Method and device for determining a vehicle inertial position

[0002] The present invention relates to a method and arrangement for determining an inertial state of a vehicle.

[0003] German Patent Application No. 196 09 717 describes an arrangement for detecting roll-over occurrences in vehicles. In case roll-over of a vehicle occurs, all passenger protection devices installed in the vehicle must be triggered promptly, including, for example, roll bars, belt tighteners and various airbags. To enable prompt triggering of all of these protection devices, it must be detected as soon as possible whether rotations of the vehicle about its vertical axis, its longitudinal axis or its transverse axis lead to a roll-over. Incorrect decisions on a roll-over occurrence must be ruled out insofar as possible so that the restraint devices are not triggered when, for example, the vehicle is on a steep slope or undergoes slow rotational occurrences during travel through curves. To prevent incorrect decisions from occurring in the roll-over sensing, the inertial state, i.e., the initial state of the vehicle relative to the earth-based coordinate system, must be known. Dynamic vehicle movements such as travel through curves or braking or acceleration procedures can have disruptive effects in determining the inertial state.

[0004] An object of the invention is to specify a method and an arrangement for determining the inertial state of a vehicle, disruptive effects due to dynamic vehicle movements being ruled out to the greatest possible extent.

[0005] Initially, the accelerations of the vehicle in the direction of its longitudinal, transverse and vertical axes are measured. Then, the magnitude of an acceleration vector resulting from the three acceleration components is formed and this magnitude is compared through threshold decisions with a window that is delimited by a threshold lying above and a threshold lying below the gravitational acceleration. The current positional angle of the vehicle with respect to its longitudinal axis and/or the current course angle with respect to its transverse axis is then determined only if the magnitude of the acceleration vector lies inside the window. However, if the magnitude of the acceleration vector lies outside the window, then a previously determined course angle with respect to the longitudinal axis and/or course angle with respect to the transverse axis is retained.

[0006] In this method or rather a corresponding arrangement for carrying out this method, dynamic acceleration components of the vehicle are ruled out with the aid of the window function when determining the inertial state. The course angles of the inertial state can then be correctly measured with the three acceleration sensors if the vehicle is either not moving at all or is moving uniformly. If the vehicle is subject to dynamic state changes, then no new current course angles are determined; instead, one falls back on previously determined course angles that are uninfluenced by dynamic position changes.

[0007] It is advantageous that the upper threshold of the window is about 10% greater and the lower threshold about 10% less than gravitational acceleration. Interfering quantities in determining the inertial state can be suppressed even more effectively if the threshold decisions are performed with respectively two acceleration vectors formed one after another in time. By forming the current course angle recursively from a component of a previously determined course angle and a component of the course angle derived from the currently measured accelerations, short-term disruptions can be suppressed very well in an advantageous manner.

[0008] Roll-over occurrences are very fast state changes of the vehicle that can be best detected with rate of rotation measurements. Based on the measured rates of rotation, the course angles are then derived through integration and it is decided based on these course angles whether a roll-over of the vehicle is occurring or not. So that dynamic vehicle movements uncritical to a roll-over occurrence do not also enter into the integration of the measured rates of rotation and the resulting course angles do not lead to an incorrect decision concerning a roll-over occurrence, it is useful to not begin integration of the rates of rotation prior to the availability of the inertial course angles derived according to the present invention.

[0009] FIG. 1 shows an earth-based and a vehicle-based coordinate system.

[0010] FIG. 2 shows a functional diagram for deriving the inertial state of the vehicle according to the present invention.

[0011] FIG. 3 shows a window function.

[0012] In FIG. 1, an earth-based coordinate system x, y, z is sketched which is oriented such that gravitational acceleration g acts in the direction of the z axis. Also sketched in FIG. 1 is a vehicle-based coordinate system x′, y′, z′ of a vehicle which exhibits a certain banking with respect to the earth-based coordinate system. In the vehicle-based coordinate system, x′ is the longitudinal axis, y′ the transverse axis and z′ the vertical axis of the vehicle. In the vehicle, there are acceleration sensors which measure the acceleration components of the vehicle in the direction of its longitudinal axis x′, its transverse axis y′ and its vertical axis z′. Gravitational acceleration g is divided among the individual acceleration components ax′, ay′ and az′ depending on the position of the vehicle-based coordinate system x′, y′, z′. If, due to dynamic vehicle movements, one or more further acceleration components are superimposed on gravitational acceleration g, it is no longer possible to determine error-free based on the measured acceleration components ax′, ay′ and az′ the actual course angles of the vehicle relative to the earth-based coordinate system. As a general rule, the course angle φx, which is the rotational angle of the vehicle-based coordinate system about the x axis of the earth-based coordinate system, and the course angle φy, which is the rotational angle of the vehicle-based coordinate system about the y axis of the earth-based coordinate system, are needed to characterize the inertial state of the vehicle. As will be explained based on FIG. 1, the two course angles φx and φy can be derived using trigonometric functions from the acceleration components ax′, ay′ and az′ and gravitational acceleration g. If interfering quantities due to dynamic vehicle movements are now superimposed on the acceleration components ax′, ay′, az′, this leads to corrupted course angles φx and φy of the vehicle.

