DE102008042277A1 - Method for determining frictional connection or force transfer, particularly controlling units, particularly for person, involves detecting frictional connection or force transfer of defined size by sensor - Google Patents

Abstract

The method involves detecting frictional connection or force transfer of defined size by a sensor (14). Multiple models are used for evaluating measurements of frictional connection or force transfer at different times and time intervals. An independent claim is included for a device for determining a frictional connection or force transfer, particularly controlling units, particularly for person.

Description

State of the art

The The invention relates to an apparatus and a method for detection the contact state at a well or an axis after the genus of the independent claims.

As part of the progressive spread of driver assistance systems in the motor vehicle, the driver is partly relieved of his vehicle management task to such an extent that he might be inclined to remove his hands from the steering handle (steering wheel). Since this could pose a security risk, one has considered ways to detect whether the driver keeps his hands on the steering wheel. For example, from the Scriptures DE 196 31 502 C1 a steering wheel with an optoelectronic sensor known which detects the elastic deformation of the casing of the steering wheel due to the embrace by the driver. For this purpose, a special sensor is provided, which is attached to this, or exclusively for this purpose.

Another procedure shows the font DE 10 2007 039 332 A1 on, where the steering movement of the steering handle with a torque sensor and an angle sensor is detected. By means of modeling, a distinction can be made between the free steering movement of the steering handle and a steering torque which the driver exerts, and thereby the contact state can be detected. In this document, sensors are used which do not have the sole purpose of detecting the contact state and therefore offer synergy potential. The effort to obtain such a model for a steering system is very high, since many boundary conditions have to be taken into account, because the steering must be measured type-dependent in order to determine the parameters of the model.

A mathematical description by means of a torque equation for estimating the steering torque of the driver is for example in the US 6,219,603 disclosed. In this method, however, contact states can not be clearly detected with small driver steering moments.

Disclosure of the invention

The This invention relates generally to a method and apparatus to determine whether a traction, or a power transmission between a controlling unit and a stored shaft or an axis takes place. Not limiting becomes the invention described in particular using the example of a steering handle, wherein the controlling unit a person, in particular the driver of a motor vehicle can be.

The inventive method and the device having the features of the independent claims In contrast, the advantage that the method with a Sensor does not work out as an exclusive task has the detection of the contact state. This allows Cost savings, space and other synergies can be used.

Advantageous is still carrying out several measurements and that Evaluate multiple readings that are so in a physical Related, that a more detailed assessment of the Contact state can be performed.

Of the The underlying idea of this method is detection an influence on the free play of the steering apparatus by an outer Source. As such source usually comes only the steering power of the driver in question. Would the driver do that Do not touch the steering wheel, this would turn the steering wheel move like those picked up by the chassis Forces on the steering column and the steering wheel act. Theoretically, if the driver's request was exactly the one from the Chassis coming from steering movement coincides and the driver thus exerts no force on the steering handle, it can not be determined whether a contact state prevails. Just a theoretically ideal driver can achieve it, no forces to exercise on the steering wheel. A real driver practices whenever he has contact with the steering wheel, low steering torque (not noticed by the driver) on the steering wheel. Also in ideal straight-ahead driving. However, these steering moments so small that these are due to the limited sensor resolution no longer distinct from the suggestions of the roadway can be. Because in reality vehicles out Drift the track (road incline, crosswind, different Tire pressure, etc.) will change after a short time (depending on the speed) at least a minimum correction by the Driver needed, which is so recognizable (filtering). Therefore is the idea in the manner described a steering force, or a To detect and evaluate the steering torque for the purpose good usable.

In this case, the sensor does not have to directly record the physical magnitude of the force or the moment, but the variable measured by the sensor must be suitable in order to be able to determine the force or the moment from it, if appropriate after further processing. Optionally, several sensors can be used, in particular if z. B. several physical quantities must be measured to calculate the desired force or moment. Also, the sensor does not need to be directly on the shaft / axis be brought, as long as he the physical size, z. B. by means of a device measures.

It Must be at least one size available that is influenced by contact so that it is another shows characteristic behavior than without contact. The to be used Size must be for a case to be recognized a characteristic behavior / pattern / finite number of patterns train (Applies to n = 2). For several to distinguish In each case, each case has to develop its own characteristic.

