DE102021206185A1 - Method and control device for activating at least one occupant protection device for a vehicle in the event of a collision of the vehicle and system for occupant protection for a vehicle - Google Patents

Abstract

The invention relates to a method for activating at least one occupant protection device (120) for a vehicle (100) in the event of a collision. An acceleration signal (105, 133) is read in from an interface (131) to an acceleration sensor (102) of the vehicle (100). Furthermore, a model time (135) related to a predefined collision model is determined as a quotient of a kth integral and a (k−1)th integral of the acceleration signal (105, 133) at the time of detection, optionally scaled with a further factor. In addition, the model time (135) is applied to at least one prediction rule for predicting at least one signal property of the acceleration signal (105, 133) after the time of detection and/or to at least one threshold value related to the activation of the at least one occupant protection device (120) in order to calculate a control parameter ( 137) to determine. Finally, a control signal (140) generated using the control parameter (137) is provided for controlling the at least one occupant protection device (120).

Description

State of the art

The invention is based on a device or a method according to the species of the independent claims. The subject matter of the present invention is also a computer program.

A control of means for protecting people from the consequences of traffic accidents can usually be based on the evaluation of sensor signals. A collision with another vehicle causes your own vehicle to decelerate. This delay can e.g. B. can be measured with an acceleration sensor installed in the vehicle and it can be determined predictively whether and when an activation of means for protecting occupants is necessary, and these are then controlled accordingly. Conventionally, however, static collision models can be used here, which can be unsuitable for some types of collision.

Disclosure of Invention

Against this background, with the approach presented here, a method, furthermore a control unit that uses this method, a system and finally a corresponding computer program according to the main claims are presented. Advantageous developments and improvements of the device specified in the independent claim are possible as a result of the measures listed in the dependent claims.

In particular, according to embodiments, an adaptive model-based feature prediction for improved collision detection for vehicles can be provided. Here, for example, time-dependent correction factors can be used to approximate a function term of a collision model. More specifically, a model time determined using an acceleration signal can be used for adaptive control of signal features or thresholds. Here, the model time can be used, for example, as a functional argument for time-dependent characteristics, which can bring about an adaptive correction of such characteristics, or for interpolating between different, in particular model-based, prediction methods or prediction features or between different threshold value curves.

In particular, it can be determined which macroscopic crumple zone behavior or collision model is present for a vehicle or for a given collision. Based on this, macroscopic collision features can be specifically predicted within this collision model. Deviations from the selected collision model can be taken into account adaptively. Instead of being based on a conventional empirical separation of collision types, a distinction between different collision models can be based on an analysis of the progressiveness of a crumple behavior or acceleration profile.

Advantageously, according to embodiments, in particular, a precise and robust prediction of collision features such as deceleration and occupant forward displacement can be achieved with a unified concept that works well for different collision types without having to make hard switches between these collision types, with differences between collision types being controlled gradually and adaptively instead can. Thus, for improved occupant protection, predictions about the course of the collision can be made based on methods that can be individually adapted or applied for each collision and/or the methodology of which can be changed gradually and adaptively depending on the type of collision and the course of the collision. The method can be based on a collision model, e.g. B. be derived from the spring model. Such predictions can predict macroscopic crash features for a wide variety of types of collisions with equal precision or can avoid erratic behavior that could otherwise result from suboptimal functioning of a rough crash type classification.

• Einlesen eines Beschleunigungssignals von einer Schnittstelle zu einem Beschleunigungssensor des Fahrzeugs, wobei das Beschleunigungssignal eine mittels des Beschleunigungssensors zu einem Erfassungszeitpunkt erfasste Beschleunigung des Fahrzeugs bei der Kollision repräsentiert;
• Bestimmen einer auf ein vordefiniertes Kollisionsmodell bezogenen Modellzeit als Quotient aus einem k-ten Integral und einem (k-1)-ten Integral des Beschleunigungssignals zu dem Erfassungszeitpunkt, optional skaliert mit einem weiteren Faktor;
• Anwenden der Modellzeit auf zumindest eine Prädiktionsvorschrift zum Prädizieren zumindest einer Signaleigenschaft des Beschleunigungssignals nach dem Erfassungszeitpunkt und/oder auf zumindest einen auf das Ansteuern der zumindest einen Insassenschutzeinrichtung bezogenen Schwellenwert, um einen Ansteuerparameter zu ermitteln; und
• Bereitstellen eines Ansteuersignals zum Ansteuern der zumindest einen Insassenschutzeinrichtung, wobei das Ansteuersignal unter Verwendung des Ansteuerparameters erzeugt ist.

A method for activating at least one occupant protection device for a vehicle in the event of a collision of the vehicle is presented, the method having the following steps:

• reading in an acceleration signal from an interface to an acceleration sensor of the vehicle, the acceleration signal representing an acceleration of the vehicle detected by the acceleration sensor at a detection time during the collision;
• determining a model time related to a predefined collision model as a quotient of a kth integral and a (k-1)th integral of the acceleration signal at the time of detection, optionally scaled with a further factor;
• Applying the model time to at least one prediction rule for predicting at least one signal property of the acceleration signal after the detection time and/or to at least one threshold value related to the activation of the at least one occupant protection device in order to determine an activation parameter; and
• Providing a control signal for controlling the at least one occupant protection device, wherein the control signal is generated using the control parameter.

This method can be implemented, for example, in software or hardware or in a mixed form of software and hardware, for example in a control unit. The vehicle can be a motor vehicle, in particular a passenger car, truck or other commercial vehicle. The collision can take place between the vehicle and a collision object, in which case the collision object can be a stationary object or another vehicle. The at least one occupant protection device can have at least one airbag and optionally also at least one restraint device, at least one adjustable absorber device for absorbing collision energy, a retractable steering wheel, a retractable dashboard, a movable vehicle seat and additionally or alternatively a service brake of the vehicle. In the providing step, a control signal can be provided which, when used by the at least one occupant protection device, causes the at least one occupant protection device to be activated at a suitable point in time. At least one calculation rule can be used in each case in the determination step and in the application step. Among other things, the start of a collision can be detected for determining integrals, in particular by comparing the acceleration signal with a threshold value. In order to generate the control signal, the control parameter can be compared with at least one threshold value, for example.

zu definieren. Hierbei kann I_m(t) ein Integral vom Grad m des Beschleunigungssignals zum Erfassungszeitpunkt t repräsentieren, kann t_k,n die Modellzeit repräsentieren, kann n den Modellparameter des Kollisionsmodells repräsentieren und kann t` ein vordefiniertes, auf das Ansteuern der zumindest einen Insassenschutzeinrichtung bezogenes Zeitinkrement repräsentieren. Eine solche Ausführungsform bietet den Vorteil, dass auch für unterschiedlichste Arten von Kollisionen eine zuverlässige und genaue Grundlage für eine Prädiktion des Kollisionsverlaufs geschaffen werden kann.According to one embodiment, in the application step, the model time can be applied to the prediction rule in order to use the prediction rule as a $I_m\left(t\right){\left(\frac{t_{k,n}+t\text{'}}{t_{k,n}}\right)}^{n+m}$

define. In this case, I _m (t) can represent an integral of degree m of the acceleration signal at the time of acquisition t, t _k,n can represent the model time, n can represent the model parameter of the collision model and t` can be a predefined one related to the activation of the at least one occupant protection device represent time increment. Such an embodiment offers the advantage that a reliable and precise basis for predicting the course of a collision can also be created for a wide variety of types of collisions.

