DE102022204404A1 - Method and device for evaluating movement of a bicycle - Google Patents

Abstract

The present invention relates to a method (100) for evaluating a movement of a bicycle (1), comprising: detecting (101) a movement of the bicycle (1) in three dimensions; Calculating (102) an indicator whose value depends on the degree to which the detected movement of the bicycle (1) is a correlated movement; and determining (103) whether a movement of interest has occurred by the bicycle (1) based on the calculated value of the indicator.

Description

State of the art

The present invention relates to a method and a device for evaluating movement of a bicycle.

In the case of bicycles, movement of the bicycle is often detected or measured using a motion sensor. A three-axis gyroscope, for example, is used as such a motion sensor. The movement of the bicycle is evaluated for different purposes. It is often necessary to distinguish a movement of the bicycle from noise emitted by the sensor, which is caused, for example, by unwanted contact with the bicycle or a wobbling of the bicycle.

In particular, it is desirable to distinguish permitted movements of the bicycle from unauthorized movements of the bicycle, which, for example, enables theft detection. It is particularly important to detect unwanted movements of the bicycle, which are caused, for example, when a person accidentally bumps into the bicycle or moves the bicycle in its current position, for example to move another bicycle from an adjacent bicycle stand.

Disclosure of the invention

The method according to the invention for evaluating a movement of a bicycle includes detecting a movement of the bicycle in three dimensions, calculating an indicator whose value depends on the degree to which the detected movement of the bicycle is a correlated movement, and determining whether a Movement of interest by the bike is made based on the calculated value of the indicator.

The movement of the bicycle in three dimensions is detected in particular by means of a motion sensor, in particular by means of a gyroscopic sensor. The motion sensor is set up to detect an acceleration or speed of a movement of the bicycle along three orthogonal axes, with linear movements along the axes or angular movements around the axes being detected. The motion sensor is arranged on the bicycle for this purpose. The movement or acceleration of the bicycle is recorded particularly over time.

The movement is preferably described by a rotation speed, which is detected by a gyroscopic sensor.

An indicator is calculated, the value of which depends on the degree to which the detected movement of the bicycle is a correlated movement, in particular a correlated rotation or rotation of the bicycle. A degree of correlation of the movement of the bicycle is thus determined. A movement is called correlated movement if it involves predictable or predictable movement sequences. A movement of the bicycle is viewed as a correlated movement in particular if individual successive movement steps are correlated, i.e. similar to one another. In other words, this means that the degree of correlation describes the degree to which a movement changes randomly over time. A movement is also considered to be a correlated movement if individual movement components of the movement that are aligned orthogonally to one another are similar to one another. For example, carrying a bicycle leads to a low degree of correlation because the bicycle is randomly moved in different directions. In contrast, a random collision of the bicycle leads to a comparatively high degree of correlation, since the bicycle is only pushed in a single direction and the orthogonally aligned movement components of the movement will deflect simultaneously and are therefore similar to one another. The only direction is also determined by the fact that the bike's movement is restricted by its contact with the ground.

The indicator describes how large a degree of correlation is. This means in particular that the indicator shows to what extent individual movement steps from a movement of the bicycle are dependent on one another or are similar to one another. The indicator is therefore suitable for displaying a correlation of a movement of the bicycle over time. The indicator also shows the degree to which the detected movement is random.

A determination is made as to whether a movement of interest has occurred by the bicycle based on the calculated value of the indicator. The indicator is a reliable indicator of whether the bicycle was only exposed to local stimulation, such as an accidental collision with the bicycle, or whether the bicycle was moved away from its location. By evaluating the indicator accordingly, it is possible to conclude that the bicycle is moving or that it is Probability of an existing movement can be concluded. The method therefore makes it possible to detect whether the bicycle is being moved away from its current location.

The subclaims show preferred developments of the invention.

The movement is preferably a correlated movement if it is a movement about an axis of rotation. An indicator is thus calculated, the value of which depends on the degree to which the detected movement of the bicycle is a movement about an axis of rotation. The axis of rotation is in particular an axis of rotation of the bicycle, which is preferably directed along a longitudinal axis of the bicycle and is further preferably defined by two contact points of the bicycle between the wheels of the bicycle and the ground. In particular, a movement around an axis of rotation occurs if the movement is caused by unwanted contact with the bicycle. In this case, the bicycle is on the ground with both tires and is therefore moved in a rotational movement around the two bearing points. Because the movement takes place around the axis of rotation, there is no rotational movement around the transverse axis of the bicycle, so there is a movement with a limited degree of freedom, which leads to an increased degree of correlation in the movement.

