PREDICTION OF DRIVER-SPECIFIC CRUISE SPEED USING DYNAMIC MODELING

• US 20130253797A1
• Filed: 03/20/2012
• Published: 09/26/2013
• Est. Priority Date: 03/20/2012
• Status: Active Grant

• Current Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Original Assignee: Toyota Motor Engineering & Manufacturing North America Incorporated (Toyota Motor Corporation)
• Inventors: MCNEW, John Michael
• Granted Patent: US 9,002,612 B2
• Time in Patent Office: Days
• US Class Current: 701/98
• CPC Class Codes: B60K 2310/244 changing target speed or setting a new target speed, e.g. changing algorithms
• CPC Class Codes: B60W 2050/0029 Mathematical model of the driver
• CPC Class Codes: B60W 2050/0083 Setting, resetting, calibration
• CPC Class Codes: B60W 2540/30 Driving style
• CPC Class Codes: B60W 2552/05 Type of road, e.g. motorways, local streets, paved or unpaved roads
• CPC Class Codes: B60W 2720/10 Longitudinal speed
• CPC Class Codes: B60W 2720/103 Speed profile
• CPC Class Codes: B60W 30/143 Speed control B60W30/16 takes precedence
• CPC Class Codes: B60W 50/0097 Predicting future conditions
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: Y02T 10/84 Data processing systems or methods, management, administration

Abstract

A controller for predicting cruising speeds of a vehicle includes a processor and an extracting unit to extract feature data from segments of a prior trajectory of the vehicle, the feature data including cruising speeds of the vehicle and predictive feature data. The controller also includes a model generator to generate a probabilistic model associating the predictive feature data with the cruising speeds of the vehicle and a predicting unit to predict a cruising speed of the vehicle for a target segment, which is an upcoming cruising segment of the vehicle, by conditioning the probabilistic model on real-time predictive feature data of segments of a current trajectory.

Claims

1. A controller for predicting cruising speeds of a vehicle, the controller including a processor and comprising:

an extracting unit configured to extract feature data from segments of a prior trajectory of the vehicle, the feature data including cruising speeds of the vehicle and predictive feature data;

a model generator configured to generate a probabilistic model associating the predictive feature data with the cruising speeds of the vehicle; and

a predicting unit configured to predict a cruising speed of the vehicle for a target segment, which is an upcoming cruising segment of the vehicle, by conditioning the probabilistic model on real-time predictive feature data of segments of a current trajectory.

2. The controller according to claim 1, wherein

the predicting unit is configured to condition the probabilistic model on a launch model of an initial segment of the segments of the current trajectory, and

the target segment follows the initial segment.

3. The controller according to claim 2, further comprising:

a collecting unit configured to collect trajectory data of the vehicle;

a segmenting unit configured to segment the collected trajectory data into segments; and

a selecting unit configured to select one or more predictive features, wherein

the extracting unit is configured to extract feature data corresponding to the selected one or more predictive features from the segments of the current trajectory as the real-time predictive feature data, and

the model generator is configured to generate the probabilistic model based on the feature data corresponding to the selected one or more predictive features.

4. The controller according to claim 3, further comprising an updating unit configured to update the probabilistic model based on the real-time predictive feature data and an actual cruising speed of the vehicle for the target segment.

5. The controller according to claim 4, further comprising:

a driver identification unit configured to identify a driver of the vehicle and associate the generated probabilistic model with the identified driver; and

a transmitter configured to transmit a message in response to a determination that the actual cruising speed is different by a predetermined threshold from a predicted cruising speed for the target segment based on the probabilistic model associated with the identified driver.

6. The controller according to claim 5, wherein the predetermined threshold is a statistically significant difference between the actual cruising speed and the predicted cruising speed based on the probabilistic model associated with the identified driver.

7. The controller according to claim 4, further comprising:

a vehicle identification unit configured to identify the vehicle and associate the generated probabilistic model with the vehicle; and

a transmitter configured to transmit a message in response to a determination that the actual cruising speed is different by a predetermined threshold from a predicted cruising speed for the target segment based on the probabilistic model associated with the vehicle.

8. The controller according to claim 4, wherein the updating unit is further configured to identify outlier feature data corresponding to unusual traffic patterns or unusual driving patterns and remove the identified outlier feature data from the real-time predictive feature data used to update the probabilistic model.

9. The controller according to claim 2, wherein the predicting unit is configured to predict the cruising speed of the vehicle for the target segment by conditioning the probabilistic model on a sequence number of a trajectory including the target segment relative to sequential trajectories forming a navigation path of the vehicle.