In addition, the quotient can be scaled with a further factor in the step of determining. The additional factor can correspond to the sum of k and a model parameter of the collision model. Such an embodiment offers the advantage that a model parameter that fits best can be determined in a particularly exact and reliable manner.

In the application step, the model time can also be applied to the threshold value in order to generate a time-dependently adapted threshold value with which an integral of degree m of the acceleration signal at the time of detection can be compared in order to determine the control parameter. Such an embodiment offers the advantage that the control parameter can be determined easily and with minimal effort.

Furthermore, in the application step, at least one characteristic curve or at least one characteristic diagram can be used as a prediction rule. Additionally or alternatively, in the application step, between predefined prediction rules and additionally or alternatively predefined threshold values depending on the model time can be interpolated in order to determine the control parameter. Such an embodiment offers the advantage that the control can be carried out quickly and inexpensively or in a resource-saving manner with regard to computing power.

In addition, the model parameters can be redetermined at least once in the step of determining. For this purpose, a parameter candidate can be determined from a plurality of parameter candidates, which causes a minimal difference between the model time and a collision time. The collision time, which is a period of time since represents or measures the beginning of the collision, can also be referred to as real time. Such an embodiment offers the advantage that a collision can be individually and precisely taken into account with regard to its collision parameters.

The model parameters can also be determined at least once anew in the step of determining. A newly determined model parameter can be selected to be larger than the model parameter if the model time determined using the model parameter is smaller than a collision time, which represents a period of time since the beginning of the collision, due to a progressive acceleration curve. A newly determined model parameter can be selected to be smaller than the model parameter if the model time determined using the model parameter is greater than the collision time due to a declining acceleration curve. Such an embodiment offers the advantage that the collision model can be adapted to the actual course of the collision as a result of the new determination or update of the model parameter.

According to one embodiment, a collision model based on an acceleration profile of the form a _n (t)=c*t ⁿ can be used as the collision model in the determination step. Here, a can represent the detected acceleration of the vehicle during the collision, n can represent the model parameter, t can represent a collision time that represents a period of time since the beginning of the collision, and c can represent at least one characteristic collision parameter, in particular speed, mass and/or rigidity of the vehicle. Also, as the collision model, an integral of the above acceleration history can be used. Such an embodiment offers the advantage that a collision model that is optimized in particular for the vehicle or own vehicle can be used as the base model that can be adapted via the model time and the collision parameter.

n=1 can describe a linearized spring model, n>1 can describe a progressive acceleration curve, 0<n<1 can describe a degressive acceleration curve, n=0 can describe a constant acceleration and n<0 can describe a falling acceleration curve. Thus, by suitably determining the model parameter, different acceleration curves or collision curves can be simulated in an exact approximation in the collision model.

The approach presented here also creates a control device that is designed to carry out, control or implement the steps of a variant of a method presented here in corresponding devices. The object on which the invention is based can also be achieved quickly and efficiently by this embodiment variant of the invention in the form of a control unit.

For this purpose, the control device can have at least one computing unit for processing signals or data, at least one memory unit for storing signals or data, at least one interface to a sensor or an actuator for reading in sensor signals from the sensor or for outputting control signals to the actuator and/or or have at least one communication interface for reading in or outputting data that are embedded in a communication protocol. The arithmetic unit can be, for example, a signal processor, a microcontroller or the like, with the memory unit being able to be a flash memory, an EEPROM or a magnetic memory unit. The communication interface can be designed to read in or output data wirelessly and/or by wire, wherein a communication interface that can read in or output wire-bound data can, for example, read this data electrically or optically from a corresponding data transmission line or can output it to a corresponding data transmission line.

In the present case, a control device can be understood to mean an electrical device that processes sensor signals and outputs control and/or data signals as a function thereof. The control unit can have an interface that can be designed in terms of hardware and/or software. In the case of a hardware design, the interfaces can be part of what is known as a system ASIC, for example, which contains a wide variety of functions of the control device. However, it is also possible for the interfaces to be separate integrated circuits or to consist at least partially of discrete components. In the case of a software design, the interfaces can be software modules which are present, for example, on a microcontroller alongside other software modules.

In an advantageous embodiment, the control unit triggers at least one occupant protection device, in particular at least one airbag, and optionally also a restraint temittel, a service brake or the like. For this purpose, the control device can, for example, access sensor signals such as acceleration signals, as well as optionally additional image signals, model data, default data and, additionally or alternatively, predefined values stored in the vehicle. The control takes place via actuators such as airbag triggering devices, as well as optionally additional belt tensioners, service brakes and the like.

• eine Ausführungsform des vorstehend genannten Steuergerätes; und
• die zumindest eine Insassenschutzeinrichtung, wobei die zumindest eine Insassenschutzeinrichtung signalübertragungsfähig mit dem Steuergerät verbindbar oder verbunden ist.

A system for occupant protection for a vehicle is also presented, the system having the following features:

• an embodiment of the above control device; and
• the at least one occupant protection device, wherein the at least one occupant protection device can be or is connected to the control unit in a manner capable of transmitting signals.

Thus, in the system, an embodiment of the above-mentioned control unit can advantageously be used or used to activate the at least one occupant protection device in the event of a collision. At least one device or interface for receiving or reading in the acceleration signal can also be provided, in which case the at least one interface can be embodied as part of the control unit or can be or can be connected to the control unit in a signal-transmitting manner.

A computer program product or computer program with program code, which can be stored on a machine-readable carrier or storage medium such as a semiconductor memory, a hard disk memory or an optical memory and for carrying out, implementing and/or controlling the steps of the method according to one of the embodiments described above, is also advantageous is used, in particular when the program product or program is executed on a computer, controller or device.