Advantageous is the use of a machine learning algorithm, which is known to be through a learned knowledge base, allows regularities to recognize in the learning data, thus later also unknown To be able to judge data. Unlike for example Model-based procedures do not have to detail each one Steering recorded, defined and measured, but the overall system can be measured quickly once in the end.

Farther Advantageous is the possibility of the machine learning system to train with multiple steering / steering types, using the same classification system; in a modeling would have a model for each steering type are provided because the physics must be known exactly. Another positive to the machine learning procedure inherent property is that a record can be used for many steering, even if this was learned only at a steering, because machine learning systems tend to generalize and thus also unknown data classifiable do. Of course, learning on several increases Systems accuracy. Patterns are taught and not, as with models, exactly the steering modeled. Leading also to a tolerance to measurement errors, or failed Measurements, e.g. In case of failure of a sensor, sensor drift or production spreads, whereas a calculating (model) system is hard to do to calculate something missing.

The The system according to the invention is thus taught first, that means there are, for example, measuring experiments and test series typical measured values or measured value profiles, For example, the steering torque manually classified and in the System introduced. The classification corresponds to the simplest Case of distinction between contact state made or Contact state not established. With several such series of measurements or data sets can train such a classification system become. The expert is a suitable machine learning algorithm to select, here z. B. neural networks or decision trees are used after the Training phase can independently classify measured value series.

For example, the determination of the measured data can be as follows:
Over a specific time interval Δt, measured data are acquired and re-evaluated after a new measured value or a measuring trigger has become available. In the method according to the invention, the information about the contact state is generally not contained in the individual measured value, but in the course of the measured value, for example of the steering torque signal.

Advantageous is the use of multiple classifiers and / or additional Signal processing and / or a Klassifikatornetzwerkes. Dependent the application case can be used to achieve a more accurate detection. Furthermore, depending on the topology, computation time can be saved (eg throughput time through filters). It is conceivable to switch of classifiers, depending on boundary conditions during operation, z. B. the use of different classifiers depends from the vehicle speed, as speed-dependent Vibrations on the part of the road surface the characteristic of the input signals affected.

Farther conceivable is a structure of the classification system adapted to the application, eg the use of different single classifiers for different measured values, or groups of measured values, whose results be merged into another classifier. This can also other parameters, such. Take into account measured values. Measurement signals or classified signals may also undergo signal processing steps (eg fast Fourier transformation, averaging, standard deviation).

Advantageous it is if only by means of a force or torque measuring device the measurement is performed (eg steering torque sensor), as so other sensors can be saved, in particular a steering angle sensor, as in the prior art for this Purpose is still used. The renunciation of this holds the Advantage that a source of error (since only 1 sensor) less available is.

Instead of An axial force or torque sensor could also be Tensile, bending or torsion sensor are used, the voltages and Force in the shaft / axis measures, for example, a strain gauge. Likewise, such a strain gauge can also be used in storage the shaft / axle be arranged to tensile, bending and torsional forces to measure, as the forces of the steering column on transfer the storage.

Advantageous is the use of a filter, which in the transition areas, in which the contact with the steering wheel is made or interrupted, provides a defined contact state signal. For this purpose is the signal from the evaluation unit, ie the classification, or speed-dependent filtered so as to debounce to effect. Depending on the application, the parameters vary of the filter (for example, time constants).

Advantageous is the use of a steering actuator or actuator, which zu Test purposes, a steering torque or a steering force on the steering handle to test whether an opposing force / moment present from the hands of the driver got to. This method is ideally applied in which Uncertainties in the classification occur. This can be, for example have physical causes, depending on the steering type or steering torque sensor or force sensor, as well as at certain Driving situations such as straight-ahead driving occur in which the driver the steering wheel only loosely holds in the hand. To be in such a Case the classification to check or at all To perform the steering actuator would be a steering torque to exercise the steering handle, which experience shows for the driver is usually imperceptible, whereupon the moment sensor exerted, even by loose hands Counterforce / Moment can measure. Depending on the application and design Such test moments / forces can be considered in the short term Test pulses or permanent small force to be formed.

