Road grade auto-mapping

• US 8,965,597 B2
• Filed: 02/26/2013
• Issued: 02/24/2015
• Est. Priority Date: 02/26/2013
• Status: Expired due to Fees

• Current Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Original Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Inventors: Filev, Dimitar P., Michelini, John O., Szwabowski, Steven J., Yanakiev, Diana, Abou-Nasr, I, Mahmoud
• Primary Examiner(s): TISSOT, ADAM D
• Application Number:
  US13/777,100
• Publication Number:
  US 20140244067A1
• Time in Patent Office: 728 Days
• Field of Search: 701/409, 701/1, 701/101, 701/102, 701/532, 701/65, 701/80, 701/93, 731/46, 180/170, 340/438, 340/995.18, 340/995.24
• US Class Current: 701/1
• CPC Class Codes: B60W 10/00 Conjoint control of vehicle sub-units of different type or different function for propulsion of purely electrically-propelled vehicles with power supplied within the vehicle B60L50/00 - B60L58/00
• CPC Class Codes: B60W 2552/15 Road slope, i.e. the inclination of a road segment in the longitudinal direction
• CPC Class Codes: G01C 21/3822 Road feature data, e.g. slope data
• CPC Class Codes: G01C 21/3826 Terrain data
• CPC Class Codes: G06F 17/00 Digital computing or data processing equipment or methods, specially adapted for specific functions information retrieval, database structures or file system structures therefor G06F16/00
• CPC Class Codes: G06F 17/16 Matrix or vector computation , e.g. matrix-matrix or matrix-vector multiplication, matrix factorization matrix transposition G06F7/78

Abstract

Road grade is modeled over a region in which a vehicle is driven on roadways having a grade that varies over a plurality of predetermined grade ranges. A succession of grade values are generated while operating the vehicle at a predetermined rate, wherein each grade value identifies a respective grade range then being encountered. A Markov chain road-grade model is updated in response to the succession of grade values, wherein the model represents respective elements of probability in a matrix of transition events from each predetermined grade range to a respective next-in-succession grade range. Each element of the matrix has a value π_i,j representing a weighted frequency of transition events from a first respective grade value to a second respective grade value divided by a weighted frequency of transition events initiating from the first respective grade value, so that the matrix successively approximates the road grade of the region.

Claims

1. A method of modeling a road characteristic over a region in which a host vehicle is driven, comprising the steps of:

operating the host vehicle on roadways having a characteristic that varies over a plurality of predetermined ranges;

generating a succession of characteristic values while operating the host vehicle at a predetermined rate, wherein each characteristic value identifies a respective predetermined range then being encountered by the vehicle;

updating a Markov chain road-characteristic model stored in an optimizer controller in response to the succession of characteristic values, wherein the model represents respective elements of probability in a matrix of transition events from each predetermined range to a respective next-in-succession range, wherein the optimizer controller stores a current state and a previous state of the matrix;

periodically testing a convergence of the current state of the matrix with the previous state of the matrix using KL divergence;

if convergence is found then using the current state of the matrix to identify a powertrain control policy optimized for the characterized road characteristic; and

adjusting operation of a powertrain of the host vehicle using the identified control policy within a powertrain controller in the host vehicle;

wherein each element of the matrix has a value π

_i,j representing a weighted frequency of transition events from a first respective characteristic value to a second respective characteristic value divided by a weighted frequency of transition events initiating from the first respective characteristic value, so that the matrix successively approximates the road characteristic of the region.

2. The method of claim 1 wherein the powertrain control policy is identified by the steps of:

comparing the converged matrix with a plurality of pre-learned reference matrices to search for a matching reference matrix; and

choosing a powertrain control policy corresponding to the matching reference matrix, wherein the chosen powertrain control policy has been optimized according to the road characteristic characterized by the matching reference matrix.

