Circuit and method for determining the orientation of a magnet and joystick

Abstract

A circuit and method for determining the orientation of a magnet and a joystick are disclosed herein. An embodiment of a circuit (150) for determining information about the orientation of a magnet includes an input interface (152) configured to receive components of a magnetic field generated by a magnet. Evaluation logic (154) corresponds to at least one trained neural network (156) and is configured to determine the information about the orientation of the magnet based on the received components.

Description

Technical Field

[0001] Embodiments relate to circuits and methods for determining the orientation of a magnet and to a joystick having a magnet. Background Art

[0002] For example, the position or orientation of a magnet is detected in a joystick to determine which orientation the joystick is currently in. Joysticks are used, for example, to control aircraft, where they typically have and control three degrees of freedom of movement. Tilting the joystick forward or backward can be used to control the pitch angle of the aircraft, wherein a swing to the left or right can be used to control the roll angle of the aircraft. Similarly, rotation of the joystick along its longitudinal axis can control the yaw angle of the aircraft. In addition to aircraft, similar joysticks are also used for agricultural commercial vehicles, cranes, excavators, etc. In order to determine the degrees of freedom, magnets are used that are attached to the joystick and generate a magnetic field, the components of which are measured and used to calculate the orientation of the joystick.

[0003] For this purpose, for example, a magnet magnetized perpendicular to the longitudinal axis can be used to rotate the joystick along its longitudinal axis, the magnetic field of which is measured by means of a magnetic field sensor. If the direction of the magnetic field is known, the geometric orientation of the magnet and therefore the rotation of the joystick are also known. Independently of this, for example, the swing of the joystick to the left or right or forward or backward can be determined by means of a magnet magnetized in the direction of the longitudinal axis of the joystick. If the direction of the magnetic field is known, the angle of the magnet generating the magnetic field and therefore the joystick relative to the magnetic field sensor can be calculated, i.e. at what angle the joystick is tilted to the left / right and forward / backward.

[0004] Since the two measurements involve independent determinations, the simultaneous determination of the three mentioned degrees of freedom usually requires a high mechanical outlay. For example, first a system would have to be created in which the joystick can be rotated about the longitudinal axis relative to a fixed reference, which allows the yaw angle to be determined. However, in order to be able to simultaneously determine the inclination about the two other axes, another mechanical solution must be found, which enables the joystick to be moved together with the entire system for measuring the yaw angle relative to another fixed reference. Such a system is mechanically complex and therefore expensive, and can be relatively susceptible to mechanical damage. There is therefore a need to determine the orientation of a magnet in space in a different way and at a lower cost. Summary of the invention

[0005] An embodiment of a circuit for determining the orientation of a magnet, the circuit having an input interface configured to receive components of a magnetic field generated by a magnet, is achieved by having the circuit having an evaluation logic corresponding to at least one trained neural network, which determines information about the orientation of the magnet based on the received components. With the aid of the trained neural network, field groups that cannot be analytically calculated can also be learned, so that from these field groups the orientation of the magnet generating the magnetic field, i.e., for example, its position and orientation, can be inferred.

[0006] Thus, an embodiment of a method for determining information about the orientation of a magnet comprises measuring components of a magnetic field generated by a magnet, and processing the measured components using a trained neural network to determine information about the orientation of the magnet.

[0007] An embodiment of a joystick comprises a support, by means of which the joystick is supported relative to the housing in a manner that it can move with at least three degrees of freedom. The joystick has a part that can be moved by a user and an inner part, which are opposite each other on different sides of the support in the longitudinal direction. A magnet is arranged on the joystick. At least one magnetic field sensor is arranged in a fixed position relative to the housing and is configured to measure a magnetic field component of a magnetic field generated by the magnet. According to some embodiments, this simple mechanical arrangement may be sufficient to measure information about the desired degrees of freedom of movement of the joystick by means of only one magnet (for example, by means of an embodiment of the circuit). BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Some examples of the apparatus and / or method are further described below with reference to the accompanying drawings, which show:

[0009] Figure 1 An embodiment of a joystick and an embodiment of a circuit for determining information about the orientation of a magnet are shown;

[0010] Figure 2 shows examples of possible orientations of magnets in space;

[0011] Figure 3 Another embodiment of a joystick is shown;

[0012] Figure 4 An example showing the topology of an embodiment of a neural network; and

[0013] Figure 5 A flow chart showing an embodiment of a method for determining information about the orientation of a magnet is shown. DETAILED DESCRIPTION

[0014] Various examples will now be described in more detail with reference to the accompanying drawings. In the accompanying drawings, the thickness of lines, layers and / or regions may be exaggerated for clarity. Other examples may have various modifications and alternative forms to those shown below, and therefore cover various modifications, correspondences and alternatives that fall within the scope of the present disclosure.