• 1 eine schematische Darstellung eines Ausführungsbeispiels eines Systems zum Insassenschutz in einem Fahrzeug;
• 2 ein Ablaufdiagramm eines Ausführungsbeispiels eines Verfahrens zum Ansteuern zum Ansteuern zumindest einer Insassenschutzeinrichtung für ein Fahrzeug;
• 3 schematische Vergleichsdiagramme hinsichtlich Kollisionsverläufen in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2;
• 4 ein schematisches Vergleichsdiagramm hinsichtlich eines Kollisionsverlaufs in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2;
• 5 ein schematisches Vergleichsdiagramm hinsichtlich eines Kollisionsverlaufs in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2;
• 6 ein schematisches Vergleichsdiagramm hinsichtlich eines Kollisionsverlaufs in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2;
• 7 ein schematisches Vergleichsdiagramm hinsichtlich eines Kollisionsverlaufs in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2;
• 8 ein schematisches Vergleichsdiagramm hinsichtlich eines Kollisionsverlaufs in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2; und
• 9 ein schematisches Vergleichsdiagramm hinsichtlich eines Kollisionsverlaufs in Zusammenhang mit dem System aus 1 und/oder mit dem Verfahren aus 2.

Exemplary embodiments of the approach presented here are shown in the drawings and explained in more detail in the following description. It shows:

• 1 a schematic representation of an embodiment of a system for occupant protection in a vehicle;
• 2 a flow chart of an exemplary embodiment of a method for controlling for controlling at least one occupant protection device for a vehicle;
• 3 schematic comparison diagrams regarding collision processes in connection with the system 1 and/or with the procedure 2 ;
• 4 Fig. 12 shows a schematic comparison diagram regarding a collision history in connection with the system 1 and/or with the procedure 2 ;
• 5 Fig. 12 shows a schematic comparison diagram regarding a collision history in connection with the system 1 and/or with the procedure 2 ;
• 6 Fig. 12 shows a schematic comparison diagram regarding a collision history in connection with the system 1 and/or with the procedure 2 ;
• 7 Fig. 12 shows a schematic comparison diagram regarding a collision history in connection with the system 1 and/or with the procedure 2 ;
• 8th Fig. 12 shows a schematic comparison diagram regarding a collision history in connection with the system 1 and/or with the procedure 2 ; and
• 9 Fig. 12 shows a schematic comparison diagram regarding a collision history in connection with the system 1 and/or with the procedure 2 .

Before exemplary embodiments are described in more detail below, the basics and backgrounds with regard to exemplary embodiments should first be briefly explained by way of example.

Vehicle collisions are detected in airbag control devices for controlling restraint devices, in particular by evaluating acceleration signals. For this purpose, acceleration sensors can be installed in the airbag control unit itself or externally. Processing of the acceleration signals usually includes a single or double time integration, which reduces the speed or corresponds to the change in path due to the collision. For central acceleration sensors in positions close to the center of gravity of the vehicle or in the interior, the path changes determined from them correspond to the forward displacement of the occupants in the collision, since these can be assumed to be free-flying masses in the period relevant for collision detection.

Since airbag deployment decisions should be made very early in the collision, especially at times when deceleration and forward displacement are still very small, deployment algorithms always contain certain explicitly or only implicitly predictive components that underline the value of features such as deceleration and forward displacement later in the course to assess the collision. For example, the usual required triggering time of an airbag is defined by the fact that an occupant in a normal position has moved forward by about 5 inches about 35 milliseconds later, after the airbag has inflated, and is thus optimally cushioned by the airbag.

Basic macroscopic properties of vehicle collisions, such as average acceleration level, collision duration, rough course of deceleration and forward displacement and the like, depending on parameters such as collision speed, masses and stiffness of the other party involved in the accident can be well understood in a simple spring model for the vehicle crumple zone. In particular, a determination of the optimal deployment time for airbags based on a prediction of the occupant's forward displacement within the framework of the spring model is known. It is also known that a ratio of forward displacement and speed reduction has the dimension of time and, within the framework of the spring model, is proportional to the collision time or time span since the beginning of the collision or since the time of contact between the vehicle and the other party involved in the accident.

Spring model predictions are recapitulated below. Within the context of the spring model mentioned above, a vehicle is described by a spring with spring constant D and a mass point m behind it. This model describes relatively well a force level in a crumple zone that increases linearly with increasing deformation. If such a vehicle collides with the speed vo with an obstacle assumed to be infinitely heavy and infinitely stiff, the solution of the associated differential equation for the acceleration profile during the collision as a function of the collision time or time since the beginning of the collision (sign convention: positive in the front crash ) form $$a\left(t\right)=v_0\sqrt{\frac{D}{m}}\text{sin}\left(\sqrt{\frac{D}{m}t}\right).$$

wie sie für die Kollisionsdetektion relevant sind, kann der Sinus linearisiert werden, und man erhält die linear mit der Zeit ansteigende Beschleunigung $$a\left(t\right)=v_0\frac{D}{m}t,$$

wobei in der Steigung $c=v_0\frac{D}{m}t,$

hierbei kollisionsrelevante Parameter enthalten sind.For little times $t<<T=2\pi \sqrt{m/D},$

as they are relevant for collision detection, the sine can be linearized and the acceleration that increases linearly with time is obtained $$a\left(t\right)=v_0\frac{D}{m}t,$$

where in the slope $c=v_0\frac{D}{m}t,$

collision-relevant parameters are included.

From the point of view of vehicle 1, a collision between two vehicles (D1, m1) and (D2, m2) with a relative speed v _rel can also be caused by a collision of a spring model with an effective spring constant D = D1*D2/(D1+D2) and reduced mass m = m1*m2/(m1+m2) can be described with a speed of v _rel * m2/(m1+m2) with an obstacle assumed to be infinitely heavy and infinitely stiff. Even if the spring model cannot describe a precise acceleration curve of a vehicle with a discretely constructed crumple zone, there is an averaged or low-pass filtered acceleration curve as well as the curve of the first integral - with regard to a reduction in speed - and the second integral - with regard to forward displacement of the Occupants - the acceleration reflected in a variety of collisions.

In the following description of favorable exemplary embodiments of the present invention, the same or similar reference symbols are used for the elements which are shown in the various figures and have a similar effect, with a repeated description of these elements being dispensed with.

1 shows a schematic representation of an exemplary embodiment of a system 110 for occupant protection in a vehicle 100. Vehicle 100 is a motor vehicle for example a passenger car or a commercial vehicle. Between the vehicle and a collision object OBJ it comes in the representation of 1 to a collision. The collision object OBJ is, for example, another vehicle or a stationary object.

