Adaptive filter model for motor vehicle sensor signals

• US 7,158,866 B2
• Filed: 04/12/2001
• Issued: 01/02/2007
• Est. Priority Date: 04/12/2000
• Status: Expired due to Term

• Current Assignee: Nira Dynamics AB (Volkswagen AG)
• Original Assignee: Nira Dynamics AB (Volkswagen AG)
• Inventors: Gustafsson, Fredrik, Forssell, Urban, Ahlqvist, Stefan, Persson, Niclas, Drevo, Marcus
• Primary Examiner(s): Nguyen; Thu V.
• Application Number:
  US10/257,596
• Publication Number:
  US 20040199300A1
• Time in Patent Office: 2,091 Days
• Field of Search: 701/1, 701/36, 701/45, 701/48, 701/58, 701/82, 701 71- 72, 701/207, 701216-217, 701220-221
• US Class Current: 701/1
• CPC Class Codes: B60C 23/061 by monitoring wheel speed measuring distance traversed on the ground by vehicles G01C22/00
• CPC Class Codes: B60C 23/062 Frequency spectrum analysis of wheel speed signals, e.g. using Fourier transformation
• CPC Class Codes: B60T 2210/13 Aquaplaning, hydroplaning
• CPC Class Codes: B60T 2240/03 Tire sensors
• CPC Class Codes: B60T 2250/06 Sensor zero-point adjustment; Offset compensation
• CPC Class Codes: B60T 8/1706 for single-track vehicles, e.g. motorcycles
• CPC Class Codes: B60T 8/1725 Using tyre sensors, e.g. Sidewall Torsion sensors [SWT] for tyre pressure and temperature detection B60C23/00
• CPC Class Codes: B60T 8/1755 Brake regulation specially adapted to control the stability of the vehicle, e.g. taking into account yaw rate or transverse acceleration in a curve road vehicle drive control systems for control of driving stability otherwise than by controlling a particular sub-unit B60W30/02

Abstract

A sensor system for combining first and second sensor signals, and generating a physical parameter values dependent on said sensor signals used in autocalibrating sensors improving the performance and quality of existing sensor signals and virtual sensors realizing new sensors by combining and integrating in adaptive filter models sensor signals representing same or different types of physical parameters.

Claims

1. A method of calculating a physical parameter value (X) of a wheeled vehicle indirectly using at least two directly measurable input signals, comprising the steps of:

(a) receiving a first input sensor signal (y1(t)) that is dependent on a first physical input parameter affecting a condition of the wheeled vehicle;

(b) receiving a second input sensor signal (y2(t)) that is dependent on a second physical input parameter affecting a condition of the wheeled vehicle, wherein the second physical input parameter can be any one of the same as physical parameter value (X), the same as the first physical input parameter, and different from the first physical input parameter;

(c) recursively filtering the first and second physical input parameters by means of an adaptive filter based on a predetermined model of the condition of the wheeled vehicle, wherein the input signals to said model include the first physical input parameter (y1(t)), the second physical input parameter (y2(t)), a first offset scaling factor (c1(t)) for the first sensor signal, and a second offset scaling factor (c2(t)) for the second sensor signal, and wherein the model output signals include a first offset error signal (b1(t)) and a second offset error signal (b2(t)), and the physical parameter value (X);

(d) estimating the values of said first offset error (b1(t)) and the second offset error (b2(t)) by means of said adaptive filter, wherein at least one offset error relates to the wheel radius; and

(e) calculating the physical parameter value (X) using the adaptive filter from the input sensor signals y1(t) and y2(t) such that offset errors in the calculation are compensated for by means of the first and second offset error parameters (b1(t)) and (b2(t)) and the offset scaling factors c1(t) and c2(t).

2. The method of claim 1, wherein the model is based on relations between measured values (y(t)) of said physical parameter (x(t)) and offsets (b1,b2) for said first and second sensor signals corresponding to the algebraic expression

y1(t)=x(t)+c1(t)b1

y2(t)=x(t)+c2(t)b2,

where y1(t) is a measurement of said physical parameter x(t) detected by means of a first sensor and represented by said first sensor signal having said first offset error b1 and a first offset scaling c1(t) according to a predetermined function of time;

and

where y2(t) is a measurement of said physical parameter x(t) detected by means of a second sensor and represented by said second sensor signal having said second offset error b2 and a second offset scaling c2(t) according to a predetermined function of time.

3. The method of claim 1, wherein the model is based on relations between measured values (yi(t)) of said physical parameter (x(t)) and offsets (bi(t)) for a number i of sensor signals corresponding to the algebraic expression

yi(t)=x(t)+ci(t)bi(t),

where yi(t) is a measurement of said physical parameter X(t) detected by means of a sensor number i and represented by a sensor signal having an offset error bi and an offset scaling ci(t) according to a predetermined function of time.

