Method for ascertaining the position of the center of gravity of a vehicle

Abstract

A method in which the position of the center of gravity of a moving motor vehicle is ascertained, wherein at least one set of related input variables is taken into consideration, and the set of input variables includes at least a longitudinal acceleration of the motor vehicle, a lateral acceleration of the motor vehicle, a yaw rate of the motor vehicle and at least one wheel rotational speed, in particular four wheel rotational speeds, wherein the set of input variables is ascertained during a steady-state driving maneuver, and a quantity of possible center of gravity positions is defined as classes and, by a learning-based classification method, on the basis of the set of input variables, a class is selected which indicates an estimated center of gravity position. A control unit for carrying out the method is also disclosed.

Claims

1. A method in which a position of a center of gravity of a moving motor vehicle is ascertained, the method comprising: determining, by a motor vehicle processor, at least one set of input variables received from vehicle sensors during one of a plurality of different types of steady-state driving maneuvers including a steady-state left-hand corner maneuver, a steady-state right-hand corner maneuver and a steady-state straight-ahead driving maneuver, the at least one set of input variables comprising at least a longitudinal acceleration of the motor vehicle, a lateral acceleration of the motor vehicle, a yaw rate of the motor vehicle, and at least one wheel rotational speed; selecting, by the motor vehicle processor, a selected one of a plurality of classifiers based on a corresponding one of the plurality of different types of steady-state driving maneuvers, the selecting including: a) selecting a steady-state left-hand corner classifier when the set of input variables correspond to a steady-state left-hand corner maneuver, the steady-state left-hand corner classifier providing a first probability distribution for center of gravity positions during the steady-state left-hand corner maneuver, the steady-state left-hand corner classifier indicating an estimated center of gravity position of the motor vehicle based on the at least one set of input variables and the first probability distribution, b) selecting a steady-state right-hand corner classifier when the set of input variables correspond to a steady-state right-hand corner maneuver, the steady-state right-hand corner classifier providing a second probability distribution for center of gravity positions during the steady-state right-hand corner maneuver, the steady-state right-hand corner classifier indicating an estimated center of gravity position of the motor vehicle based on the at least one set of input variables and the second probability distribution, and c) selecting a steady-state straight-ahead driving classifier when the set of input variables correspond to a steady-state straight-ahead driving maneuver, the steady-state straight-ahead driving classifier providing a third probability distribution for center of gravity positions during the straight-ahead driving maneuver, the steady-state straight-ahead driving classifier indicating an estimated center of gravity position of the motor vehicle based on the at least one set of input variables and the third probability distribution; calculating, by the motor vehicle processor, a plurality of estimated center of gravity positions based on the at least one set of input variables and based on the selected one of the plurality of classifiers; and controlling, by the motor vehicle processor, an operation of at least one of a braking system or a steering system of the motor vehicle based on the plurality of estimated center of gravity positions.

2. The method as claimed in claim 1, wherein a non-linear assignment between sets of input variables and the plurality of center of gravity positions is learned using simulation data of a model of the motor vehicle.

3. The method as claimed in claim 1, wherein the at least one set of input variables comprises a steering angle and/or the at least one set of input variables comprises an estimated value of a total mass of the motor vehicle and/or the at least one set of input variables comprises a measured or estimated roll angle.

4. The method as claimed in claim 1, wherein a steady-state driving maneuver is identified if a vehicle speed and also the lateral acceleration and/or the yaw rate and/or the steering angle are constant over a predefined period of time, wherein: the vehicle speed is regarded as constant if a variance of the vehicle speed lies below a first threshold value, the lateral acceleration is regarded as constant if a variance of the lateral acceleration lies below a second threshold value, the yaw rate is regarded as constant if a variance of the yaw rate lies below a fourth threshold value, and the steering angle is regarded as constant if a variance of the steering angle lies below a third threshold value.

5. The method as claimed in claim 1, wherein, by a classification method, at least two intermediate results are ascertained on the basis of different sets of input variables, and the estimated center of gravity position is calculated on the basis of the at least two intermediate results.

6. The method as claimed in claim 1, wherein at least two sets of input variables are taken into consideration to determine each estimated center of gravity position of the plurality of estimated center of gravity positions, wherein a first set of input variables is ascertained during a first steady-state driving maneuver and a second set of input variables is ascertained during a second steady-state driving maneuver, and the first steady-state driving maneuver differs in type from the second steady-state driving maneuver.