[0013] Based on the functional diagram shown in FIG. 2, a method will now be described for determining, based on the acceleration components ax′, ay′ and az′ measured in the vehicle, the course angles φx and φy, which are uninfluenced to the greatest possible extent by interfering quantities due to dynamic vehicle movements. In functional block 1, the measured acceleration components ax′, ay′ and az′ are subjected to filtering. The filtering serves to filter out small disturbances of the individual acceleration components. A suitable filter is, e.g., a median filter or some other digital filter with a low-pass characteristic. In a median filter, each acceleration component ax′, ay′, az′ is sampled over a certain time interval and all of the sampled values are subdivided into several data tupels. For each data tupel, the average sampled value is determined. Assuming there are i data tupels from each acceleration component ax′, ay′ and az′, the filtered acceleration components ax′(i), ay′(i) and az′(i) are present at the outputs of the filter 1.

[0014] In the second functional block 2, a resulting acceleration vector is formed from the filtered acceleration components ax′(i), ay′(i) and az′(i) and from this the magnitude:

[0015] |a′(i)|={square root over (ax′+L (i+L )²+ay′+L (i+L )²+az′+L (i+L )²+L )}  (1)

[0016] In connection block 3, the magnitude of the acceleration vector |a′(i)| is subjected to a threshold decision. As shown in FIG. 3, a window function is involved here. This window has an upper threshold co and a lower threshold cu. The upper threshold co is about 10% greater than gravitational acceleration g and the lower threshold cu about 10% less than gravitational acceleration g. With this window function, it is thus determined whether the magnitude of the acceleration vector is equal more or less to that of gravitational acceleration g. If, namely, the magnitude of the acceleration vector deviates by a certain degree, which is specified by the thresholds co and cu, from gravitational acceleration g, then one must assume that the acceleration components ax′, ay′ and az′ measured in the vehicle have interfering components superimposed on them due to dynamic vehicle movements. Diagram block 3 sets its output signal h to 1 if the magnitude of the acceleration vector lies inside the window and sets its output signal h to 0 if the magnitude of the acceleration vector lies outside the window. One can also, as is shown in FIG. 3, observe respectively two acceleration vectors present one after another in time on diagram block 3 with regard to their position in the specified window. In other words, only if the magnitudes of the acceleration vectors present at instant i and at instant i—1 both lie inside the specified window, the signal h is set to 1, and otherwise to 0. $\begin{array}{cc}h=\{\begin{array}{cc}1,& \mathrm{if}\left|a^{\prime }\left(i\right)\right|\mathrm{and}\left|a^{\prime }\left(i-1\right)\right|\mathrm{within}\mathrm{the}\mathrm{window}\\ 0& \mathrm{otherwise}\end{array}& \left(2\right)\end{array}$

[0017] In functional block 4, the course angles φx and φy are derived with the aid of the computational procedure described hereafter from the acceleration components ax′ and ay′. In equation (3), the trigonometric relationship between the acceleration components ax′, ay′, az′ measured in the vehicle and the acceleration components with respect to the earth-based coordinate system x, y, z is shown. Since gravitational acceleration g acts only in the direction of the z axis of the earth-based coordinate system, acceleration ax and ay in the direction of the x axis and the y axis of the earth-based coordinate system are 0. Therefore: $\begin{array}{cc}\left[\begin{array}{c}a{x}^{\prime }\\ a{y}^{\prime }\\ a{z}^{\prime }\end{array}\right]=\left[\begin{array}{ccc}\cos \varphi y& 0& -\sin \varphi y\\ \sin \varphi x\sin \varphi y& \cos \varphi x& \sin \varphi x\cos \varphi y\\ \cos \varphi x\sin \varphi y& -\sin \varphi x& \cos \varphi x\cos \varphi y\end{array}\right]\left[\begin{array}{c}0\\ 0\\ -g\end{array}\right]& \left(3\right)\\ \left[\begin{array}{c}a{x}^{\prime }\\ a{y}^{\prime }\\ a{z}^{\prime }\end{array}\right]=\left[\begin{array}{c}\sin \varphi y\\ -\sin \varphi x\cos \varphi y\\ -\cos \varphi x\cos \varphi y\end{array}\right]& \left(4\right)\\ \varphi y=\arcsin \frac{a{x}^{\prime }}{g}& \left(5\right)\\ \varphi x=-\arcsin \frac{a{y}^{\prime }}{g\cos \varphi y}& \left(6\right)\end{array}$

[0018] If the window signal h=1, however, errors can still arise when computing the course angles if, e.g., a resulting vector from interfering accelerations and the gravitational acceleration coincidentally assumes a magnitude of 1 g. To prevent such an error from occurring, it is useful to compute each new course angle φx_newand φy_newrecursively from a component of a course angle φx_oldand φy_oldcomputed earlier and a component of the course angles φx and φy derived from the currently measured accelerations:

[0019] φx_new=c1φx_old+c2φx  (7)

[0020] φy_new=c1φy_old+c2φy  (8)

[0021] The weighting factors c1 and c2 in the equations (7) and (8) must be determined experimentally. They lie between 0 and 1 and have low-pass properties.

[0022] In case roll-over occurrences of the vehicle are to be sensed, a functional block 5 is provided which determines the rotational angle αx and αy of the vehicle about the earth-based x and y axes by integrating measured rates of rotation ωx′, ωy′, ωz′ about the longitudinal axis x′, transverse axis y′ and vertical axis z′ of the vehicle. To prevent slight dynamic positional changes of the vehicle from also entering into the integration, the integration is begun with the previously determined course angles αx and αy, the reason being that these course angles αx and αy are largely uninfluenced by disruptive dynamic vehicle movements (e.g., travel through curves, acceleration and braking processes).