Method for detecting anomalies in an inertial measurement unit

Abstract

The present invention uses the angular velocity (ω A ,ω B ,ω C ) and specific force (f A ,f B ,f C The present invention relates to a method for detecting an abnormality in an inertial measurement unit (4, 5, 6) used in a vehicle to measure a vehicle speed, the method comprising: at least three inertial measurement units (4, 5, 6) each having a plurality of sensors including an accelerometer and a gyro sensor; a first inertial measurement unit (4) being used as a master inertial measurement unit; and a second inertial measurement unit (5) and a third inertial measurement unit (6) having lower performance reliability than the first inertial measurement unit (4) being used as slave inertial measurement units; a measurement value from the master inertial measurement unit (4) being used as a reference value; and an error value relative to the master inertial measurement unit (4). Regarding the estimation of the model parameters, the measurements of the slave inertial measurement units (5, 6) are corrected, thereby detecting an abnormality in one of the three sensor signals based on three corresponding sensor signals by 2-out-of-3 voting, and one of the three inertial measurement units (4, 5, 6) whose sensor signal has an interval from the median (M, D) of the three sensor signals exceeding a threshold is identified as defective, and due to the abnormality of the master inertial measurement unit (4), switching is made to the signal of one of the slave inertial measurement units (5, 6) that is still functioning, taking into account the pre-calculated pre-correction.

Description

[Technical field]

[0001] The invention relates to a method for detecting anomalies in an inertial measurement unit according to claim 1. [Background technology]

[0002] An inertial measurement unit (IMU) is an arrangement of accelerometers and gyroscopes, used to measure the specific force (acceleration) and rotational speed of an object in space. For the designation of the accelerometer signal, the term specific force and the symbol f have been established in German navigation literature (see document citation 2). For the gyroscope signal, the symbol ω is used here. The term inertial measurement unit is represented in the following by the abbreviation IMU. Inertial measurement units are used to determine the position, attitude and speed of a vehicle, especially during autonomous driving. Examples of this type of use of IMUs can be found in document citations (3) to (6) or in US 9,753,144 B1 (DE 10 2017 102 269 A1).

[0003] Figure 1 shows an example of the use of an inertial measurement unit 1 for modern motion sensing and navigation. The inertial measurement unit 1 transmits rotation rate signals from a gyroscope and specific force signals from an acceleration motor. In a sensor fusion unit 2, motion values ​​such as azimuth angle α, velocity v, and / or attitude P are calculated by integrating these values ​​and are corrected using additional sensors 3 such as an odometer, magnetometer, barometric altimeter, and / or GNSS receiver. Current technology uses Kalman filters for sensor fusion.

[0004] Inertial measurement units are available in different quality levels. Their quality and price are mainly determined by the sensor technology used. In automotive serial applications, cost-effective sensors based on MEMS (micro-electromechanical sensor) technology are most commonly used. However, the accuracy of this type of cost-effective IMU sensor is limited by stochastic sensor noise and systematic errors such as bias, sensitivity errors, non-linearity and mismatch. Such errors are properties that affect the sensor in normal operation. Systematic errors must be detected and corrected by suitable methods to ensure high accuracy of the calculated movement values. Examples of correction methods can be found in the technical literature or also in DE10 2017 102 269 A1. The stochastic and systematic errors of IMU sensors are described using models. For example, the value of the specific force measured by the sensor f im is described by the following model:

[0005]

number

[0006]

number

[0007] In addition to random and systematic errors, IMU sensors can also have anomalies, which cause unspecified measurement procedures, since the sensor loses the specified (and therefore pre-set by design) association between the physical measurements and the emitted sensor signals, making the affected IMU sensor unusable.

[0008] Sensor anomalies that propagate undetected during travel calculations or operation can introduce significant errors or significant false values ​​into the operational signals, thus posing a safety risk in safety-relevant applications.

