METHOD AND DEVICE FOR CLASSIFYING AN ACCIDENT INVOLVING A TWO-WHEEL CHARGE

Description

[0001] The invention relates to a method and a device for classifying or verifying the plausibility of an accident involving a two-wheeler, in particular an accident involving an electric bicycle. State of the art

[0002] A series of sensors attached to a two-wheeler, such as a bicycle and especially an e-bike, can be used to monitor both the operation of the two-wheeler and its riding condition. Speed ​​sensors can record the wheel speed and, from this, the overall speed of the two-wheeler. Furthermore, tilt sensors, or acceleration and yaw rate sensors, can detect the tilt or inclination on the road surface, and sensors on the pedal crank of a bicycle can detect the riding activity.

[0003] The analysis of these recorded sensor values ​​can detect critical situations or even accidents, for example, by monitoring the temporal behavior of these sensor values. These detected critical situations or accidents can then be used to automatically request assistance or at least to forward information to third parties.

[0004] Document EP 2 632 772 A1 discloses the features of the preamble of claim 1.

[0005] The object of the present invention is to classify an accident so that an accident event in the form of a critical situation can be distinguished from a wheel falling over without a direct collision. Disclosure of the invention

[0006] The present invention relates to a method for classifying an accident involving a two-wheeled vehicle, in particular a bicycle. The method according to the invention can be implemented in a device with an evaluation unit as an algorithm to indicate a collision or a fall of the two-wheeled vehicle to the driver or a third party by means of corresponding generated and / or transmitted information. The device can be used with a two-wheeled vehicle, such as a bicycle or, in particular, an e-bike. Of course, its use is also possible with a motorcycle or other single-track vehicle.

[0007] To classify an accident involving a two-wheeler or a comparable single-track vehicle, the acceleration of the two-wheeler in its plane of motion is recorded. Optionally, the movement towards the elevated surface can also be recorded. For example, two acceleration directions orthogonal to each other can be recorded to capture the acceleration in the plane of motion. This recording can be continuous, providing a constant supply of new sensor data, or it can be time-controlled within a predefined timeframe or at predetermined intervals. Alternatively, the recording can be controlled based on predefined or specifically identified operating conditions. Subsequently, an initial integration variable is calculated or generated based on the temporal integration of a first sensor value; this variable represents, for example, the acceleration in the longitudinal direction or direction of travel of the two-wheeler.Accordingly, a second integration parameter is formed or generated as a function of a second sensor parameter, which, for example, represents the acceleration in the lateral direction relative to the longitudinal direction or direction of travel of the two-wheeler. The core of the invention lies in classifying an accident, which is detected, for example, by a further method, according to whether it is a serious accident or merely a relatively unproblematic fall of the two-wheeler without expected personal injury to the rider. For this purpose, the classification and thus the detection of a serious accident is carried out as a function of the first and second integration parameters.

[0008] The classification can be carried out by comparing the first and / or second integration variable with at least one common or individually assigned threshold value. Alternatively or additionally, the result of the classification can also be made dependent on a direct comparison of the two integration variables.

[0009] The method according to the invention can detect a collision, and thus an accident with suspected personal injury, if the first integration parameter, which, for example, represents the acceleration in the longitudinal direction of the two-wheeler, or its magnitude, exceeds a first threshold value. Optionally or additionally, a collision can also be detected if the first integration parameter, or its magnitude, is greater than the second integration parameter, particularly by a multiple. In this case, it could, for example, be indicated that the (negative) acceleration or deceleration in the longitudinal direction is abrupt and significantly greater than the tipping of the two-wheeler to the side. If a collision is detected, the method can generate information that is forwarded to the driver or to an external party to summon help.

[0010] The method according to the invention can also detect an accident situation in which no personal injury is expected, for example, in an incident where the two-wheeler only tips over sideways. For this purpose, the method checks during classification whether the second integration parameter or its magnitude exceeds a second threshold value. Optionally or additionally, the method can also check whether the second integration parameter is larger than the first integration parameter, in particular by a multiple. Such behavior would imply that the two-wheeler undergoes a lateral tipping motion and does not move substantially forward or decelerate longitudinally.

[0011] According to the invention, the method also captures the roll angle and pitch angle of the two-wheeler, or their change over time in terms of rotation rate. A classification of the two-wheeler's movement can also be derived from the roll angles or pitch angles thus captured, or from the associated rotation rates.

