the window of time;

process, by the sensor data processing device, the supplemental sensor data collected from the sensors associated with the vehicle to remove duplicate data points; and

apply, by the braking event classification device, the classification machine learning algorithm to the raw sensor data, the processed sensor data, and the supplemental sensor data to determine that the window should be classified as a braking event.

5. The system of claim 1, further including instructions that, when executed by the at least one processor, cause the system to:

apply, by the braking event classification device, the classification machine learning algorithm to determine a probability that the raw sensor data and the processed sensor data associated with the window is a braking event; and

determine, by the braking event classification device, that the probability is greater than a probability threshold.

6. The system of claim 1, further including instructions that, when executed by the at least one processor, cause the system to:

transmit, by the braking event notification device, the notification via a short-range communication protocol.

7. The system of claim 1 wherein the classification machine learning algorithm comprises a random forest model.

8. A method comprising:

collecting, by a sensor data collecting device of a braking event detection system, raw sensor data from sensors associated with a mobile device within a vehicle during a window of time using a polling frequency, wherein the sensors comprise a GPS receiver, an accelerometer, and a gyroscope, wherein the sensor data comprises information relating to a location, a speed, and an acceleration of the vehicle, and wherein the polling frequency is determined based on at least one of;

the speed of the vehicle, a battery status of the mobile device, traffic information within a first threshold radius of the vehicle, and weather information within a second threshold radius of the vehicle;

processing, by a sensor data processing device of the system, the raw sensor data collected from the sensors associated with the mobile device to remove duplicate data points and generating processed sensor data;

applying, by a braking event classification device of the system, a classification machine learning algorithm to the raw sensor data and processed sensor data to determine that a window should be classified as a braking event, wherein the classification machine learning algorithm is stored in a braking event classification model of the system; and

generating and transmitting, by a braking event notification device of the system, a notification to at least one of;

a second mobile device within a first predetermined distance of the vehicle and a second vehicle within a second predetermined distance of the vehicle relating to the braking event of the vehicle.

9. The method of claim 8, further comprising:

replacing a duplicate data point with an average value of a first data point and a second data point, wherein the first data point immediately precedes the duplicate data point, and wherein the second data point immediately follows the duplicate data point.

10. The method of claim 8, further comprising:

performing an alignment of a first axis, a second axis, and a third axis of the raw sensor data with an x-axis, a y-axis, and a z-axis of a reference frame of the vehicle based, at least in part, on a vector indicating acceleration due to gravity and a detection of large magnitudes of the speed and the acceleration of the vehicle along one of the first axis, the second axis, and the third axis.

11. The method of claim 8, further comprising:

collecting, by the sensor data collection device, supplemental sensor data from sensors associated with the vehicle during the window of time;

processing, by the sensor data processing device, the supplemental sensor data collected from the sensors associated with the vehicle to remove duplicate data points; and

applying, by the braking event classification device, the classification machine learning algorithm to the raw sensor data, the processed sensor data, and the supplemental sensor data to determine that a window should be classified as a braking event.

12. The method of claim 8, further comprising:

applying, by the braking event classification device, the classification machine learning algorithm to determine a probability that the raw sensor data associated with the window is a braking event; and

determining, by the braking event classification device, that the probability is greater than a probability threshold.

13. The method of claim 8, further comprising:

transmitting, by the braking event notification device, the notification via a short-range communication protocol.

14. The method of claim 8, wherein the classification machine learning algorithm comprises a random forest model.

15. A computer-assisted method of detecting braking events comprising:

collecting, by a sensor data collecting device of a braking event detection system, raw sensor data from sensors associated with a mobile device within a vehicle during a window of time using a polling frequency, wherein the sensors comprise a GPS receiver, an accelerometer, and a gyroscope, wherein the sensor data comprises information relating to a location, a speed, and an acceleration of the vehicle, and wherein the polling frequency is determined based on at least one of;

the speed of the vehicle, a battery status of the mobile device, traffic information within a first threshold radius of the vehicle, and weather information within a second threshold radius of the vehicle, wherein the first threshold radius is smaller than the second threshold radius;

processing, by a sensor data processing device of the system, the raw sensor data collected from the sensors associated with the mobile device to remove duplicate data points by replacing a duplicate data point with an average value of a first data point and a second data point and generating processed sensor data, wherein the first data point immediately precedes the duplicate data point, and wherein the second data point immediately follows the duplicate data point;

applying, by a braking event classification device, a classification machine learning algorithm to the raw sensor data and processed sensor data to determine that a window should be classified as a braking event, wherein the classification machine learning algorithm is stored in a braking event classification model of the system; and

generating and transmitting, by a braking event notification device of the system, a visual notification to at least one of;

a second mobile device within a first predetermined distance of the vehicle and a second vehicle within a second predetermined distance of the vehicle relating to the braking event of the vehicle.

16. The computer-assisted method of claim 15, further comprising:

performing an alignment of a first axis, a second axis, and a third axis of the raw sensor data with an x-axis, a y-axis, and a z-axis of a reference frame of the vehicle based, at least in part, on a vector indicating acceleration due to gravity and a detection of large magnitudes of the speed and the acceleration of the vehicle along one of the first axis, the second axis, and the third axis.

17. The computer-assisted method of claim 15, further comprising:

collecting, by the sensor data collecting device, supplemental sensor data from sensors associated with the vehicle during the window of time;

processing, by the sensor data processing device, the supplemental sensor data collected from the sensors associated with the vehicle to remove duplicate data points; and

applying, by the braking event classification device, the classification machine learning algorithm to the raw sensor data, the processed sensor data, and the supplemental sensor data to determine that a window should be classified as a braking event.

18. The computer-assisted method of claim 15, further comprising:

applying, by the braking event classification device, the classification machine learning algorithm to determine a probability that the raw sensor data associated with the window is a braking event; and

determining, by the braking event classification device, that the probability is greater than a probability threshold.

19. The computer-assisted method of claim 15, further comprising:

using short-range communication protocols to send the visual notification to nearby vehicles.

20. The computer-assisted method of claim 15, wherein the classification machine learning algorithm comprises a random forest model.