Method for determining a state of motion of a vehicle

Abstract

A method for determining the state of motion of a vehicle (10) by means of an observer (23, 100) which is designed as a non-linear Luenberger observer with a non-linear vehicle model, wherein the observer (23, 100) has an input vector (u(t)), an observer output vector (ŷ(t)), a measurement output vector (y(t)), an observer state vector (x̂ Obs (t)), a model state vector (x̂ Mdl ) of the non-linear vehicle model and a model output vector (ỹ Mdl ) is described.

Description

[0001] The invention relates to a method for determining a state of motion of a vehicle.

[0002] EP 1 452 353 A2 shows a method for determining a state of motion of a vehicle.

[0003] US 2007 / 0 067 085 A1 shows a method for determining a state of motion of a vehicle.

[0004] US 2017 / 0 247 038 A1 shows a method for determining a state of motion of a vehicle.

[0005] CN 1 16 691 677 A shows a method for determining a state of motion of a vehicle.

[0006] US 2021 / 0 139 028 A1 shows a system for identifying a previous longitudinal velocity and for receiving data from an inertial measuring unit, wherein the system can determine a roll rate and a pitch rate.

[0007] DE 11 2014 001 807 B4 shows a system for use in a vehicle for estimating a vehicle tilt angle and a road gradient angle in real time, with a sensor configured to measure a vehicle tilt rate.

[0008] DE 11 2014 001 809 B4 shows a system for use in a vehicle for estimating a vehicle roll angle and a road bank angle in real time, with a sensor designed to measure a vehicle roll rate.

[0009] It is therefore an object of the invention to provide a new method for determining a state of motion of a vehicle.

[0010] This problem is solved by the subject matter of claim 1.

[0011] A method for determining the motion state of a vehicle by means of an observer which is designed as a non-linear Luenberger observer with a non-linear vehicle model, wherein the observer has an input vector, an observer output vector, a measurement output vector, an observer state vector, a model state vector of the non-linear vehicle model and a model output vector, the method comprising the following steps: A) First data from at least one first sensor are received, wherein the first data comprises the following first information: - measured yaw rate, - measured pitch rate, - measured roll rate, - measured longitudinal acceleration, - measured lateral acceleration, - measured vertical acceleration, - speed values ​​or rotational speed values ​​of at least two wheels, and - drive torque; B) depending on the first data, the input vector for the observer is generated, whereby the input vector has the following second information: - Longitudinal acceleration, - lateral acceleration, - global pitch rate, - global roll rate, and - Rate of change of a global friction coefficient; C) a time derivative of the current observer state vector is determined depending on: - the input vector, - a matrix-vector product of a Coriolis matrix with the current observer state vector, - a gravitational force matrix, and - a product of a Jacobian matrix, a state feedback gain matrix, a state gain matrix and a total control difference, wherein the total control difference is composed of a first control difference between the observer output vector and the corresponding values ​​of the measurement output vector and a second control difference between the model output vector of the non-linear vehicle model and the corresponding values ​​of the measurement output vector; D) the current observer state vector is updated in an integrator by integrating the calculated time derivative of the current observer state vector; E) an updated observer output vector is calculated depending on an output matrix and the current observer state vector; and F) A model output vector is determined depending on a model output matrix of a non-linear vehicle model and a model state vector.

[0012] The method enables very accurate determination of the observer state vector with high observer quality. Furthermore, the computational effort is significantly lower than with an extended Kalman filter.

[0013] According to a preferred embodiment, when generating the input vector, at least one correction is carried out depending on the first data from a correction group consisting of: - compensation of an offset error due to a temperature change at the at least one first sensor, and - Compensation of an offset error due to an elastokinematic behavior of the longitudinal system.

[0014] These compensations have led to a significant improvement in the observer.