Vehicle 100 has, for example, only one acceleration sensor 102 and system 110 for occupant protection or occupant protection system 110 . Furthermore, there is only one vehicle occupant OCC in vehicle 100, for example.

Acceleration sensor 102 is designed to provide an acceleration signal 105 . Acceleration signal 105 represents an acceleration of vehicle 100 detected by acceleration sensor 102 at a detection time during the collision. For this purpose, acceleration sensor 102 is designed to detect an acceleration acting on vehicle 100 . A characteristic acceleration acts on vehicle 100 during the collision. Acceleration sensor 102 is designed to detect an acceleration acting on vehicle 100 with respect to at least one spatial direction or axis.

The occupant protection system 110 has at least one occupant protection device 120 and a control unit 130 . According to the 1 In the exemplary embodiment illustrated, the occupant protection system 110 has, for example, only one occupant protection device 120 in the form of an airbag. The occupant protection device 120 and the control unit 130 are connected to one another in a manner capable of transmitting signals.

The occupant protection device 120 is designed to protect the vehicle occupant OCC in the event of a collision between the vehicle 100 and the collision object OBJ. According to one exemplary embodiment, at least one further occupant protection device 120 can also be provided. Such a further occupant protection device can have at least one restraint device, at least one adjustable absorber device for absorbing collision energy, a retractable steering wheel, a retractable dashboard, a movable vehicle seat and/or a service brake of vehicle 100.

Control unit 130 is connected to at least acceleration sensor 102 and occupant protection device 120 in a manner capable of transmitting signals. Control unit 130 is designed to activate passenger protection device 120 in the event of a collision, for example in the event of a collision between vehicle 100 and collision object OBJ. In this case, control unit 130 has a reading device 132 , a determination device 134 , an application device 136 and a provision device 138 .

Reading device 132 is designed to read acceleration signal 105 from an interface 131 to acceleration sensor 102 . The interface 131 is designed as part of the control unit 130, for example. Alternatively, the interface 131 can be provided separately from the control unit 130 . The read-in device 132 is designed to output or forward the acceleration signal 105 in the form of an acceleration signal 133 that has been read in. The read-in acceleration signal 133 represents a read-in and/or pre-processed variant of the acceleration signal 105 detected by means of the acceleration sensor 102.

Determination device 134 is designed to determine a model time 135 related to a predefined collision model, specifically as a quotient of a kth integral and a (k-1)th integral of acceleration signal 105 or 133 at the time of acquisition, optionally scaled with a further factor, in particular multiplied by the sum of k and a model parameter of the collision model. Determination device 134 is also designed to output or forward a determination signal that represents at least model time 135 .

Determination device 136 is designed to apply model time 135 to at least one prediction rule for predicting at least one signal property of acceleration signal 105 or 133 after the time of acquisition and/or to at least one threshold value related to the activation of the at least one occupant protection device 120, in order to calculate an activation parameter 137 to determine. Determination device 136 is also designed to output or forward control parameter 137 .

Provision device 138 is designed to provide an actuation signal 140 for activating the at least one occupant protection device 120 . In this case, control signal 140 is or will be generated using control parameter 137 . According to one embodiment, the Already Setting device 138 and / or another or further device of the control unit 130 is formed to generate the control signal 140. Control signal 140 is suitable for minimizing or reducing an occupant energy of vehicle occupant OCC that is to be dissipated by the end of the collision by means of at least one occupant protection device 120 . For this purpose, control signal 140 is particularly suitable for activating the at least one occupant protection device 120 . Provision device 138 and/or another or further device of control unit 130 is designed to output actuation signal 140 via a further interface 139 to the at least one occupant protection device 120.

2 shows a flow chart of an embodiment of a method 200 for activating at least one occupant protection device for a vehicle in the event of a collision of the vehicle. By executing method 200 for activation, at least one occupant protection device for a vehicle can be activated in the event of a collision of the vehicle. The method 200 for driving is here using the control device 1 or a similar control unit and/or in connection with the occupant protection system 1 or a similar occupant protection system executable. The method 200 for controlling comprises a step 210 of reading in, a step 220 of determining, a step 230 of application and a step 240 of providing.

In step 210 of reading in, an acceleration signal is read in from an interface to an acceleration sensor of the vehicle. The acceleration signal represents an acceleration of the vehicle detected by the acceleration sensor at a time of detection during the collision. Subsequently, in step 220 of determining, a model time related to a predefined collision model is determined as the quotient of a kth integral and a (k−1)th integral of the acceleration signal at the time of detection, optionally scaled by a further factor, in particular multiplied by the sum of k and a model parameter of the collision model. In step 230 of applying, the model time is applied to at least one prediction rule for predicting at least one signal property of the acceleration signal after the time of acquisition and/or to at least one threshold value related to the activation of the at least one occupant protection device in order to determine an activation parameter. Subsequently, in step 240 of providing, a control signal for controlling the at least one occupant protection device is provided. The control signal is generated using the control parameter.

With reference to those described above 1 and 2 In the following, exemplary embodiments and advantages of exemplary embodiments are summarized and presented or presented in other words.

First, a prediction in a general collision model is discussed. While a spring model can describe vehicles and collisions with a linearly increasing force level well, it shows deviations for vehicles and collisions with a different crumple zone characteristic. On the one hand, the crumple zone of a given vehicle 100 can structurally have a force curve that increases on average linearly with a deformation caused by a collision. On the other hand, even with an essentially linear crumple zone course, certain types of collision can nonetheless show other characteristics. For example, in a so-called small overlap crash, in which the colliding vehicles overlap only slightly, z. B. 25%, the structures in the crumple zone are not deformed in a way well described by the spring model - the result is a stagnant force level in the crumple zone after an initial increase, which also results in a stagnant acceleration curve.

angenommen werden. Im Hinblick auf das Steuergerät 130 aus 1 und/oder das Verfahren 200 aus 2 basiert gemäß einem Ausführungsbeispiel das Kollisionsmodell auf einem solchen Beschleunigungsverlauf. Der Fall n=1 beschreibt das Federmodell bzw. linearisierte Federmodell, n>1 beschreibt einen progressiven bzw. konvexen Beschleunigungsverlauf, 0<n<1 einen degressiven bzw. konkaven Beschleunigungsverlauf, n=0 eine konstante Beschleunigung und der (theoretische) Fall n<0 einen abfallenden Beschleunigungsverlauf. Kollisionscharakteristika bzw. Crashcharakteristika, wie Geschwindigkeit, Masse, Steifigkeit etc., sind in dem Vorfaktor c enthalten.As a general collision model, an acceleration curve according to a power law can be used $$a_n\left(t\right)=c\cdot t^n$$

be accepted. With regard to the controller 130 off 1 and/or the method 200 2 According to one exemplary embodiment, the collision model is based on such an acceleration curve. The case n=1 describes the spring model or linearized spring model, n>1 describes a progressive or convex acceleration curve, 0<n<1 a degressive or concave acceleration curve, n=0 a constant acceleration and the (theoretical) case n< 0 a falling acceleration curve. Collision characteristics or crash characteristics, such as speed, mass, rigidity, etc., are contained in the pre-factor c.

