Method for operating a sensor system, device for processing sensor data, and sensor system

Abstract

The invention relates to a method for operating and calibrating an inertial sensor system. In particular, a concept is proposed which enables the implementation of parameters of a correction model for compensating for sensor errors during operational operation. In particular, the parameters can be obtained by analyzing the change process of the evaluation speed. The invention also relates to a device for processing sensor data from an inertial sensor system and to a sensor system.

Description

Technical Field

[0001] The present invention relates to a method for operating a sensor system, in particular an inertial sensor system. Furthermore, the present invention also relates to a device for processing sensor data, in particular sensor data of an inertial sensor system, and such a sensor system. Background Art

[0002] Numerous modern applications, such as navigation systems, require very precise determination of position data and trajectories. This position data can be obtained, for example, by a Global Navigation Satellite System (GNSS), such as GPS, Galileo system, etc. In addition, solutions based on inertial sensor systems are also known, which can determine the movement of an object with the help of acceleration sensors and rotational rate sensors, and can deduce the position change and thus the movement path based on these sensor data.

[0003] In addition, systems that can implement the determination of position and movement direction based as much as possible only on inertial sensor data are also increasingly being put into use.

[0004] Publications DE 10 2022 205 457 A1 and respectively describe a sensor system, a computing unit, and a method for analyzing and evaluating sensor data, in which the drift of the sensor data is obtained and removed from the measurement data. Summary of the Invention

[0005] The present invention creates a method for operating a sensor system, in particular a method for operating an inertial sensor system, a method for processing sensor data, a device for processing sensor data, in particular a device for processing sensor data from an inertial sensor system, and a sensor system. Further advantageous embodiments are described below.

[0006] Therefore, it is stipulated that:

[0007] A method for operating a sensor system, in particular an inertial sensor system, the method having the following steps. In a first step, detection of a stable motion state is carried out. The sensor data of the sensor system can be used to determine whether the motion state is stable. In addition, the method further includes the step of "determining the current velocity and the current orientation". Here, the sensor data of the inertial sensor system can be used to determine the velocity and the orientation. In addition, the method further includes the step of "calculating the average motion direction and the average orientation of the sensor system". The obtained values can be calculated for a predetermined time interval or time window. The obtained current velocity and orientation are used to calculate the average values. In addition, the method further includes the step of "calculating the difference between the change in the motion direction and the change in the orientation". In addition, the method further includes the step of "adjusting the parameters of a calibration model for compensating for sensor errors of the sensor system". In particular, minimizing the difference between the change in the motion direction and the change in the orientation can be used to determine the parameters of the calibration model. In particular, if a stable motion state has been detected, then the parameters of the calibration model are matched.

[0008] It is further stipulated that:

[0009] A device for processing sensor data, in particular sensor data from an inertial sensor system. For this purpose, the device includes a processing facility. The processing facility is designed to receive and process the sensor data of the sensors of the inertial sensor system. The processing facility is in particular designed to implement the method according to the invention.

[0010] Finally, it is stipulated that:

[0011] A sensor system having an acceleration sensor, a rotational rate sensor, and a device for processing sensor data according to the invention. The device is designed to receive and process the sensor data from the acceleration sensor and the rotational rate sensor.

[0012] Advantages of the Invention

[0013] The idea of the present invention at this time is to create a concept by which a sensor system based on inertial sensors, in particular an acceleration sensor and a rotational rate sensor, can be made more efficient and accurate. On the one hand, the concept according to the invention enables efficient determination of the parameters for compensating for sensor errors. In addition, through this concept, the determination of the parameters for compensating for sensor errors can be carried out during operational operation, for example during the determination of motion data or a motion path.

[0014] This improves the accuracy of the information determined regarding the detected movement or position. In particular, the ability to dynamically adapt the parameters for compensating for sensor errors allows for a good accuracy of the determined information even over longer time periods. Furthermore, the proposed concept enables particularly efficient and therefore resource-saving processing of sensor data for position determination and for determining parameters for error compensation.