[0009] In safety-critical applications such as autonomous driving, sensor anomalies need to be detected and then back-up solutions need to be triggered in order to eliminate the associated errors from the navigation solution and ensure that the system remains fully functional (error-tolerant systems).

[0010] This applies in particular to anomalies in the sensors of the inertial measurement unit. The solution proposed for this purpose according to the invention is shown in FIG.

[0011] Other solutions based on the prior art are described below.

[0012] (prior art) US9,568,321B2 describes the detection of anomalies in individual IMUs without redundancy. It is proposed to integrate the detection in the form of a health monitoring block into a doubly redundant navigation system (Inertial Navigation System - INS). In the health monitoring block, estimation errors appearing during the motion estimation are evaluated. This solution has the drawback that no direct conclusions can be made about defective sensor elements due to the influence of the sensor anomalies on the navigation system. Furthermore, the integration of the IMU signals used in the INS leads to long error detection times. Furthermore, the error detection is based on the use of additional sensors such as GNSS, magnetometer, barometer, odometer, etc. These sensors are generally more error-sensitive and less accurate than the IMU sensors. The above arrangement does not allow a clear distinction between anomalies of these additional sensors and anomalies of the IMU. In summary, it can be seen that the described process allows the detection of anomalies but not the identification of defective sensors. This is, however, a necessary requirement of the error-tolerant system method.

[0013] The shortcomings of the above process have been known for a long time, and are detailed further in, for example, (3) Chapter 17, which explicitly prescribes the need for redundant hardware in safety-critical INS applications.

[0014] The use of redundant Inertial Measurement Units (IMUs) is the cutting edge of aircraft technology. It involves incorporating high quality Inertial Measurement Units with identical characteristics in a redundant sensor configuration.

[0015] In this regard, for example, the prior art uses an Air Data Inertial Reference Unit (ADIRU) with six accelerometers and six rotational rate sensors arranged in a non-orthogonal pattern to achieve maximum redundancy with a minimum number of sensors (see references 1 and 2).

[0016] Here, for example, six specific force signals f 1 ,...,f 6 The orthogonal component of the specific force [f x ,fy ,f z ] is solved to obtain a vector of results.

[0017]

number

[0018] Monitoring and averaging can be done in parallel while solving the regression equation. Up to two accelerometers in the system can fail without compromising its functionality. For functional safety, at least four sensors need to be available.

[0019] Another solution is proposed in US 5,184,304 A, which has two INSs with their respective inertial measurement units rotated by 60° relative to each other, in which the error statistics emitted by each system are evaluated by the other system, thereby identifying defective sensors.

[0020] Sensors suitable for air travel, especially non-orthogonal arrangements, are less suitable for use in automobiles due to the large amount of installation space required.

[0021] In much more cost-sensitive automotive applications, the use of redundant, high quality inertial sensors of this type is economically infeasible.

[0022] A description of the use of MEMS sensors in a multiple redundant configuration is given in US8,825,436B2, where the multiple configuration is used exclusively to reduce measurement noise and systematic sensor errors (run-to-run bias and temperature dependent bias). It does not include a strategy to detect and eliminate sensor anomalies.

[0023] (6) describes an inertial measurement unit for automobiles with triple redundancy, but it suffers from the disadvantage that it is built into the structural elements and is therefore not adequately protected from power problems and common cause errors.

[0024] Also known from the literature (see reference 2) is a master-slave arrangement of inertial measurement units or navigation systems. The basic idea of ​​this is to estimate parameters from the slave system, such as alignment angles, offsets, etc., based on navigation data from the master system using a Kalman filter. The corresponding process is called transfer alignment. This method is commonly used in weapon systems that include low-cost navigation systems, while the guidance system or carrier includes high-quality navigation systems.