If one now integrates this acceleration m times, one obtains the mth integral I _m,n (t) according to $$I_{m,n}\left(t\right)=\frac{c}{\left(n+1\right)\left(n+2\right)\mathrm{...}\left(n+m\right)}\cdot t^{n+m}$$

$$=I_{m,n}\left(t\right){\left(\frac{t+t\text{'}}{t}\right)}^{n+m}$$

Of particular interest are the first integral, which describes the reduction in speed during the collision, and the second integral, which describes a forward displacement of the occupant OCC in the collision. For the value of the m-th integral at a time in the future from the instantaneous time t or acquisition time t by the time span t′, the following then applies directly within the model $$I_{m,n}\left(t+t\text{'}\right)=\frac{c}{\left(n+1\right)\left(n+2\right)\mathrm{...}\left(n+m\right)}\cdot {\left(t+t\text{'}\right)}^{n+m}$$

$$=I_{m,n}\left(t\right){\left(\frac{t+t\text{'}}{t}\right)}^{n+m}$$

As already mentioned in connection with the spring model, real acceleration curves α(t) usually deviate from simple model assumptions as in Eq. (3) on, e.g. B. by a superposition of oscillations. However, these are largely averaged out by filtering or integration. This means that the behavior according to the model in the integrated accelerations according to Eq. (4) is even better expressed than in Eq. (3). Provided that the basic model assumptions are correct, the deviations between real acceleration curves α(t) and the approximate model-based acceleration curve according to Eq. (3) exist, decreasing with increasing integration or averaging, and the simplified model-based relationship in Eq. (4) describes the course of the actual m-th integral I _m (t) of the acceleration α(t) quite well.

$$=I_{m,n}\left(t\right){\left(\frac{t+t\text{'}}{t}\right)}^{n+m}$$

In order to obtain a prediction pI _m (t) of the expected value of the mth integral at a time t+t' in the future by the time span or time increment t' at the time or detection time t in the event of a collision, ie i.e. to estimate I _m (t + t') as well as possible, the model-based predictions from Eq. (5a) and (6a) adopted for the real course, $$p{I}_m\left(t\right)=I_{m,n}\left(t+t\text{'}\right)=\frac{c}{\left(n+1\right)\left(n+2\right)...\left(n+m\right)}\cdot {\left(t+t\text{'}\right)}^{n+m}$$

$$=I_{m,n}\left(t\right){\left(\frac{t+t\text{'}}{t}\right)}^{n+m}$$

In Eq. (6b) the modeled integral I _m,n (t) from Eq. (6a) is again replaced by the actual integral at the time or acquisition time t, ie I _m (t).

These considerations are now to be transferred to real collision signals or acceleration signals 105 of a specific vehicle 100 with a specific crumple zone characteristic. To do this, it is first necessary to determine the exponent n for the collision model that is suitable for this vehicle 100 . This can e.g. B. on an empirical basis based on available collision data or acceleration data {α(t)} so that the deviation from Eq. (3) to the real collision data is smallest or the deviation of the real m-th integral to the model curve according to Eq. (4) becomes the smallest. This can also take place outside of control unit 130 , in which case only the parameter or model parameter n determined in this way is then stored in control unit 130 .

In the event of the collision, the predicted integral is then determined in control unit 130 in accordance with Eq. (5b) or (6b). A so-called crash timer, which records the time t from the start of the collision, e.g. B. on the basis of a threshold exceeding the acceleration α(t) detects or measures, is already present in conventional crash detection algorithms. In the variant according to Eq. (5b), in which the predicted integral is expressed as a function of the collision parameter or crash parameter c and the collision time t or the time of detection, where t' is fixed and not variable, the collision-relevant parameter c is also determined. In the linearized spring model (n=1), this value corresponds to the slope of the acceleration (1) and can be determined by determining the gradient from an i. A. processed acceleration can be determined. This parameter may be more difficult to determine in non-linear collision models. Therefore the variant according to Eq. (6b) advantageous, in which the predicated integral as a function of the instantaneous integral value I _m (t) and the collision time t or the detection time is determined. The instantaneous integral value is also available in the usual algorithms and is determined by m-fold integration of the possibly pre-processed acceleration α(t) since the start of the collision.

Since the power term in Eq. (5b) or (6b) is computationally expensive, it can, according to one exemplary embodiment, be implemented on a control unit 130 by a time-dependent factor which, for example, B. is stored by a characteristic as a function of time t, can be approximated. In this realization, deviations from the exact progression in Eq. (5b) or (6b) can be realized, which then have different model curves than in Eq. (3) represent.

In many variants, this predicted integral can be included in a triggering decision when the at least one occupant protection device 120 is actuated. A simple application example is the estimation of the future occupant forward displacement using the second integral in order to determine an optimal deployment time (TTF) of an airbag according to the so-called 5-inch rule, pI ₂ (t = TTF) > Thd _Disp . In this case, t′ describes the inflation time of the airbag and the forward displacement threshold Thd _Disp is approximately 5 inches in the normal position of the occupant OCC, but can also assume other values for other occupant sizes and seat positions.