[0015] Various approaches can be used to evaluate the change in average direction of motion and average orientation. For example, the difference between the average change in velocity direction and the average change in the sensor's orientation can be determined over a predetermined time interval. Alternatively, the difference between the average value of the change in velocity direction and the average value of the change in the sensor's orientation can be determined. Furthermore, the difference between the change in velocity direction and the change in the sensor's orientation can also be accumulated, for example, over a predetermined time interval. Unless otherwise specified, for the purposes of the present invention, the direction of motion and orientation are preferably used as averaged values.

[0016] According to one embodiment, the parameters of the correction model are only adapted when the speed is greater than a predetermined threshold value, thereby ensuring that a sufficiently significant movement is present for calibrating the correction model.

[0017] According to one embodiment, the method includes the step of "determining the inaccuracies of the detected stable motion state, the average motion direction, and / or the change in orientation." In this case, these inaccuracies can also be taken into account when adapting the parameters of the correction model. Such inaccuracies can be derived, for example, based on known tolerances of sensor values. Furthermore, inaccuracies of current data or states during operation can also be derived, for example, based on available information. For example, at the start or after initialization, a high inaccuracy can be initially assumed, which is then refined over time to a higher accuracy based on increasing information.

[0018] According to one embodiment, the correction model for compensating for sensor errors comprises a probabilistic filter, in particular a Kalman filter. The Kalman filter can be, for example, a nonlinear Kalman filter, such as an extended Kalman filter or a cubic Kalman filter, in particular a filter configured as a square-root cubic Kalman filter.

[0019] According to one embodiment, available sensor data is used to detect a stable motion state. As an additional option or alternative, data from other sensors can also be utilized. For example, patterns or similar rules can be identified in the available data, and such patterns indicate a stable motion state, such as a uniform motion change process. In addition, of course, any other method for identifying a stable motion state is also possible. Here, the corresponding speed or motion direction can also be estimated for the stable motion state. For example, the following motion change process can be regarded as a stable motion state: in such a motion change process, a pre-given parameter, such as a velocity vector or a similar parameter, is within a tolerance range within a determined time period.

[0020] According to one embodiment, a pre-trained neural network is used to detect a stable motion state. Thereby, complex data structures can be analyzed, classified, and evaluated very simply in a relatively simple manner.

[0021] According to one embodiment, the calculation of the change in average speed and / or azimuth is performed respectively for a time window of one to three seconds. Such a time window within the range of one to three seconds has proven to be very suitable for averaging. In addition, of course, any other time window for forming an average can be used according to the application situation.

[0022] According to one embodiment, the calculation of the change in average speed and / or azimuth is performed respectively using pre-given weightings for the individual speed values and / or azimuth values. For example, weighting factors can be assigned separately for each sensor value within the time window of observation. In this way, the dynamics when calculating the averaged values can be matched.

[0023] The above configurations and expansion options can be arbitrarily combined with each other as long as it makes sense. Other configurations, expansion options, and implementations of the present invention also include combinations of features of the present invention that are not explicitly mentioned, described in the foregoing or in the following with respect to the embodiments. In particular, those skilled in the art will also add individual aspects as improvements or supplements to the corresponding basic forms of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features and advantages of the present invention will be explained subsequently with the aid of the drawings. Shown here are:

[0025] Figure 1 A schematic diagram of a sensor system according to one embodiment;

[0026] Figure 2 A flowchart serving as the basis for a method for calibrating a sensor system according to one embodiment;

[0027] Figure 3A flow chart as a basis for a method for calibrating a sensor system according to another embodiment;

[0028] Figure 4 a flow chart as a basis for a method for processing sensor data according to one embodiment; and

[0029] Figure 5 A flow chart serves as a basis for a method for processing sensor data according to another specific embodiment. DETAILED DESCRIPTION

[0030] Figure 1 A schematic diagram of a sensor system 1 according to one embodiment is shown. Sensor system 1 may include multiple inertial sensors, such as at least one rotation rate sensor 11 and at least one acceleration sensor 12. Inertial sensors 11 and 12 may, for example, be sensors based on microelectromechanical systems (MEMS). However, the principles described here can, in principle, also be applied to any other inertial sensors.

