DE102017209747A1 - Determining a slope of a roadway - Google Patents

Abstract

A method is used for determining a gradient (a) of a roadway (FB), in which a speed (v) of a vehicle (1) is determined, by means of an acceleration sensor an inertial acceleration (a _x ) of the vehicle (1) in the direction of travel (x) the slope (a) is calculated using stochastic filtering, the velocity (v) and the inertial acceleration (a _x ) being used as input parameters of the stochastic filtering. A vehicle (1) has a speed sensor for determining a speed (v) of the vehicle (1), an acceleration sensor for determining an inertial acceleration (a _x ) of the vehicle (1) in the direction of travel (x) and an evaluation device for carrying out the method. The invention is particularly advantageous applicable to two-wheeled vehicles, especially motorcycles.

Description

The invention relates to a method for determining a slope of a roadway. The invention also relates to a vehicle that is set up to carry out this method. The invention is particularly advantageous applicable to two-wheeled vehicles, especially motorcycles.

An estimation of a slope of a road is important for a calculation of a wheel force, since a (positive or negative) slope of the roadway can significantly affect a driving resistance. Various methods for estimating a road gradient are already known:

For example, uses Boniolo I, Corbetta S, Savaresi SM: "Attestation estimation of a motorcycle in a Kalman filtering framework"; 6th IFAC Symposium Advances in Automotive Control; 2010, July 12th-14th, Munich, Germany, pp. 779-784 , a six-degree-of-freedom ("6-DOF-IMU") inertial accelerometer to estimate states of a motorcycle using Euler angles. In addition, an extended Kalman filter is used to estimate vehicle conditions.

Vahidi A, Stefanopoulou A, Peng H .: "Recursive least squares with forgetting for online estimation of vehicle mass and road grade: theory and experiments"; Veh. Syst. Dyn. 2005; vol. 43 (1), pp. 31-55 , estimates a slope of a roadway along with the mass of a heavy duty vehicle using a recursive least squares method with a memory factor.

Lingman P, Schmidtbauer B .: "Road Slope and Vehicle Mass Estimation Using Kalman Filtering"; Veh. Syst. Dyn. 2002; vol. 37 (1), pp. 12-23 discloses another example of estimating a slope using extended Kalman filtering.

The above methods use drag and longitudinal dynamics to estimate the slope.

Corno M, Spagnol P, Savaresi SM. Road Slope Estimation in Bicycles without Torque Measurements: IFAC Proceedings Volumes; 2014 Aug. 24-29, Cape Town, South Africa: IFAC; 2014, pp. 6295-6300 , describes an estimate of a gradient for bicycles based on a Kalman filter.

In the known methods, it is disadvantageous that a mass of the vehicle is or must be estimated for this purpose.

It is the object of the present invention to at least partially overcome the disadvantages of the prior art, and in particular to provide a particularly simple and robust method for estimating or determining a slope of a roadway under a vehicle.

This object is achieved according to the features of the independent claims. Preferred embodiments are in particular the dependent claims.

The object is achieved by methods for determining a slope α a roadway under a vehicle where a speed v the vehicle is determined in the direction of travel, by means of an inertial acceleration sensor acceleration ("inertial acceleration") a _x of the vehicle is measured in the direction of travel and by means of a stochastic filtering the slope α is calculated, the speed v and the inertial acceleration a _x be used as input parameters of the stochastic filtering.

This method has the advantage that it can be dispensed with a separate pitch or pitch sensor, which allows a particularly inexpensive construction. Another advantage is that the slope α can be calculated independently of a mass or mass estimate, so the mass of the vehicle is not an input parameter for the process and also need not be estimated. This greatly facilitates the implementation of the method and also increases accuracy. In addition, the speed v of the vehicle usually already determined, so that this input parameter is readily available. By using stochastic filtering, the calculation results are particularly robust and reliable.

It is an embodiment that the vehicle is a two- or three-wheeled vehicle. This provides the advantage that by dispensing with a gradient sensor, a particularly high relative cost savings is possible. The vehicle may in particular be a motorcycle.