[0012] In a further development of the invention, a mass-specific energy quantity is derived from the first and second integration quantities, which can also be used for classification. For this purpose, a common tilting quantity is additionally derived from the roll angle and the pitch angle, which is also included in the classification. To detect a collision, it is checked whether the energy quantity exceeds an energy threshold and the tilting quantity exceeds a tilting threshold. If the energy threshold is exceeded first, and then the tilting quantity threshold, the method generates information indicating that a collision with likely personal injury has occurred. However, if both the energy value and the tilting quantity remain below their respective thresholds, the method generates information representing only a fall over, particularly a sideways tip-over, or even a bicycle that remains upright.To differentiate between a side collision and a rollover, an additional rollover threshold can be used. Typical rollover values ​​for an upright bicycle range from 0 to 45°.

[0013] Optionally, it can be provided that, based on an energy value which is above a corresponding threshold and a tipping point which is below a corresponding threshold, the procedure generates information that represents a collision of the two-wheeler.

[0014] According to the invention, the energy threshold for detecting a collision is modulated depending on the roll magnitude, roll angle, and / or pitch angle, particularly their absolute values. It can be provided that the energy threshold is chosen to be low for small values ​​of the roll magnitude, roll angle, and / or pitch angle, and that it initially increases slowly and then, for example, exponentially or quadratically as the values ​​of these values ​​increase. A collision is detected when this energy threshold is exceeded. Alternatively, it can be provided that an energy threshold for detecting an impact is used that is initially set high for small values ​​of the roll magnitude, roll angle, and / or pitch angle, and that it initially decreases slowly and then also exponentially or quadratically as the values ​​of these values ​​increase.

[0015] Further advantages arise from the following description of exemplary embodiments or from the dependent patent claims. Brief description of the drawings

[0016] Figure 1 The diagram schematically shows a two-wheeler with a mobile device and a coordinate system in which the two-wheeler moves during normal riding. The block diagram of the Figure 2 shows a possible implementation of a device according to the invention. The flowchart of the Figure 3 describes a possible embodiment of the method according to the invention. Embodiments of the invention

[0017] The invention is described as a two-wheeled vehicle in the form of a bicycle 10, although other single-track vehicles such as electric bicycles, motorcycles, e-scooters, scooters, or even motor vehicles could also be equipped with this invention. In the present case of the bicycle 10, a smartphone 20 is provided as a mobile device, which is attached to the handlebars of the bicycle 10 and is configured for acquiring and processing sensor data. The smartphone 20 serves, for example, as a navigation instrument, as a display of driving dynamics parameters, and / or for controlling the drive of the bicycle 10. To carry out the method according to the invention, the smartphone 20 accesses the sensor data provided by the sensors available in the smartphone 20. Alternatively or optionally, the smartphone 20 can also access sensors that are attached to the bicycle 10.Possible sensors for this purpose include acceleration sensors on one of the wheels and / or the frame, as well as angle sensors and yaw rate sensors. The bicycle typically moves forward along the road surface in the longitudinal direction x in the x-axis. Turning and cornering extend the x-axis to include a lateral direction y. Movement along the vertical axis z occurs, for example, when climbing or descending. Further movements along the vertical axis are generated by rotations along the x-axis or by uneven road surfaces. All these movements along the z-axis can be distinguished by characteristic magnitudes of their rate of change, i.e., dz / dt.For example, riding uphill or downhill involves a longer time constant than the bicycle tipping sideways, primarily along the x-axis. In contrast, road surface irregularities are characterized by very small movements in the z-direction within a short time.

[0018] The sensors within the smartphone 20 are aligned to their own coordinate system. Therefore, mounting the smartphone on the handlebars necessitates recalibration to align it with the coordinate system determined by the movement. This allows the spatially resolved sensor readings from the gyroscope and accelerometer in the smartphone to be used to acquire this defined coordinate system during normal riding. One method for performing this calibration, to align the sensor readings with the xy plane of motion of the bicycle, is, for example, the Euler angle estimation method.