Advantageous it is when the evaluation and classification takes place only if it's necessary. That way, resources can be (Computing time) can be saved. Thus, if necessary, the embedding also succeeds into a foreign controller, which can save costs. For example can be turned off in a turn the system, because then It is clear anyway that the driver is steering the steering wheel. Possibly. can Measurement (without evaluation) be continued, so that the Memory is filled when a contact detection should take place again. Particularly important is a contact detection For example, if a Lane Keeping System (LKS) is active. For safety reasons the driver is forbidden to give up the driving, why he might because of the comfortable driving feel would be tempted.

Farther an evaluation can also take place in particular if it it is necessary to check the contact state, which is especially true will be the case when a steering movement is imminent. So, for example a video sensor or a navigation system notice that in the direction of travel ahead of a turn, then in time for the turn, or Lane change, to check the contact state and if necessary Initiate actions or warnings. Different There are other driving conditions (straight ahead) that are not critical are because letting go of the steering wheel is no other driving movement would cause. But here are the usual Safety regulations requiring the use of this procedure allow or forbid to take into account.

advantageously, can also be checked with which Quality the measurability is guaranteed, or at all the necessity of a measurement is present in order to obtain a degree of quality for the classification and a higher one To obtain meaningfulness.

Definitions

Machine learning (German: machine learning) is a generic term for the artificial generation of knowledge from experience. One artificial system learns from examples and can after completion of the Generalize learning phase. That is, it does not just learn memorizing the examples, but it recognizes laws in the learning data. This allows the system to assess unknown data. This can be procedures of supervised as well unsupervised learning, where for the present invention is likely to be the choice of a supervised one Learning system seems more appropriate. This is done in a training or learning phase records the correct classification given to the system from which the system learns later too for unknown records a correct classification perform. Different models, algorithms (eg neural networks, EM algorithms, Hidden Markov models, decision trees, random forest (or derivatives thereof), C4.5, Bayesian networks).

The term steering handle or steering wheel describes the functional effect of a device for transmitting a steering force to a steering. In the present case, however, the mention of a steering handle serves as an example, since the method according to the invention can always be used when it is to be checked whether a power transmission or an adhesion between a system - which for example a driver or a driver's hand or a robot or another Device may - and a shaft for the transmission of torque, or an axis for the transmission of tensile and compressive forces, or combinations thereof, takes place. This method may for example also find application in the steering of any vehicles, or z. As the control of vehicle parameters, such as the speed, possibly also via axial transducers, as is the Use for checking the contact state of an operator on a machine conceivable. For the use of the method safety-critical aspects may play a role, such as driving vehicles or operating machinery, as well as the successful non-positive or force-transmitting connecting two systems, such as a tap with a filler neck.

Of the Use case to determine a contact state reflects physically in the test whether a power transmission takes place, or can take place or comparable a force conclusion is present. In the case of using a steering handle in one Vehicle is under contact state to understand that the driver put his hands on the steering wheel so that he has one Could perform steering movement. Modifications, like steering with the legs or otherwise, are from this invention not differentiated.

figure description

embodiments The invention is illustrated in the drawings and in the following Description explained in more detail.

It demonstrate

1 three different embodiments of force effects that can be treated by the method according to the invention.

2 a block diagram of the components involved,

3 exemplary measurement series.

In the 1a becomes a physical active component a torque 13 used which on a shaft 12 is present, for example, part of a steering handle and a steering wheel 11 can be. There is a torque sensor for measuring the torque 14 ,

In 1b the physical action principle is an axial force 17 , An axis 15 moves here, for example, axially in a guide, the force exerted by the movements 17 through a sensor 16 is recorded. Such a device can be used, for example, to control the speed or other variables that are adjustable via an axial encoder. Depending on the application, sensors can 16 also be designed to shear and bending forces on the axis 15 to eat.

In 1c a simplified system is shown which also has an axial force 21 used as a physical working principle, which via a rope 18 is transmitted. In the drawing is with 19 a spring is shown which causes a restoring force. A sensor 20 can now measure the force, or the stop of the spring, if no force 21 acts and on the basis of the characteristic of the force curve, for example, classify which load moves the rope. So it could be checked whether at a ski lift (tow lift) a person, a fallen person who is entrained hangs, or the hanger grinds in the snow.