[0015] According to a preferred embodiment, when generating the input vector depending on the first data, at least one adaptation is carried out from an adaptation group consisting of: - Transformation of longitudinal acceleration, lateral acceleration and vertical acceleration to the corresponding movement of the center of gravity, - Euler transformation of yaw rate, pitch rate and roll rate, - Transformation of the kinematic steering wheel angle into an elastokinematic wheel angle, and - Transforming the peripheral speed of the wheels into the center of gravity. These adjustments are beneficial for the quality of the observer.

[0016] According to a preferred embodiment, the first data comprises at least one third piece of information from an information group consisting of: - direction of rotation of the wheels, - braking torque, - brake pressure, - steering wheel angle, - Steering angle of a rear wheel steering, - degree of locking of a limited-slip differential, - Ride height (ground clearance), - Wheel coupling accelerations, and - Status information of an active damping system, in particular active damping forces.

[0017] This information allows for improvement of the model and the observer.

[0018] According to a preferred embodiment, the observer state vector comprises at least a fourth piece of information from a state information group consisting of: - longitudinal speed of the vehicle, - lateral speed, - Float angle at the center of gravity, - road gradient angle, - Road bank angle, - Friction factor relative to the center of gravity, - Vehicle roll angle, - Vehicle pitch angle, - longitudinal slip, and - Slip angle.

[0019] This information can be used to control a variety of vehicle systems.

[0020] According to a preferred embodiment, the measurement output vector comprises at least a fifth piece of information from a measurement output vector information group consisting of: - Reference longitudinal speed, - Longitudinal acceleration, and - Lateral acceleration.

[0021] According to a preferred embodiment, the gravitational force matrix includes roll angle and pitch angle compensation with respect to the effect of the vehicle's roll angle and pitch angle. This improves observation in situations with inclined paths.

[0022] The observer state vector is used in at least one vehicle system from a vehicle system group consisting of: - Electronic Stability Program (ESP), - Traction control (TCS), - All-wheel drive (AWD), - Torque vectoring (TV), - Steering assistance, - Trajectory planning in a driver assistance system, - Adaptive cruise control in a driver assistance system, - Control of damper forces in a semi-active or active damper system, - Spring rate control, - Control of the vehicle level, and - Roll stabilization.

[0023] In these vehicle systems, the state parameters generated by the observer can be used to advantage, and a fast and accurate response is possible.

[0024] According to a preferred embodiment, a global friction coefficient is determined by an additional linear observer. This determination has led to a significant improvement in the observer, particularly for a comparatively low friction coefficient.

[0025] According to a preferred embodiment, the state gain matrix is ​​adaptively selected depending on the detection of longitudinal excitation and lateral excitation. In critical driving situations, such as anti-lock braking system deployment, the observer functions better with this measure.

[0026] According to a preferred embodiment, a two-track vehicle model is used as the non-linear vehicle model. Such a vehicle model is precise and harmonizes well with the observer.

[0027] Further details and advantageous developments of the invention will become apparent from the exemplary embodiments described below and illustrated in the drawings, which are in no way to be understood as limiting the invention, as well as from the dependent claims. It is understood that the features mentioned above and those to be explained below can be used not only in the respective combinations specified, but also in other combinations or on their own, without departing from the scope of the present invention. It shows: Fig. 1 in a schematic representation an observer, Fig. 2 in detailed embodiment the observer of Fig. 1, Fig. 3 an input vector u(t), Fig. 4 a measurement output vector y(t), Fig. 5 a state vector x̂ Obs (t) of the observer, Fig. 6 a model state vector x̂Mdl , Fig. 7 a vector F V , Fig. 8 a Coriolis matrix M C , Fig. 9 a gravitational force matrix F G , Fig. 10 a calculation of an observer output vector ŷ Obs (t), Fig. 11 an output matrix C Obs, Fig. 12 a calculation of a model output vector ŷ Mdl , Fig. 13 a model output matrix Ĉ Mdl , Fig. 14 a first variant of a state gain matrix L µ (t), Fig. 15 a second variant of a state gain matrix L(t), Fig. 16 a Jacobian matrix, Fig. 17 a matrix with normalization factors, Fig. 18 and Fig. 19 Jacobi terms, Fig. 20 and Fig. 21 normalization factors, Fig. 22 Investigations of a road gradient, Fig. 23 Investigations of a road slope, Fig. 24 Errors of a pitch angle determination using different methods, and Fig. 25 a vehicle with a control arrangement.