The base model described above would be appropriate in the event that all collision modes stress the crush zone in a similar manner. As already mentioned, this does not necessarily have to be the case for individual types of collisions such as a small overlap crash, and depending on the properties of the host vehicle and the enemy vehicle, more or less large deviations from the “characteristic of the host vehicle” characterized by a fixed n can be observed. base model”. In order to arrive at an optimal estimate of future integrator values even for such deviating types of collision, an adaptive prediction is discussed below. In this way, a high degree of robustness against deviations from the assumed basic model and/or an adaptive adjustment of the selected collision model to the observed course of the collision can be undertaken, as explained below.

$$I_{k-\mathrm{1,}n}\left(t\right)=\frac{c}{\left(n+1\right)\left(n+2\right)\dots \left(n+k-1\right)}\cdot t^{n+k-1},$$

so entspricht dies bis auf einen modellabhängigen Skalierungsfaktor gerade der „Echtzeit“ bzw. Kollisionszeit t bzw. dem Erfassungszeitpunkt, $$t=\left(n+k\right)\frac{I_{k,n}\left(t\right)}{I_{k-\mathrm{1,}n}\left(t\right)}.$$

If one considers the ratio of the kth to the (k-1)th integral within the selected collision model or model parameter n with $$I_{k,n}\left(t\right)=\frac{c}{\left(n+1\right)\left(n+2\right)...\left(n+k\right)}\cdot t^{n+k}$$

$$I_{k-\mathrm{1,}n}\left(t\right)=\frac{c}{\left(n+1\right)\left(n+2\right)...\left(n+k-1\right)}\cdot t^{n+k-1},$$

apart from a model-dependent scaling factor, this corresponds to the "real time" or collision time t or the time of acquisition, $$t=\left(n+k\right)\frac{I_{k,n}\left(t\right)}{I_{k-\mathrm{1,}n}\left(t\right)}.$$

The "real time" or collision time is reproduced with sufficient accuracy within triggering algorithms by the crash timer. Based on this relationship, a “model time of degree k in crash model n” can be defined within the selected basic model n by the ratio of the real kth integral to the real (k-1)th integral according to $$t_{k,n}\left(t\right)=\left(n+k\right)\frac{I_k\left(t\right)}{I_{k-1}\left(t\right)}.$$

$$=I_m\left(t\right){\left(\frac{t_{k,n}+t\text{'}}{t_{k,n}}\right)}^{n+m},$$

wodurch sich für eine ins Modell passende Kollision lediglich unbedeutende Änderungen ergeben. Praktischerweise können der Grad k der Modellzeit und der Grad m der Integration identisch gewählt werden. In 4 sind nachfolgend Prädiktionen für das zweite integral gemäß Gl. (6b) (Kurve 401) und (6c) (Kurve 402) für einen Rigid Barrier Crash dargestellt, die sich nur wenig unterscheiden.For a collision that is well described by the chosen base model and thus I _k (t) and I _k-1 (t) well by Eq. (7) and (8) are reproduced, the model time according to Eq. (10) then because of Eq. (9) agree very well with the real time, t _k,n (t) ≈ t - this is subsequently in 3 shown above (300A) for a so-called rigid barrier crash or a collision with a rigid obstacle, with curve 301 representing real time and curve 302 representing model time. Therefore, for adaptive prediction formulas, the real time in Eq. Replace (5b) and (6b) with the model time of degree k in the crash model n, $$p{I}_m\left(t_{k,n}\right)=I_{m,n}\left(t_{k,n}+t\text{'}\right)=\frac{c}{\left(n+1\right)\left(n+2\right)...\left(n+m\right)}\cdot {\left(t_{k,n}+t\text{'}\right)}^{n+m}$$

$$=I_m\left(t\right){\left(\frac{t_{k,n}+t\text{'}}{t_{k,n}}\right)}^{n+m},$$

resulting in only insignificant changes for a collision that fits into the model. Practically, the degree k of the model time and the degree m of the integration can be chosen to be identical. In 4 are below predictions for the second integral according to Eq. (6b) (curve 401) and (6c) (curve 402) for a rigid barrier crash, which differ only slightly.

However, if a given collision deviates significantly from the base model or the selected model parameter n, the corresponding model time in model n also deviates significantly from real time. If the collision fits better into a model with exponent or model parameter n`, instead of the model time in the model n Eq. (10) the model time in the collision model or with the model parameter n' agrees well with the real time, $$t_{k,n\text{'}}\left(t\right)=\left(n\text{'}+k\right)\frac{I_k\left(t\right)}{I_{k-1}\left(t\right)}\approx t.$$

With Eq. (10) and (11) $$t_{k,n}\left(t\right)\approx \frac{\left(n+k\right)}{\left(n\text{'}+k\right)}t.$$

If the acceleration curve in the considered collision is more progressive than according to the base model, i.e. it is better described by a model with n'> n, then the model time in the model n is smaller than the real time, t _k,n (t) < t. Since the power term describing the time dependence in Eq. (6b) and (6c) for the usual case n+m>0 increases with decreasing argument, this is expressed in Eq. (6c) by a comparison to eq. (6b) increased prediction value. However, if the acceleration curve in the collision under consideration is more degressive than according to the base model, i.e. it is better described by a model with n'< n, then the model time in the model n is greater than the real time, t _k,n > t. this is in 3 for a small overlap crash or a collision with a small amount of overlap or coverage - curve 301 represents the real time and curve 303 represents the model time. Since the power term describing the time dependency in Eq. (6b) and (6c) for the usual case n+m>0 decreases with increasing argument, this is expressed in Eq. (6c) by a comparison to eq. (6b) decreased prediction value. In 5 below are the predictions for the second integral according to Eq. (6b) (curve 501) and (6c) (curve 503) are compared for a small overlap crash. In both cases, by using the model time in Eq. (6c) instead of real time in Eq. (6b) a correction of the integral prediction can be achieved in a direction that corresponds better to the actual course of the collision.

Except for the one given by Eq. In addition to the adaptive prediction described in (6c), there are other options for adaptively adapting the prediction to the properties of a collision that have been determined. In this way, the basic model can be abandoned and instead, according to one exemplary embodiment, each individual collision can be treated individually. For this purpose, for each individual collision during the course according to Eq. (11) dynamically determines which model parameter n' best characterizes the collision. The model-based predictions (5b) and (6b) are carried out with this value. The value n' can change dynamically during the course of the collision.

As already described above, it can be advantageous, according to an exemplary embodiment, to dispense with the calculation of the power term in an implementation on control unit 130 and to approximate it using a time-dependent characteristic curve for a given model parameter n or to deliberately take deviations from the model into account. In order to take into account collisions with different model parameters n, a few characteristic curves can be stored, e.g. B. for the base model n, collisions with a more progressive acceleration profile n1 > n, collisions with a degressive acceleration profile n2 < n etc. Depending on the collision for the present collision according to Eq. (11) dynamically (newly) determined model parameters n′, one of these characteristic curves can then be selected or dynamically interpolated between these characteristic curves. An interpolation between the models n=1 and n=0 for the second is integral hereinafter in 6 for a rigid barrier crash and in 7 for a small overlap crash: the curves 601 and 701 describe the predictions in the model n=1, the curves 602 and 702 the predictions in the model n=0 and the curves 603 and 703 finally the interpolation between the models n =1 and n=0 based on the determined value n'.