The acceleration sensor may be an inertial translation acceleration sensor for measuring an inertial acceleration in the direction of travel. For example, the acceleration sensor may be an inertial five-degree-of-freedom ("5-DOF-IMU") acceleration sensor.

The speed v of the vehicle is determined by means of a sensor other than the acceleration sensor. It is a training that speeds v the vehicle is determined by means of a wheel sensor. The wheel sensor can, for example, a rotation of a wheel (for example, a speed of a wheel, eg a rear wheel). In turn, it can be concluded on the speed of the vehicle or the speed v of the vehicle.

That the speed v and the inertial acceleration a _x may be used as input parameters of the stochastic filtering, may in particular include that advantageously only the speed v and the inertial acceleration a _x need to be used as input parameters of the stochastic filtering.

It is an embodiment that the slope α is calculated by Kalman filtering. This is particularly easy to implement and robust.

It is still an embodiment that the slope is calculated by means of a linear Kalman filtering. This allows a particularly simple implementation of the method, for example compared to the use of an extended Kalman filtering.

der Kalman-Filterung ein Vektor $$\overrightarrow{x_s}=\left(\begin{array}{c}v\\ \text{sin }\alpha \end{array}\right)$$

mit v der Geschwindigkeit des Fahrzeugs und α der Steigung verwendet wird. Es ist noch eine Weiterbildung, dass als ein Messvektor ${\overrightarrow{z}}_s$

die Geschwindigkeit v des Fahrzeugs verwendet wird. Es ist auch eine Weiterbildung, dass als ein Störungsvektor ${\overrightarrow{u}}_s$

die Inertialbeschleunigung a_x des Fahrzeugs verwendet wird. Als ein Resultat der Kalman-Filterung wird ein Schätzwert für sin α ausgegeben. Die Nutzung von sin α gibt im Vergleich zu einer Nutzung von α in dem Zustandsvektor ${\overrightarrow{x}}_s$

eine zuverlässigere Berechnung.It is a training that as a state vector ${\overrightarrow{x}}_s$

Kalman filtering a vector $$\overrightarrow{x_s}=\left(\begin{array}{c}v\\ \text{sin}\alpha \end{array}\right)$$

with v the speed of the vehicle and α the slope is used. There is still a training that as a measurement vector ${\overrightarrow{z}}_s$

the speed v the vehicle is used. It is also a training that as a disturbance vector ${\overrightarrow{u}}_s$

the inertial acceleration a _x the vehicle is used. As a result of Kalman filtering, an estimate for sin α output. The use of sin α gives in comparison to a use of α in the state vector ${\overrightarrow{x}}_s$

a more reliable calculation.

It is another embodiment that calculates the slope α is suspended when an invalid driving condition is detected. Thus, advantageously, a distortion of the calculation of the slope α due to certain ("invalid") driving conditions to which linear Kalman filtering is not readily applicable.

It is yet another embodiment that calculating the slope α is continued with pre-suspension values of the input parameters when the invalid drive state is terminated. As a result, ending the invalid driving state results in a virtually instantaneous calculation of the gradient α allows.

It is also an embodiment that calculates the slope α is continued only with values of the input parameters measured after the completion of the invalid driving state. This corresponds to a reset or re-establishment of the procedure from the termination of the invalid driving state. Thus, a falsification of the calculation of the slope α be prevented particularly reliable.

It is a training that calculates the slope α is continued with values of the input parameters present before the suspension, if a start of an invalid driving state is not older than a predetermined period of time, and that the calculation of the slope α is continued only with values of the input parameters, if a start of an invalid driving state is older than a predetermined period of time. The predetermined period of time may be, for example, 1 s, 2 s, 5 s, etc.

It is also an embodiment that an invalid driving state includes cornering, braking and / or acceleration of the vehicle. These driving conditions can calculate the slope α especially badly distorted. That a cornering, a braking and / or acceleration is classified as invalid, can be detected by the fact that an associated critical value is reached or exceeded. For example, an associated critical value of a turn may be a yaw rate, a yaw rate, or a yaw rate. A critical value of braking and / or acceleration may be, for example, a predetermined sign-sensitive or absolute value of braking and / or acceleration.