3. The method of claim 2 wherein if no matching reference matrix is found then:

storing the converged matrix as a new reference matrix by appending with the pre-learned reference matrices;

determining a new control policy optimized to the road characteristic which is characterized by the new reference matrix; and

adjusting operation of the powertrain of the host vehicle using the new control policy.

4. The method of claim 1 wherein each element of the transition probability matrix is estimated using a relationship:

5. The method of claim 4 wherein the matrix is recursively updated as a probability distribution:

P(k)=diag(F₀(k))<sup>−

1</sup>F(k)

and wherein f_i,j(k) is replaced by its vector-counterpart τ

(k)γ

(k)^T and f_i(k) is replaced by its vector-counterpart τ

(k)γ

(k)^T e.

6. The method of claim 4 wherein φ

is a constant value forgetting factor such that 0<

φ

≦

1.

7. The method of claim 4 wherein φ

has a value determined in response to 1/k for at least some of the updating.

8. The method of claim 1 further comprising the steps of:

continuing to periodically test convergence of updated states of the matrix with previous states of the matrix; and

if convergence is not found then discontinuing use of the identified powertrain control policy and waiting until convergence is again found in order to identify a respective powertrain control policy.

9. The method of claim 1 wherein the road characteristic is a road grade, and wherein the predetermined ranges are ranges of road grade.

10. Apparatus for a host vehicle that is driven in a region over roadways exhibiting a road characteristic that varies over a plurality of predetermined ranges, comprising:

a powertrain controller for adjusting control parameters of powertrain components of the host vehicle;

a road monitor generating a succession of characteristic values while operating the host vehicle at a predetermined rate, wherein each characteristic value identifies a respective range then being encountered by the host vehicle; and

an optimizer including a Markov chain road-characteristic model that is updated in response to the succession of characteristic values, wherein the model represents respective elements of probability in a matrix of transition events from each predetermined range to a respective next-in-succession range, wherein the optimizer stores a current state and a previous state of the matrix, and wherein each element of the matrix has a value π

_i,j representing a weighted frequency of transition events from a first respective characteristic value to a second respective characteristic value divided by a weighted frequency of transition events initiating from the first respective characteristic value, so that the matrix successively approximates the road characteristic of the region;

wherein the optimizer periodically tests a convergence of the current state of the matrix with the previous state of the matrix, wherein if convergence is found then the optimizer uses the matrix to identify a powertrain control policy optimized for the characterized road characteristic, and wherein the powertrain controller adjusts the control parameters using the identified control policy.

11. The apparatus of claim 10 further comprising:

a library of learned reference matrices and corresponding control policies;

wherein the optimizer compares the converged matrix with the reference matrices to search for a matching reference matrix and chooses a control policy corresponding to the matching reference matrix, wherein the chosen control policy has been optimized according to the road characteristic that is characterized by the matching reference matrix.

12. The apparatus of claim 11 wherein if no matching reference matrix is found then the optimizer stores the converged matrix in the library as a new reference matrix, determines a new control policy optimized to the road characteristic which is characterized by the new reference matrix, and adjusts the control parameters using the new control policy.

13. The apparatus of claim 10 wherein each element of probability satisfies a relationship:

14. The apparatus of claim 13 wherein the matrix is recursively updated as a probability distribution:

P(k)=diag(F₀(k))<sup>−

1</sup>F(k)

and wherein f_i,j(k) is replaced by τ

(k)γ

(k)^T and f_i(k) is replaced by τ

(k)γ

(k)^T e.

15. The apparatus of claim 13 wherein φ

is a constant value such that 0<

φ

≦

1.

16. The apparatus of claim 13 wherein φ

has a value determined in response to 1/k for at least some of the updating.

17. The apparatus of claim 10 wherein the optimizer continues to periodically test convergence of updated states of the matrix with previous states of the matrix, and if convergence is not found then discontinuing use of the identified powertrain control policy and waiting until convergence is again found in order to identify a respective powertrain control policy.

18. The apparatus of claim 10 wherein the road characteristic is a road grade, and wherein the predetermined ranges are ranges of road grade.