In 2 is the embodiment of 1a shown in the system context in which the sensor signal 22 a preprocessing filter 23 , or a preprocessing is supplied to then in Messdatenzhischenspeicher 24 to be filed. The evaluation unit or the classifier 25 Read from the memory a number of measurements and pass them to the evaluation algorithm, so that classification results 26 (eg yes / no) that indicate whether the driver has his hands on the steering wheel or not. It is now optionally possible to use an already described filter 27 which debounces the classification results and includes characteristics of a Schmitt trigger. The resulting output result 28 can now be used to issue warnings to the driver or to inform other driver assistance systems and, if necessary, to initiate (re) actions. In the extreme case, for example, this may be an emergency braking or stopping a system, for example of the radio, to alert the driver, or lowering the triggering threshold of an airbag if it is to be assumed that an increased safety risk exists.

In the 3a and 3b measurement series are numerically or graphically plotted, which should illustrate a classification algorithm. The exact operation of these classification algorithms is not a core component of this invention, but is known in the art. The measurement series denoted by 1, 2 and 3 represent learning data sets with which the system was trained, the classification, here shown as 0 or 1, being called for teaching. If, after the learning phase, an unknown data record appears, represented in the graph as a characteristic with circles, a classification is possible in the system. Looked at visually, a key feature in the graph could be the monotonicity rather than the monotony of the curves, and the classification could be done by it. In this case, a data record is acquired over a time interval with a specific duration .DELTA.t and depends, for example, on the resolution of the sensor and in dependence on the cyclic sampling time of the measured values, since an optimum exists for the .DELTA.t depending on the ambient conditions. This is idea It may be empirically determined in experiments. Furthermore, this value may vary depending on the machine learning algorithm used.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - DE 19631502 C1 [0002]
• - DE 102007039332 A1 [0003]
• - US 6219603 [0004]

Claims (13)

Method for detecting a frictional connection and / or a force transmission - between a controlling unit, in particular a person, - and a stored shaft ( 12 ) or an axis ( 15 . 18 ), - wherein a sensor for detecting ( 14 ) at least one of the traction and / or power transmission ( 13 . 17 . 21 ) determinable size ( 22 ) - and doing more than one measurement - at different times - makes within a time interval, characterized in that - from the more than one measurement, a pattern is evaluated, - which allows the determination. Method according to claim 1, characterized in that that - To evaluate the pattern at least one machine learning Algorithm is applied. Method according to claim 2, characterized in that that - Additional classifiers and / or signal processing in combination with the at least one machine-learning algorithm be applied. Method according to one of the preceding claims, characterized in that - the measurements are made exclusively by means of a force ( 16 . 20 ) or torque measuring device ( 14 ) be performed. Method according to one of the preceding claims, characterized in that - additionally a time- or speed-dependent filtering ( 27 ) is used. Method according to one of the preceding claims, characterized marked that - a steller a test force and / or torque on the shaft or axle. Method according to one of the preceding claims, characterized marked that - the one evaluation only then takes place, - if the measurability of a frictional connection and / or Power transmission is expected and / or necessary. Method according to one of the preceding claims, characterized marked that - the evaluation and / or determination being affected, - depending on the measurability or the need to measure a traction and / or power transmission. Device for detecting a frictional connection and / or a force transmission - between a controlling unit, in particular a person, - and a stored shaft ( 12 ) or axis ( 15 . 18 ), - with a sensor ( 14 . 16 . 20 ) for detecting at least one of the traction and / or power transmission determinable size ( 22 ) - and an evaluation unit ( 25 ), characterized in that - the evaluation unit ( 25 ) is so pronounced, - by means of the recorded quantities ( 22 ) - to make the determination possible. Device according to claim 9, characterized in that that - The sensor as a force or torque measuring device is trained. Device according to one of claims 9, 10, characterized in that - the supported shaft ( 12 ) or axis ( 15 . 18 ) Part of a steering handle ( 11 ) is for a vehicle. Device according to one of claims 9-11, characterized in that - a filter unit ( 23 . 27 ) additionally performs a time- or speed-dependent filtering. Device according to one of claims 9-12, characterized in that - an actuator is present, - a test force and / or torque on the shaft ( 12 ) or axis ( 15 . 18 ) exercises.