Additionally or alternatively, according to one embodiment, a combination of the above-described interpolation between different model predictions with the temporarily described use of the respective model time as a function argument in the prediction formula according to Eq. (6c) take place. By using the model time, as described, an adaptation of the prediction for crashes that deviate from the model is already achieved within the base model. The interpolation between the base model and other models that may be a better match then only needs to make minor corrections. This is below in 8th shown for the second integral of a small overlap crash. Curve 801 describes the prediction in the model n=1, curve 802 the predictions in the model n=0 and curve 803 finally the interpolation between the models n=1 and n=0 on the determined value n'.

According to a further exemplary embodiment, an interpolation between other prediction formulas that are different for different collision models than the predictions derived directly from the model according to Eq. (5b/c) and (6b/c) respectively. Examples of this include Taylor expansions for the mth integral at time t by the increment t' to estimate the value of I _m (t + t'). Depending on the model assumptions, various Taylor developments work better here. In the model n=0 it is generally sufficient to use an m-th order Taylor expansion for the m-th integral. In the model n=1, a (m+1)-th order Taylor expansion makes more sense. The interpolation between the two variants can now take place dynamically based on the determined value n'.

äquivalent zu $$I_m\left(t\right)>Thd\left(t\right)=Th{d}_0{\left(\frac{t}{t+t\text{'}}\right)}^{n+m}.$$

A consideration of the prediction in the threshold value is briefly discussed below. If the predicted integrals according to Eq. (6b) or (6c) are simply fed to a threshold value comparison, instead of comparing the predicted integral against a threshold value, the instantaneous integral can also be compared against a threshold value that varies as a function of time. So e.g. B. with Eq. (6b) $$p{I}_m\left(t\right)=I_m\left(t\right){\left(\frac{t+t\text{'}}{t}\right)}^{n+m}>TH{\mathrm{i.e}}_0$$

equivalent to $$I_m\left(t\right)>TH\mathrm{i.e}\left(t\right)=TH{\mathrm{i.e}}_0{\left(\frac{t}{t+t\text{'}}\right)}^{n+m}.$$

Correspondingly, the variants mentioned above can also be used in this form according to Eq. (14) can be implemented: replacing the power term or the entire threshold value with a time-dependent characteristic, replacing the real time with the model time, see also the one described below 9 , individual treatment of each individual collision by determining the appropriate model parameter n for each individual collision, interpolation between different basic characteristics, e.g. B. for base model n, for convex model n1>n, concave model n2<n etc., based on the determined value of the model parameter n' of a given collision.

The correct factor (n+k) with which the quotient of the kth integral and the (k-1)th integral of the acceleration signal at the time of acquisition is used in the calculation of the model time, Eq. (10), is important in some application examples, e.g. B. when using the exact analytical prediction rule, see Eq. (6a) or (6b), or in determining the best fitting model parameter n'. For other application examples, in particular if the exact formulas are or will be replaced by characteristic curves, the exact factor is not important; it can be or can be taken into account implicitly in the characteristic curve.

3 FIG. 12 shows schematic comparison charts 300A and 300B regarding collision histories associated with the system from FIG 1 and/or with the procedure 2 . The comparison diagrams 300A and 300B show a real time 301 obtained by a conventional crash timer compared to a model time of degree 2 in the model n=1 or linearized spring model, with a first comparison diagram 300A for a so-called rigid barrier crash with full overlap at 26 km/h, a first model time 302 and a second comparison diagram 300B for a so-called small overlap crash at 64 km/h a second model time 303 are shown, each as a function of the absolute time or over a time from barrier contact. The real time 301 and the model times 302 and 303 are given in units of 2 cycles per millisecond.

While first model time 302 and real time 301 match fairly well in the rigid barrier crash illustrated by comparison chart 300A, i. H. the crash is well described by the spring model n=1, in the case of the small overlap crash illustrated by the circle diagram 300B, the second model time 303 from t > 10 ms is significantly greater than the real time 301. This means that the crash is better described by a model with smaller model parameters n' would be described. The rough factor 1.5, by which the second model time 303 in the triggering-relevant area is 40 ms greater than the real time 301, indicates a model parameter of approximately n`=0, i. H. to a constant acceleration model.

4 FIG. 4 shows a schematic comparison diagram 400 regarding a collision history in connection with the system from FIG 1 and/or with the procedure 2 . The comparison diagram 400 shows a predicted 2nd integral in the model n=1 or linearized spring model for a so-called rigid barrier crash at 26 km/h in the “real-time” variants according to Eq. (6b) as a first graph 401 and "model time of degree 2 in the model n=1" according to Eq. (6c) as a second graph 402. Since the rigid barrier crash after 3 above (300A) is well described by the spring model n=1, there is little deviation between the predictions within the spring model in real time according to Eq. (6b) or model time according to Eq. (6c). A possible 5-inch forward displacement threshold is exceeded here, for example, after about 19 milliseconds.

5 FIG. 5 shows a schematic comparison diagram 500 regarding a collision history in connection with the system from FIG 1 and/or with the procedure 2 . The comparison diagram 500 shows a predicted 2nd integral in the model n=1 or linearized spring model for a so-called small overlap crash in the “real-time” variants according to Eq. (6b) as a first graph 501 and "model time of degree 2 in the model n=1" according to Eq. (6c) as a second graph 503. Since the small overlap crash after 3 below (300B) is not well described by the spring model, using the model time, which describes the deviation from the spring model, results in a clear correction of the predicted integral towards smaller values. A possible 5-inch forward displacement threshold is exceeded by the first graph 501 or, for example, after about 11 milliseconds when using real time, ie too early, but by the second graph 503 or when using the model time, for example after about 37 milliseconds.

6 FIG. 6 shows a schematic comparison diagram 600 regarding a collision history in connection with the system from FIG 1 and/or with the procedure 2 . The comparison diagram 600 shows a difference between predictions according to Eq. (6b) for the models n=1 or linearized spring model as the first graph 601 and n=0 or model with constant acceleration as the second graph 602, interpolated predicted 2nd integral as the third graph 603 for a rigid barrier crash. The interpolation is based on the value determined from the model time for the model parameter n' of the crash. Since the according to Eq. (12) optimal model parameter n' for the rigid barrier crash is in the range of 1, the adaptive interpolation will deliver a predicted integral which is close to the predicted integral in the model n=1. A possible 5-inch forward displacement threshold is exceeded here, for example, after about 20 milliseconds.