[0028] In the following, identical or functionally identical parts are provided with the same reference symbols and are usually described only once. The description builds on each figure to avoid unnecessary repetition.

[0029] Fig. Figure 1 shows the basic, simplified structure of an observer. Such an observer, in its linear form, is also called a Luenberger observer. In this case, however, it is a nonlinear observer.

[0030] The upper part represents the real world with an observable system. An input vector u(t) has measurable input variables of the system, and the system produces an output vector y(t) with measurable output variables.

[0031] Typically, not all input variables are measurable. Therefore, an observer 100 is used, which is configured to generate an internal observer state vector x̂(t) such that the observer's behavior corresponds to the behavior of the real system above, and with the observer, an observer output vector ŷ(t) with the corresponding observer output variables is created.

[0032] With the observer state variables of the observer state vector x̂(t), other systems of the vehicle can then be advantageously operated.

[0033] The basic equation of the observer is in the simplified example x^=A x^(t)+B u(t)+ub(t) with ub(t)=L(y(t)−y^(t))

[0034] It is a controller and the feedback is provided by the difference between the measured output vector y(t) and the observer output vector ŷ(t).

[0035] The matrix L is called the Luenberger matrix or observer gain matrix.

[0036] As long as the observer functions, the model corresponds to the real system. The observer is described in more detail below.

[0037] Fig. Figure 2 shows a more detailed representation of the observer 100.

[0038] The basic equation of the observer, also called the main observer scheme, is x^˙(t)=Fν(t−1)⋅u(t)Mc⋅x^(t−1)+FG(t−1)(x^Obs(t−1),x^νert(t−1)) +L(t)⋅ξ(t)⋅Λ(t)⋅[(Yν⋅ymeas(t)−y^Obs(t−1))+(Ya⋅ymeas(t)−y^Mdl(t−1))]

[0039] Here are - x^˙(t) a time derivative of the current observer state vector x̂ Obs (t), - F v (t - 1) a matrix for the appropriate assignment of the input vector u(t), - Mc a Coriolis matrix, - F G a gravitational force matrix, - x vert (t - 1) a state vector that provides information about the observer pitch angle θ̂ B and observer roll angle Φ̂ B of the vehicle and the observer-road gradient angle Θ̂ G and the observer-road bank angle Φ̂ G has, - L(t) a state gain matrix - ξ(t) is a Jacobian matrix, - Λ(t) is a state feedback gain matrix, - Y, a selection matrix for the measured reference speed, - y meas (t) an output vector with measured states, - Y a a selection matrix for measured accelerations, - ŷMdl (t - 1) a vector with initial states of a nonlinear vehicle model.

[0040] The summator 102 of the individual terms produces a value x^˙(t) for the time derivative of the current observer state vector x̂ Obs (t).

[0041] The value x^˙(t) is fed to an integrator 104, which calculates the observer state vector x̂ Obs (t) updated.

[0042] An updated observer output vector ŷ Obs (t) is dependent on an output matrix C Obs of the observer and the current observer state vector x̂ Obs (t) is calculated.

[0043] Depending on a model output matrix Ĉ Mdl a non-linear vehicle model and a model state vector x̂ Mdl becomes a model output vector ŷ Mdl calculated.

[0044] A control difference ŷ is calculated in a comparator 106. Obs between the observer output vector ŷ Obs and the selection matrix Y v selected, measured states y meas (t) is calculated.

[0045] A control difference ỹ is calculated in a comparator 108. Mdl between the output vector ŷ Mdl of the nonlinear vehicle model and the selection matrix Y a selected, measured states y meas (t) is calculated.

[0046] In an adder 109, the control difference ỹ Obs and the control difference ỹ Mdl added to a total control difference ỹ.