10. The controller according to claim 2, wherein the predicting unit is configured to predict the cruising speed of the vehicle for the target segment by conditioning the probabilistic model on a cruising speed of a cruising segment of the current trajectory, the current trajectory being a trajectory prior to a trajectory including the target segment.

11. The controller according to claim 2, wherein the predicting unit is configured to predict the cruising speed of the vehicle for the target segment by conditioning the probabilistic model on a deceleration model of a last segment of the current trajectory, the current trajectory being a trajectory prior to a trajectory including the target segment.

12. The controller according to claim 2, wherein, when an actual cruising speed for the target segment is reached by the vehicle, the controller designates an upcoming cruising segment as a next target segment for cruising speed prediction.

13. The controller according to claim 10, wherein the deceleration model is a time-based polynomial model, a state-space-based linear model, or a state-space-based non-linear model.

14. The controller according to claim 2, wherein the launch model is a time-based polynomial model, a state-space-based linear model, or a state-space-based non-linear model.

15. The controller according to claim 1, wherein

the vehicle includes a Global Positioning System (GPS) and a map database,

the feature data includes additional features, including environmental features, road geometry features, road condition features, and a road speed limit, and

the predicted cruising speed is based on a speed limit of the target segment.

16. The controller according to claim 15, wherein

the controller is configured to receive image data from a camera mounted to the vehicle, and

the controller extracts the speed limit from speed limit signs captured by the camera.

17. The controller according to claim 15, wherein the additional features include one or more of weather, road surface conditions, time of day, light level, atmospheric temperature, road inclination, radius of road curvature, and road composition.

18. The controller according to claim 1, wherein

the vehicle includes a Global Positioning System (GPS), and

the model generator generates the probabilistic model for each new road segment driven by the vehicle and associates the probabilistic model with the respective road segment.

19. The controller according to claim 1, wherein the predicted cruising speed is a probability density function.

20. The controller according to claim 18, further comprising a warning unit configured to integrate the probability density function to determine a cumulative distribution function, determine a cruising speed value at a user-selected certainty threshold of the cumulative distribution function, and issue a warning in response to the actual cruising speed exceeding the determined cruising speed value.

21. The controller according to claim 1, wherein the predicted cruising speed is a value.

22. The controller according to claim 1, further comprising an eco-acceleration cruise controller that receives the predicted cruising speed and determines an acceleration sequence to achieve the predicted cruising speed and maximize fuel efficiency.

23. The controller according to claim 1, wherein the model generator performs kernel density estimation, least absolute shrinkage and selection operator (LASSO) linear regression, or boosted tree regression to generate the probabilistic model.

24. The controller according to claim 1, wherein the predicting unit is further configured to predict the cruising speed for the target segment as a mean of actual cruising speeds of prior segments when no feature data has been extracted in the current trajectory.

25. The controller according to claim 1, wherein the controller outputs the predictive cruising speed to a client application.

26. A vehicle including the controller according to claim 1.

27. A method of predicting cruising speeds of a vehicle, the method comprising:

extracting feature data from segments of a prior trajectory of the vehicle, the feature data including cruising speeds of the vehicle and predictive feature data;

generating a probabilistic model associating the predictive feature data with the cruising speeds of the vehicle; and

predicting a cruising speed of the vehicle for a target segment, which is an upcoming cruising segment of the vehicle, by conditioning the probabilistic model on real-time predictive feature data of segments of a current trajectory.

28. A computer readable storage medium storing computer readable instructions thereon which, when executed by a computer, cause the computer to perform a method of predicting cruising speeds of a vehicle, the method comprising:

extracting feature data from segments of a prior trajectory of the vehicle, the feature data including cruising speeds of the vehicle and predictive feature data;

generating a probabilistic model associating the predictive feature data with the cruising speeds of the vehicle; and

predicting a cruising speed of the vehicle for a target segment, which is an upcoming cruising segment of the vehicle, by conditioning the probabilistic model on real-time predictive feature data of segments of a current trajectory.

29. A cruising speed prediction system comprising:

a controller aboard a vehicle, the controller being configured to collect trajectory data of the vehicle; and

a remote processing server configured to predict cruising speeds of the vehicle based on the trajectory data of the vehicle, the remote processing server including;

an extracting unit configured to extract feature data from segments of a prior trajectory of the vehicle, the feature data including cruising speeds of the vehicle and predictive feature data;

a model generator configured to generate a probabilistic model associating the predictive feature data with the cruising speeds of the vehicle; and

a predicting unit configured to predict a cruising speed of the vehicle for a target segment, which is an upcoming cruising segment of the vehicle, by conditioning the probabilistic model on real-time predictive feature data of segments of a current trajectory, wherein

the controller transmits the collected trajectory data to the remote processing server and receives from the remote processing server the predicted cruising speed of the vehicle for the target segment.