[0019] In the Figure 2A schematic representation of a device according to the invention, which carries out the method according to the invention, is shown. The device may include an evaluation unit 100 with a memory 110, while the sensors 120, 130, and 140 are configured separately, so that the corresponding sensor parameters can be read by the evaluation unit 100. Optionally, however, the device may also include, in addition to the evaluation unit 100 and the memory 110, one or more sensors and thus itself acquire at least some of the sensor parameters necessary for the method according to the invention. In this case, reference is made, for example, to a smartphone 20 mentioned at the outset as a device according to the invention, which already has accelerometers 120, angle sensors 130, and / or gyroscopes 140.The assigned memory location 110 can store both the acquired sensor values ​​and the derived values, such as the integration values, the energy value, and the tipping point. Furthermore, the corresponding threshold values ​​can be stored in memory location 110.

[0020] The evaluation unit 100 uses at least one first sensor 120 to detect an acceleration value representing the movement of the two-wheeler. It is particularly advantageous if the first sensor 120 detects the acceleration in at least both directions x and y of the plane of motion in the form of first and second sensor values. Alternatively, it can also be provided that a separate acceleration sensor detects corresponding first and second sensor values ​​for each direction of movement.

[0021] In further embodiments, additional sensors can be provided to enable or validate the classification of the accident process. For example, an angle sensor 130 can be provided to detect the roll angle and pitch angle of the two-wheeler. Alternatively, two angle sensors can be provided, each independently detecting the roll angle and pitch angle. In addition to the angle sensor 130, or alternatively, the evaluation unit 100 can also detect the sensor values ​​of a yaw rate sensor 140. For classifying the accident process, a first yaw rate ωx about the longitudinal axis x or the direction of travel of the two-wheeler, as well as a second yaw rate ωy about the y-direction perpendicular to the longitudinal axis x, are particularly relevant.As with the accelerometer 120 or the angle sensor 130, it can also be provided for the gyroscope 140 that separate gyroscopes 140x detect the first and second gyroscope rates ω x and ω y, respectively.

[0022] If the system detects that the accident involved a collision or personal injury, the evaluation unit 100 generates a notification to indicate this. In the simplest case, the driver can be informed via a corresponding visual or audible indicator 160. However, it is also possible for this information to be sent to an external receiver 150 to summon assistance. This could be a central control center, a pre-selected contact person, or emergency services in general. Optionally, it can be provided that further units 170 in the device 100 or on the two-wheeler are activated if a collision is detected. For example, the drive unit can be switched off, a power source can be de-energized, or a pre-prepared emergency signal can be sent.

[0023] Similar information can also be provided if the system only detects a minor fall. For example, information about the severity of the fall can be shared to provide advance support to emergency services. This can also influence the extent of intervention required on the components and equipment of the two-wheeler.

[0024] A possible sequence of the process according to the invention will be shown below using the flowchart of the Figure 3This method, or rather the associated algorithm, can differentiate between a frontal or side collision of the two-wheeler with another vehicle, a person, an animal, or an object, and a sideways fall. This allows for a classification of the accident severity as well as an assessment of the accident sequence or its progression. Both the classification and the assessment of the accident sequence can be used directly or in conjunction with other methods to initiate emergency response measures.

[0025] After the process is initiated, at least 300 acceleration values ​​ax and ay are recorded in a first step, representing the acceleration of the two-wheeler in the x and y directions. These two directions essentially comprise the direction of movement or longitudinal direction x of the two-wheeler and a y-direction orthogonal to this x-direction in the typical plane of motion of the two-wheeler. The required acceleration sensor can be permanently installed on the two-wheeler or be part of a display and control device that the rider can temporarily attach to the bicycle or carry with them to record the movement while riding. For example, sensors in a mobile device, such as a smartphone, or a human-machine interface (HMI) could be used to record these and other sensor values.Human Machine Interface (HMI), which is used, for example, to control the drive system of an electric vehicle.

[0026] Optionally, additional sensor parameters can be acquired in step 300, which can be used for further embodiments. For example, yaw rate sensor parameters ωx and ωy can be acquired, representing the yaw rate around the x and y directions, respectively, and thus detecting the (lateral) tilting of the two-wheeler or its forward rollover over the handlebars. Additionally or alternatively, the roll angle φ and / or the pitch angle θ can also be acquired using a suitable sensor. Instead of independently acquiring the roll angle and pitch angle, they can also be derived from the corresponding yaw rate sensor parameters.

[0027] The acquisition of the sensor parameters required for the process can optionally also be carried out in a separate process, so that the inventive process constantly has current sensor parameters available.