[0047] The state feedback gain matrix Λ(t), the Jacobian matrix ξ(t) and the state gain matrix L(t) are applied to the total control difference ỹ, and the result is fed to the adder 102.

[0048] The Matrix F v(t - 1) is fed to the input vector u(t).

[0049] The matrices Mc and F G the current observer state vector x̂ Obs (t) is supplied.

[0050] The gravitational force matrix F G the state vector x̂ vert supplied.

[0051] The state gain matrix L(t) is dependent on a flag µ Flag (t) either independent of an adaptive friction factor µ̂ G or as L µ (t) depends on the adaptive friction factor µ̂ G , as explained below.

[0052] The system is, for example, a vehicle. Measurable input variables include, for example, longitudinal acceleration a x and a roll rate Φ̇ meas Output variables are determined via the observer and the nonlinear vehicle model. These output variables can be compared with corresponding measured variables such as the yaw rate ψ̇meas or the longitudinal velocity v x,ref This allows the variables of the observer state vector to be influenced so that they correspond to the actual variables. This is done via comparators 106, 108, adder 109, and feedback via the adaptive state gain matrix.

[0053] Fig. Figure 3 shows the input vector u(t). In the example, this contains the following information: - Longitudinal acceleration a x,corr , - Lateral acceleration a y,corr , - global pitch rate Θ̇ G , - global roll rate Φ̇ G , and - Rate of change μ^˙G a global friction coefficient.

[0054] The rate of change μ^˙G of the global friction coefficient is intended to characterize the error of the longitudinal and lateral motion dynamics. From the rate of change μ^˙G a value for the global friction factor µ̂ G by numerical integration. A low-pass filter is preferably used to detect a rapid change in the global friction factor µ̂ G to prevent, using a speed-dependent frequency. The preferred friction factor is µ̂ G in a range of values ​​0.05 < µ̂ G < 1.2. The global friction factor µ̂ is preferred. G determined using a separate, linear observer.

[0055] Fig. 4 shows the measurement output vector y meas (t). In the example, this contains the following information: - Reference longitudinal speed v x,ref , - Longitudinal acceleration a x,corr , and - Lateral acceleration a y,corr .

[0056] The second position is zero, since the lateral velocity v y cannot be measured directly.

[0057] Fig. 5 shows the observer state vector x̂ Obs (t). In the example, this contains the following information in the observer: - Longitudinal speed of the vehicle (v̂ x ), - lateral velocity (v̂ y ), - Road gradient angle (Θ̂ G ), - Road bank angle (Φ̂ G ), and - Friction factor (µ̂ G ) with respect to the center of gravity.

[0058] Preferably, additional information is available at least in part: - Yaw rate in the observer (ψ^˙Obs), - Float angle at the center of gravity (β̂ COG ), - Vehicle roll angle (Φ̂), - vehicle pitch angle (Θ̂), - longitudinal slip, and - Slip angle.

[0059] Fig. 6 shows the model state vector x̂ Mdl . In the example, this contains the following information: - Longitudinal acceleration â x from the observer, and - Lateral acceleration â y from the observer.

[0060] Fig. 7 shows the matrix F V , which is designed as an identity matrix and thus passes the input vector along with the input vector u(t). Alternatively, u(t) can also be written directly.

[0061] Fig. 8 shows the Coriolis matrix M C . She has the elements m 21 = ψ̇ meas (measured yaw rate) m 12 = -ψ̇ meas (negative measured yaw rate) The remaining elements have the value 0.

[0062] Fig. 9 shows the gravitational force matrix F G , which has one column. The values ​​lead to a compensation of the offset by the pitch angle Θ̂ B and the roll angle Φ̂ B in relation to the vehicle's center of gravity, the road gradient angle Θ̂G and the road bank angle Φ̂ G . g is the gravitational acceleration.

[0063] Fig. 10 shows the calculation of the observer output vector ŷ Obs (t) from a product of the output matrix C Obs of the observer and the current observer state vector x̂ Obs (t).