[0025] The following documents describe various concepts regarding error tolerance in redundant inertial measurement units:

[0026] -Kohn et al., “Fail-operational in safety-related automotive multi-core systems”, 10th IEEE International Symposium on Industrial Embedded Systems (SIES), 2015, pp. 1-4, -T. Ishigooka, S. Honda and H. Takada, “Cost-Effective Redundancy Approach for Fail Operational Autonomous Driving System”, 2018 IEEE 21st International Symposium on Real-Time Distributed Computing (ISORC), 2018, pp. 107-115, - M. Li and L. Eckstein, “Fail-Operational Steer-By-Wire System for Autonomous Vehicles”, 2019 IEEE International Conference on Vehicular Electronics and Safety (ICVES), 2019, pp. 1-6, and -T. Schmid et al., “A Safety Argumentation for Fail-Operational Automotive Systems in Compliance of ISO 26262”, 2019 4th International Conference on System Reliability and Safety (ICSRS), 2019, Paragraphs 484-493 [Prior art documents] [Patent documents]

[0027] [Patent Document 1] US9,568,321B2 Systems and Methods for Determining Inertial Navigation System FaultsBharadwaj et al. 02 / 2017 [Patent Document 2] US8,065,074B1 Configurable Inertial Navigation System with Dual Extended Kalman Filter Modes Licardo, 11 / 2011 [Patent Document 3] US5,184,304A Fault-Tolerant Inertial Navigation SystemHuddle 02 / 1993 [Patent Document 4] US8,825,436B2 Inertial Sensing with Spatially Distributed Sensor Array and Two-Dimensional Data ProcessingZhang et al. 09 / 2014 [Patent Document 5] US9,753,144B1 Bias and Misalignment Compensation for 6-DOF IMU Using GNSS / INS DataJafari et al. 05 / 2017 [Non-patent literature]

[0028] [Non-Patent Document 1] MLSheffels, “A Fault-Tolerant Air Data / Inertial Reference Unit”, IEEE AES Systems Magazine, 8840 Evergreen Blvd,MN51-1355,Minneapolis,MN 5543,1993;MDMcIntyre and CAGossett, “The Boeing 777 Fault Tolerant Air Data Inertial Reference System-A New Venture In Working Together”, Seattle, WA 98124:Boeing Commercial Airplane Group,1995 [Non-Patent Document 2] Jan Wendel, lntegrierte Navigationssysteme [Integrated Navigation Systems], Oldenbourg 2007 [Non-Patent Document 3] Paul D. Groves, “Principles of GNSS, Inertial and Multisensor Integrated Navigation Systems”, Artech House, Boston, London 2013 [Non-Patent Document 4] Willy Klier, Andreas Reim and Dietmar Stapel, Sae Technical Paper Series, No. 2008-01-0582, “Robust Estimation of Vehicle Sideslip Angle an Approach w / o Vehicle and Tire Models”, Apr. 14-17, 2008, (7 pages) [Non-Patent Document 5] Andreas Reim, Alexander Steinbach, Oliver Oettgen and Dietmar Stapel, 2009 SAE International, No. 2009-01-0430, “Central Sideslip Angle Estimation on a Software Integration Platform,” 2009, (5 pages) [Non-Patent Document 6] http: / / www.aceinna.com Triple redundant IMU targets autonomous vehicles,09 / 2019 Summary of the Invention

[0029] It is an object of the present invention to provide an improved method for detecting anomalies in inertial measurement units (IMUs) used in vehicles to measure angular velocity and forces or acceleration.

[0030] This object is achieved according to the invention by means of a method having the features of claim 1.

[0031] Advantageous embodiments of the invention are the subject of the dependent claims.

[0032] According to the present invention, a method for detecting anomalies in an inertial measurement unit used in a vehicle to measure angular velocity and specific force includes at least three inertial measurement units, each equipped with a number of sensors including an accelerometer and a gyro sensor. According to the present invention, a first inertial measurement unit is used as a master inertial measurement unit, and a second inertial measurement unit and a third inertial measurement unit, which may have lower performance reliability than the first inertial measurement unit, are used as slave inertial measurement units, and measurements from the master inertial measurement unit are used as reference values ​​to correct measurements from the slave inertial measurement units through estimation of error model parameters (e.g., as in reference (1)) and in comparison with the master inertial measurement unit, whereby an abnormality can be detected in one of the three sensor signals based on three corresponding sensor signals and using 2-out-of-3 voting, and one of the three inertial measurement units whose sensor signal has a distance from the median of the three sensor signals that exceeds a threshold value is identified as defective, and due to an abnormality in the master inertial measurement unit, the signal is switched to one of the slave inertial measurement units that are still functioning and a pre-calculated pre-correction is applied.