[0015] Throughout the description of the figures, identical or similar reference numerals denote identical or similar elements, which may be implemented in an identical or modified manner while providing the same or similar functionality when compared to one another.

[0016] It should be understood that when an element is referred to as being "connected" or "coupled" to another element, these elements may be connected or coupled directly, or connected or coupled through one or more intermediate elements. When "or" is used to combine two elements A and B, this should be understood to disclose all possible combinations, i.e., only A, only B, and A and B, unless otherwise defined explicitly or implicitly. Alternative expressions for the same combination are "at least one of A and B" or "A and / or B". The same applies to combinations of more than two elements.

[0017] The terms used to describe specific examples herein are not intended to limit other examples. When using the singular form (e.g., "one", "the one"), it is not explicitly or implicitly defined that the use of only a single element is mandatory, and other examples can also use plural elements to achieve the same function. When the functions implemented in the case of using multiple elements are described below, other examples can achieve the same function in the case of using a single element or a single processing entity. It should also be understood that when used, the terms "include", "comprise", "have" and / or "contain" specify the presence of given features, integers, steps, operations, processes, elements, parts and / or groups thereof, but do not exclude the presence or addition of one or more other features, integers, steps, operations, processes, elements, parts and / or groups thereof.

[0018] Unless otherwise defined, all terms (including technical and scientific terms) used herein are used with their common meanings in the art to which the examples belong.

[0019] Figure 1 An embodiment of a joystick and an embodiment of a circuit for determining information about the position of a magnet are schematically shown. Information about the position of a magnet in the joystick is determined by way of example, wherein according to other embodiments, information about the position of other magnets that are supported or freely located in space can also be determined.

[0020] Joystick 100 has joystick 102, and this joystick extends through support 104 on the longitudinal direction 106 of joystick, so that joystick 102 has part 102a and inner part 102b that can be operated by user outside the shell that is not shown here.For the sake of clarity, the support connected with the shell of joystick is not shown.At the end of inner part 102b, magnet 108 is arranged, and this magnet is magnetized perpendicular to longitudinal axis 106, that is, is magnetized radially with respect to magnet itself.Spherical support 104 realizes magnet 108 to move with 3 degrees of freedom, that is, rotates with first rotation angle 110 around longitudinal axis 106, rotates with second rotation angle 112 around first axis 113 extending through support 104, and rotates with third rotation angle 114 by second axis 115 extending through support 104 and perpendicular to first axis 113. In the illustrated example of a joystick, the longitudinal axis 106 rotates in each case together with the magnet 108 , so that the longitudinal axis 106 also rotates relative to the other two axes.

[0021] By Figure 1 , the position of the magnet 108 in space is changed by the user's interaction with the user-movable part 102a of the operating lever. Here, the position of the magnet in space and any parameter set that can unambiguously describe the arrangement of the magnet in space, such as the position of the center of gravity of the magnet and its orientation, i.e., the axis defined, for example, by the direction of the strongest magnetic field lines, can be understood as the position of the magnet 108 in space.

[0022] In order to measure the magnetic field generated by the magnet 108, an embodiment of the joystick 100 has at least one magnetic field sensor 120, which is arranged in a fixed position relative to the housing. The support of the support 104 is also fixed in position relative to the housing, so that when the joystick 102 moves, the magnet 108 moves and rotates relative to the magnetic field sensor 120, that is, the orientation of the magnet 108 relative to the magnetic field sensor 120 changes. Figure 1 In the exemplary embodiment shown in FIG. 1 , a further optional magnetic field sensor 122 is additionally shown, which, however, can also be omitted according to some exemplary embodiments.

[0023] Magnetic field sensor 120 measures the component of the magnetic field generated by magnet 108, with the aid of circuit 150 for determining the orientation of magnet 108. Figure 1108 to evaluate these components in order to determine information about the orientation of the magnet 108. For this purpose, the circuit 150 has an input interface 152 to receive components of the magnetic field generated by the magnet 108, i.e., those components measured by the magnetic field sensor 120, for example. The circuit 150 also has an evaluation logic 154 corresponding to at least one trained neural network 156. Through the function of the trained neural network, the selected logic 154 is constructed or can determine information about the orientation of the magnet 108 based on the components of the received magnetic field.