[0028] If sensor data, such as acceleration measurements, is acquired using a detachable mobile device, it may be necessary to calibrate the sensor orientation. This is because the spatial directions defined in the mobile device may not correspond to the direction of movement due to its mounting and orientation. Calibration using an Euler angle estimation is one suitable method for this. Optionally, the system can also compensate for the gravitational acceleration component in the acquired acceleration measurements.

[0029] In the next step 310, the recorded acceleration sensor values ​​are processed according to Δ v x = ∫ a x dt Δ v y = ∫ a y dt The system integrates data accordingly to detect changes in velocity in both the x and y directions. To detect a collision, it is sufficient to consider only the two spatial directions in which the bicycle's movement is primarily intended. The movement along the z-axis mainly involves events caused by uneven terrain or by inclines and declines. The integration limits are advantageously chosen based on the time required to acquire the sensor data. For example, at a sampling rate of 100 Hz, integrating 5 to 10 sensor values ​​at a time might suffice, resulting in an integration time of 50 to 100 ms. This timeframe is also sufficient to distinguish a collision from a (sideways) fall over ("ground hit").

[0030] In step 330, the following occurs due to the change in velocity Δ vx or Δ you It is determined in which spatial direction the principal change in velocity occurred. If it is determined that the change in velocity occurred essentially in the y-direction, the change in velocity Δ you If a threshold SW y has been exceeded in the y-direction, or if the change in the y-direction predominates, especially by a multiple, a (lateral) tipping of the two-wheeler is detected, so that in step 350 the driver can be informed or, more generally, information can be generated that the two-wheeler has fallen over.

[0031] Optionally, by selecting the appropriate threshold values ​​for SW x and SW y, a distinction could be made between a frontal (or rearal) or a side collision. For example, a very large value of Δ could... youA side collision is detected because a side collision is expected to result in a greater change in speed than a sideways fall of the two-wheeler. However, if the system detects that the change in speed Δ you If the threshold is very small, for example by comparing it to another, smaller threshold SW y, it can be recognized that there has been neither a collision nor an impact on the ground.

[0032] If step 330 detects that a frontal or side collision has occurred, for example by a higher change in velocity Δ in the x-direction vxOnce an object has been detected in the y-direction, the method recognizes a collision (longitudinal or lateral). Exceeding a corresponding threshold value SW x can also be included to further validate the collision. Different threshold values ​​SW x,n can be used to classify the severity of the collision.

[0033] After the collision is detected in step 330, information regarding the collision can be generated in a further step 380, which can then be forwarded to third parties. For example, a mobile device can be used to call for help via radio or simply to inform a pre-defined contact. Based on a classification of the collision severity, further information can also be forwarded to facilitate the care of the injured driver. The mobile device used can also be the same device that acquired the necessary sensor data.

[0034] After the detection of a fall or collision, the procedure can be terminated or repeated starting from step 300.

[0035] In a further embodiment, after a collision of the two-wheeler is detected, at least one component of the two-wheeler can be activated in an optional step 390 to prevent further damage to the rider and / or the two-wheeler. For example, the drive system of an e-bike could be switched off. However, it is also conceivable that a visual and / or audible warning device could be activated to inform other road users and / or call for assistance. Similarly, upon detection of the two-wheeler falling over, a comparable activation of components of the two-wheeler can be provided in an optional step 360. For example, the rider could be alerted to the fall over if they have parked the two-wheeler and are not nearby.

[0036] As already mentioned at the beginning, it is possible to acquire further sensor parameters to make the detection of a collision more plausible and / or to improve the differentiation from a simple fall over of the two-wheeler. For example, in step 310, the detected roll angle φ and pitch angle θ can be used according to... γ = √ θ 2 + φ 2 A tipping parameter γ can be derived, representing the orientation of the two-wheeler. This tipping parameter γ can be combined with an estimated energy parameter Ekin according to... E kin , xy = Δ v x 2 + Δ v y 2 depending on the change in velocity Δ vx and Δ youStep 330 uses the tipping parameter γ to validate the detection of a collision. It is assumed that a high tipping parameter γ (compared to a tipping threshold SW K) followed by a high energy value E kin (compared to an energy threshold SW E) indicates a collision of the two-wheeler. Conversely, a fall or tipping of the two-wheeler, particularly a sideways tip, is detected when the tipping parameter γ is below the tipping threshold SW K and the energy value E kin is below the energy threshold SW E. The case where one threshold is exceeded by the corresponding value and the other value is not exceeded would need to be validated with additional sensor parameters to detect a collision or a tipping of the two-wheeler. Such and further validations could also be performed in step 330.