Preferably, when determining whether a movement of interest has occurred, a movement of interest is detected based on the indicator if the indicator indicates a degree of correlation of the movement that is less than a threshold value, and it becomes a movement of little Interest detected when the indicator indicates a degree of correlation of the movement that is greater than a threshold value. A movement of interest is a movement in which the bicycle is moved away from its current location. A movement of low interest is a movement in which the bicycle remains in its current location but is moved locally. If the movement is a correlated movement, it can be assumed that the bicycle has both wheels on the ground and is therefore only tilted at an angle. However, if the vehicle is moved at the same time, for example by being rolled or carried, a further movement will also occur that does not take place around this axis of rotation, since such a movement typically involves a movement around another axis of rotation. When the bicycle comes to a standstill and makes an unwanted movement, it is typically only tilted to the right and left around a longitudinal axis. However, if the bicycle is moved, the bicycle is also tilted from front to back about a transverse axis. In particular, when the bicycle is carried, a very uncorrelated movement occurs, which is close to a random movement.

For example, the movement is classified as a movement of interest or a movement of low interest, depending on the calculated value of the indicator. Non-directional movements of the bicycle are preferably movements of little interest, since these typically do not occur in the event of theft, for example. Conversely, directed movements of the bicycle are preferably movements of interest, since these typically occur when the bicycle is moved in the event of theft. This means that a detected movement of the bicycle is preferably classified into at least one of two groups. The further interpretation of the movement in the sense of “permitted” or “unauthorized” is fundamentally freely selectable. However, in the context of theft detection, it is advantageous if a signal is provided when the movement is of interest and is therefore classified as an unauthorized movement, for example to trigger an alarm or notice. The movement is classified depending on the calculated value of the indicator. Thus, the movement is classified into different categories depending on whether the movement is a correlated movement or an uncorrelated movement. A movement of interest is preferably an unauthorized movement and is in particular a movement which does not lead to the bicycle moving from a current position. A movement of little interest is preferably a permitted movement and is in particular a movement in which the bicycle is moved away from its current position. However, it should be noted that depending on the use of the procedure, there may be different definitions as to which movements are considered as permitted movement and which movements are considered as prohibited movement. The method is preferably carried out when the bicycle is parked. In particular, the bicycle is locked.

Depending on whether the indicator is rising or falling with increasing correlation, the indicator can be compared to a threshold to distinguish a move of interest from a move of low interest. For example, a high correlation of the movement leads to a low value of the indicator and a comparatively low correlation of the movement leads to a comparatively higher value of the indicator. This results in a particularly high value being calculated for the indicator and the movement as a movement of interest and therefore can be classified as prohibited movement if the indicator is higher than a threshold value. In particular, the threshold for classifying the motion as a motion of interest is the same threshold used for classifying the motion as a motion of low interest. Depending on whether the indicator is above or below the threshold, a permitted or prohibited movement is detected or the movement is classified as such.

The following relationship preferably applies: A low degree of correlation leads to a high value of the indicator and suggests a movement of interest. In this case, a comparably large space is spanned by a recursive covariance matrix. A high degree of correlation leads to a low value of the indicator and suggests a movement of low interest. In this case, a comparably small space is spanned by a recursive covariance matrix.

Preferably, the movement is only recognized as a movement of interest when the degree of correlation of the movement indicated by the indicator falls below the threshold value for a predefined period of time. This means that short peaks in the time course of the indicator do not lead to false detections.

Further preferably, when detecting the movement of the bicycle in three dimensions, the movement is described in a three-dimensional movement vector, the three-dimensional movement vector describing three angular speeds. Such a vector can be provided by a sensor unit arranged on the bicycle. Preferably, three-dimensional motion vectors are provided by the sensor unit in time sequence, which describe the movement of the bicycle over time. An angular speed is a rotational speed about an associated axis, with the three angular speeds preferably being detected about three axes that are orthogonal to one another and described in the three-dimensional movement vector.