[0064] Fig. 11 shows the output matrix C Obs of the observer. In the example, it is a 4x5 matrix. It has the number 1 in element c11 and the number 0 in the rest of the matrix. Thus, the longitudinal velocity v̂ x used.

[0065] Fig. 12 shows the calculation of the model output vector ŷ Mdl from a product of the model output matrix Ĉ Mdl of the non-linear vehicle model and the model state vector x̂ Mdl .

[0066] Fig. 13 shows the model output matrix Ĉ Mdlof the non-linear vehicle model. In the example, it is a 4x4 matrix. It has the number 1 in the elements c 33 and c 44 , incidentally the number 0. Thus, longitudinal acceleration â x and the lateral acceleration â y used.

[0067] Fig. 14 shows the state gain matrix L µ (t) in case the friction observer flag µ Flag (t) is set, and Fig. 15 shows the state gain matrix L(t) for the case that the friction observer flag µ Flag (t) is not set.

[0068] Both matrices have gain factors that depend on the friction observer flag µ Flag (t) differ.

[0069] Both embodiments have in element I 11 the gain factor L vx , in element I 13 the gain factor L ax and in Element I 24the gain factor L vy .

[0070] The state gain matrix L(t) has in element I 31 the gain factor L θ , in element I 33 the gain factor L θ,ax and in Element I 44 the gain factor -L Φ . For the state gain matrix L µ (t) these values ​​differ, and this is indicated by the suffix µ.

[0071] The remaining elements are each 0.

[0072] If the friction observer flag is set due to a longitudinal or lateral excitation, the matrix of Fig. 14 is used. If, however, neither longitudinal nor lateral excitation is present, or only a very small amount of excitation is present, the matrix of Fig. 15. This is also called adaptive gain.

[0073] Lateral excitation can be determined by μ^˙G=ay⋅Λay⋅KμFy⋅(ay,corr−a^y).

[0074] The longitudinal excitation can be determined via μ^˙G=Kνx,ref⋅KμFx⋅(μFxAνg−μ^˙G).

[0075] Here, μFxAνg the averaged pre-estimate based on the longitudinal wheel forces. KμFy and KμFx are observer rewards of the longitudinal and lateral friction fault dynamics. K vx,ref is a confidence factor for the reference speed.

[0076] The differences in the state gain matrices L(t) lead to a better observation result.

[0077] Fig. 16 shows the Jacobian matrix ξ.

[0078] The element x 11 is 1. In element x 33 there is a Jacobian value ξ ax and in element x 44 there is a Jacobian value ξ ay . The remaining elements are 0.

[0079] Fig. 17 shows the matrix Λ.

[0080] In Element I 11 there is a trust factor K vx,ref for the preliminary estimation of the longitudinal velocity v x,ref .

[0081] In the Elements I 33 Normalization factors Λ ax and Λ ay for the Jacobian values.

[0082] Fig. 18 shows the Jacobian value ξax=∂a^x∂ν^x−λx, which results from a partial derivative of the estimated longitudinal acceleration with respect to the estimated longitudinal velocity minus the eigenvalue of the Jacobian matrix with respect to the stabilization of the estimated longitudinal acceleration.

[0083] Fig. 19 shows the Jacobi value ξay=∂a^y∂ν^y−λy, which results from a partial derivative of the estimated lateral acceleration with respect to the estimated lateral velocity minus the eigenvalue of the Jacobian matrix with respect to the stabilization of the estimated lateral acceleration.

[0084] Fig. 20 shows the normalization factor Λax=1a^x2+ξax2

[0085] Fig. 21 shows the normalization factor Λay=1a^y2+ξay2

[0086] The normalization factors stabilize the Jacobian terms.

[0087] Fig. 22 to Fig. Figure 24 shows example measurements with a vehicle driving along an uphill and downhill road with a changing cross slope. The gradient of the road in degrees in Fig. 22 was determined using an extended Kalman filter, a reference measurement unit with differential GPS (ADMA), and the presented observer, and the results are close to each other. This is also the case for the determination of the road's cross slope in degrees in Fig. 23.