7 FIG. 7 shows a schematic comparison diagram 700 regarding a collision history in connection with the system from FIG 1 and/or with the procedure 2 . The comparison diagram 700 shows a comparison between predictions according to Eq. (6b) for the models n=1 or linearized spring model as the first graph 701 and n=0 or model with constant acceleration as the second graph 702, interpolated predicted 2nd integral as the third graph 703 for a small overlap crash. The interpolation is based on the value determined from the model time for a model parameter n' of the crash. Since the according to Eq. (12) optimal model parameter n' for the small overlap crash is around 0, the adaptive interpolation will deliver a predicted integral which is close to the predicted integral in the model n=0. A possible 5-inch forward displacement threshold is exceeded here, for example, after about 40 milliseconds.

8th FIG. 8 shows a schematic comparison diagram 800 regarding a collision history in connection with the system from FIG 1 and/or with the procedure 2 . The comparison diagram 800 shows a difference between the predictions according to Eq. (6c) for the models n=1 or linearized Spring model with associated model time n=1 as first graph 801 and n=0 or model with constant acceleration and associated model time n=0 as second graph 802, interpolated predicted 2nd integral as third graph 803 for a small overlap crash. The interpolation is based on the value determined from the model time for a model parameter n' of the crash. Here, by using the model time in the model n=1, a clear correction of the corresponding predicted integral for n=1 was achieved. The use of the model time in the model n=0, on the other hand, results in only little change for the predicted integral in the model n=0, which already fits well. Only a small difference needs to be compensated for by interpolating between the two models using the optimal model parameter n'. A possible 5-inch forward displacement threshold is exceeded here, for example, after about 40 milliseconds.

9 FIG. 9 shows a schematic comparison diagram 900 regarding a collision history in connection with the system from FIG 1 and/or with the procedure 2 . The comparison diagram 900 shows a representation of the instantaneous 2nd integral of a small overlap crash as a first graph 901 in comparison to a time-dependent threshold Thd(t) according to Eq. (14) with a base threshold Thd ₀ = 5 inches in model n=1 in the variants "real time" as a second graph 902 and "degree 2 model time in model n=1" as a third graph 903. The use of model time from Model n=1, which is well above real time for the small overlap crash, see also 3 below (300B), leads to a clear upward correction of the threshold value, which leads to a later exceeding, for example at approx. 37 milliseconds, by the instantaneous 2nd integral. 9 is functionally identical to 5 , only the time-dependent prediction term according to Eq. (13) and (14) moved to threshold side.

If an embodiment includes an "and/or" link between a first feature and a second feature, this should be read in such a way that the embodiment according to one embodiment includes both the first feature and the second feature and according to a further embodiment either only that having the first feature or only the second feature.

Claims (13)

Method (200) for activating at least one occupant protection device (120) for a vehicle (100) in the event of a collision of the vehicle (100), the method (200) having the following steps: Reading in (210) an acceleration signal (105, 133) from an interface (131) to an acceleration sensor (102) of the vehicle (100), the acceleration signal (105) being an acceleration of the vehicle ( 100) represented at the collision; Determining (220) a model time (135) related to a predefined collision model as a quotient of a kth integral and a (k-1)th integral of the acceleration signal (105, 133) at the time of detection, optionally scaled with a further factor; Applying (230) the model time (135) to at least one prediction rule for predicting at least one signal property of the acceleration signal (105, 133) after the time of detection and/or to at least one threshold value related to the activation of the at least one occupant protection device (120) by an activation parameter (137) to determine; and Providing (240) a control signal (140) for controlling the at least one occupant protection device (120), wherein the control signal (140) is generated using the control parameter (137). Method (200) according to claim 1 , characterized in that in the step (230) of applying the model time (135) is applied to the prediction rule in order to use the prediction rule as $I_m\left(t\right){\left(\frac{t_{k,n}+t\text{'}}{t_{k,n}}\right)}^{n+m}$

to be defined, where I _m (t) represents an integral of degree m of the acceleration signal (105, 133) at the time of acquisition, t _k,n represents the model time, n represents the model parameter of the collision model and t' represents a predefined, on the activation of the at least represents a time increment related to an occupant protection device (120). Method (200) according to one of the preceding claims, characterized in that in the step (220) of determining the quotient is scaled with a further factor, the further factor corresponding to the sum of k and a model parameter of the collision model. Method (200) according to one of the preceding claims, characterized in that in the step (230) of applying the model time (135) is applied to the threshold value in order to determine a time lapse to generate a dependently adapted threshold value with which an integral of degree m of the acceleration signal (105, 133) is compared at the time of detection in order to determine the control parameter (137). Method (200) according to one of the preceding claims, characterized in that in the step (230) of applying at least one characteristic curve or at least one characteristic map is used as the prediction rule and/or in the step (230) of applying between predefined prediction rules and/or threshold values in Depending on the model time (135) is interpolated to determine the control parameter (137). Method (200) according to one of the preceding claims, characterized in that in the step (220) of determining the model parameters, the model parameters are determined anew at least once, with a parameter candidate being determined from a plurality of parameter candidates for this purpose, which has a minimal difference between the model time (135) and a collision time, which represents a period of time since the beginning of the collision. Method (200) according to one of the preceding claims, characterized in that in the step (220) of determining the model parameters, the model parameters are determined at least once anew, with a newly determined model parameter being selected to be greater than the model parameter if, due to a progressive acceleration curve, the model parameter determined model time (135) is less than a collision time, which represents a period of time since the beginning of the collision, wherein a newly determined model parameter is selected to be smaller than the model parameter if the model time determined using the model parameter (135) is greater than due to a declining acceleration curve is the collision time. Method (200) according to one of the preceding claims, characterized in that in the step (220) of determining a collision model based on an acceleration profile of the form a _n (t) = c t ⁿ is used as the collision model, where a is the detected acceleration of the vehicle (100) during the collision, n represents the model parameter, t represents a collision time, which represents a period of time since the start of the collision, and c represents at least one characteristic collision parameter, in particular speed, mass and/or rigidity of the vehicle (100) . Method (200) according to claim 8 , characterized in that n=1 describes a linearized spring model, n>1 describes a progressive acceleration curve, 0<n<1 describes a degressive acceleration curve, n=0 describes a constant acceleration and n<0 describes a falling acceleration curve. Control unit (130) set up to carry out steps of the method (200) according to one of the preceding claims in corresponding units (132, 134, 136, 138). System (110) for occupant protection for a vehicle (100), the system (110) having the following features: the control unit (130) according to claim 10 ; and the at least one occupant protection device (120), wherein the at least one occupant protection device (120) can be or is connected to the control unit (130) in a manner capable of transmitting signals. Computer program that is set up to the method (200) according to one of Claims 1 until 9 to execute. Machine-readable storage medium on which the computer program claim 12 is saved.

Images