It is also advantageous if calculating the indicator includes determining a recursive covariance matrix based on the three-dimensional motion vector and calculating a determinant of the recursive covariance matrix, using the determinant as the indicator. The covariance matrix can be used to describe the correlation in the detected movement. In particular, it can be determined whether the movement takes place in one plane or in different planes. The determinant therefore describes a low value in particular when the bicycle is stationary or is only moved about one axis of rotation. Accordingly, the determinant describes a comparatively large value if the bicycle is additionally moved, for example pushed or carried. A computational path is thus provided which indicates a correlation in the movement of the bicycle and can be carried out in real time by a computing unit. With the determinant, the determined correlation is summarized in a single value, which enables an assessment of whether the movement is a permitted or an illegal movement of the bicycle. Using the recursive covariance matrix and the determinant, the movement detected by the sensor is both evaluated and converted into a single value, which indicates whether the movement of the bicycle is a correlated movement, or to what degree the detected movement of the Bicycle is a correlated movement. A covariance matrix is referred to as a recursive covariance matrix if it is determined from the three-dimensional motion vector that is recorded at different sampling times. The recursive covariance matrix takes into account a current movement of the bicycle as well as a past movement of the bicycle. In other words, the three-dimensional motion vector is considered at at least two different sampling times in order to determine the recursive covariance matrix. In particular, to generate the recursive covariance matrix in an iteration k, a covariance matrix determined in a previously executed iteration k-1 is used.

Calculating the indicator preferably includes determining a motion vector adapted to a target value from the three-dimensional motion vector, with the recursive covariance matrix being determined based on the adapted motion vector. This means that an amount of the motion vector is adjusted to the target value. The amount is adjusted to the target value by bringing it closer to the target value and limiting it to this. This means that the movement vector is limited within a given width. This means that motion vectors with a particularly high magnitude are reduced in magnitude so that they do not have too much of an influence on the indicator. Stronger movements therefore have a lower weighting than comparatively weaker movements in order to avoid false movement detections. This ensures, for example, that a violent collision of the bicycle is not viewed as an uncorrelated movement, even though it is a movement around the axis of rotation. At the same time, careful carrying or pushing is achieved of the bicycle is still recognized as an uncorrelated movement, since, for example, detected movements around the transverse axis are amplified.

A so-called clamping is therefore carried out for the motion vector. An influence of the magnitude of the motion vectors is reduced and an influence of the orientation of the motion vector or of the course of the movement described by the motion vector is increased.

wobei $\overrightarrow{x}$

der dreidimensionalen Bewegungsvektor ist, $\overrightarrow{\nu }$

der angepasste Bewegungsvektor ist, und b ein konfigurierbarer Wert ist, der den Zielwert definiert.It is also advantageous if the adjusted motion vector is calculated from the three-dimensional motion vector based on the following formula: $$\overrightarrow{\nu }=\frac{\overrightarrow{x}}{\sqrt{1+\overrightarrow{x}\cdot \frac{\overrightarrow{x}}{b^2}}},$$

where $\overrightarrow{x}$

is the three-dimensional motion vector, $\overrightarrow{\nu }$

is the adjusted motion vector, and b is a configurable value that defines the target value.

wobei cov_k die rekursive Kovarianzmatrix für eine k-te Iteration ist, cov_k-1 die rekursive Kovarianzmatrix für eine (k-1)-te Iteration ist, ${\overrightarrow{\nu }}_k$

der angepasste Bewegungsvektor für die k-te Iteration, und a eine konfigurierbare Filterkonstante ist.It is also advantageous if the recursive covariance matrix is calculated based on the following formula: $$cO{\nu }_k=\left(1-a\right)\ast cO{\nu }_{k-1}+a\ast {\overrightarrow{\nu }}_k\otimes {\overrightarrow{\nu }}_k$$

where cov _k is the recursive covariance matrix for a k-th iteration, cov _k-1 is the recursive covariance matrix for a (k-1)-th iteration, ${\overrightarrow{\nu }}_k$

is the adjusted motion vector for the kth iteration, and a is a configurable filter constant.