[0088] in case of error in pitch angle determination according to Fig. However, significant differences emerge in Figure 24. The error of the proposed observer is represented by line 110, and in many places it is a factor of 5 to 10 smaller than the error of the extended Kalman filter, represented by line 112.

[0089] Further tests have shown that the determination of the road gradient and the transverse slope of the roadway, for example during normal driving in the linear range, is significantly better for the imagined observer than for an extended Kalman filter.

[0090] The differences are particularly evident when using an anti-lock braking system on wet tiles with acceleration and braking. Here, a pitch angle error in the range of 0.0° to 0.25° was achieved with the proposed observer, while the error with an extended Kalman filter was in the range of 0.50° to 1.80°. The error for the determined roll angle in this measurement with the proposed observer was also between 0.0° and 0.22°, while it was between 0.0° and 0.62° with the extended Kalman filter.

[0091] The observer presented has proven to be very robust.

[0092] Fig. 25 shows an example of a vehicle 10 with four wheels 11, 12, 13,14.

[0093] A sensor 22 is provided. The sensor 22 preferably comprises an inertial measurement unit (IMU) that measures the following: - Yaw rate ψ̇ eas (English: yaw rate), - Pitch rate θ̇ meas (English: pitch rate), - Roll rate Φ̇ meas (English: roll rate), - Longitudinal acceleration a x,meas (English: longitudinal acceleration), - Lateral acceleration a y,meas (English: lateral acceleration), - Vertical acceleration a z,meas (English: vertical acceleration)

[0094] The sensor 22 preferably additionally has information about - speed values ​​or values ​​of the rotational speeds of at least two wheels, usually available for an electronic stability program, and - Drive torque, usually provided by the drive.

[0095] The sensor data are transmitted to a control unit 24, which executes the observer 23. The data of the observer state vector x̂ Obs (t) are used in at least one vehicle system 26 from a vehicle system group consisting of: - Electronic Stability Program (ESP), - Traction control (TCS), - All-wheel drive (AWD), - Torque vectoring (TV), - Steering assistance, - Trajectory planning in a driver assistance system, - Adaptive cruise control in a driver assistance system, - Control of damper forces in a semi-active or active damper system, - Spring rate control, - Control of the vehicle level, and - Roll stabilization.

[0096] In summary, a method for determining the state of motion of a vehicle 10 by means of an observer 23, 100 which is designed as a non-linear Luenberger observer with a non-linear vehicle model, wherein the observer has an input vector u(t), an observer output vector ŷ Obs (t), a measurement output vector y meas (t), an observer state vector x̂ Obs (t), a model state vector x̂ Mdlof the non-linear vehicle model and a model output vector ŷ Mdl has the following steps: A) First data from the at least one first sensor 22 are received, the first data comprising the following first information: - measured yaw rate ψ̇ meas , - measured pitch rate θ̇ meas , - measured roll rate Φ̇ meas , - measured longitudinal acceleration a x,meas , - measured lateral acceleration a y,meas , - measured vertical acceleration a z,meas , - speed values ​​or rotational speed values ​​of at least two wheels, and - drive torque; B) depending on the first data, the input vector u(t) is generated for the observer 23, 100, wherein the input vector u(t) has the following second information: - Longitudinal acceleration (a x,corr ), - Lateral acceleration (a y,corr ), - global pitch rate (Θ̇ G ), - global roll rate (Φ̇ G ), and - Rate of change of a global friction coefficient (μ^˙G); C) a time derivative x^˙(t) of the current observer state vector x̂ Obs (t - 1) is determined depending on: - the input vector u(t), - a matrix-vector product of a Coriolis matrix Mc with the current observer state vector x̂(t - 1), - a gravitational force matrix (F G ), and - a product of a Jacobian matrix ξ(t), a state feedback gain matrix Λ(t), a state gain matrix L(t) and a total control difference (ỹ), where the total control difference is composed of a first control difference (ỹ Obs ) between the observer output vector (ỹ Obs) and the corresponding values ​​of the measurement output vector (y meas (t)) and a second control difference (ỹ Mdl ) between the model output vector (ỹ Mdl ) of the non-linear vehicle model and the corresponding values ​​of the measurement output vector (y meas (t)); D) the current observer state vector x̂ Obs (t) is calculated in an integrator by integrating the calculated temporal change of state x^˙(t) updated; E) an updated observer output vector ŷ(t) is calculated depending on an output matrix C Obs and the current observer state vector x̂ Obs (t) is calculated; and F) Depending on a model output matrix Ĉ Mdl a non-linear vehicle model and a model state vector x̂ Mdl becomes a model output vector ŷ Mdl determined.