It is a training that out of the estimated slope α a height profile of a worn by the vehicle track is determined. This can be done, for example, by linking the slope to the speed v or the distance traveled under the slope.

The object is also achieved by a vehicle, comprising a speed sensor for determining a speed v of the vehicle, an acceleration sensor for determining an inertial acceleration a _x of the vehicle in the direction of travel and an evaluation device for carrying out the method as described above. The vehicle can be designed analogously to the method and has the same advantages.

The acceleration sensor may be an inertial five-degree-of-freedom ("5-DOF-IMU") acceleration sensor. The evaluation device may be a control unit of the vehicle.

It is an embodiment that the vehicle is a two- or three-wheeled vehicle. This provides the advantage that by dispensing with a gradient sensor, a particularly high relative cost savings is possible. The vehicle may in particular be a motorcycle.

• 1 zeigt ein Fahrzeug, das sich auf einer schrägen Fahrbahn befindet;
• 2 zeigt eine Auftragung einer Steigung α gegen eine Länge eines Fahrwegs; und
• 3 zeigt eine Auftragung einer Höhe h über dem Meeresspiegel gegen eine Länge eines Fahrwegs.

The above-described characteristics, features and advantages of this invention, as well as the manner in which they are achieved, will become clearer and more clearly understood in connection with the following schematic description of an embodiment which will be described in detail in conjunction with the drawings.

• 1 shows a vehicle that is on a sloping roadway;
• 2 shows a plot of a slope α against a length of a driveway; and
• 3 shows a plot of a height H over the sea level against a length of a driveway.

1 shows a vehicle in the form of a motorcycle 1 that is on a sloping roadway FB located. The roadway FB has a rising slope α on. The motorcycle 1 is considered stiff, so that too the motorcycle 1 inclined with a corresponding pitch angle or pitch angle. The motorcycle 1 can its speed v along the roadway FB (ie, along an x-direction in one on the roadway FB referenced Cartesian coordinate system), eg by means of a wheel sensor (not shown).

The motorcycle 1 also has a 5-DOF IMU as an inertial acceleration sensor, by means of which an inertial acceleration a _x along the roadway FB (ie, in the x-direction) is measurable. The inertial acceleration a _x corresponds to horizontal carriageway FB with a slope α = 0 ° one of the measured speed v derived actual acceleration V ,

On a sloping road FB In contrast, the inertial acceleration sensor additionally measures a proportion of the gravitational acceleration G acting along the x-direction. With increasing slope α increases the amount of a _x , Consequently, $$\text{sin}\alpha =\frac{\left(a_x-\dot{v}\right)}{G}$$

The road gradient or incline α can directly from the inertial acceleration a _x and the actual acceleration determined from a wheel speed V calculated according to: $$\alpha =\text{arcsin}\frac{\left(a_x-\dot{v}\right)}{G}$$

Since the input signals or input parameters are highly susceptible to noise due to their measurement and possibly their differentiation, use is of the slope α by means of a direct calculation from Eq. ( 3 ) strongly faulty.

weist als Komponenten die Geschwindigkeit v sowie eine Größe Φ auf: $$x_s=\left(\begin{array}{l}v\\ \mathrm{\Phi }\end{array}\right),\mathrm{\Phi }=\text{sin}\alpha$$

To suppress or eliminate the noise, therefore, a linear Kalman filtering is used to control the slope α to calculate or estimate. Kalman filtering can be understood as a recursive data processing algorithm that statistically minimizes a random error. To the slope α By means of linear Kalman filtering it is necessary to establish a discrete linear difference equation. For this purpose, the system to be solved has been formulated in a state space. The state vector ${\overrightarrow{x}}_s\in {\Re }^n$

indicates the speed as components v and a size Φ on: $$x_s=\left(\begin{array}{l}v\\ \mathrm{\Phi }\end{array}\right).\mathrm{\Phi }=\text{sin}\alpha$$

