Apparatus for deciding a shift pattern suitable for a driver's driving habit using neural network operation and fuzzy inference and a control method thereof

• US 6,078,857 A
• Filed: 12/23/1997
• Issued: 06/20/2000
• Est. Priority Date: 10/19/1995
• Status: Expired due to Fees

• Current Assignee: Hyundai Motor Company (Hyundai Motor Group)
• Original Assignee: Hyundai Motor Company (Hyundai Motor Group)
• Inventors: Kang, Hun, Jung, Soo-Yong
• Primary Examiner(s): Hafiz, Tariq R.
• Assistant Examiner(s): Brown, Edward G.
• Application Number:
  US08/996,700
• Time in Patent Office: 910 Days
• Field of Search: 706/20, 706/25, 701/48
• US Class Current: 701/59
• CPC Class Codes: B60W 2050/0057 Frequency analysis, spectral techniques or transforms
• CPC Class Codes: F16H 2061/0081 Fuzzy logic
• CPC Class Codes: F16H 2061/0084 Neural networks
• CPC Class Codes: F16H 2061/0227 Shift map selection, i.e. methods for controlling selection between different shift maps, e.g. to initiate switch to a map for up-hill driving
• CPC Class Codes: F16H 59/18 dependent on the position of the accelerator pedal
• CPC Class Codes: F16H 59/54 dependent on signals from the brakes, e.g. parking brakes
• CPC Class Codes: F16H 59/58 dependent on signals from the steering
• CPC Class Codes: F16H 61/0213 characterised by the method for generating shift signals
• CPC Class Codes: G05B 13/0285 using neural networks and fuzzy logic

Abstract

An apparatus for deciding a shift pattern suitable for a driver'"'"'s habit using a neural network operation and fuzzy inference and a control method thereof which perform a neural network operation by inputting a driver'"'"'s driving operation quantity as a deciding condition of a shift pattern, and decide an optimal shift pattern by performing fuzzy inference from the output from the a neural network operation.

Claims

1. A shift pattern deciding apparatus, comprising:

a driving operation quantity sensor for sensing a driver'"'"'s driving operation quantity and outputting a driving operation quantity vector X[k];

a moving-average calculator for receiving the vector X[k] and calculating moving-average vector M[k];

a network operator for receiving the vector X[k], performing a neural network operation, and outputting a network output vector NET, wherein the network operator outputs the network output vector NET by multiplying the driving operation quantity vector X[k] by an update weighting coefficient matrix W;

a neural network learning algorithm unit for receiving an error vector E, wherein the error vector E corresponds to the vector M[k] minus the vector NET, and wherein the neural network learning algorithm performs neural network learning and feeds back a learned result to the network operator, and wherein the neural network learning algorithm unit obtains an update weighting coefficient matrix dW by multiplying the error vector E by a learning rate η

, and adds the update weighting coefficient matrix dW to the weighting coefficient matrix W by feeding back the update weighting coefficient matrix dW to the network operator; and

a fuzzy inference unit for receiving the network output vector NET, performing fuzzy inference, and deciding an optimal shift pattern suitable for a driving habit of the driver, and

wherein the moving-average vector M[k] is obtained by the following expression, using the driving operation quantity vector X[k] and an n delay driving operation quantity vector X[k-n] which the driving operation vector X[k] is delayed by n;

space="preserve" listing-type="equation">M[k]=M[k-1]+1/n(X [k]-X[k-n]).

2. The shift pattern deciding apparatus of claim 1, wherein the driving operation quantity vector includes accelerator operation quantity x1[k], brake operation quantity x2[k], and steering wheel operation quantity x3[k].

3. The shift pattern deciding apparatus of claim 2, wherein the accelerator operation quantity x1[k] is defined by a sum of throttle opening and a throttle opening speed;

the brake operation quantity x2[k] is defined by a vehicle deceleration according to the brake operation; and

the steering wheel operation quantity x3[k] is defined by an angle of the steering wheel and angular velocity of the steering wheel.