[0033] This process may generally be extended to include majority voting for more than three inertial measurement units.

[0034] In one embodiment of the present invention, the error model parameters may be estimated using a Kalman filter or a recursive least squares method.

[0035] In one embodiment of the present invention, an error model including all model parameters from the slave inertial measurement unit may be used to estimate the error model parameters, and at least the time-varying relative sensor offsets from the master inertial measurement unit are estimated online.

[0036] In one embodiment of the invention, a relative force from the master inertial measurement unit may be converted to the measured position of the slave inertial measurement unit.

[0037] In one embodiment of the present invention, the estimated parameters may be limited to a maximum allowed value.

[0038] In one embodiment of the present invention, in detecting the significance of the signal deviation, the covariance of the estimation error estimated by the Kalman filter may be used in estimating the sensor signal deviation.

[0039] In one embodiment of the present invention, all sensors in all inertial measurement units may be filtered and time synchronized using their respective filter units.

[0040] In one embodiment of the present invention, online estimated error model parameters may be time delayed and used on the measurements of the slave inertial measurement unit.

[0041] In one embodiment of the invention, the slave inertial measurement unit may be pre-corrected using a master-slave configuration, where a time delay is applied to the estimated sensor output and if an error is detected in the first inertial measurement unit within the time delay, the estimated sensor error in the slave inertial measurement unit is frozen at a value for the end of the time delay.

[0042] In one embodiment of the invention, a detection logic is applied to the signal from the first inertial measurement unit and the pre-corrected signal from the slave inertial measurement unit to detect the signal that deviates most from the average or median of the three signals, and if the normalized signal difference from the average or median exceeds a value of 1, it is deemed to be a sensor error.

[0043] The invention can provide an error-tolerant determination of vehicle motion values ​​in the event of an IMU anomaly (operational failure) using 2-out-of-3 voting, without any special requirements on geometry or sensor quality. However, the invention is also applicable to cost-effective heterogeneous IMU hardware that is already partially available in the vehicle.

[0044] The method according to the invention makes it possible to protect the signal from the master inertial measurement unit against sensor errors in a way that ensures the ability to estimate the correct motion of the vehicle (failure continuity), taking into account the increasing safety requirements for automated vehicle operation.

[0045] This specification focuses on the use of heterogeneous (not identical) Inertial Measurement Units (IMUs) specifically installed at different locations within the vehicle.

[0046] Error tolerance (failure continuity) in calculating the vehicle's displacement values ​​is achieved by the solution according to the invention, which places only low requirements on the structural elements. The available infrastructure of the vehicle (ESP sensors, undercarriage sensors) may be used for it, resulting in a cost-effective solution.

[0047] Examples of the invention are explained in more detail below with reference to the drawings. [Brief description of the drawings]

[0048] [Figure 1] FIG. 1 is a schematic diagram of a typical example of using an inertial measurement unit (IMU) to estimate vehicle displacement. [Diagram 2] FIG. 1 is a schematic diagram of a monitoring system according to the present invention combined with displacement estimation; [Diagram 3] FIG. 4 is a schematic diagram of a preliminary data processing operation (pre-compensation) of a rotation speed sensor. [Figure 4] FIG. 1 is a schematic diagram of the preliminary data processing operation (pre-compensation) of the accelerometer. [Diagram 5] FIG. 2 is a schematic diagram of the error detection logic and decision-making logic. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0049] Similar items are identified by the same names in all the drawings.