4. The method of claim 1, further comprising the steps:

evaluating the observability of said physical parameter; and

updating offset measurements if a predetermined observability criterion met.

5. The method of claim 4, further comprising the step of using external information in the observability evaluation.

6. The method of claim 4, further comprising the steps of:

initializing the sensor and filter system;

receiving as an input the next measurement or sample of said sensor signals;

checking the observability of said physical parameter;

if observability is met, then updating offset estimates;

after updating of offset estimates or if observability is not met, then compensate sensor signal with the offset output from the adaptive filter.

7. The method of claim 1, further comprising the steps of receiving a third sensor signal dependent on a third physical parameter affecting the condition of the wheeled vehicle and recursively filtering said third sensor signal together with said first and second sensor signals by means of said adaptive filter, which is based on a model that also is dependent on said third physical parameter and on the offset error for said third sensor signal.

8. The method of claim 1, wherein said filter is based on a model that also is dependent on measurement noise in the sensor signals respectively and further comprises the step of calculating and eliminating the measurement noise of the calculated physical parameter value signal.

9. The method of claim 1, wherein said filter is based on a recursive least means square algorithm (RLS).

10. The method of claim 1, wherein said filter is based on a least means square algorithm (LMS).

11. The method of claim 1, wherein said filter is based on a Kalman filter.

12. The method of claim 11, wherein said Kalman filter is specified by means of a state space equation of the form:

x(t+1)=Ax(t)+Bv(t)

y(t)=Cx(t)+e(t)

where the covariance matrices of v(t) and e(t) are denoted Q and R, respectively, and wherein the unknown quantities in a state vector x(t) are estimated by a recursion;

{circumflex over (x)}(t+1)=A{circumflex over (x)}(t)+K(t;

A,B,C,Q,R)(v(i)−

C{circumflex over (x)}(t))

where the filter gain K(t;

A,B,C,Q,R) is given by predetermined Kalman filter equations.

13. The method of claim 12, wherein said Kalman filter incorporates a model for the variation of the true physical parameter value of the wheeled vehicle.

14. The method of claim 1, further comprising the steps of:

examining the data quality according to predetermined rules;

producing statistical matrices for Kalman filtering dependent on said rules.

15. The method of claim 14, wherein said rules are:

if (velocity estimate <

LOW_LEVEL) (then increase values in Kalman measurement covariance matrix R and

if (velocity estimate=ABSOLUTE_ZERO) (then decrease values in Kalman measurement covariance matrix R.

16. The method according to claim 1, further comprising applying Kalman filter equations by means of the steps of:

time updating the Kalman filter

xhat=F*xhat;

Phat=F*Phat*F′

+G*Q*G′

;

measurement updating the Kalman filter

K=Phat*H′

*inv(H*Phat*H′

+R);

e=y−

H*xhat;

xhat=xhat+K*e;

Phat=Phat−

K*H*Phat′

;

Phat=0.5*(Phat+Phat′

);

where xhat is the current state estimate and Phat is the current Kalman error covariance matrix, y is the current measurement, H is measurement matrix, F is state space model update matrix and G is noise update matrix.

17. The method of claim 1, further comprising pre-processing raw sensor data by scaling sample sensor signals to physical constants according to the model such that each physical constant is set equal to a nominal scale factor multiplied with the raw sensor data value minus a nominal sensor offset.

18. The method of claim 17, further comprising pre-processing raw sensor data by low-pass filtering sensor signal values in order to reduce quantisation errors and noise errors.

19. The method of claim 1, further comprising pre-processing raw sensor data by rotationally synchronising clock stamps from cog-wheels by calculating wheel angular velocity to avoid cog deformity error effects.

20. The method of claim 1, further comprising the step of providing an offset estimate to said sensor signals for use in diagnosis functions.

21. The method of claim 1, further being adapted for determining the yaw rate of said wheeled vehicle,

wherein said first sensor signal is a yaw rate signal from a yaw rate gyro mounted in the wheeled vehicle;

wherein said second sensor signals are angular wheel velocity signals from angular wheel velocity sensors mounted in the wheeled vehicle for sensing the angular velocity of the wheels respectively;

wherein said model is dependent on said yaw rate signal, said angular wheel velocity signals, the offset error for the yaw rate signal and the offset errors for the angular wheel velocity signals; and

wherein said delivered physical parameter value signal is a calculated yaw angle and a calculated yaw rate signal.