7. The method as claimed in claim 1, wherein, by a classification method, at least one probability distribution of the plurality of estimated center of gravity positions is determined, which assigns a probability value to each of the plurality of estimated center of gravity positions.

8. The method as claimed in claim 1, wherein, in order to calculate each estimated center of gravity position of the plurality of estimated center of gravity positions, firstly a lateral coordinate and a longitudinal coordinate of the respective estimated center of gravity position are determined, and subsequently a vertical coordinate of the respective estimated center of gravity position is determined on the basis of the lateral and longitudinal coordinates.

9. The method as claimed in claim 1, wherein a set of input variables comprising the roll angle is taken into consideration for determining a vertical coordinate of each estimated center of gravity position of the plurality of estimated center of gravity positions.

10. An electronic control unit, for a motor vehicle, comprising: a vehicle processor configured to: determine at least one set of input variables received from vehicle sensors during one of a plurality of different types of steady-state driving maneuvers including a steady-state left-hand corner maneuver, a steady-state right-hand corner maneuver and a steady-state straight-ahead driving maneuver, the at least one set of input variables comprising at least a longitudinal acceleration of the motor vehicle, a lateral acceleration of the motor vehicle, a yaw rate of the motor vehicle, and at least one wheel rotational speed; select a selected one of a plurality of classifiers based on a corresponding one of the plurality of different types of steady-state driving maneuvers, the selecting including: a) selecting a steady-state left-hand corner classifier when the set of input variables correspond to a steady-state left-hand corner maneuver, the steady-state left-hand corner classifier providing a first probability distribution for center of gravity positions during the steady-state left-hand corner maneuver, the steady-state left-hand corner classifier indicating an estimated center of gravity position of the motor vehicle based on the at least one set of input variables and the first probability distribution, b) selecting a steady-state right-hand corner classifier when the set of input variables correspond to a steady-state right-hand corner maneuver, the steady-state right-hand corner classifier providing a second probability distribution for center of gravity positions during the steady-state right-hand corner maneuver, the steady-state right-hand corner classifier indicating an estimated center of gravity position of the motor vehicle based on the at least one set of input variables and the second probability distribution, and c) selecting a steady-state straight-ahead driving classifier when the set of input variables correspond to a steady-state straight-ahead driving maneuver, the steady-state straight-ahead driving classifier providing a third probability distribution for center of gravity positions during the straight-ahead driving maneuver, the steady-state straight-ahead driving classifier indicating an estimated center of gravity position of the motor vehicle based on the at least one set of input variables and the third probability distribution; calculate a plurality of estimated center of gravity positions based on the at least one set of input variables and based on the selected one of the plurality of classifiers; and control an operation of at least one of a braking system or a steering system of the motor vehicle based on the plurality of estimated center of gravity positions.

11. The method as claimed in claim 1, wherein in order to calculate each estimated center of gravity position of the plurality of estimated center of gravity positions, at least two intermediate results are ascertained on the basis of different sets of input variables, and the respective estimated center of gravity position is calculated on the basis of the at least two intermediate results by a minimum mean square error method.

12. The method as claimed in claim 1, wherein, by a random forest or import vector machine method, at least one probability distribution of the plurality of estimated center of gravity positions is determined, which assigns a probability value to each of the plurality of estimated center of gravity positions.

13. The method as claimed in claim 1, wherein in order to calculate each estimated center of gravity position of the plurality of estimated center of gravity positions, firstly a lateral coordinate and a longitudinal coordinate of the respective estimated center of gravity position are determined, and subsequently a vertical coordinate of the respective estimated center of gravity position is determined on the basis of the lateral and longitudinal coordinates, by a linear classification.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Advantageous refinements of aspects of the invention will be discussed below with reference to figures. In the figures:

(2) FIG. 1 shows an exemplary schematic sequence of a method,

(3) FIG. 2 shows an internal structure of an exemplary classifier,

(4) FIG. 3 is a three-dimensional illustration of the information content of exemplary steady-state driving maneuvers,

(5) FIG. 4 shows an evaluation of a center of gravity estimation example.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

(6) FIG. 1 schematically illustrates a sequence of a method with input variables ω, ψ, δ, a.sub.x, a.sub.y, m.sub.load, φ, automatic identification of a valid driving maneuver, and averaging of the noisy signals and storage of the spatial probability distributions in a buffer 5.