[0031] The rotation rate sensor 11 can provide sensor data, for example, for a rotational movement about a spatial axis (yaw rate). In particular, the rotation rate sensor 11 can also provide sensor data for the rotation rates of, for example, two or three mutually orthogonal spatial axes. The sensor data are typically provided relative to a reference system of the respective sensor and, therefore, relative to a reference system of the sensor system 1 including the sensor.

[0032] Likewise, acceleration sensor 12 can provide sensor data for, for example, translational acceleration along a spatial axis. Here, acceleration sensor 12 can also provide acceleration values for, for example, two or three mutually orthogonal spatial axes. These acceleration values are typically also provided relative to the reference system of the corresponding sensor and, therefore, relative to the reference system of sensor system 1.

[0033] The sensory values detected by inertial sensors 11, 12 can be provided, for example, at input interface 21 of processing device 20. The sensor values can be provided in the form of analog or digital values. In particular, the sensor values can be provided and / or detected by input interface 21 at a predetermined first sampling rate.

[0034] Received sensor values can be temporarily stored if necessary. Preprocessing, such as filtering or the like, can also be performed if necessary. This can be performed either already in the input interface 21 or, alternatively, in a downstream processing facility 22. In particular, unless otherwise specified, the direction of motion and orientation are preferably used as averaged values for the embodiments described herein.

[0035] The processing facility 22 can process sensor data from, for example, one or more acceleration sensors 12 in order to derive information about motion, direction of motion or speed therefrom. For example, the acceleration values can be integrated (aufintegriert) for the individual spatial directions in order to calculate the values of the corresponding velocity components thereby. By integrating the velocity components again, the traveled path length can also be determined.

[0036] Similarly, the processing facility 22 can also process sensor data from one or more rate-of-rotation sensors 11 and obtain therefrom information about, for example, the orientation / heading of the sensor 11 and thus about the orientation / heading of the sensor system 1. For this purpose, for example, the sensor values of the rate-of-rotation sensors 11 can also be integrated.

[0037] Due to inaccuracies or sensor errors, in particular due to offsets / biases or other errors (such as axis errors and sensitivity errors), the received sensor values can be erroneous. The above integration of the sensor values results here in that, as the integration time increases, the data determined - such as velocity, path length and orientation - can deviate more and more from the actual motion of the sensor system.

[0038] Therefore, processing for detecting and correcting or compensating for such inaccuracies can be provided in the processing facility 22 of the processing device 20.

[0039] Although traditional concepts for correcting or compensating for biases are usually based on the use of additional information, such as additional information from a satellite navigation system (GNSS) or a similar system, the concept described below enables, during operation of the sensor system 1 when it is in motion and in particular also during operation for determining navigation information (such as velocity, path length, direction of motion, etc.), the determination and adaptation of parameters for correcting or compensating for systematic errors or sensor errors.

[0040] In a possible implementation, the compensation or correction of sensor values can be performed, for example, based on the analysis of speed. The basic principle of this solution can be based on, for example: identifying a state of the sensor system 1 in which the sensor system 1 is moving at least approximately at a constant speed, and then comparing the average speed during this phase with the current speed development. Based on this comparison, parameters can be derived to match the correction model for compensating sensor errors. On the one hand, the correction model can be used to compensate for possible errors in the raw data from sensors 11 and 12. In addition, the previously obtained position data or motion data can be matched or corrected with the help of the correction model. Therefore, the previously obtained position data or motion data based on the data in the inertial sensors 11 and 12 can be modified with the help of this correction model to compensate for offsets, drifts or other deviations.