22. The method of claim 21, further comprising the steps of

receiving a third sensor signal dependent on a third physical parameter affecting the condition of the wheeled vehicle and recursively filtering said third sensor signal together with said first and second sensor signals by means of said adaptive filter, which is based on a model that also is dependent on said third physical parameter and on the offset error for said third sensor signal;

wherein said third sensor signal is an acceleration signal from an accelerometer mounted in the wheeled vehicle for sensing a lateral acceleration and wherein said adaptive filter is based on a model that is also dependent on said lateral acceleration signal and on the offset error for said acceleration signal.

23. The method of claim 21, further comprising the steps of calculating and eliminating measurement noise from the calculated yaw rate signal.

24. The method of claim 21, wherein said Kalman filter incorporates a model for the variation of the true yaw rate of the wheeled vehicle.

25. The method of claim 21, further comprising the steps of calculating offset errors for said angular wheel velocity signals and calculating the relative difference in wheel radii between right and left wheels dependent on said offset errors for wheel velocity signals.

26. The method of claim 21, further comprising calculating filter input signals and parameters in an error model by the steps of:

calculating an inverse curve radius estimate from front axle wheels of the wheeled vehicle;

calculating an inverse curve radius estimate from rear axle wheels of the wheeled vehicle;

calculating vehicle velocity estimate from the angular wheel velocity sensors;

calculating yaw rate estimates from front and rear wheels;

calculating an error propagation function for wheel radius offsets for a rear axle of the wheeled vehicle;

calculating an error propagation function for wheel radius offsets for a front axle of the wheeled vehicle.

27. The method of claim 21, further comprising the step of providing a fast yaw rate which is the current measured low-pass filtered yaw rate minus the estimated offset for use in time critical control systems.

28. The method of claim 21, further comprising the step of providing a filtered yaw rate which is the current yaw rate state estimate of the Kalman filter.

29. The method of claim 21, further comprising the step of providing relative wheel radii between left and right wheels on rear and front wheelage to be used in a tire pressure estimation system.

30. The method of claim 21, further comprising the step of providing a yaw rate gyro offset estimate to be used in a diagnosis function.

31. The method of claim 1, further adapted for determining longitudinal movement of said wheeled vehicle,

wherein said first sensor signal is an acceleration signal from an accelerometer mounted in the wheeled vehicle;

wherein said second sensor signals are angular wheel velocity signals from angular wheel velocity sensors mounted in the wheeled vehicle for sensing the angular velocity of the respective wheels;

wherein said model is dependent on said acceleration signal, said angular wheeled velocity signals, the offset error for the acceleration signal and the offset errors for the angular wheel velocity signals; and

wherein said delivered physical parameter value signal is a calculated longitudinal velocity signal and a calculated longitudinal acceleration signal.

32. The method of claim 31, wherein the adaptive filter is an extended Kalman filter.

33. The method of claim 31, further comprising the steps of:

receiving a third sensor signal dependent on a third physical parameter affecting the condition of the wheeled vehicle and recursively filtering said third sensor signal together with said first and second sensor signals by means of said adaptive filter, which is based on a model that also is dependent on said third physical parameter and on the offset error for said third sensor signal;

wherein said third sensor signal is a yaw rate signal for the wheeled vehicle and wherein said adaptive filter is based on a model that is dependent also on said yaw rate signal and on the offset error for said yaw rate sensor signal.

34. The method of claim 31, further comprising calculation of filter input signals and parameters in an error model by the steps of:

calculating a vehicle velocity estimate from angular vehicle velocity sensors;

calculating the parameters in the error model;

calculating current matrices for an extended Kalman filter (F,G,H,Q).

35. The method of claim 31, further comprising the step of providing a velocity estimate for a control system and an MMI (Man Machine Interface).

36. The method of claim 31, further comprising the step of providing wheel slip values.

37. The method of claim 31, further comprising the step of providing an accelerometer offset estimate for use in a diagnosis function.

38. The method of claim 1, further being adapted for determining fuel level and fuel consumption of said wheeled vehicle:

wherein said first sensor signal is a fuel volume signal from a tank meter of the wheeled vehicle;

wherein said second sensor signal is a fuel injection sensor signal from the engine of the wheeled vehicle;

wherein said model is dependent on said fuel volume signal, said fuel injection signal, the offset error for the fuel volume signal and the offset error for the fuel offset signal; and

wherein said delivered physical parameter value signal is a calculated fuel level and a calculated fuel consumption signal.

39. The method of claim 38, further comprising calculation of filter input signals and parameters in an error model by the steps of:

calculating a fuel level estimate from the fuel level sensor signal;

calculating the fuel consumption from the fuel injection signal.

40. The method of claim 38, further comprising the steps of:

examining the data quality according to the predetermined rules;

using statistical matrices for Kalman filtering dependent on said rules.