(7) The input variables are for example the signals of average value of the wheel rotational speeds ω, yaw rate ψ, steering angle δ, longitudinal and lateral acceleration a.sub.x, a.sub.y, estimated value of the vehicle mass m.sub.load and roll angle ϕ.

(8) An automatic identification 4 of a valid driving maneuver is performed. The driving state of the motor vehicle is monitored on the basis of the input variables of yaw rate ψ, longitudinal acceleration a.sub.x and lateral acceleration a.sub.y. If the yaw rate ψ, longitudinal acceleration a.sub.x and lateral acceleration a.sub.y are constant over a predefined time interval, a first steady-state driving maneuver is identified. Preferably, a steady-state driving maneuver is identified if the longitudinal acceleration a.sub.x is zero or approximately zero, whereby a constant speed can be inferred.

(9) Additionally, the steering angle δ may also be monitored, and a driving maneuver identified as being steady-state only if the steering angle δ is also constant. Alternatively, the monitoring of the steering angle δ may replace the monitoring of the yaw rate ψ.

(10) Here, a value is regarded as constant if the variance of the value remains below a threshold value over the predefined period of time.

(11) On the basis of the values of the yaw rate/i and lateral acceleration a.sub.y, it is determined what type of steady-state driving maneuver is present: whether straight-ahead driving, a left-hand corner or a right-hand corner. Depending on the result, a classifier is selected from a straight-ahead driving classifier 1, a right-hand corner classifier 2 or a left-hand corner classifier 3, and the method proceeds with the selected classifier.

(12) For example, the average wheel rotational speed ω, the yaw rate ψ and the steering angle δ are entered into the straight-ahead driving classifier 1.

(13) For example, the lateral acceleration a.sub.y, the estimated value of the vehicle mass m.sub.load and the roll angle φ are also entered into the right-hand corner classifier 2 and left-hand corner classifier 3 in addition to the average wheel rotational speed ω, yaw rate ψ and steering angle δ.

(14) By means of the selected classifier 1, 2 or 3, a probability distribution P.sub.i(x, y, z) is calculated as an intermediate result and stored in the buffer 5. The index i corresponds to the selected classifier 1, 2 or 3.

(15) As soon as the first steady-state driving maneuver has ended, the monitoring of the driving state of the motor vehicle is continued and further steady-state driving maneuvers are identified. For each detected steady-state driving maneuver, a new probability distribution P.sub.i(x, y, z) is calculated and stored in the buffer 5.

(16) Preferably, for each type of steady-state driving maneuver, a maximum number k of probability distributions P.sub.i(x, y, z) is stored in the buffer 5. For example, it is in particular the case that, for each type of steady-state driving maneuver, exactly one probability distribution P.sub.i(x, y, z) is stored in the buffer 5.

(17) If a probability distribution is calculated for a new steady-state driving maneuver, for the type of which the maximum number of probability distributions is already stored in the buffer 5, then a probability distribution that has already been stored is replaced with the new probability distribution. For example, it is thus the case that a maximum of three different steady-state driving maneuvers are used, one of each type of driving maneuver.

(18) From the probability distributions P.sub.1(x, y, z), P.sub.2(x, y, z), P.sub.3(x, y, z) stored in the buffer 5, an estimated center of gravity position (x, y, z) is calculated by means of a minimum square error calculation 6.

(19) The probability distributions P.sub.1(x, y, z), P.sub.2(x, y, z), P.sub.3(x, y, z) in the buffer 5 are combined with one another in accordance with the formula of a Bayesian filter. The minimum square error estimated value is calculated from the resulting common probability distribution P(x, y, z).

(20) A recalculation of the estimated center of gravity position (x, y, z) is preferably performed after every newly identified steady-state driving maneuver.

(21) The method is advantageously ended, and no new calculation of the estimated center of gravity position (x, y, z) performed, if a termination criterion is met.

(22) The internal structure of the classifiers 1, 2, 3 listed in FIG. 1 is illustrated in FIG. 2. Each of the classifiers 1, 2, 3 has such a structure, for example.

(23) For example, the detected values of the wheel rotational speed ω, the yaw rate ψ, the steering angle δ and the lateral acceleration a.sub.y are each averaged during a detected steady-state driving maneuver, and the respective averaged value is used as an input variable.