[0097] Preferably, when generating the input vector u(t) depending on the first data, at least one correction is carried out from a correction group consisting of: - compensation of an offset error due to a temperature change at the at least one first sensor, and - Compensation of an offset error due to an elastokinematic behavior of the longitudinal system.

[0098] These corrections have proven to be very beneficial for the functioning of the observer 23, 100.

[0099] Preferably, when generating the input vector u(t) depending on the first data, at least one adaptation is carried out from an adaptation group consisting of: - Transformation of longitudinal acceleration, lateral acceleration and vertical acceleration to the corresponding movement of the center of gravity, - Euler transformation of yaw rate, pitch rate and roll rate, - Transformation of the kinematic steering wheel angle into an elastokinematic wheel angle, and - Transformation of the circumferential speed of the wheels into the center of gravity.

[0100] This increases the quality of the observer.

[0101] Preferably, the first data additionally contain at least one third piece of information from an information group consisting of: - direction of rotation of the wheels, - braking torque, - brake pressure, - steering wheel angle, - Steering angle of a rear wheel steering, - degree of locking of a limited-slip differential, - ride height, - Wheel coupling accelerations, and - Status information of an active damping system, in particular active damping forces.

[0102] This information enables improvement of the observer or the vehicle model.

[0103] Preferably, a global friction coefficient is determined by an additional linear observer.

[0104] Preferably, the state gain matrix is ​​determined adaptively depending on the global friction coefficient.

[0105] A two-track vehicle model is preferably used as a non-linear vehicle model.

[0106] Naturally, various variations and modifications are possible within the scope of the present invention.