Both the value b, which can be configured for the target value, and the value a, which can be configured for the filter constant, enable the calculation to be adapted to an individual system.

A device for evaluating a movement of the bicycle, which is set up to carry out the method according to the invention, has all the advantages of the method.

Brief description of the drawings

• 1 ein Ablaufdiagramm eines erfindungsgemäßen Verfahrens zum Bewerten einer Bewegung eines Fahrrades,
• 2 eine schematische Darstellung eines Fahrrades mit einer erfindungsgemäßen Vorrichtung zum Bewerten einer Bewegung eines Fahrrades,
• 3 ein Diagramm, welches einen Grad einer Korrelation einer Bewegung des Fahrrades für unterschiedliche Szenarien darstellt, und
• 4a einen von einer rekursiven Kovarianzmatrix aufgespannten Raum bei einer korrelierten Bewegung des Fahrrades, und
• 4b einen von einer rekursiven Kovarianzmatrix aufgespannten Raum bei einer nicht korrelierten Bewegung des Fahrrades.

Exemplary embodiments of the invention are described in detail below with reference to the accompanying drawing. In the drawing is:

• 1 a flowchart of a method according to the invention for evaluating a movement of a bicycle,
• 2 a schematic representation of a bicycle with a device according to the invention for evaluating a movement of a bicycle,
• 3 a diagram depicting a degree of correlation of movement of the bicycle for different scenarios, and
• 4a a space spanned by a recursive covariance matrix with a correlated movement of the bicycle, and
• 4b a space spanned by a recursive covariance matrix when the bicycle moves in an uncorrelated manner.

Embodiments of the invention

1 shows a flowchart of a method 100 according to the invention for evaluating a movement of a bicycle 1. The method 100 is carried out by a device 2 which is arranged on the bicycle 1. The method 100 is carried out in particular by an electronic control unit which is arranged on the bicycle 1. The arrangement of the device 2 on the bicycle 1 is exemplified in 2 shown.

If the method 100 is initiated, a first step 101 is first carried out. In the first step 101, a movement of the bicycle 1 is detected in three dimensions. In order to detect the movement of the bicycle 1 in three dimensions, a gyroscopic sensor is arranged on the bicycle 1, for example a MEMS sensor.

beschrieben. Der dreidimensionale Bewegungsvektor ist dabei durch eine X-Komponente, eine Y-Komponente und eine Z-Komponente definiert. Die X-Komponente beschreibt dabei eine in eine X-Richtung gerichtete und von dem gyroskopischen Sensor erfasste Bewegung des Fahrrades 1. Die Y-Komponente beschreibt dabei eine in eine Y-Richtung gerichtete und von dem gyroskopischen Sensor erfasste Bewegung des Fahrrades 1. Die Z-Komponente beschreibt dabei eine in eine Z-Richtung gerichtete und von dem gyroskopischen Sensor erfasste Bewegung des Fahrrades 1. Die X-Richtung, die Y-Richtung und die Z-Richtung stehen dabei orthogonal zueinander. Die Bewegungen in X-Richtung, Y-Richtung und/oder Z-Richtung können dabei auch Bewegungen um eine X-Achse, um eine Y-Achse und um eine Z-Achse sein. Die Bewegungen sind dabei durch Geschwindigkeiten beschrieben. Der dreidimensionale Bewegungsvektor $\overrightarrow{x}$

wird von dem gyroskopischen Sensor an die Steuereinheit übertragen, durch welche eine weitere Verarbeitung erfolgt.In detecting the movement of the bicycle in three dimensions in the first step 101, the movement of the bicycle is in a three-dimensional motion vector $\overrightarrow{x}$

described. The three-dimensional motion vector is defined by an X component, a Y component and a Z component. The X component describes a movement of the bicycle 1 directed in an X direction and detected by the gyroscopic sensor. The Y component describes a movement of the bicycle 1 directed in a Y direction and detected by the gyroscopic sensor -Component describes a movement of the bicycle 1 directed in a Z direction and detected by the gyroscopic sensor. The X direction, the Y direction and the Z direction are orthogonal to one another. The movements in the X direction, Y direction and/or Z direction can also be movements around an X axis, around a Y axis and around a Z axis. The movements are described by speeds. The three-dimensional motion vector $\overrightarrow{x}$

is transmitted from the gyroscopic sensor to the control unit, through which further processing takes place.