[0037] Alternatively, a side collision can also be detected in step 330. For this, a (lateral) rollover or tipping is first detected with a large roll angle φ, e.g., by comparison with a threshold value SW φ, possibly in conjunction with an increased change in velocity in the y-direction. However, if a small roll angle φ is detected, at which there is simultaneously a high kinetic energy E kin,y in the y-direction (for example, via comparison with a threshold value SW Ekin,y), a side collision is detected instead of a simple lateral rollover.

[0038] A further provision in step 330 may include the requirement that a temporal relationship must exist between the individual detected angles or the exceeding of one or more threshold values ​​in order to detect a frontal or side collision or a fall of the two-wheeler. For example, a frontal collision can be detected if a pronounced pitch angle is detected first, followed shortly thereafter by a roll angle. This would indicate that the two-wheeler struck an obstacle and subsequently the rider tipped over with the two-wheeler. Similarly, the sequential decelerations in the longitudinal direction x and the lateral direction y can also indicate a frontal collision.

[0039] Optionally, plausibility checks can also be performed after each detection in step 330 to verify the detection in step 330. In a separate, subsequent step 340, the detection of a fall can be checked, and in step 370, the detection of a collision can be verified. For this purpose, additional sensor parameters or their changes over time can be used. For example, within a short time after the detection of a collision, it could be determined whether the two-wheeler continues to move longitudinally at an explainable speed, thus characterizing a collision as unlikely. In this case, the procedure could be terminated or repeated with step 300. Similarly, the rider righting the two-wheeler in step 340, or the detection of a 180° turn for a maintenance situation, could explain the two-wheeler falling over. In the latter case, the locking mechanism or...Deactivating the drive in step 360 is helpful for changing the chain. However, if step 340 detects that there is no tipping situation, the procedure can be terminated or the system can proceed to step 300 to acquire new sensor data for reassessment.

[0040] In a further embodiment, a preliminary step 320 can be performed before the classification or detection of the collision or the tip-over situation in step 330. In this step, the threshold values ​​used, e.g., SW K, SW E, SW x, or SW y, are defined based on operational parameters of the two-wheeler and / or depending on driving parameters or environmental parameters, e.g., depending on the variance of the detected yaw rate in the x and / or y direction. It is also conceivable that several threshold values ​​are defined for a sensor parameter or a determined or generated value, for example, to classify the severity of the collision. Optionally, at least one of the energy threshold values ​​E kin,x, or E kin,y can be made variable. For example, the corresponding energy threshold could be made to increase or decrease with the tipping magnitude and / or the absolute roll or pitch angles.An energy threshold for collision detection can be chosen such that it is set low for small values ​​of roll magnitude, roll angle, and / or pitch angle, while the energy threshold is also increased as the roll magnitude, roll angle, and / or pitch angle increase, for example, exponentially or quadratically. If the detected or calculated energy exceeds the energy threshold, a collision is detected. Alternatively or additionally, an energy threshold can be used to detect an impact with the ground. In this case, a high energy threshold can initially be set for low values ​​of roll magnitude, roll angle, and / or pitch angle, which is then reduced as the roll magnitude, roll angle, and / or pitch angle increase, for example, exponentially or quadratically.In this case, an impact can be detected if the energy threshold is exceeded.

[0041] The use of probability models instead of thresholds, or in the derivation of thresholds, is also conceivable. This could involve, for example, accessing internal or external databases, the individual sensor measurements, or the derived values. For instance, evaluation unit 100 could have corresponding databases stored in memory 110, or it could access such a database, particularly via a wireless connection.

[0042] In step 330, the detection or classification can also be performed directly from the individually recorded roll angle φ, pitch angle θ, rotation rate parameters, in particular their variance, and / or the directed energy parameters E kin,x or E kin,y. Appropriately assigned threshold values ​​can be provided for this purpose.