[0024] The information about the orientation need not necessarily be the complete information required to fully and univocally describe the position and orientation of the magnet in space, although this is the case in some embodiments. It is also not necessary to determine the information about the orientation so that the orientation exists directly in a predetermined coordinate system, for example in the form of a vector for the position of the magnet's center of gravity and for its orientation. On the contrary, it is sufficient that the information about the orientation is determined so that the position and orientation of the magnet can be derived univocally from specific information. It can be advantageous to determine information about an orientation that does not directly correspond to a position coordinate or a sought parameter, for example, in order to improve the prediction accuracy of a neural network. In the case shown for an application in a joystick, for example, determining three angles, i.e., a first rotation angle 110, a second rotation angle 112, and a third rotation angle 114, is sufficient as information about the orientation.

[0025] In this case, for example, complete information about all three angles can be determined using a single neural network. Figure 1 In the case shown in , an embodiment of the circuit 150 can also have one of the evaluation logics 154, which corresponds to a trained neural network that determines only a subset of this information. The evaluation logic 154 is designed so that it corresponds to a first neural network 156, which determines the second rotation angle 112 and the third rotation angle 114 based on the components of the magnetic field. The functionality of the second trained neural network 158 for determining the first rotation angle 110 can be in Figure 1 However, this is to be understood as optional. Therefore, further embodiments may each have only one of the two neural networks.

[0026] exist Figure 1 In the case of a magnetization of magnet 108 shown in FIG. 1 that is perpendicular to longitudinal axis 106, second rotation angle 112 and third rotation angle 114 cannot be calculated analytically with sufficient accuracy from the measured components of the magnetic field, especially since magnet 108 can also be rotated about longitudinal axis 106. In contrast, second rotation angle 112 and third rotation angle 114 can be determined with sufficient accuracy using a suitably trained neural network.

[0027] To this end, neural networks of different types and topologies may be used. For example, the table below shows an example of the results of a systematic study of different network types and topologies for determining a first rotation angle 110 (α), and a second rotation angle 112 (tilt Y) and a third rotation angle 114 (tilt X).

[0028]

[0029] For the purposes of the study, a fully connected feed-forward network with three hidden layers was studied, which should not imply that other network topologies could not also lead to similar or possibly even better prediction results and therefore be used in other embodiments. The studied feed-forward network is suitable for the specific application within a joystick because it requires a relatively low computational complexity compared to other network types, which is advantageous when the neural network is to be implemented within the controller of the joystick.

[0030] In order to determine the second rotation angle 112 and the third rotation angle 114, different network topologies were investigated, each of which has a regression layer as the last layer, ie, a regression network. Examples of the topologies of the investigated networks are given in Figure 4 . In order to determine the first rotation angle 110, both a classification network and a regression network are investigated. In the first column of the table shown above, it is respectively stated how many nodes the network used has in each of the three layers and which network type is involved. The remaining columns of the table indicate how large the maximum observed prediction error is for the third angle 114, the second angle 112 and the first angle 110. The last column of the table indicates how many parameters the hidden layers of the two networks have in total to be learned.

[0031] According to the study summarized in the second row of the table, the classification network of the topology described in the table (50 nodes in the first layer, 30 nodes in the second layer and 20 nodes in the third layer) is used to determine the first rotation angle 110. A vector with 90 entries is used as the output vector of the network, which is used for the category to be determined. This produces a theoretical angular resolution of 4° in the case of a possible 360° rotation. As can be seen from the table, the observed error of 40° is obviously large, which indicates that the model is too complex during the training phase and thus leads to overfitting. On the contrary, the small number of nodes in the regression network for determining the second rotation angle 112 and the third rotation angle 114 can infer that the complexity of the model is too low. Nevertheless, it should be noted that not only the classification network for determining the first rotation angle 110 but also the regression network for determining the second rotation angle 112 and the third rotation angle 114 provide usable results in principle, and the network topology that is not optimized itself is suitable for analyzing the magnetic field and determining the orientation of the magnet causing the magnetic field in space.

[0032] The third row of the table shows the results of the study when a classification network with reduced complexity (lower number of nodes in the individual layers) was used for the first rotation angle 110 and a regression network with higher complexity was used for the second and third rotation angles. As can be seen, this makes it possible to increase the accuracy in all application cases, in which discernible deviations still occur when determining the first rotation angle 110.