41. The method of claim 40, wherein said rules for examining the data quality is:

if high load on engine then increase part of values in the Kalman measurement covariance matrix R.

42. The method of claim 38, further comprising the step of providing the fuel level estimate for an MMI system (Man Machine Interface).

43. The method of claim 38, further comprising the step of providing a fuel consumption estimate for a control system and an MMI system (Man Machine Interface).

44. The method of claim 1, wherein analytical redundancy is arranged for the adaptive filter by generating a number of equations to process in said filter.

45. An apparatus for calculating a physical parameter value (X) of a wheeled vehicle indirectly using at least two directly measurable input signals, comprising:

(a) a first input sensor signal (y1(t)) that is dependent on a first physical input parameter affecting a condition of the wheeled vehicle;

(b) a second input sensor signal (y2(t)) that is dependent on a second physical input parameter affecting a condition of the wheeled vehicle, wherein the second physical input parameter can be any one of the same as physical parameter value (X), the same as the first physical input parameter, and different from the first physical input parameter;

(c) means for recursively filtering the first and second physical input parameters by means of an adaptive filter based on a predetermined model of the condition of the wheeled vehicle, wherein the input signals to said model include the first physical input parameter (y1(t)), the second physical input parameter (y2(t)), a first offset scaling factor (c1(t)) for the first sensor signal, and a second offset scaling factor (c2(t)) for the second sensor signal, and wherein the model output signals include a first offset error signal (b1(t)) and a second offset error signal (b2(t)), and the physical parameter value (X);

(d) means for estimating the values of said first offset error (b1(t)) and the second offset error (b2(t)) by means of said adaptive filter, wherein at least one offset error relates to the wheel radius; and

(e) means for calculating the physical parameter value (X) using the adaptive filter from the input sensor signals y1(t) and y2(t) such that offset errors in the calculation are compensated for by means of the first and second offset error parameters (b1(t)) and (b2(t)) and the offset scaling factors c1(t) and c2(t).

46. The apparatus of claim 45, further comprising a processing unit (110,112,114,116,117) devised for carrying out the functions related to said means.

47. A computer program directly loadable into the internal memory of a digital computer, comprising program code for calculating a physical parameter value (X) of a wheeled vehicle indirectly using measurable input signals, the program code comprises sets of instructions for:

(a) receiving a first input sensor signal (y1(t)) that is dependent on a first physical input parameter affecting a condition of the wheeled vehicle;

(b) receiving a second input sensor signal (y2(t)) that is dependent on a second physical input parameter affecting a condition of the wheeled vehicle, wherein the second physical input parameter can be any one of the same as physical parameter value (X), the same as the first physical input parameter, and different from the first physical input parameter;

(c) recursively filtering the first and second physical input parameters by means of an adaptive filter based on a predetermined model of the condition of the wheeled vehicle, wherein the input signals to said model include the first physical input parameter (yl(t)), the second physical input parameter (y2(t)), a first offset scaling factor (c1(t)) for the first sensor signal, and a second offset scaling factor (c2(t)) for the second sensor signal, and wherein the model output signals include a first offset error signal (b1(t)) and a second offset error signal (b2(t)), and the physical parameter value (X);

(d) estimating the values of said first offset error (b1(t)) and the second offset error (b2(t)) by means of said adaptive filter, wherein at least one offset error relates to the wheel radius; and

(e) calculating the physical parameter value (X) using the adaptive filter from the input sensor signals y1(t) and y2(t) such that offset errors in the calculation are compensated for by means of the first and second offset error parameters (b1(t)) and (b2(t)) and the offset scaling factors c1(t) and c2(t).

48. A method of calculating a physical parameter value (X) of a wheeled vehicle indirectly by recursively filtering at least two measurable input sensor signals, comprising the steps of:

processing a plurality of input sensor signals (y1(t),y2(t)) that are dependent on first and second physical input parameters respectively that affect a condition of the wheeled vehicle;

recursively filtering the first and second physical input parameters by means of an adaptive filter based on a predetermined model of the condition of the wheeled vehicle, wherein the input signals to said model include the first physical input parameter (y1(t)), the second physical input parameter (y2(t)), a first offset scaling factor (c1(t)) for the first sensor signal, and a second offset scaling factor (c2(t)) for the second sensor signal, and wherein the model output signals include a first offset error signal (b1(t)) and a second offset error signal (b2(t)), and the physical parameter value (X);

estimating the values of said first offset error (b1(t)) and said second offset error (b2(t)) by means of recursively filtering with an adaptive filter, wherein at least one offset error relates to the wheel radius;

calculating one or more physical parameter value (X) signals using the adaptive filter; and

using the first and second offset errors (b1(t),b2(t)) to adjust the physical parameter values (X) to a value that is more representative of its actual value.