The index s indicates that the state vector is formulated so that with its help the slope α is estimated. The replacement of Φ = sin α results in a linear formulation of the system, allowing the use of linear Kalman filtering.

mit hier m = 1 ist durch die gemessene Geschwindigkeit v definiert und wird typischerweise bei Motorrädern bereitgestellt. Eine Transformation von Gl. (4) ergibt die Zustandsbeschreibung des Systems: $${\dot{x}}_s=\left(\begin{array}{c}{a}_x-g\mathrm{\Phi }\\ 0\end{array}\right),z_s=v$$

A measurement vector ${\overrightarrow{z}}_s\in {\Re }^m$

with here m = 1 is by the measured speed v is defined and typically provided on motorcycles. A transformation of Eq. ( 4 ) gives the state description of the system: $${\dot{x}}_s=\left(\begin{array}{c}{a}_x-G\mathrm{\Phi }\\ 0\end{array}\right).z_s=v$$

In order to apply linear Kalman filtering, the explicit discrete time invariant formulation of the system is now derived. This is achieved by using the explicit Eulerian forward integration, where s is the time step size: $$x_{s|k}=\left(\begin{array}{l}{v}_k\\ {\mathrm{\Phi }}_k\end{array}\right)=\left(\begin{array}{c}{v}_{k-1}+s\left[a_{x|k-1}-G{\mathrm{\Phi }}_{k-1}\right]+q_{s_1|k-1}\\ {\mathrm{\Phi }}_{k-1}+q_{s_2|k-1}\end{array}\right)$$

Process and measurement noise are determined by the sizes q _s or rs shown. It is assumed that these quantities are independent and uncorrelated, with a probability distribution corresponding to white noise according to: $$\begin{array}{l}p\left(q_s\right)~N\left(0{\text{Q}}_{\text{s}}\right).\\ p\left(r_s\right)~N\left(0{\text{R}}_{\text{s}}\right),\end{array}$$

und zwar mit der Messgleichung (9): $$z_{\text{s}|k}=\underset{{\text{H}}_{\text{s}}}{\underbrace{\left[\begin{array}{cc}1& 0\end{array}\right]}}\underset{x_{s|k}}{\underbrace{\left(\begin{array}{l}{v}_k\\ {\mathrm{\Phi }}_k\end{array}\right)}}+r_{\text{s}|k}$$

Qs represents the covariance matrix of the process noise and Rs the covariance matrix of the measurement noise. Equation (YY) is reformulated into the linear stochastic difference equation (8): $$x_{s|k}=\underset{{\text{A}}_{\text{s}}}{\underset{\}}{\left[\begin{array}{cc}1& -s\mathrm{<figure-callout\; id="0"\; label="G"\; filenames="DE102017209747A1\_0027.png"\; state="\{\{state\}\}">G</figure-callout>}\\ 0& 1\end{array}\right]}}\underset{x_{s|k-1}}{\underset{\}}{\left(\begin{array}{l}{v}_{k-1}\\ {\mathrm{\Phi }}_{k-1}\end{array}\right)}}+\underset{B_s}{\underset{\}}{\left[\begin{array}{l}s\\ 0\end{array}\right]}}\underset{u_{s|k-1}}{\underset{\}}{a_{x|k-1}}}+\underset{q_{s|k-1}}{\underset{\}}{\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right]\left(\begin{array}{l}{q}_{s_1|k-1}\\ {\mathrm{<figure-callout\; id="2"\; label="q\; s"\; filenames="DE102017209747A1\_0028.png"\; state="\{\{state\}\}">q\; </figure-callout>}}_{s_{}}\end{array}\right)}}$$

with the measurement equation (9): $$z_{\text{s}|k}=\underset{{\text{H}}_{\text{s}}}{\underset{\}}{\left[\begin{array}{cc}1& 0\end{array}\right]}}\underset{x_{s|k}}{\underset{\}}{\left(\begin{array}{l}{v}_k\\ {\mathrm{\Phi }}_k\end{array}\right)}}+r_{\text{s}|k}$$