[0033] Therefore, for the results in the fourth row of the table, the regression network is also used for determining the first rotation angle 110 , whereby the accuracy of the prediction is significantly improved.

[0034] The use of a regression network that is particularly suitable for processing a series of input values ​​may be advantageous for applications in joysticks, since in this application the magnet can only move on a stable track, since it is rigidly connected to the support 104. Therefore, a component of the continuous measurement of the magnetic field is a series of input values ​​for which only certain curves are possible, which can be well predicted with the aid of the regression network.

[0035] Thus, according to some embodiments, the evaluation logic corresponds to a neural network whose last layer is a regression layer.

[0036] It can further be seen from the above considerations that the complexity of the neural network is appropriately selected if, for example, the number of nodes of the neural network is greater than 30 and less than 100. Such complexity provides sufficient prediction accuracy in the application of determining the orientation of a magnet based on the magnetic field generated by the magnet without causing overfitting.

[0037] Furthermore, a choice of adaptation problem for a network topology that yields good prediction accuracy is given, for example for an embodiment with between 10 and 30 nodes in the first layer of the neural network. The choice of a fully connected feed-forward network with three layers enables the network to be run with radio quality on simple hardware. Thus, for example, the functionality can be implemented in a controller directly integrated in a joystick, so that the controller can directly generate measurement values ​​for each degree of freedom of movement of the joystick, such as Figure 1 The rotation angles 110, 112 and 114 in FIG.

[0038] Although according to Figure 1 As described, measuring a component of the magnetic field at only one position by means of a single magnetic field sensor 120 is sufficient to determine information about the orientation of the magnet by means of a neural network, but according to some embodiments, the accuracy can be increased by additionally measuring a further component of the field generated by the magnet at the second position at the second position, for example by means of a Figure 1 A second magnetic field sensor 122 is optionally shown.

[0039] The starting example of a joystick can therefore have a first magnetic field sensor 120 arranged in a fixed position relative to the housing and a second magnetic field sensor 122 arranged in a fixed position relative to the housing. The position of the magnetic field sensors can be arbitrary, for example the magnetic field sensors can also be arranged at an acute angle or 90° to the undeflected longitudinal axis 106, which leads to good prediction results.

[0040] Correspondingly, a further embodiment of a circuit for determining information about the orientation of a magnet may have an input interface which is designed to receive a first component of a magnetic field generated by the magnet at a first position and additionally receive a second component of the magnetic field generated by the magnet at a second position.

[0041] According to some embodiments, the direction of the magnetic field is determined comprehensively, ie the circuit obtains three linearly independent components of the magnetic field from one sensor or from two sensors, respectively.

[0042] According to further embodiments and application scenarios, it may be sufficient to process only two components of the magnetic field.A further starting example as a neural network alternative also only determines the absolute value of the field strength.

[0043] Although it is possible in principle to generate all the information which unambiguously describes the position of the magnet in a particular application using a single trained neural network, Figure 1 An exemplary embodiment is shown in which an optional second trained neural network 158 is used for the first rotation angle 110 and a first neural network 156 is used for the second rotation angle 112 and the third rotation angle 114. The first neural network 156 is trained to determine a vector having two components, the first component of which is the second rotation angle 112 relative to the first axis 113 and the second component of which is the third rotation angle 114 relative to the second axis 115. The second neural network 158 is trained to determine a vector having two components, the first component of which is the sine of the first rotation angle 110 of the rotation of the magnet about the longitudinal axis 106, wherein the second component of the vector is the cosine of the first rotation angle 110. The rotation angle 110 itself yields the inverse tangent as the quotient of the two components.

[0044] The first rotation angle 110 is not directly output by the neural network to be trained, which may enable the neural network to be successfully trained, which may otherwise be hindered by the ambiguity of the angle. Otherwise, the network will learn two angles that differ by 360° as equally good outputs. This ambiguity is implicitly eliminated by selecting the periodic function to be trained, allowing the neural network to be successfully trained.

[0045] According to further embodiments, the neural network can also be trained directly for the prediction of the first rotation angle 110. Likewise, a neural network trained for this purpose can also be used to predict the sine or cosine of the relevant angle for one or both of the second rotation angle 112 or the third rotation angle 114. In the case where a single neural network is used to predict all three rotation angles 110, 112 and 114, it may be advantageous to predict both the sine and the cosine for all angles so that the variables to be learned within the network do not differ too much, which would otherwise make learning difficult.