Claims

[1] Method for determining the state of motion of a vehicle (10) by means of an observer (23, 100) which is designed as a non-linear Luenberger observer with a non-linear vehicle model, wherein the observer (23, 100) has an input vector (u(t)), an observer output vector (ŷ Obs (t)), a measurement output vector (y meas (t)), an observer state vector (x̂ Obs (t)), a model state vector (x̂ Mdl ) of the non-linear vehicle model and a model output vector (ỹ Mdl ), the method comprising the following steps: A) First data from at least one first sensor (22) are received, the first data comprising the following first information: - measured yaw rate (ψ̇ meas ), - measured pitch rate (θ̇ meas ), - measured roll rate (Φ̈ meas ), - measured longitudinal acceleration (a x,meas ), - measured lateral acceleration (a y,meas ), - measured vertical acceleration (a z,meas ), - speed values ​​or rotational speed values ​​of at least two wheels, and - drive torque; B) depending on the first data, the input vector (u(t)) for the observer (23, 100) is generated, wherein the input vector (u(t)) has the following second information: - Longitudinal acceleration (a x,corr ), - Lateral acceleration (a y,corr ), - global pitch rate (Θ̇ G ), - global roll rate (ϕ̇ G ), and - Rate of change (μ^˙G) a global friction coefficient; C) a time derivative (x^˙(t)) of the current observer state vector (x̂ Obs (t - 1)) is determined depending on: - the input vector (u(t)), - a matrix-vector product of a Coriolis matrix (M C ) with the current observer state vector (x̂ Obs (t - 1)), - a gravitational force matrix (F G ), and - a product of a Jacobian matrix (ξ(t)), a state feedback gain matrix (Λ(t)), a state gain matrix (L(t)) and a total control difference (ỹ), where the total control difference is composed of a first control difference (ỹ Obs ) between the observer output vector (ỹ Obs ) and the corresponding values ​​of the measurement output vector (y meas (t)) and a second control difference (ỹ Mdl ) between the model output vector (ỹ Mdl ) of the non-linear vehicle model and the corresponding values ​​of the measurement output vector (y meas (t)); D) the current observer state vector (x̂ Obs(t)) is calculated in an integrator (104) by integrating the calculated time derivative (x^˙(t)) of the current observer state vector (x̂ Obs (t)) updated; E) an updated observer output vector (ŷ Obs (t)) is calculated depending on an output matrix (C Obs ) and the current observer state vector (x̂ Obs (t)); and F) Depending on a model output matrix (Ĉ Mdl ) of a non-linear vehicle model and a model state vector (x̂ Mdl ) an updated model output vector (ỹ Mdl ), where the observer state vector (x̂ Obs( t)) is used in at least one vehicle system from a vehicle system group consisting of: - Electronic Stability Program (ESP), - Traction control (TCS), - All-wheel drive (AWD), - Torque vectoring (TV), - Steering assistance, - Trajectory planning in a driver assistance system, - Adaptive cruise control in a driver assistance system, - Control of damper forces in a semi-active or active damper system, - Spring rate control, - Control of the vehicle level, and - Roll stabilization. [2] Method according to claim 1, wherein, in the generation of the input vector (u(t)) as a function of the first data, at least one correction is carried out from a correction group consisting of: - compensation of an offset error due to a temperature change at the at least one first sensor, and - Compensation of an offset error due to an elastokinematic behavior of the longitudinal system. [3] Method according to one of the preceding claims, in which, in the generation of the input vector (u(t)) as a function of the first data, at least one adaptation is carried out from an adaptation group consisting of: - Transformation of longitudinal acceleration, lateral acceleration and vertical acceleration to the corresponding movement of the center of gravity, - Euler transformation of yaw rate, pitch rate and roll rate, - Transformation of the kinematic steering wheel angle into an elastokinematic wheel angle, and - Transformation of the circumferential speed of the wheels into the center of gravity. [4] Method according to one of the preceding claims, in which the first data comprises at least one third information from an information group consisting of: - direction of rotation of the wheels, - braking torque, - brake pressure, - steering wheel angle, - Steering angle of a rear wheel steering, - degree of locking of a limited-slip differential, - ride height, - Wheel coupling accelerations, and - Status information of an active damping system, in particular active damping forces. [5] Method according to one of the preceding claims, in which the observer state vector (x̂ Obs (t)) has at least a fourth piece of information from a state information group consisting of: - Longitudinal speed of the vehicle (v̂ x ), - lateral velocity (v̂ y ), - Float angle at the center of gravity (β̂ COG ), - Road gradient angle (Θ̂ G ), - Road bank angle (Φ̂ G ), - Friction factor (µ̂ G ) in relation to the center of gravity, - Vehicle roll angle (Φ̂), - vehicle pitch angle (Θ̂), - longitudinal slip, and - Slip angle. [6] Method according to one of the preceding claims, in which the measurement output vector (y meas (t)) has at least a fifth piece of information from a measurement output vector information group consisting of: - Reference longitudinal speed (v x,ref ), - Longitudinal acceleration (a x,corr ), and - Lateral acceleration (a y,corr ). [7] Method according to one of the preceding claims, in which the gravitational force matrix (F G ) has a compensation of the roll angle and pitch angle with respect to the effect of the vehicle roll angle and the vehicle pitch angle. [8] Method according to one of the preceding claims, in which a global friction coefficient is determined by an additional linear observer. [9] Method according to one of the preceding claims, in which the state gain matrix is ​​adaptively determined in dependence on a flag (µ Flag (t)) is selected. [10] Method according to claim 9, wherein the flag (µ Flag (t)) depends on the detection of a longitudinal excitation and a lateral excitation. [11] Method according to one of the preceding claims, in which a two-track vehicle model is used as the non-linear vehicle model.

Images