Method for assessing models of vehicle driving style or vehicle usage model detector

• US 7,383,154 B2
• Filed: 12/14/2005
• Issued: 06/03/2008
• Est. Priority Date: 12/14/2005
• Status: Expired due to Fees

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations Incorporated (Motors Liquidation Co. GUC Trust)
• Inventors: Tate, Edward D. Jr.
• Primary Examiner(s): Barlow; John
• Assistant Examiner(s): KHUU, HIEN DIEU THI
• Application Number:
  US11/302,675
• Publication Number:
  US 20070136040A1
• Time in Patent Office: 902 Days
• Field of Search: 702/181
• US Class Current: 702/181
• CPC Class Codes: B60W 2050/0028 Mathematical models, e.g. for simulation
• CPC Class Codes: B60W 2050/0051 combined with averaging
• CPC Class Codes: B60W 40/09 Driving style or behaviour

Abstract

The present invention relates to a method of determining the probability that a statistical model describes the driving style or usage of a vehicle observed. By identifying an appropriate statistical model, the driving style can be detected and vehicle operation can be optimized. The method may be applied to both hybrid and non-hybrid vehicles. The method may also be augmented to determine the most probable statistical model of the usage of the vehicle. This allows detection of driving styles that can be classified as urban, highway, aggressive, etc. For example, the augmented method can be used to distinguish between highway and urban driving with a very high certainty.

Claims

1. A method of determining the probability that a statistical model describes the driving style of a vehicle observed from time step k to present, said method comprising:

(a) establishing at least one statistical model of driving style, wherein said at least one statistical model is a conditional probability model that describes probability of conditions at a next time step k+1, given conditions at said time step k;

(b) determining initial probability distribution of said at least one statistical model of driving style,

 

wherein said (b) determining the initial probability distribution of said at least one statistical model comprises setting vector Π

₀ as the initial probability distribution wherein said vector Π

₀ is re-determined for each said at least one statistical model;

(c) determining probability of an observed transition being predicted by said at least one statistical model of driving style;

(d) calculating probability of said at least one statistical model explaining driving style of said vehicle observed at said next time step k+1; and

wherein said (c) determining the probability of said observed transition and said (d) calculating probability of said at least one statistical model are iterated for all time steps until present.

2. The method of claim 1, wherein said statistical model of driving style is a conditional probability model that describes the probability P at a time step k+1, of having a vehicle velocity V_k+1, given a vehicle velocity V_k at said time step k.

3. The method of claim 2:

wherein said statistical model of driving style is formed from a sample of driving that represents said driving style; and

wherein said sample comprises velocity of said vehicle over time.

4. The method of claim 1, wherein said statistical model of driving style is a conditional probability model that describes the probability P at a time step k+1, of a vehicle with specific velocity, torque, acceleration, road grade, physical location, vehicle load, passenger count and seating, transmission gear, driver identification, key, or special vehicle mode, given the respective factors at said time step k.

5. The method of claim 1, wherein there are N statistical models of driving style:

wherein each element of said vector Π

₀ set to 1/N, whereby said statistical model has an equal probability of matching driving style of said vehicle observed.

6. The method of claim 1, further comprising:

(e) controlling at least one operating parameter of the vehicle based upon said at least one statistical model of driving style in a manner to improve fuel economy at said next time step k+1.

7. The A method of determining the probability that a statistical model describes the driving style of a vehicle observed from time step k to present, said method comprising:

(a) establishing at least one statistical model of driving style, wherein said at least one statistical model is a conditional probability model that describes probability of conditions at a next time step k+1, given conditions at said time step k;

(b) determining initial probability distribution of said at least one statistical model of driving style;

(c) determining probability of an observed transition being predicted by said at least one statistical model of driving style;

(d) calculating probability of said at least one statistical model explaining driving style of said vehicle observed at said next time step k+1;

wherein said (c) determining the probability of said observed transition and said (d) calculating probability of said at least one statistical model are iterated for all time steps until present; and

wherein said (c) determining the probability of the observed transition being predicted by said statistical model of driving style includes;

employing a reference table look up, wherein said table look-up is a two dimensional matrix developed for each said statistical model, wherein there are N said statistical models of driving style, wherein said table look-up has three independent indices;

an index identifying said statistical model, current velocity and previous velocity of said vehicle observed; and

employing a pre-determined matrix M₁ to define probability of transitioning from one driving style to another at each sample, wherein M₁ is a Markov matrix.

8. The method of claim 7, wherein said (c) calculating the probability of said statistical model explaining driving style of said vehicle observed at time step k+1 comprises:

employing a pre-determined matrix M₂ to define probability of transitioning from one driving style to another at each sample, wherein M₂ is a Markov matrix.

9. The method of claim 8, wherein there is no change in driving model expected:

wherein M₁ is an identity matrix; and

wherein M₂ is an identity matrix.

10. A method of determining the most probable statistical model of driving style of a vehicle observed from time step k to present, said method comprising:

(a) establishing at least two statistical models of driving style, wherein said at least two statistical models are conditional probability models that describe probability of conditions at a next time step k+1, given conditions at said time step k;

(b) determining initial probability distribution of said at least two statistical models of driving style,

 

wherein said (b) determining the initial probability distribution of said at least two statistical models comprises setting vector Π

₀ as the initial probability distribution, wherein said vector Π

₀ is pre-determined for each of said at least two statistical models;

(c) determining probability of an observed transition being predicted by said at least two statistical models of driving style;

(d) calculating probability of said at least two statistical models explaining driving style of said vehicle observed at said next time step k+1; and

wherein said (c) determining the probability of said observed transition and said (d) calculating probability of said at least two statistical models are iterated for all time steps until present; and

(e) determining which of said at least two statistical models has the highest probability of explaining the driving style observed.

11. The method of claim 10, wherein said statistical model of driving style is a conditional probability model that describes the probability P at a time step k+1, of having a vehicle velocity V_k+1, given a vehicle velocity V_k at said time step k.

12. The method of claim 10:

wherein said statistical model of driving style is formed from a sample of driving that represents said driving style; and

wherein said sample comprises velocity of said vehicle over time.

13. The method of claim 10, wherein said statistical model of driving style is a conditional probability model that describes the probability P at a time step k+1, of a vehicle with specific velocity, torque, acceleration, road grade, physical location, vehicle load, passenger count and seating, transmission gear, driver identification, key, or special vehicle mode, given the respective factors at said time step k.

14. The method of claim 10, wherein there are N statistical models of driving style:

wherein each element of said vector Π

₀ set to 1/N, whereby each statistical model has an equal probability of matching driving style of said vehicle observed.

15. The method of claim 10, wherein said (c) determining the probability of the observed transition being predicted by each statistical model of driving style comprises:

employing a reference table look up, wherein said table look-up is a two dimensional matrix developed for each said statistical model, wherein there are N statistical models of driving style, wherein said table look-up has three independent indices;

an index identifying said statistical model, current velocity and previous velocity of said vehicle observed; and

employing a pre-determined matrix M₁ to define the probability of transitioning from one driving style to another at each sample, wherein M₁ is a Markov matrix.

16. The method of claim 15, wherein said (c) calculating the probability of said statistical model explaining driving style of said vehicle observed at time step k+1 comprises:

employing a pre-determined matrix M₂ to define the probability of transitioning from one driving style to another at each sample, wherein M₂ is a Markov matrix.