[0046] The training of the neural network can be carried out, for example, according to the principle of supervised learning, by virtue of the fact that the network is provided with training data for which the desired network output is known a priori. Figure 2 A plurality of such training groups for a given arrangement are shown by way of example. The arrangement is essentially defined by the position of the support 104 relative to the magnet 108 and the position of the sensors 120 and 122. In the case specifically shown, the first magnetic field sensor is located at position X=8 mm, Y=0 mm, Z=0 mm, and the second magnetic field sensor is located at position X=-8 mm, Y=0 mm, Z=0 mm. The magnet is cylindrical and has a height of 4 mm and a diameter of 5 mm. The fulcrum, i.e. the point through which all the axes of rotation extend, is located at position X=0 mm, Y=0 mm, Z=14 mm, and the length of the arm between the fulcrum and the magnet is 10 mm.

[0047] However, it should be noted here that the relative position between, for example, the sensor and the magnet is not limited to this configuration or similar configurations, but the training of the neural network must be performed for a position selected once. Figure 2 , the upper left figure shows the magnet of the joystick in a static position, i.e. without deflection, wherein the three figures in the right column each show a rotation of 90° around the longitudinal axis 160, i.e. in these figures the first rotation angle 110 is 90°. In addition, the individual training data sets are distinguished by different offsets or rotations about the axes 113 and 115. The training data required by the neural network for the individual offsets are the components of the magnetic field measured at the position of one or more sensors. This can be generated in two ways. On the one hand, the components can be determined by calculation for different positions within the scope of numerical simulation and used as training input. On the other hand, the components of the magnetic field can also be directly measured on the existing hardware of the joystick.

[0048] Figure 3 Another embodiment of a joystick 300 having three degrees of freedom of motion is shown, however, the joystick differs from Figure 1 The examples discussed in Figure 1Unlike the case shown in , the third degree of freedom is not a further rotation about an axis perpendicular to the first axis 113, but the third degree of freedom is the possibility of lateral movement along the axis 113. In other words, the third degree of freedom of the joystick 300 is not a rotation about the axis 113, but a lateral movement of the operating rod 102 in a direction defined by the support 104 adapted to the operating rod, for example in the direction of the axis 113.

[0049] Since it has been described above that with the aid of a neural network, any orientation of a magnet in space can be determined based on the measured components of the magnetic field generated by the magnet, Figure 3 The embodiment shown in Figure 1 The embodiment discussed in detail in the embodiment differs only in that the same Figure 3 The three degrees of freedom of the embodiment of the invention are adapted to the training data of the trained neural network. Therefore, for other details of the discussion, in order to avoid repetition, reference is made to Figure 1 and Figure 2 , which can be directly transferred to Figure 3 on the embodiments of the present invention.

[0050] Figure 4 A specific example of a network topology of a neural network is shown, which can be used to determine the orientation of a magnet based on the components of the magnetic field generated by the magnet. In particular Figure 4 The topology of a neural network 400 is shown, which is the basis of the study discussed above with reference to the table. The neural network has an input layer 402, which delivers the three components of the magnetic field to a three-layer feedforward network for each of the two measuring points. There are 20 nodes in the first layer 404 of the network 400, 15 nodes in the second layer 406 and 10 nodes in the third layer 408 of the network. The last layer of the neural network is a regression layer 410 with two nodes for the output value to be predicted, and a vector is passed to the output 412 of the network 400, the components of which are the trained parameters. According to the embodiments discussed above, these are the sine and cosine of the first rotation angle 110, or the second rotation angle 112 and the third rotation angle 114. As already mentioned, for further embodiments, any other neural network with a completely different topology or using a different mapping between the layers of the network can be used.

[0051] Just for completeness reasons, Figure 5 An embodiment of a method for determining information about the orientation of a magnet is summarized in a simple flow chart of . The method first comprises measuring 502 a component of a magnetic field 502 generated by a magnet. The method further comprises processing 504 the measured component by means of a trained neural network.

[0052] By means of such a method, for example, a first rotation angle, a second rotation angle and a third rotation angle can be determined, each indicating a rotation of the magnet about one of three axes that are not parallel to one another in pairs.

[0053] However, embodiments are not limited to this specific type of parameters describing the orientation of the magnet. In fact, the neural network can be trained for any other set of parameters that enable the orientation of the magnet in space to be determined unambiguously, such as position and orientation.