[0050] One method of detecting anomalies in the inertial measurement units used in a vehicle to measure angular velocity and specific forces uses at least three inertial measurement units 4, 5, 6, each equipped with multiple accelerometers and gyro sensors, as shown in FIG. 2.

[0051] According to the invention, a first inertial measurement unit 4 is used as a master inertial measurement unit, and a second inertial measurement unit 5 and a third inertial measurement unit 6, which can have a lower performance reliability than the first inertial measurement unit 4, are used as slave inertial measurement units, where messages from the master inertial measurement unit 4 are used in a detection device 7 to estimate the systematic error parameters of the slave inertial measurement units 5, 6 relative to the master inertial measurement unit 4 in the model and detect anomalies of the individual sensors by 2-out-of-3 voting on the corrected signals. In the 2-out-of-3 voting, the two signals with the smallest difference are considered to be error-free and a third signal is tested against them. If the difference between the third signal and the signal considered to be error-free is too large, a potential anomaly of the evaluated third signal is assumed.

[0052] The drawing also shows three inertial measurement units 4, 5, 6, designated IMU A, IMU B, IMU C, each of which has three angular rate signals vector

[0053]

number

[0054]

number

[0055]

number

[0056] The details of the detection device 7 are shown in FIGS.

[0057] 3 shows an example of pre-correction of the rotational speed sensors, in which filter devices 9, 10, 11 are provided, via which the sensor signals of the master inertial measurement unit 4 and the two slave inertial measurement units 5, 6 are corrected, time-synchronized and filtered taking into account known mismatches.

[0058] Time synchronization is necessary because time synchronism of the measurements cannot be assumed for a non-uniform arrangement of the inertial measurement units: during signal transmission, differences in communication latency, for example, may occur and pre-filtering of the signals in the inertial measurement units 4, 5, 6 may be different.

[0059] Then, the parameters of the error models of both slave inertial measurement units 5, 6 are determined for the master inertial measurement unit 4. To do this, the data from the master inertial measurement unit 4 for the error models 13 and 14 are used to calculate the estimated sensor outputs ω Bestimated ,ω Cestimated Each of the error models 13, 14 is calculated based on the rotational speed ω A This involves adjusting the individual rotation rate sensors to the vector of ω, which is determined, for example, during end-of-line calibration. Deviations from the target motion are then identified as sensor errors in the slave inertial measurement units 5, 6 and corrected by estimating parameters of an error model. In the embodiment shown in FIG. 3, the sensor bias Bias_ω B ,Bias_ω Care determined exclusively in the estimation blocks 15 and 16. However, additional parameters such as bias drift, sensitivity deviation, mismatch, etc. may also be estimated. As is known from the literature, the parameter estimation in the estimation blocks 15 and 16 may be performed using recursive least squares or Kalman filters. For properly functioning inertial measurement units 4, 5, 6, these estimated parameters change only very slowly over time and therefore remain valid for a long period of time. Thus, the correction of the slave inertial measurement units 5, 6 remains correct even if it is performed with a time delay error parameter. The corresponding time delay blocks are indicated as 17 and 18, where the delay time must be greater than the error detection time of the detection device 7, for example, and equal to 100 ms.

[0060] The preprocessed signal 20 from the master inertial measurement unit 4 and the corrected signals 21, 22 from the two slave inertial measurement units 5 and 6 are then sent to the detection unit 7 for interpretation. If an anomaly occurs in the master inertial measurement unit 4, the estimated parameters of the slave inertial measurement units 5, 6 will change significantly. However, if this anomaly in the master inertial measurement unit 4 is recognized by the detection unit 7, the time delay blocks 17 and 18 prevent the parameter change from affecting the detection. Instead, the erroneous parameters from the slave inertial measurement units 5, 6 are frozen at their last valid values. In this way, correction and detection can be decoupled from each other and a failure of the master inertial measurement unit 4 can be reliably distinguished from a failure of the slave inertial measurement units 5, 6.