[0041] Figure 2 A flowchart is shown, which can be used as the basis for a method for operating the sensor system 1, in particular an inertial sensor system, according to an embodiment.

[0042] First, sensor data can be received from one or more inertial sensors 11, 12, in particular at least one acceleration sensor 12. Preferably, sensor data can be received from at least one acceleration sensor 12 and a rotational rate sensor 11.

[0043] In step S11, the data from one or more sensors, in particular the sensor data of the sensor system 1, can be analyzed and evaluated to detect a stable motion state of the sensor system 1. In this embodiment, this stable motion state can in particular be a state in which the average orientation of the sensor system is at least approximately constant with respect to the average motion direction of the sensor system. For this purpose, for example, the motion direction and orientation of the sensor system 1 can be observed and averaged over a pre-given time period. In principle, for this analysis of sensor values, any suitable and available sensor data can be analyzed and evaluated. In addition, any suitable method, such as frequency analysis of the available sensor data or a similar method, can also be used to analyze and evaluate an at least approximately constant motion direction. As an additional solution or as an alternative, a trained neural network or other suitable means can also be used to analyze the change process of the motion direction.

[0044] In step S12, the current motion, in particular the current vector velocity of the sensor system 1, can be determined based on the received sensor data. For this purpose, any suitable scheme can be used, in particular by integrating the acceleration values. In addition, for example, the sensor values of the rotational rate sensor 11 can also be used to determine the orientation of the sensor system 1.

[0045] In step S13, the (vector) velocity obtained by the sensor system 1 can be monitored in order to thereby determine the direction of movement varying over time. Thereby, the average direction of movement within the corresponding time interval can be determined. Similarly, the orientation of the sensor system 1 can be analyzed and evaluated in order to thereby determine the average orientation of the sensor system 1 within the observed time interval. For example, a time interval between 1 second and 3 seconds can be used as the time interval.

[0046] In step S14, the difference between the change in the direction of movement and the change in the orientation can be calculated based on the values calculated in this way. This difference corresponds to, for example, the change in the orientation of the sensor system 1 relative to the (averaged) direction of movement.

[0047] If there is at least an approximately stable state of motion according to the analysis performed in step S11, then in step S15, the parameters of the correction model for compensating for sensor errors can be determined. For this purpose, for example, a scheme can be used that matches the parameters of the correction model for compensating for sensor errors such that the difference between the previously calculated change in the direction of movement and the value of the average change in the orientation of the sensor system 1 is minimized, i.e., preferably approaches zero.

[0048] The correction model for compensating for sensor errors can be, for example, a mathematical model based on a probability filter. The correction model can in particular include, for example, a so-called Kalman filter. In principle, however, any other suitable correction model or filtering scheme is also possible.

[0049] Due to system-related characteristics and interference effects, the sensor values present from the inertial sensors 11 and 12 and the results in the calculation of the velocity values can all have uncertainties. Therefore, this uncertainty can likewise affect the matching of the parameters of the correction model. Therefore, for example, inaccuracies or tolerances can be assigned to the input data (such as acceleration, rotation rate, or similar data), and the resulting uncertainty or accuracy can be derived therefrom. As an additional scheme or as an alternative, uncertainty or accuracy values can also be assigned to the resulting variables (such as velocity, orientation, etc.). In addition, for example, at the beginning or during initialization, due to the lack of a data basis, a large uncertainty can be assumed first. Subsequently, during further operation, corresponding to the acquired data, the uncertainty of the values can decrease, or the accuracy of the underlying values can increase. For this purpose, in principle, any suitable scheme for evaluating uncertainty or the accuracy of the basis is possible. This information can likewise affect the determination of the parameters of the correction model. For example, the individual variables can be weighted corresponding to the underlying inaccuracy.