[0054] In the case of a joystick, for example, a vector having two components can be determined with the aid of a first trained neural network, the first component of which describes the sine of a first rotation angle of a rotation of the magnet about the longitudinal axis, and the second component of which describes the cosine of the first rotation angle. Alternatively or simultaneously, a vector having two components can be determined with the aid of a second trained neural network, the first component of which describes a second rotation angle of the magnet for a rotation of the longitudinal axis about a first axis, and the second component of which describes a third rotation angle of the magnet for a rotation of the longitudinal axis about a second axis perpendicular to the first axis.

[0055] In the previous paragraphs, embodiments of methods and circuits for determining the orientation of magnets for magnets arranged in a joystick have been discussed. For such a joystick, a series of advantages are obtained by applying a neural network. Compared with conventional solutions, the joystick only requires a unique magnet. That is, some embodiments of the joystick have a unique magnet. If all three degrees of freedom can be evaluated together, especially if such a joystick only requires a unique support, the mechanism of the system is greatly simplified. Therefore, in principle, any standardized joystick layout can be used. The measurement is contactless and is therefore not susceptible to wear and contamination. The first rotation angle of the rotation around the longitudinal axis is not limited by the measuring principle, providing a complete 360-degree range. Due to the learning ability of the neural network, the joystick is not limited to a specific mechanical design. In fact, there is a large degree of design freedom when constructing the joystick. The precise positioning of the sensor, the precise size and shape of the magnet, the size of the air gap between the magnet and the sensor, and the length or geometry of the operating rod are not restricted, because the trained neural network can learn all possible groups during training. In particular, for example, the relative arrangement of multiple sensors for determining the magnetic field and the number of sensors themselves can also be changed arbitrarily. Thus, other exemplary embodiments may have not only one or two sensors, but also, for example, 3, 4 or 5 sensors for determining the components of the magnetic field at different positions.

[0056] Of course, the embodiments described above are not limited to use in joysticks. In fact, any other application (where the orientation of a magnet is to be determined or where a magnet is used to determine the movement of any type of inspection instrument) can use embodiments of the methods or circuits described herein. For example, other embodiments can be a multi-function input device for a motor vehicle, an input device for an electronic device, or a graphic input device for a computer.

[0057] The aspects and features described in conjunction with one or more of the above detailed examples and drawings may also be combined with one or more of the other examples to replace similar features of the other examples or to otherwise introduce the features into the other examples.

[0058] The example may also be or relate to a computer program with a program code, and when the computer program is executed on a computer or processor, the program code is used to perform one or more of the above methods. The steps, operations or processes of the above various methods can be performed by a programmed computer or processor. The example may also cover a program storage device, such as a digital data storage medium, which is machine-readable, processor-readable or computer-readable, and encodes a machine-executable, processor-executable or computer-executable instruction program. The instructions perform some or all of the steps of the above methods or cause the execution of some or all of the steps of the above methods. The program storage device may, for example, include or may be, for example, a digital memory, a magnetic storage medium such as a disk and a tape, a hard drive or an optically readable digital data storage medium. Other examples may also cover a computer, a processor or a control unit programmed to perform the steps of the above methods or a (field) programmable logic array ((F)PLA) or a (field) programmable gate array ((F)PGA) programmed to perform the steps of the above methods.

[0059] Only the principles of the present disclosure are described through the specification and the drawings. In addition, all examples recorded herein are in principle explicitly intended to be used only for illustrative purposes to help readers understand the principles of the present disclosure and the concepts contributed by (one or more) inventors to further development of the technology. All statements herein about the principles, aspects and examples of the present disclosure and their specific examples include their equivalents.

[0060] A functional block referred to as a "means for" performing a specific function may refer to a circuit configured to perform the specific function. Thus, a "means for something" may be implemented as a "means configured for or suitable for something", such as a structural element or circuit configured for or suitable for a corresponding task.

[0061] The functions of the various elements shown in the figure (including any functional blocks referred to as "devices", "devices for providing signals", "devices for generating signals", etc.) can be implemented in the form of dedicated hardware (such as "signal providers", "signal processing units", "processors", "control units", etc.) and can be implemented as hardware capable of executing software in association with related software. When provided by a processor, the functions can be provided by a single dedicated processor, by a single shared processor, or by multiple separate processors, some or all of which can be shared. However, the term "processor" or "control unit" is broadly not limited to hardware capable of executing only software, but can include digital signal processor hardware (DSP hardware; DSP = digital signal processor), network processor, application specific integrated circuit (ASIC), field programmable gate array (FPGA), read-only memory (ROM), random access memory (RAM) and non-volatile storage devices for storing software. Other conventional and / or user-specific hardware may also be included.