zu einem vorhergehenden Zeitschritt oder Zeitpunkt k-1 mit dem Zustandsvektor ${\overrightarrow{x}}_s$

zum aktuellen Zeitschritt k. Die lineare (n x I) Eingangsmatrix Bs verknüpft einen Störungsvektor („control input vector“) ${\overrightarrow{u}}_s\in {\Re }^l$

mit dem Zustandsvektor ${\overrightarrow{x}}_s,$

während die Inertialbeschleunigung a_x als die Störung verwendet wird, also ${\overrightarrow{u}}_s={\text{a}}_{\text{x}}$

gilt. Die lineare (m x n) Messmatrix Hs verknüpft den Zustand oder Zustandsvektor ${\overrightarrow{x}}_s$

mit der Messung bzw. dem Messvektor ${\overrightarrow{z}}_s.$

The linear (nxn) system matrix A _s combines the state or state vector ${\overrightarrow{x}}_s$

at a previous time step or time k-1 with the state vector ${\overrightarrow{x}}_s$

to the current time step k. The linear (nx I) input matrix Bs combines a disturbance vector ("control input vector") ${\overrightarrow{u}}_s\in {\Re }^l$

with the state vector ${\overrightarrow{x}}_s.$

during the inertial acceleration a _x when the error is used, so ${\overrightarrow{u}}_s={\text{a}}_{\text{x}}$

applies. The linear (mxn) measurement matrix Hs combines the state or state vector ${\overrightarrow{x}}_s$

with the measurement or the measurement vector ${\overrightarrow{z}}_s,$

Compared to the direct calculation of the slope according to Eq. ( 3 ) Kalman filtering in the described formulation advantageously requires no time derivative of the measured velocity v to the slope α to calculate. The covariance matrix Rs of the measurement noise becomes from previous measurements of the velocity v determined or estimated and can be set to Rs = 0.01, for example. The covariance matrix Qs of the process noise can be adjusted empirically, which is known practice in Kalman filtering.

Unlike a perfectly rigid motorcycle, a real motorcycle has one degree of freedom about the y-axis (nick). An on / off logic may be implemented to detect an influence of pitching on an estimate of the slope α reduce or virtually eliminate it.

Since a motorcycle for carrying out the above method does not require or use a pitch or pitch signal, the longitudinal inertial acceleration may a _x used to detect acceleration and deceleration processes. In particular, the above algorithm may be suspended when a predetermined inertial acceleration threshold a _x is reached, exceeded or exceeded, in particular an absolute value of the inertial acceleration a _x reaches or exceeds a predetermined threshold. This is equated with the occurrence of an "invalid" driving condition.

Further, the assumptions for linear Kalman filtering are valid only in the case where the motorcycle does not make any noticeable turn or has any appreciable steering deflection. Consequently, the above algorithm can be suspended even when a measured yaw rate reaches or exceeds a predetermined (threshold) value.

Overall, therefore, in particular the boundary conditions for applying the estimate of the slope α using the Kalman filtering described above, in which a ) an absolute value of the yaw rate is less than a predetermined first threshold value and / or ( b ) an absolute value of the longitudinal inertial acceleration a _x is less than a predetermined second threshold. If at least one condition is not met, the slope estimation algorithm is halted until both conditions are met again.

2 shows a plot of a test ride with a motorcycle 1 estimated slope α in% of a worn lane against a length L of a travel path in m compared to an actual slope AFB the roadway for different speeds v , The slope estimated for both speeds v = 25 km / h and v = 45 km / h α is in very good agreement with the actual slope AFB ,

3 shows a plot of a height H above the sea level of a motorcycle 1 during a test drive against a length L a guideway for one by means of the estimated slope α derived altitude Ha and for a height determined by a GPS system HGPS , The from the estimated slope α derived altitude Ha is in very good agreement with the altitude determined by the GPS system HGPS ,

Of course, the present invention is not limited to the embodiment shown.