[0061] Additionally, or in other embodiments of the invention, the estimates of the parameters of the estimation blocks 15 and 16 may be limited to maximum allowed values, thereby improving detection sensitivity.

[0062] Additionally or in other embodiments of the invention, estimated parameters that reach or exceed a maximum allowable value may be used for error detection of the affected sensor.

[0063] Additionally, or in other embodiments of the invention, weighted residuals from the pre-correction may be calculated and evaluated in the estimation blocks 15,16.

[0064] Additionally, or in other embodiments of the invention, the calculated covariance for the estimation error can be used in the estimation blocks 15, 16 to quantify the signal deviation during detection.

[0065] Figure 4 shows the rotation speed ω A、 ω B、 ω C (also called angular velocity) corresponds to the accelerometer signal (specific force f A ,f B ,f c ) is shown. If the three inertial measurement units 4, 5, 6 are spatially spread on the vehicle, the rotational speed ω A、 ω B、 ω C Unlike the specific force f A , f B , f C This is done when generating the target values ​​from the master inertial measurement unit 4 signals in the error models 23 and 24, where the specific force f A ,f B ,f C are the measured positions of the respective slave inertial measurement units 5, 6 using the following equations known from the literature:

[0066]

number

[0067]

number

[0068] 5 shows an embodiment of error detection by 2-out-of-3 voting principle. Three individual signals for comparison and pre-processing from a master inertial measurement unit 20 and two slave inertial measurement units 21, 22 are designated A, B, C.

[0069] A, B and C are the rotational speeds ω A, ω B, ω C or specific force f A ,f B ,f C In total, one detection is used for each degree of freedom of the inertial measurement units 4, 5, 6.

[0070] Furthermore, in block 27, the medians M and D of the signals and their deviations are calculated. Alternatively (even for three signals equivalent here), instead of the medians M, D, the average of the two signals with the smallest distance between them may be used.

[0071] Then, we calculate the difference between each individual signal and the median,M,D,and normalize it to a threshold,S,. As the threshold,S, we use a combination of the variance,σ,of the signal-noise process and a component,T,representing the uncertainty of the time synchronization.

[0072]

number

[0073] The detection logic 28 tests whether the normalized signal differences ΔA, ΔB, ΔC from the median values ​​M, D exceed the value 1. If so, an anomaly ERROR A, ERROR B, ERROR C is assumed for the sensor signal in question.

[0074]

number

[0075] Because the error model parameters for the affected slave inertial measurement unit 5 have been determined based on the master inertial measurement unit 4, the slave inertial measurement unit 5 is processed using the same bias, drift, sensitivity, etc. as the master inertial measurement unit 4, and subsequent movement calculations can be performed without taking into account the switching process.

[0076] As already explained, in the event of an anomaly in the master inertial measurement unit 4, it is necessary to isolate the parameter estimates for the slave inertial measurement units 5, 6. This is done by signal 19.

[0077] If an abnormality, ERROR B or ERROR C, is detected in either of the slave inertial measurement units 5, 6, it is reported as a fault.

[0078] It is desirable to find an optimal compromise between sensitivity to sensor errors and resistance to false positives due to signal noise, the effects of road damage, and other uncertainties. [Explanation of symbols]