[0062] For example, a block diagram may represent a rough circuit diagram that implements the principles of the present disclosure. In a similar manner, a flow chart, a process diagram, a state transition diagram, a pseudo code, etc. may present various processes, operations, or steps, which are, for example, substantially represented in a computer-readable medium and thus executed by a computer or processor, regardless of whether such a computer or processor is explicitly shown. The methods disclosed in the specification or in the claims may be implemented by structural elements having means for performing each of the individual steps of these methods.

[0063] It should be understood that the disclosure of multiple steps, processes, operations or functions disclosed in the specification or claims should not be interpreted as being in a specific order unless otherwise explicitly or implicitly stated, for example, for technical reasons. Therefore, the disclosure of multiple steps or functions will not limit these to a specific order unless such steps or functions are not interchangeable due to technical reasons. In addition, in some examples, a single step, function, process or operation may include multiple sub-steps, sub-functions, sub-processes or sub-operations and / or may be decomposed into multiple sub-steps, sub-functions, sub-processes or sub-operations. Such sub-steps may be included and are part of the disclosure of the single step unless such sub-steps are explicitly excluded.

[0064] In addition, the attached claims are hereby incorporated into the detailed description, wherein each claim can be independently used as a separate example. Although each claim can be independently used as a separate example, it should be noted that although a dependent claim can refer to a specific combination with one or more other claims in a claim, other examples can also include a combination of a dependent claim with the subject matter of each other dependent claim or independent claim. Such a combination is explicitly proposed herein unless it is stated that a specific combination is not expected. In addition, the features of a claim are also included in any other independent claim, even if the claim is not directly subordinate to the independent claim.

Claims

1. A circuit (150) for determining information about the orientation of a magnet, comprising: an input interface (152) configured to receive a component of the magnetic field generated by the magnet; and evaluation logic (154) corresponding to the trained neural network and configured to determine the information about the orientation of the magnet based on the received components, The neural network includes at least one of the following: a first neural network that determines the second rotation angle and the third rotation angle based on components of the magnetic field, and a second neural network that determines the first rotation angle, The first rotation angle, the second rotation angle and the third rotation angle respectively indicate the rotation of the magnet around one of three axes that are not parallel to each other.

2. The circuit (150) of claim 1, wherein the input interface (152) is configured to receive a first component of a magnetic field generated by the magnet at a first position and a second component of a magnetic field generated by the magnet at a second position.

3. The circuit (150) of claim 1 or 2, wherein the evaluation logic (154) corresponds to a neural network, the last layer of the neural network being a regression layer.

4. The circuit (150) according to claim 1 or 2, wherein the evaluation logic (154) corresponds to a neural network, the number of nodes of which is greater than 30 and less than 100.

5. The circuit (150) of claim 1 or 2, wherein the evaluation logic (154) corresponds to a neural network, the number of nodes of the neural network in a first layer of the neural network being greater than 10 and less than 30.

6. The circuit (150) of claim 1 or 2, wherein the evaluation logic (154) corresponds to a neural network, the neural network being a fully connected feed-forward network having three layers.

7. The circuit (150) of claim 2, wherein the input interface (152) is configured to receive three linearly independent components of the magnetic field for the first position and the second position, respectively.

8. A circuit (150) according to claim 1 or 2, wherein the evaluation logic (154) is configured to determine a vector having two components, a first component of the vector describing the sine of a first rotation angle (110) of the rotation of the magnet around the longitudinal axis (106), and a second component of the vector describing the cosine of the first rotation angle (110).

9. A circuit (150) according to claim 1 or 2, wherein the evaluation logic (154) is configured to determine a vector having two components, a first component of the vector describing a second rotation angle (112) of the magnet about a first axis (113), and a second component of the vector describing a third rotation angle (114) of the magnet about a second axis (115) perpendicular to the first axis.