Generally, "on", "an", etc. may be taken to mean a singular or a plurality, in particular in the sense of "at least one" or "one or more" etc., unless this is explicitly excluded, e.g. by the expression "exactly one", etc.

Also, a number may include exactly the specified number as well as a usual tolerance range, as long as this is not explicitly excluded.

LIST OF REFERENCE NUMBERS

1

motorcycle

a _x

inertial acceleration

Department

roadway

G

acceleration of gravity

H

Above sea level

Ha

From the estimated slope determined height above sea level

HGPS

From a GPS system certain altitude above sea level

L

Length of a driveway

α

Estimated slope

AFB

Slope of a roadway

v

Measured speed

V

Acceleration derived from the measured velocity

x

Movement direction of the vehicle / x-direction in the coordinate system of the roadway

y

y-direction in the coordinate system of the roadway

z

z-direction in the coordinate system of the roadway

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• Boniolo I, Corbetta S, Savaresi SM: "Attestation estimation of a motorcycle in a Kalman filtering framework"; 6th IFAC Symposium Advances in Automotive Control; 2010, July 12th-14th, Munich, Germany, pp. 779-784 [0003]
• Vahidi A, Stefanopoulou A, Peng H .: "Recursive least squares with forgetting for online estimation of vehicle mass and road grade: theory and experiments"; Veh. Syst. Dyn. 2005; vol. 43 (1), pp. 31-55 [0004]
• Lingman P, Schmidtbauer B .: "Road Slope and Vehicle Mass Estimation Using Kalman Filtering"; Veh. Syst. Dyn. 2002; vol. 37 (1), pp. 12-23 [0005]
• Corno M, Spagnol P, Savaresi SM. Road Slope Estimation in Bicycles without Torque Measurements: IFAC Proceedings Volumes; 2014 Aug. 24-29, Cape Town, South Africa: IFAC; 2014, pp. 6295-6300 [0007]

Claims (9)

Method for determining a gradient (a) of a roadway (FB), in which - a speed (v) of a vehicle (1) is determined, - an inertial acceleration (a _x ) of the vehicle (1) in the direction of travel (x) by means of an acceleration sensor the gradient (a) is calculated by means of a stochastic filtering, whereby the velocity (v) and the inertial acceleration (a _x ) are used as input parameters of the stochastic filtering. Method according to Claim 1 in which the slope (a) is calculated by Kalman filtering. Method according to Claim 2 in which the slope (a) is calculated by means of a linear Kalman filtering, wherein - as a state vector $\overrightarrow{x_s}$

Kalman filtering a vector $$\overrightarrow{x_s}=\left(\begin{array}{c}v\\ \text{sin}\alpha \end{array}\right)$$

is used with v the speed of the vehicle (1) and α of the slope, - as a measuring vector $\overrightarrow{z_s}$

the speed (v) of the vehicle (1) is used and - as a disturbance vector $\overrightarrow{u_s}$

the inertial acceleration (a _x ) of the vehicle (1) is used. Method according to one of the preceding claims, wherein a calculation of the slope (a) is suspended when an invalid driving condition is detected. Method according to Claim 4 in which the calculation of the slope (a) is continued with pre-suspension values of the input parameters (v, a _x ) when the invalid driving condition is terminated. Method according to one of Claims 4 or 5 in which the calculation of the slope (a) is continued only with values of the input parameters (v, a _x ) measured after the invalid driving condition has ended. Method according to one of Claims 4 to 6 in which an invalid driving condition includes cornering, braking and / or acceleration. Vehicle (1), comprising a speed sensor for determining a speed (v) of the vehicle (1), an acceleration sensor for determining an inertial acceleration (ax) of the vehicle (1) in the direction of travel (x) and an evaluation device for carrying out the method according to one of previous claims. Vehicle (1) to Claim 8 , wherein the vehicle (1) is a two-wheeled vehicle, in particular motorcycle.

Images