4. The shift pattern deciding apparatus of claim 1, wherein the moving-average calculator is formed by a ring buffer.

5. A shift pattern deciding apparatus, comprising:

a driving operation quantity sensor for sensing a driver'"'"'s driving operation quantity and outputting driving operation quantity vector X[k];

a moving-average calculator for receiving the vector X[k] and calculating a moving-average vector M[k];

a network operator for receiving the vector X[k], performing a neural network operation, and outputting a network output vector NET, and wherein the network operator outputs the network output vector NET by multiplying the driving operation quantity vector X[k] by an update weighting coefficient matrix W;

a threshold logic unit for receiving the network output vector NET, and outputting logic vector Y of 0 or 1 after comparing the inputted network output vector NET with a predetermined threshold value;

a neural network learning algorithm unit for receiving an error vector E, wherein the error vector E corresponds to the vector M[k] minus the vector Y, wherein the neural network learning algorithm unit performs a neural network learning and feeds back a learned result to the network operator, and wherein the neural network learning algorithm unit obtains an update weighting coefficient matrix dW by multiplying the error vector E by a learning rate η

, and adds the update weighting coefficient matrix dW to the weighting coefficient matrix W by feeding back the update weighting coefficient matrix dW to the neural network operator; and

a fuzzy inference unit for receiving the vector Y, performing fuzzy inference, and deciding an optimal shift pattern suitable for a driving habit of the driver, and

wherein the moving-average vector M[k] is obtained by the following expressions using the driving operation quantity vector X[k] and an n delay driving operation quantity vector X[k-n] which the driving operation X[k] is delayed by n;

space="preserve" listing-type="equation">M[k]=M[k-1]+1/n(X[k]-X[k-n]).

6. The shift pattern deciding apparatus of claim 5, wherein the driving operation quantity vector includes accelerator operation quantity x1[k], brake operation quantity x2[k], and steering wheel operation quantity x3[k].

7. The shift pattern deciding apparatus of claim 6, wherein the accelerator operation quantity x1[k] is defined by a sum of a throttle opening and a throttle opening speed;

the brake operation quantity x2[k] is defined by a vehicle deceleration according to the brake operation; and

the steering wheel operation quantity x3[k] is defined by an angle of the steering wheel and angular velocity of the steering wheel.

8. The shift pattern deciding apparatus of claim 5, wherein the moving-average calculator is formed by a ring buffer.

9. A method for deciding a shift pattern, comprising the steps of:

sensing a driver'"'"'s driving operation quantity and outputting a driving operation quantity vector X[k];

receiving the vector X[k] and calculating a moving-average vector M[k];

outputting the network output vector NET by multiplying the driving operation quantity vector X[k] by an update weighting coefficient matrix W;

outputting logic vector Y of 0 or 1 after comparing the inputted network output vector NET with a predetermined threshold value;

obtaining an error vector E which subtracts the vector Y from the vector M[k], and obtaining an update weighting coefficient matrix dW by multiplying the error vector E by a learning rate η

;

modifying the weighting coefficient matrix by adding the update weighting coefficient matrix dW to the weighting coefficient matrix W; and

receiving the logic vector Y, performing fuzzy inference, and deciding an optimal shift pattern suitable for a driving habit of the driver.

10. The method for deciding a shift pattern of claim 9, wherein the moving-average vector M[k] is obtained by the following expression using the driving operation quantity x[k] and n delay driving operation quantity vector X[k-n] which the driving operation vector x[k] is delayed by n:

##EQU9##

11. The method for deciding a shift pattern of claim 10, wherein the driving operation quantity vector includes accelerator operation quantity x1[k], brake operation quantity x2[k], and steering wheel operation quantity x3[k].

12. The method for deciding a shift pattern of claim 11, wherein the accelerator operation quantity x1[k] is defined by a sum of a throttle opening and a throttle opening speed;

the brake operation quantity x2[k] is defined by a vehicle deceleration according to the brake operation; and

the steering wheel operation quantity x3[k] is defined by an angle of the steering wheel and angular velocity of the steering wheel.