Claims

1. Method for classifying an accident event involving a two-wheeled vehicle, in particular a bicycle, wherein the method at least • detects first and second sensor variables (ax, ay) that represent the acceleration of the two-wheeled vehicle (10) in two directions in the plane of movement of the two-wheeled vehicle (10), and • generates a first integration variable (vx) as a function of an integration of the first sensor variable (ax) with respect to time, and • generates a second integration variable (vy) as a function of an integration of the second sensor variable (ay) with respect to time, and carries out a classification of the accident event as a function of the first and the second integration variable (vx, vy), characterized in that the method • detects a roll angle (φ) that represents the rotation of the two-wheeled vehicle about the longitudinal axis, and • detects a pitch angle (θ) that represents the rotation of the two-wheeled vehicle about the transverse axis, and • carries out the classification of the accident event additionally as a function of the roll angle and the pitch angle, and / or the method • detects a first rate of rotation (ωX) that represents the rotation of the two-wheeled vehicle about the longitudinal axis (x), and • detects a second rate of rotation (ωy) that represents the rotation of the two-wheeled vehicle about the transverse axis (y), and carries out the classification of the accident event additionally as a function of the first and the second rate of rotation (ωX, ωy).

2. Method according to Claim 1, characterized in that the method carries out the classification by comparing • the first and / or the second integration variable (vx, vy) with at least one associated threshold value (SWx, SWy), and / or • the first integration variable (vx) with the second integration variable (vy) •3. Method according to either of the preceding claims, characterized in that the method • detects, as the first sensor variable (ax), an acceleration variable that represents an acceleration of the two-wheeled vehicle in the longitudinal direction (x), and • detects, as the second sensor variable (ay), an acceleration variable that represents an acceleration of the two-wheeled vehicle in the transverse direction (y) in relation to the longitudinal direction (x).

4. Method according to Claim 3, characterized in that the method, during the classification, generates an item of information that represents a collision of the two-wheeled vehicle if • the first integration variable (vx) exceeds a first threshold value (SW1) and / or • the first integration variable (vx) overshoots the second integration variable (vy), in particular by a multiple.

5. Method according to Claim 3 or 4, characterized in that the method, during the classification, generates an item of information that represents a situation of the two-wheeled vehicle falling over, in particular laterally, if • the second integration variable (vy) exceeds a second threshold value (SW2) and / or • the second integration variable (vy) overshoots the first integration variable (vx), in particular by a multiple.

6. Method according to one the preceding claims, characterized in that provision is made for the classification to take into account overshooting of the corresponding threshold value (SWφ, SWθ) by the roll angle and / or the pitch angle, wherein in particular overshooting of the threshold values is taken into account in a predetermined time interval.

7. Method according to Claim 6, characterized in that the method • generates an energy variable (E) as a function of the first and the second integration variable, and • generates a tilting variable (γ) as a function of the roll angle and the pitch angle, and • carries out the classification of the accident event additionally as a function of the energy variable (Ekin) and the tilting variable (y), wherein the method • generates an item of information that represents a collision of the two-wheeled vehicle if the energy variable exceeds an energy threshold value (SWE) and the tilting variable (γ) exceeds a tilting threshold value (SWK), or • generates an item of information that represents a situation of the two-wheeled vehicle falling over, in particular laterally, if the energy variable (Ekin) does not exceed the energy threshold value (SWE) and the tilting variable (γ) does not exceed the tilting threshold value (SWK).

8. Method according to one of the preceding claims, characterized in that provision is made for the method • to generate an item of information that represents a situation of the two-wheeled vehicle falling over, in particular laterally, if the first rate of rotation (ωX) exceeds a first rate-of-rotation threshold value (SWωx), or • to generate an item of information that represents a collision of the two-wheeled vehicle if the second rate of rotation (ωY) exceeds a second rate-of-rotation threshold value (SWωx).

9. Apparatus for classifying an accident event involving a two-wheeled vehicle, in particular a bicycle, wherein the apparatus has an evaluation unit (100) which executes a method according to one of Claims 1 to 8, wherein the evaluation unit (100) at least • detects first and second sensor variables that represent the acceleration of the two-wheeled vehicle (10) in two directions in the plane of movement of the two-wheeled vehicle (10), and • generates a first integration variable as a function of an integration of the first sensor variable with respect to time, and • generates a second integration variable as a function of an integration of the second sensor variable with respect to time, and • carries out a classification of the accident event as a function of the first and the second integration variable, in particular by outputting an item of information that represents a collision or a situation of the two-wheeled vehicle falling over, in particular laterally.

10. Two-wheeled vehicle, in particular a bicycle (10), comprising an apparatus according to Claim 9.