If the movement of the bicycle 1 was recorded in three dimensions, it takes place in a second one Step 102 of the method involves calculating an indicator whose value depends on the degree to which the detected movement of the bicycle 1 is a correlated movement. The second step 102 is, for example, divided into several sub-steps 102a to 102c. A first sub-step 102a, then a second sub-step 102b and then a third sub-step 102c are carried out.

Der Zielwert definiert dabei insbesondere einen Betrag, welchen der dreidimensionale Bewegungsvektor wobei $\overrightarrow{x}$

bevorzugt annehmen soll. Um den auf den Zielwert angepassten Bewegungsvektor $\overrightarrow{\nu }$

zu ermitteln wird die folgende Formel angewandt: $$\overrightarrow{\nu }=\frac{\overrightarrow{x}}{\sqrt{1+\overrightarrow{x}\cdot \frac{\overrightarrow{x}}{b^2}}},$$

In the first sub-step 102a, a motion vector adapted to a target value is determined from the three-dimensional motion vector $\overrightarrow{x}.$

The target value defines in particular an amount which the three-dimensional movement vector $\overrightarrow{x}$

should be preferred. About the motion vector adjusted to the target value $\overrightarrow{\nu }$

The following formula is used to determine: $$\overrightarrow{\nu }=\frac{\overrightarrow{x}}{\sqrt{1+\overrightarrow{x}\cdot \frac{\overrightarrow{x}}{b^2}}},$$

der dreidimensionale Bewegungsvektor, $\overrightarrow{\nu }$

der angepasste Bewegungsvektor und b ein konfigurierbarer Wert, welcher den Zielwert definiert.There is $\overrightarrow{x}$

the three-dimensional motion vector, $\overrightarrow{\nu }$

the adjusted motion vector and b a configurable value that defines the target value.

ermittelt, so wird der zweite Unterschritt 102b ausgeführt. In diesem Unterschritt 102b erfolgt ein Ermitteln einer rekursiven Kovarianzmatrix basierend auf dem angepassten Bewegungsvektor $\overrightarrow{\nu }$

und somit indirekt basierend auf dem dreidimensionalen Bewegungsvektor $\overrightarrow{x}.$

Die rekursiven Kovarianzmatrix wird bevorzugt basierend auf der folgenden Formel berechnet: $$co{\nu }_k=\left(1-a\right)\ast co{\nu }_{k-1}+a\ast {\overrightarrow{\nu }}_k\otimes {\overrightarrow{\nu }}_k$$

Became the adjusted motion vector $\overrightarrow{\nu }$

determined, the second sub-step 102b is carried out. In this sub-step 102b, a recursive covariance matrix is determined based on the adapted motion vector $\overrightarrow{\nu }$

and thus indirectly based on the three-dimensional motion vector $\overrightarrow{x}.$

The recursive covariance matrix is preferably calculated based on the following formula: $$cO{\nu }_k=\left(1-a\right)\ast cO{\nu }_{k-1}+a\ast {\overrightarrow{\nu }}_k\otimes {\overrightarrow{\nu }}_k$$

Here cov _k is the recursive covariance matrix for a kth iteration. The value k is therefore an index of the iteration carried out. The three-dimensional motion vector and the adjusted motion vector are re-determined in each iteration. The recursive covariance matrix is also determined based on the recursive covariance matrix determined in a previous iteration, which is why it is referred to as recursive.

für die k-te Iteration, wird aus dem dreidimensionalen Bewegungsvektor ${\overrightarrow{x}}_k$

der k-ten Iteration berechnet, der der die aktuelle Bewegung des Fahrrades 1 in der k-ten Iteration beschreibt.The recursive covariance matrix is determined anew each time the movement of the bicycle 1 is detected in three dimensions and thus with each three-dimensional movement vector. A current determination of the recursive covariance matrix is the kth iteration. The (k-1)th iteration is a previous iteration. Thus, cov _k-1 describes the recursive covariance matrix for the previous (k-1)th iteration. The adjusted motion vector ${\overrightarrow{\nu }}_k,$

for the kth iteration, becomes the three-dimensional motion vector ${\overrightarrow{x}}_k$

of the kth iteration, which describes the current movement of bicycle 1 in the kth iteration.