[0050] As can be recognized from these explanations, the parameters of the correction model for compensating sensor values are determined or matched in the following operating state or in the following operating phase: In this operating state or in this operating phase, the current measured values, i.e., speed, direction of movement, orientation, etc., can be determined. Thus, the dynamic matching of the correction model does not require an interruption of the operational run.

[0051] Thus, by this continuous and dynamic matching of the correction model, stable, reliable, and relatively accurate data on speed, direction of movement, position, etc. can also be obtained and provided over a significantly longer period of time. In particular, using this correction model can also (retrospectively) correct possible deviations, such as offsets, drifts, or similar deviations, in the position data or movement data that have been determined previously.

[0052] Similar to the matching of the parameters of the correction model for compensating sensor values based on monitoring the direction of movement and orientation of the sensor system 1, as an additional option or alternative, the parameters of the correction model can also be matched by analyzing the speed or the time course of the speed. This will be explained below with reference to Figure 3 explain possible concepts.

[0053] Here, sensor data is also received from one or more inertial sensors 11, 12. Preferably, sensor data can be received from at least one acceleration sensor 12 and a rotational rate sensor 11.

[0054] In step S21, the data from one or more sensors, in particular the sensor data of the sensor system 1, can be analyzed to detect a predetermined stable operating state, such as at least approximately constant speed of the sensor system 1. For example, the course of the speed over a pre-given time period can be observed as a constant speed, within which the speed lies within a pre-given value range / tolerance band. In principle, for this analysis of sensor values for determining at least an approximately constant speed course, any suitable and available sensor data can be analyzed. In addition, any suitable method, such as a frequency analysis of the available sensor data or a similar method, can also be used to analyze at least an approximately constant speed. As an additional option or alternative, a trained neural network or other suitable means can also be used to analyze the course of the speed.

[0055] In step S22, the current speed of the sensor system 1 can be determined based on the received sensor data. For example, the values from the acceleration sensor 12 can be integrated for this purpose.

[0056] In step S23, for a pre-given time window, the average speed can be calculated based on the obtained speed value of the sensor system 1. For this purpose, any suitable time window can be used. For example, a time interval between 1 second and 3 seconds can be used as the time window.

[0057] If there is at least an approximately constant speed, then in step S24, the parameters of the correction model for compensating for sensor errors can be obtained. Here, the minimization of the difference between the average speed calculated according to step S23 and the speed estimated based on the constant speed of the sensor in step S21 can be used to match the parameters of the correction model. In particular, the method mentioned in the introduction can also be implemented.

[0058] For this purpose, for example, a scheme can be used that matches the parameters of the correction model for compensating for sensor errors such that the difference becomes minimal, i.e., preferably approaches zero.

[0059] The underlying correction model for compensating for sensor errors can also be, for example, a mathematical model based on a probability filter. The correction model can in particular include, for example, a so-called Kalman filter. However, in principle, any other suitable correction model or filtering scheme is also possible.

[0060] Similarly, data regarding the uncertainty of the sensor values and / or the resulting uncertainties of variables (such as speed, orientation, etc.) may also affect the obtaining of the parameters of the correction model.

[0061] In the two above-mentioned methods for obtaining the parameters of the correction model, in particular, the average values of speed, rotation rate, and / or orientation are used. In addition to the traditional averaging (in which all sensor values within a pre-given time interval are weighted equally), any other scheme is also possible. For example, individual weightings can be assigned to each data point corresponding to the time position of the data point in the time window and the value to be considered. In this way, the dynamics of the formed average value can be matched accordingly.

[0062] In order to determine whether the sensor system 1 is in a stable motion state, in principle, various suitable schemes are possible. For example, the changes or variations of the translational speed and the rotation rate can be compared with pre-given thresholds. Here, the translational speed or the rotation rate can be averaged, for example, over a predetermined time period, and these average values can be analyzed accordingly. In addition, any other suitable model, such as a statistical model, is also possible. Similarly, for example, a stable motion state can also be detected by means of simple machine learning methods or similar methods.