(24) For example, the detected value of the roll angle φ and/or the estimated value of the vehicle mass m.sub.load are also averaged in each case.

(25) After the averaging of the signals in the steady state, the resulting feature vector, which comprises the wheel rotational speed ω, the yaw rate ψ, the steering angle δ and the lateral acceleration a.sub.y, and an estimated value of the vehicle mass m.sub.load, are made available to a random forest partial classifier 20. The random forest partial classifier 20 ascertains a probability distribution P(x, y) of the center of gravity position in the longitudinal and lateral directions.

(26) In a processing step 23, the expected values μ.sub.x, μ.sub.y of the probability distribution P(x, y) are determined. μ.sub.x is the expected value of the distribution in the longitudinal direction, and μ.sub.y is that in the lateral direction.

(27) On the basis of the expected values μ.sub.x, μ.sub.y of the probability distribution P(x, y) and taking into consideration the lateral acceleration a.sub.y, the estimated vehicle mass m.sub.load and the roll angle φ, the probability distribution in the vertical direction P(z) is determined by means of a linear partial classifier 21. The probability distribution for a driving maneuver P(x, y, z) is obtained by combining the two results. The probability distribution obtained is stored in buffer 5.

(28) FIG. 3 graphically shows the information content of exemplary steady-state driving maneuvers. In the three-dimensional space of possible center of gravity positions, the steering angle δ, as the strongest feature, describes a line of intersection between the surfaces of possible center of gravity positions for right-hand corners (31, 32, 33) and a left-hand corner (30), which line runs exactly through the center of gravity. The center of gravity height is subsequently determined using the roll angle φ.

(29) FIG. 4 shows the evaluation of an exemplary center of gravity estimation by means of the exemplary method. A winding route is simulated using a simulation program known per se. The measurement signals are superimposed with Gaussian noise, for example with the signal-to-noise ratio SNR=20 dB. The lateral acceleration a.sub.y and the roll angle φ are shown with and without measurement noise. The time t is plotted on the abscissa of all diagrams i-vii.

(30) The first diagram (i) illustrates a simulated lateral acceleration profile 41, and the same profile with superimposed noise 42. The second diagram (ii) illustrates a simulated roll angle profile 43, and the same profile with superimposed noise 44. The state identification and the center of gravity estimation are performed on the basis of the noisy signals.

(31) The state identification 4 identifies steady-state driving maneuvers and classifies these in accordance with their type A. In the third diagram (iii), the result of the identification is plotted as profile 45. Here, the identified type A is depicted on the ordinate as a value (1: straight-ahead driving, 2: right-hand corner and 3: left-hand corner). In this example, five steady-state maneuvers are correctly identified: left-hand corner, right-hand corner, straight-ahead driving, right-hand corner, right-hand corner.

(32) While a steady-state maneuver is taking place, the signals are averaged in each case and, at the end of a maneuver, are passed to the respective classifiers 1, 2, 3 as a single feature vector of input variables. In this example, a total of five classifications are performed, corresponding to the five successive steady-state driving maneuvers identified. The MMSE estimated value (x, y, z) is used for the center of gravity estimation.

(33) The components of the MMSE estimated value x, y and z are illustrated in the diagrams iv-vi. In each time step, these result from the combination of the probability distributions P.sub.i(x, y, z) from the buffer 5. The diagrams iv-vi plot the true value of the respective coordinate (the value assumed for the example) x 46, y 49, z 52 and the profile of the value x 47, y 50, z 53 estimated by the method. The profile of the variance of the probability distribution x 48, y 51 is also plotted in each case for the longitudinal coordinate x and the lateral coordinate y.

(34) At the start of the method, the probability distributions for all coordinates in buffer 5 correspond to an equal distribution, in which all classes of center of gravity positions are assumed to be equally probable. This results in an MMSE estimated value (x, y, z) in the middle of the range of possible center of gravity positions.

(35) After the first left-hand corner in the example, the probability distribution P.sub.3(x, y, z) for this left-hand corner is stored in the buffer 5 and the x and y components 47, 50 approach the true value 46, 49. The classification results of the subsequent steady-state maneuvers are also entered into the buffer 5. An estimation error F, the profile of which is plotted in diagram vii, decreases with each new steady-state driving maneuver and the associated classification.

(36) The termination criterion for the method is preferably considered to be met if the estimation error F falls below an error threshold value.