eine Kovarianzmatrix zu berechnen wird ein dyadisches Produkt des angepassten Bewegungsvektor ${\overrightarrow{\nu }}_k$

mit sich selbst gebildet. Das bedeutet, dass der angepasste Bewegungsvektor ${\overrightarrow{\nu }}_k$

v_k mit dem transponierten angepassten Bewegungsvektor ${\overrightarrow{\nu }}_k$

multipliziert wird, um eine Matrix zu bilden. Diese Matrix wird mit der konfigurierbaren Filterkonstante a multipliziert und auf die mit 1-a multiplizierte rekursiven Kovarianzmatrix einer vorangegangenen Iteration addiert. Die rekursive Kovarianzmatrix cov_k wird somit basierend auf dem angepassten Bewegungsvektor ${\overrightarrow{\nu }}_k$

ermittelt.The value a is a configurable filter constant. To based on the adjusted motion vector ${\overrightarrow{\nu }}_k$

To calculate a covariance matrix is a dyadic product of the fitted motion vector ${\overrightarrow{\nu }}_k$

formed with itself. This means that the adjusted motion vector ${\overrightarrow{\nu }}_k$

v _k with the transposed adjusted motion vector ${\overrightarrow{\nu }}_k$

is multiplied to form a matrix. This matrix is multiplied by the configurable filter constant a and added to the recursive covariance matrix of a previous iteration multiplied by 1-a. The recursive covariance matrix cov _k is thus based on the fitted motion vector ${\overrightarrow{\nu }}_k$

determined.

In the third substep 102c, a determinant of the recursive covariance matrix is calculated. A single scalar value is thus determined. The determinant is used as an indicator and shows the degree of correlation of the detected movement of the bicycle.

und somit der Bewegung des Fahrrades 1 in einem dreidimensionalen Raum abgebildet. Liegen diese Bewegungen im Wesentlichen in einer gemeinsamen Ebene oder sind entlang einer Achse ausgerichtet, so führt dies zu einem geringeren Wert der Determinante. Zugleich zeigt die Tatsache, dass die Bewegungskomponenten der Bewegung des Fahrrades 1 im Wesentlichen in einer Ebene sind entlang einer Achse ausgerichtet liegen, auch an, dass die Bewegung eine hohe Korrelation aufweist. Wird durch das Fahrrad 1 eine vergleichsweise unkorrelierte Bewegung ausgeführt, so wird durch die Kovarianzmatrix in dem dreidimensionalen Raum eine Kugel oder Ellipsoid aufgespannt. Daraus ergibt sich, dass die Determinante der rekursiven Kovarianzmatrix einen vergleichsweise hohen Wert aufweist. Die Determinante der rekursiven Kovarianzmatrix dient daher als ein Indikator, der anzeigt, in welchem Grad die erfasste Bewegung des Fahrrades 1 eine korrelierte Bewegung ist.The recursive covariance matrix creates a distribution of occurring motion components in the different dimensions of the adapted motion vector ${\overrightarrow{\nu }}_k$

and thus the movement of the bicycle 1 is depicted in a three-dimensional space. If these movements lie essentially in a common plane or are aligned along an axis, this leads to a lower value of the determinant. At the same time, the fact that the movement components of the movement of the bicycle 1 lie essentially in a plane aligned along an axis also indicates that the movement has a high correlation. If the bicycle 1 carries out a comparatively uncorrelated movement, a sphere or ellipsoid is spanned in the three-dimensional space by the covariance matrix. This means that the determinant of the recursive covariance matrix has a comparatively high value. The determinant of the recursive covariance matrix therefore serves as an indicator that shows the degree to which the detected movement of the bicycle 1 is a correlated movement.

The space 41 spanned by the recursive covariance matrix is in 4a shown as an example of a correlated movement of the bicycle. The spanned space 41 could be described as an elongated ellipsoid or as a cigar shape. The space 42 spanned by the recursive covariance matrix is in 4b exemplary of a comparatively uncorrelated movement of the vehicle wheel shown. The spanned space 42 could be described as a wider ellipsoid. The determinant can be viewed as the volume of the space 41, 42 spanned in each case. The larger the volume of space, the lower the correlation in the movement of the bike.