[0063] In particular, it is possible to classify, for example, the motion state. For example, the group consisting of the categories of various motion states may include, for example, a first category in which the sensor system 1 is completely at rest. For example, another second category can be assigned to such a motion state in which the sensor system 1 is on average at rest within a pre-given time interval. Here, the sensor system 1 may also move within this time interval, for example, vibrate, swing back and forth or make similar movements. For example, a third category of the motion state can be assigned to such a motion in which the sensor system 1 is in a uniform motion. Finally, for example, a fourth category of the motion state can be assigned to such a motion in which the sensor system 1 moves on average uniformly, but in this case may also briefly perform intentional, random or irregular movements. This may be, for example, the case when such a sensor system 1 is mounted on the user's wrist, where the user moves forward and swings the arm, such that the sensor system on the wrist swings back and forth during the arm movement. However, the classification of the four categories listed here is only to be understood as exemplary. In addition, any other type of classification for subdividing the motion state, especially the classification of various stable motion states, is also possible.

[0064] For error compensation or the establishment of a correction model, in the case of a static state, it is possible to perform correction, for example, based on the probability of one or more of the following assumptions (where, in addition, other assumptions can also be made in principle):

[0065] a) The translational speed is at least approximately zero;

[0066] b) The rotational rate is approximately zero or corresponds to the offset of the rotational rate sensor;

[0067] c) The amplitude and / or direction of the values from the acceleration sensor correspond to the amplitude or direction of the gravitational acceleration converted into the coordinate system of the sensor.

[0068] Here, the correction can be performed according to the measurement or estimation by means of a probability filter or in any other way. In particular, if the correction is applied here to such a time window: within which a stable motion state is on average obtained, then the correction can be performed for all of the stable motion directions of the above categories.

[0069] In addition, if necessary, other correction schemes can also be applied, such as pre-given average speed, average acceleration or similar schemes.

[0070] This calibration scheme also directly or indirectly affects the position, velocity, orientation, and sensor errors obtained or corrected by the acceleration sensor and the rotational rate sensor. If, for example, calibration is performed for a dedicated velocity of zero, then the position and linear acceleration are updated here. The update of the linear acceleration results in error compensation for the acceleration and the orientation. The error compensation for the acceleration subsequently results in error compensation for the acceleration sensor, and the correction of the orientation results in error compensation for the rotational rate sensor, which in turn results in error compensation for the rotational rate sensor. Thus, all states are calibrated. The degree of influence can here be based on the cross-covariance estimated by the probability filter, which is derived from the noise propagation and the probabilities defined in the probability filter.

[0071] When processing sensor values for determining velocity, direction of motion, path of motion, orientation, position, etc., these sensor values are typically detected, provided, and processed at a relatively high first sampling rate. This sampling rate can be in the range of several hundred Hertz to several thousand Hertz. Processing all data at such a high sampling rate may require high computational power if necessary. In addition, when forwarding the corresponding data and results, a transmission path with a suitable high bandwidth must also be provided.

[0072] To further optimize and improve efficiency, the sampling rate can be reduced in a first processing step such that only subsequent steps need to be carried out at a lower second sampling rate, or rather, the data can be forwarded at this lower second sampling rate. Here, the second sampling rate can, for example, be in the range of several tens of Hertz to a few Hertz.

[0073] The following is combined with Figure 4 Describe a possible method for reducing the sampling rate. Here, the following basic principle serves as the basis for this method: First, data is received at a higher first sampling rate and integrated over multiple iterative steps. Subsequently, the integrated values are differentiated according to the integration time interval and these differentiated values are output at the second sampling rate.

[0074] At the beginning, initialization can first occur, in which all velocity values and rotational rate values are reset, for example, set to 0.

[0075] In step S31, for this purpose, sensor data is first received from sensors 11, 12 of the inertial sensor system 1 at the first sampling rate.