[0079] 1 Inertial measurement device 2. Sensor Fusion Device 3 Other sensors 4 Inertial Measurement Unit, IMU, Master Inertial Measurement Unit, Master IMU 5 Inertial Measurement Unit, IMU, Slave Inertial Measurement Unit, Slave IMU 6 Inertial Measurement Unit, IMU, Slave Inertial Measurement Unit, Slave IMU 7. Detection Equipment 8 Switching Device 9. Filter device 10 Filter device 11 Filter device 13 Error Model 14 Error Model 15 Estimated Blocks 16 Estimated Blocks 17 Time Delay Block 18 Time Delay Blocks 19 Signal 20 Preprocessed Signal 21 Corrected Signals 22 Corrected Signals 23 Error Model 24 Error Model 25 estimated blocks 26 Estimated Blocks 27 Blocks 28 Detection Logic 29 Decision-making logic A, B, C Pre-processed individual signals Bias_fB,Bias_fC Sensor bias Bias_ωB,Bias_ωC Sensor bias D Median ERROR A, ERROR B, ERROR C Abnormal d / dtωA rotational acceleration fA, fB, fC specific force fB_estimated, fC_estimated Estimated sensor output IMU A Inertial Measurement Unit, Master Inertial Measurement Unit IMU B Inertial Measurement Unit, Slave Inertial Measurement Unit IMU C Inertial Measurement Unit, Slave Inertial Measurement Unit M median P posture S Threshold T component v velocity α attitude angle ΔA, ΔB, ΔC normalized signal difference σ variance ωA, ωB, ωC rotational speed, angular velocity ωB-estimated, ωC-estimated target value

Claims

1. Angular velocity (ωA, ωB, ωC) and specific force (f Ａ , f Ｂ , f Ｃ A method for detecting an anomaly in an inertial measurement unit (4, 5, 6) used in a vehicle to measure a vehicle speed, the method comprising: A first inertial measurement unit (4) is used as a master inertial measurement unit, and a second inertial measurement unit (5) and a third inertial measurement unit (6), which may have lower performance reliability than the first inertial measurement unit (4), are used as slave inertial measurement units; a measurement value from the master inertial measurement unit (4) being used as a reference value to estimate parameters of an error model (13, 14, 23, 24) of both slave inertial measurement units relative to the master inertial measurement unit (4) as estimated sensor outputs and to correct measurements from the slave inertial measurement units (5, 6) based on the estimated sensor outputs, thereby detecting an anomaly in one of the three sensor signals based on corresponding three sensor signals and using 2-out-of-3 voting; one of the three inertial measurement units (4, 5, 6) whose sensor signal has a distance from the median (M, D) of the three sensor signals that exceeds a threshold value is identified as abnormal, and due to the abnormality of the master inertial measurement unit (4), the signal is switched to one of the slave inertial measurement units (5, 6) that is still functioning and a pre-calculated pre-correction is applied; the slave inertial measurement unit (5, 6) is pre-calibrated using a master-slave configuration; A time delay is applied to the estimated sensor outputs (f B_estimated , f C_estimated ); 4. A method according to claim 3, wherein if an error is detected in the first inertial measurement unit (4) within the time delay, the estimated sensor error in the slave inertial measurement unit (5, 6) is frozen at a value for the end of the time delay.

2. 2. The method of claim 1, wherein the parameters of the error model are estimated using a Kalman filter or a recursive least squares method.

3. an error model (13, 14, 23, 24) including all model parameters from the slave inertial measurement unit (5, 6) is used for estimating the parameters of the error model; 3. The method according to claim 1 or 2, characterized in that at least the time-varying relative sensor offsets from the master inertial measurement unit (4) are estimated online.

4. Method according to any one of claims 1 to 3, characterized in that the estimated parameters are limited to a maximum allowed value.

5. 3. The method according to claim 2, characterized in that, when detecting the significance of the signal deviation, the covariance of the estimation errors evaluated in a Kalman filter is used for the evaluation of the sensor signal deviation.

6. Method according to any one of claims 1 to 5, characterized in that all sensors in all inertial measurement units (4, 5, 6) are filtered using respective filter devices (9, 10, 11) and are time-synchronised.

7. The method according to any one of claims 1 to 6, characterized in that online estimated error model parameters are time delayed and applied to the measurements of the slave inertial measurement unit (5, 6).

8. a detection logic (28) is applied to the signal from the first inertial measurement unit (4) and the pre-corrected signal from the slave inertial measurement unit (5, 6) to detect the signal that deviates most from a mean or median (M, D) of the three sensor signals; The method according to any one of claims 1 to 7, wherein a normalized signal difference from the mean or median (M,D) exceeding a value of 1 is considered as a sensor error.

Images