The movement is in particular a correlated movement if it is a movement about a rotation axis of the bicycle 1. Such movements about an axis of rotation occur in particular when the bicycle 1 is not moved from the spot and is merely tilted over the two contact points of the tires with the ground. This is referred to here as non-directional movement. This occurs, for example, when the bicycle 1 is shaken or tilted or someone accidentally bumps into the bicycle. The axis of rotation could alternatively also be provided by a suspension point of the bicycle on a bicycle stand. However, if the bicycle 1 is moved in a direction, i.e. moved away from its location, the movement does not only take place around the axis of rotation, but the bicycle 1 also tilts forwards and backwards, i.e. the bicycle 1 rotates around its transverse axis and its vertical axis. This in turn results in the fact that the recursive covariance matrix does not create an elongated ellipsoid but rather a space due to the additional movement around the transverse axis and the vertical axis. Such movements are caused, for example, when the bicycle 1 rolls or the bicycle 1 is carried.

It can therefore be seen that shaking or tilting and accidentally hitting the bicycle 1 compared to rolling, driving, lifting or carrying the bicycle 1 lead to different values of the determinant of the recursive covariance matrix. It can therefore also be recognized from the determinant which movement is carried out by the bicycle 1. This makes it possible to determine, based on the calculated value of the indicator, whether a movement of the bicycle has occurred and also to classify this movement of the bicycle 1. This is done in a third step 103.

In the third step 103, it is determined whether the bicycle has moved based on the calculated value of the indicator. The movement is also classified as a movement of interest or a movement of little interest, depending on the calculated value of the indicator, i.e. depending on the calculated value of the determinant of the recursive covariance matrix.

The movement is classified as an undirected or permitted movement if the indicator shows a degree of correlation of the movement that is greater than a threshold value. This is the case when using the determinant as an indicator if the indicator is smaller than a threshold value. The undirected or permitted movement is a movement of little interest, since in this case no anti-theft actions are necessary. The movement is classified as a directed or illicit movement if the indicator shows a degree of correlation of the movement that is less than the threshold value. This is the case when using the determinant as an indicator if the indicator is greater than a threshold value. The directed or unauthorized movement is a movement of interest, since in this case the bicycle 1 moves away from its standing position and therefore anti-theft measures may be necessary.

It is noted that a high value of the determinant indicates a low correlation of movement and a low value of the determinant indicates a high correlation of movement. If the determinant is used as an indicator, a high value of the indicator indicates a low correlation of the movement and a low value of the indicator indicates a high correlation of the movement. The threshold value can be determined mathematically or through an experimental setup.

3 shows an exemplary time course of a value of the determinant and thus of the indicator for different scenarios 31 to 34. In a first scenario 31, the bicycle 1 is shaken, in a second scenario 32, the bicycle 1 is accidentally touched, in a third scenario 33 the bicycle 1 moves rolling and in a fourth scenario 34 the bicycle 1 is carried.

It can be seen that, particularly in the curve for the course of the determinant associated with the fourth scenario 34, comparatively high values occur compared to the other scenarios 31-33. It can also be seen that comparatively higher values occur for the third scenario 33 than for the first and second scenarios 31, 32. The threshold value could therefore, for example, be set so that it lies between the curves shown for the first and second scenarios 31, 32 and the curves shown for the third and fourth scenarios 33, 34 when compared with the value of the determinant. It can also be seen that it is advantageous if the indicator exceeds or falls below the threshold value for a predefined period of time before the movement is considered permitted or unauthorized movement is classified. This means that individual peaks in the curves cannot lead to wrong decisions.

The method 100 is therefore based on an algorithm through which in particular three axial velocities or accelerations are measured, which are measured either along axes that are orthogonal to one another or as angular acceleration or speed about the axes that are orthogonal to one another. This is done, for example, by the gyroscopic sensor, for example a MEMS gyroscope. The gyroscopic sensor is sampled, for example, with a typical frequency of 50 Hz, whereby each sampling process can be viewed as an iteration. The frequency of 50 Hz is chosen merely as an example and it is pointed out that sampling frequencies between 10 and 1000 Hz are particularly advantageous. The movement data or sensor data provided by the gyroscopic sensor are continuously processed in real time in the method 100.