[0076] Optionally, subsequently in step S32, the received sensor values are corrected according to the calibration model if necessary. For example, suitable scaling can be performed here. As an additional option or alternative, offsets can also be matched or removed.

[0077] In step S33, the sensor data received at the first sampling rate is integrated, i.e., accumulated. Here, the data can be processed, for example, in the reference system of the sensor system 1. Since the sensor system 1 may move over time and, in particular, rotation of the sensor system 1 may also occur here, the corresponding reference system of the sensor system 1 may change over time. To take into account this change in the reference system, the already integrated data can be transformed, for example, into the current reference system of the sensor system 1 respectively. Subsequently, the current sensor data can be added to the transformed data for integration. In this way, the integration result always exists in the form of a reference system according to the current orientation of the sensor system 1. As an alternative, in principle, the integration can also be carried out based on any fixed reference system. For example, the original reference system can be used as a basis when initializing the sensor system 1. Subsequently, the sensor data can be transformed onto this fixed reference system before integration. In particular, the reference system of the sensor system 1 can also be transformed into a global reference system (e.g., relative to the earth's gravitational field). Here, linear approximation can be used, for example, for integration, especially for the transformation of the reference system. Here, in particular, trigonometric functions or non-linear formulas with one or more trigonometric functions can be approximated at least piecewise by suitable linear equations.

[0078] After the sensor data has been integrated over at least two integration steps, for example, within a pre-given time interval or for a pre-given number of integration steps or iteration steps, the result of this integration can be differentiated in step S34 corresponding to the integration time interval. In this way, new data of the sensor values can be obtained corresponding to a significantly lower second sampling rate after each differentiation. If necessary, suitable linear approximation can also be used for this differentiation of the sensor values. For this purpose, various linear approximations can be used in particular according to the sensor data. For example, a different linear approximation can be used for high rotation rates than for low rotation rates. In addition, of course, any other distinction can also be made in order to select a suitable linear approximation. Through this linear approximation, simpler computational structures can also be used, which cannot implement complex, partly trigonometric functions. In addition, the required computing power and computing time can be reduced through linear approximation.

[0079] If the sensor values have been corrected according to the correction model in the optional step S32, then the compensated error can be added back to the sensor values at the lower second sampling rate at this time in the also optional step S35. In this way, the sensor values are prepared for further processing at the second sampling rate, and these sensor values still have the original error characteristics. Therefore, the sensor values provided at the second sampling rate can also be used to establish the parameters of the model for compensating the error. The sensor values provided at the second sampling rate can in particular also be used to perform the method for matching the parameters of the correction model described above.

[0080] In principle, the first sampling rate and the second sampling rate can be fixedly pre-given. However, in addition to this, the first sampling rate and / or the second sampling rate can also be dynamically matched during operation. For example, the first sampling rate can be matched according to the data rate of the sensor system 1. In addition to this, any other scheme for changing the first sampling rate is also possible. The second sampling rate can be matched or pre-given, for example, by a downstream processing facility. Thereby, data can be provided at the second sampling rate, which corresponds to the requirements of further processing. For example, the second sampling rate can be matched according to the data rate for transmitting the output data. The second sampling rate can also be matched according to the processing capacity of the downstream processing facility. In addition to this, any other criteria for dynamically matching the second sampling rate are also possible for the second sampling rate.

[0081] The above components and method elements can be used, for example, in the sensor system 1 with inertial sensors 11, 12 in order to improve accuracy, long-term stability, and processing speed. Thereby, such a sensor system can be realized: This sensor system can accurately provide position messages, movements, and movement change processes, etc. in an efficient manner.

[0082] As described above, the parameters for compensating the sensor error can be determined and matched during the operational run, that is, determined and matched in parallel with the determination of the motion data. Thus, this sensor system 1 is not limited to a static, pre-determined correction model. By integrating the compensation of the sensor error for the processing of the sensor data, the parameters of the correction model can also be dynamically matched during operation. For example, Figure 5 shows a possible flow of such a method for processing the sensor data of the sensor system 1 with multiple inertial sensors 11, 12.