10. The circuit (150) of claim 1 or 2, wherein the evaluation logic (154) corresponds to at least a first trained neural network and a second trained neural network, wherein the first neural network is trained to determine a vector having two components, a first component of the vector describing the sine of a first rotation angle (110) of a rotation of the magnet about a longitudinal axis (106), and a second component of the vector describing the cosine of the first rotation angle (110); and wherein the second neural network is trained to determine a vector having two components, a first component of the vector describing a second rotation angle (112) of the magnet (108) for a rotation of the longitudinal axis (106) about a first axis (113), and a second component of the vector describing a third rotation angle (114) of the magnet (108) for a rotation of the longitudinal axis (106) about a second axis (115) perpendicular to the first axis.

11. The circuit (150) of claim 1 or 2, wherein the evaluation logic (154) corresponds to at least one trained neural network, the at least one trained neural network being trained to determine a first rotation angle (110), a second rotation angle (112), and a third rotation angle (114), the first rotation angle, the second rotation angle, and the third rotation angle each describing a rotation of the magnet about one of three axes that are not parallel to one another in pairs.

12. A joystick (100) having the following characteristics: a support (104) by means of which an operating lever (102) is supported relative to the housing in a manner movable with at least three degrees of freedom, wherein the operating lever (102) has a portion (102a) movable by a user and an inner portion (102b), the portion (102a) movable by a user and the inner portion being opposite to each other in a longitudinal direction on different sides of the support (104); a magnet (108), the magnet being arranged on the operating rod; at least one magnetic field sensor (120) which is arranged in a fixed position relative to the housing and is configured to measure a magnetic field component of a magnetic field generated by the magnet (108); and An evaluation logic (154) corresponds to at least one trained neural network and is configured to determine information about the orientation of the magnet based on the received magnetic field components.

13. The joystick (100) of claim 12, further comprising: a first magnetic field sensor (120) arranged in a fixed position relative to the housing and configured to measure a first magnetic field component of a magnetic field generated by the magnet (108); as well as A second magnetic field sensor (122) is arranged in a fixed position relative to the housing and is configured to measure a second magnetic field component of the magnetic field generated by the magnet (108).

14. The joystick (100) according to claim 12 or 13, further comprising: At least one circuit (150) according to any one of claims 1 to 11, which is configured to generate a measurement value for each degree of freedom of the movement based on the measured magnetic field components.

15. The joystick (100) according to claim 12 or 13, wherein The first degree of freedom is the rotation of the operating rod (102) around the longitudinal direction; A second degree of freedom is the tilting of the operating rod (102) about an axis defined by the support; and The third degree of freedom is the tilting of the operating rod (102) about another axis defined by the support (104), or the lateral movement of the operating rod in a direction defined by the support (104).

16. The joystick (100) according to claim 12 or 13, wherein the magnet (108) is magnetized perpendicular to the longitudinal direction.

17. A method for determining information about the orientation of a magnet: measuring (502) components of the magnetic field generated by the magnet; and processing (504) the measured components using a trained neural network to determine the information about the orientation of the magnet, The neural network includes at least one of the following: a first neural network that determines the second rotation angle and the third rotation angle based on components of the magnetic field, and a second neural network that determines the first rotation angle, The first rotation angle, the second rotation angle and the third rotation angle respectively indicate the rotation of the magnet around one of three axes that are not parallel to each other.

18. The method of claim 17, wherein the measuring (502) comprises: measuring a first component of a magnetic field generated by the magnet at a first location; A second component of the magnetic field generated by the magnet is measured at a second location.

19. The method of claim 17 or 18, wherein the last layer of the neural network is a regression layer.

20. The method according to claim 17 or 18, wherein the number of nodes of the neural network is greater than 30 and less than 100.

21. The method of claim 17 or 18, wherein the number of nodes in the first layer of the neural network is greater than 10 and less than 30.

22. The method of claim 17 or 18, wherein the neural network is a fully connected feed-forward network having three layers.

23. The method of claim 18, wherein three linearly independent components of the magnetic field are measured at the first position and the second position, respectively.

24. A method according to claim 17 or 18, wherein the information about the orientation is the angle of rotation of the magnet around the longitudinal axis, wherein a vector having two components is determined with the aid of the trained neural network, the first component of the vector describing the sine of the angle and the second component of the vector describing the cosine of the angle.

25. A method according to claim 17 or 18, wherein the information about the orientation is the rotation of the magnet around a first axis and around a second axis perpendicular to the first axis, wherein a vector having two components is determined with the aid of the trained neural network, the first component of the vector describing the tilt angle about the first axis and the second component of the vector describing the tilt angle about the second axis. 26 . A computer program product having a program code which, when executed on a programmable processor, enables the execution of the method according to claim 17 .

Images