The method 100 is based on the fact that each bicycle 1 has two main positions over which it stands on the ground. Any movement that does not change these points can be considered normal movement for a parked bicycle and can therefore be classified as permitted movement. Movements that are limited to these two positions therefore define a movement around an axis of rotation.

A recursive covariance matrix is calculated. From this, a feature response is determined by determining a determinant of the calculated recursive covariance matrix, which correlates to a volume of the covariance matrix. The determinant ensures that comparatively low values are output as an indicator when the bicycle is stationary and/or rotates around a single axis of rotation. In a three-dimensional movement space, a thin ellipse is created by such a rotation around a single axis, as occurs, for example, when a bicycle is shaken. In contrast, in the case of an uncorrelated movement, such as rolling or carrying the bicycle 1, a comparatively high value is described by the determinant. An uncorrelated rotation in the movement space leads to a three-dimensional figure, which could be described as a sphere. Therefore, the determinant is less sensitive to shaking the bicycle than to moving the bicycle. This is due, among other things, to the fact that a length of the ellipse can be significantly larger than a diameter of the sphere in the movement space and still lead to a smaller volume.

In addition to the above revelation, explicit reference is made to the revelation of 1 until 4b pointed out.

Claims (10)

Method (100) for evaluating a movement of a bicycle (1), comprising: - Detecting (101) a movement of the bicycle (1) in three dimensions; - Calculating (102) an indicator whose value depends on the degree to which the detected movement of the bicycle (1) is a correlated movement; and - Determine (103) whether a movement of interest has occurred by the bicycle based on the calculated value of the indicator. Procedure (100) according to Claim 1 , whereby the movement is a correlated movement if it is a movement about an axis of rotation. Method (100) according to one of the preceding claims, wherein determining (103) whether a movement of interest has occurred based on the indicator: - a movement of interest is detected when the indicator indicates a degree of correlation of the movement that is less than a threshold value, and - a movement of low interest is detected if the indicator indicates a degree of correlation of the movement that is greater than a threshold value. Method (100) according to one of the preceding claims, wherein the movement is only recognized as a movement of interest when the degree of correlation of the movement indicated by the indicator falls below the threshold for a predefined period of time. Method (100) according to one of the preceding claims, wherein when detecting (101) the movement of the bicycle in three dimensions, the movement is described in a three-dimensional movement vector, the three-dimensional movement vector describing three angular speeds. Procedure (100) according to Claim 5 , wherein calculating (102) of the indicator includes: - determining (102b) a recursive covariance matrix based on the three-dimensional motion vector, - Calculate (102c) a determinant of the recursive covariance matrix, using the determinant as the indicator. Procedure (100) according to Claim 6 , wherein calculating (102) of the indicator includes determining (102a) a motion vector adapted to a target value from the three-dimensional motion vector, wherein determining (102b) the recursive covariance matrix takes place based on the adapted motion vector. Procedure (100) according to Claim 7 , where the adjusted motion vector is calculated from the three-dimensional motion vector based on the following formula: $$\overrightarrow{\nu }=\frac{\overrightarrow{x}}{\sqrt{1+\overrightarrow{x}\cdot \frac{\overrightarrow{x}}{b^2}}}$$

- where $\overrightarrow{x}$

is the three-dimensional motion vector, - where $\overrightarrow{\nu }$

is the adjusted motion vector, and - where b is a configurable value that defines the target value. Procedure (100) according to Claim 8 , where the recursive covariance matrix is calculated based on the following formula: $$cO{\nu }_k=\left(1-a\right)\ast cO{\nu }_{k-1}+a\ast {\overrightarrow{\nu }}_k\otimes {\overrightarrow{\nu }}_k$$

- where cov _k is the recursive covariance matrix for a k-th iteration, - where cov _k-1 is the recursive covariance matrix for a (k-1)-th iteration, - where ${\overrightarrow{\nu }}_k$

the adjusted motion vector for the kth iteration, and - where a is a configurable filter constant. Device (2) for evaluating a movement of a bicycle (1), which is set up to carry out the method (100) according to one of the preceding claims.

Images