[0083] In step S41, sensor data can be received from the sensors 11, 12, in particular from one or more rotational rate sensors 11 and one or more acceleration sensors 12.

[0084] In step S42, the sensor errors of the received sensor data are corrected using a dynamically adapted calibration model. The parameters of the calibration model can be dynamically adapted using the currently determined values for orientation, velocity, and / or position. To this end, the above-described method can be used, in particular, to determine or adapt the parameters of the calibration model.

[0085] In step S43, direction, speed and / or position and any other suitable parameters can be obtained as needed. In addition, position data or motion data obtained before the correction model after matching can also be modified or corrected as needed.

[0086] The determination of the parameters of the correction model and the ascertainment of the speed can also be carried out in this case, in particular, at a reduced second sampling rate, wherein, as described above, this second sampling rate is lower than the first sampling rate at which sensors 11 , 12 provide sensor data.

[0087] In summary, the present invention relates to a method for operating and calibrating an inertial sensor system. In particular, a concept is proposed that enables the parameters of a calibration model for compensating for sensor errors to be implemented during operational operation. These parameters can be determined, in particular, by evaluating the velocity profile.

[0088] Additionally or alternatively, these parameters can also be determined by evaluating the course of the movement direction.

[0089] The present invention also relates to a method for reducing the data rate of sensor data in an inertial sensor system. To this end, raw sensor data may be preprocessed and integrated at a first sampling rate. The integrated value may then be differentiated according to a second, lower sampling rate.

Claims

1. A method for operating an inertial sensor system (1), the method having the following steps: Detecting (S11) a stable motion state using sensor data of the inertial sensor system (1); Determining (S12) a current velocity and a current orientation using sensor data from the inertial sensor system (1); Calculating (S13) an average motion direction and an average orientation for a predetermined time interval using the determined current velocity and orientation; Calculating (S14) a difference between a change in the average motion direction and a change in the average orientation; If a stable motion state has been detected, then matching (S15) parameters of a correction model for compensating sensor errors of the inertial sensor system (1) using minimization of the difference between the change in the average motion direction and the change in the average orientation; 2. The method according to claim 1, wherein, The matching (S15) of the parameters of the correction model is only performed if the velocity is greater than a predetermined threshold; 3. The method according to claim 1 or 2, wherein, The method includes the steps of determining an inaccuracy of a change in the detected stable motion state, the average motion direction, and / or the orientation, and wherein the determined inaccuracy is used to perform the matching (S15) of the parameters of the correction model; 4. The method according to any one of claims 1 to 3, wherein, The correction model for compensating sensor errors includes a probability filter, in particular a Kalman filter; 5. The method according to any one of claims 1 to 4, wherein The detection (S12) of the stable motion state is performed using the sensor data and / or data of at least one additional sensor; 6. The method according to any one of claims 1 to 5, wherein The detection (S12) of the stable motion state is performed using a pre-trained neural network; 7. The method according to any one of claims 1 to 6, wherein The calculation (S13) of the average motion direction and / or the average orientation is performed for a time window of one to three seconds respectively; 8. The method according to any one of claims 1 to 7, wherein The calculation (S13) of the average motion direction and / or the average orientation is performed using pre-given weightings for respective velocity values and / or rotation rate values respectively; 9. Apparatus (20) for processing sensor data from an inertial sensor system (1), said apparatus having processing means (22) designed to receive and process sensor data from sensors (11, 12) of said inertial sensor system (1), wherein, The processing facility (22) is designed to perform the method according to any one of steps 1 to 8; 10. A sensor system (1), the sensor system having: An acceleration sensor (11); A rotation rate sensor (12); and The apparatus (20) for processing sensor data according to claim 9, wherein, The device (20) is designed to receive and process sensor data from the acceleration sensor (11) and the rotation rate sensor (12).

Images