MONITORING AND DETECTION OF SENSOR ERRORS IN INERTIAL MEASUREMENT SYSTEMS

Description

[0001] The invention relates to a method for detecting malfunctions in inertial measuring units according to the preamble of claim 1.

[0002] Inertial measurement units (IMUs) are arrangements of accelerometers and gyroscopes used to measure the specific forces (accelerations) and rotational speeds of bodies in space. Inertial measurement units play a crucial role in estimating the state of motion of vehicles. This state of motion is required in vehicle control systems, driver assistance systems, and automated driving, for example, to correctly interpret information from environmental sensors (cameras, lidar, radar). The term Vehicle Motion Observer (VMO) has been established for arrangements and methods for calculating the vehicle's own motion from the data of the inertial measurement unit in combination with additional vehicle sensors [5]. Such an arrangement isFigure 1 shown. In Figure 1shows, by way of example, the application of an inertial measurement unit 1 for motion estimation and navigation according to the prior art. The inertial measurement unit 1 sends angular velocity signals from the gyroscopes and specific force signals from the accelerometers [9]. In a sensor fusion unit 2, motion variables, for example attitude angle α, speed v and / or position P, are calculated by integration from these variables and corrected with the aid of further sensors or sensor units 3, for example odometers, magnetometers, barometric altimeters, steering angle, vehicle level and / or GNSS receivers. In the following, the result of the described data fusion will be referred to as the estimated motion state or, for short, as motion estimation. In places where the movement of the vehicle can be confused with the movement of objects in the vehicle's surroundings, the vehicle's own motion state orWe can speak of self-motion estimation.

[0003] A malfunction of an IMU sensor occurs when the specified (and thus predetermined by design) relationship between the physical measured value and the output sensor signal is lost, resulting in unspecified measurement behavior. This renders the affected IMU sensor unusable. Malfunctions in the sensors that propagate undetected into the motion calculation can lead to significant errors or grossly incorrect values ​​in the useful signals and therefore pose a safety risk in safety-relevant applications.

[0004] Vehicle control systems, driver assistance systems, and automated driving are applications with very high safety requirements. Therefore, malfunctions of the IMU sensors must be automatically detected and localized by the system, and the faulty sensor must be eliminated from the motion calculation within a specified time (FTTI - fault tolerant time interval), for example, by switching to a redundant sensor (backup).

[0005] The class of IMU systems for automotive applications considered here (automotive-grade IMUs) already includes basic monitoring functions, particularly for electrical faults. This type of monitoring is referred to below as monolateral monitoring because it is based on information from only one sensor at a time. Monolateral methods have low sensitivity and can only detect faults with relatively large amplitudes; this means that the monitoring thresholds are generally too high for driver assistance systems and automated driving. Diagnostic coverage for minor faults, which can also pose a safety risk, is low. Diagnostic coverage can be significantly improved by introducing redundant sensor elements.

[0006] A summary of the current state of the art for sensor arrays with simple redundancy is given in [4]. A related solution is presented in Figure 2The two redundant inertial measurement units 4 (IMU A) and 5 (IMU B) are first calibrated against each other with regard to their systematic errors (bias, sensitivity, alignment error, etc.) and then the corresponding sensor values ​​thus calibrated are compared. If the difference between the measured values ​​exceeds a threshold, then a malfunction in the respective sensor pair is likely. However, the malfunction cannot yet be assigned to one of the two sensors because this would require an additional reference value. In [4], this problem is realized by monitoring the innovations of the respective Kalman filters for estimating the self-motion.Because the motion states predicted with the help of the IMU are fused with the remaining vehicle sensors (wheel speeds, steering angle) in the two Kalman filters and the vehicle sensors represent a third independent source of information, an IMU error can be localized by evaluating the innovations (which are a weighted difference between prediction and measurement). A disadvantage of this approach is that the measured variables derived from the vehicle sensors are essentially limited to the translational speeds, and thus the monitoring method described in [4] is not equally selective in all degrees of freedom of the IMU. For example, the sensitivity to malfunctions that lead to deviations in the rotational degrees of freedom is low. In addition, the detection of small-amplitude malfunctions is limited by the resolution of the wheel speed sensors.In particular, for a fusion of the estimated self-motion state with environmental information as occurs in object tracking, there are sometimes higher requirements for the detection of IMU malfunctions with smaller detection thresholds and shorter error detection times, which may not be met by the described method.

[0007] Although [7] and [6] make the general claim that malfunctions in the self-motion estimation can be detected during the fusion of environmental sensors, the source does not contain any specific implementation examples. Furthermore, the claim focuses on malfunctions that generally lead to physically implausible information about the vehicle's motion state. No statement is made about malfunctions where the motion state is physically possible but nevertheless incorrect. Furthermore, the solution proposed therein contains no redundancy with regard to the inertial measurement units or the self-motion calculation. However, redundancy with regard to inertial sensors is necessary for short error detection times and high sensitivity of error diagnosis through small error thresholds.

[0008] From document

[11] , a correction of a position determined from image data of a sensor is known, whereby information from an inertial measuring unit is used to determine the orientation of the sensor.

[0009] The invention is based on the object of providing a novel method for detecting malfunctions in inertial measuring units.

[0010] The object is achieved according to the invention by a method having the features of claim 1.

[0011] Advantageous embodiments of the invention are the subject of the subclaims.

[0012] A method according to the invention for detecting malfunctions in inertial measuring units used in a vehicle for measuring angular velocities and specific forces, wherein two inertial measuring units, each with a plurality of sensors, comprising accelerometers and gyroscopic sensors, are used, makes use of a multi-hypothesis approach, for example.

[0013] According to the invention, a first inertial measuring unit is used as the master inertial measuring unit, wherein a second inertial measuring unit, the performance of which may be lower than that of the first inertial measuring unit, is used as the slave inertial measuring unit, wherein measurements of the master inertial measuring unit are used as reference values ​​in order to compensate measurements of the slave inertial measuring unit by estimating error model parameters compared to the master inertial measuring unit in order to detect a malfunction on the basis of two corresponding sensor signals of the two inertial measuring units by comparison, wherein the one of the two inertial measuring units is detected as faulty whose movement state, estimated in a respective downstream Kalman filter, leads to an incorrect prediction of the calculated positions of objects in the surroundings of the vehicle in an object tracking unit.

[0014] In one embodiment, an error condition is transmitted from a monitoring unit, to which sensor values ​​of the inertial measurement units are supplied, to the object tracking unit, so that in the event of a malfunction, several hypotheses about the motion condition can be tested, while in normal operation only one motion condition is used.

[0015] In one embodiment, a vehicle's own movement is described by a vehicle pose and transmitted to the object tracking unit.

[0016] In one embodiment, in the object tracking unit, a new object position of this object or the new object positions of the objects in vehicle-fixed coordinates are predicted from the calculated hypotheses about the vehicle pose and a previously calculated object position of an object or several previously calculated object positions of objects.

[0017] In one embodiment, the object position(s) of one or more objects measured by an environmental sensor system of the vehicle is / are compared with the object positions predicted via self-motion hypotheses of the vehicle determined by means of the Kalman filter, wherein those self-motion hypotheses are detected as false whose associated deviation between the predicted and measured object position exceeds a threshold value.

[0018] In one embodiment, in the event of a malfunction in one of the inertial measurement units, if the malfunction was detected by comparison with a redundant sensor of the other of the inertial measurement units and the self-motion hypothesis calculated with an erroneous sensor value was rejected as erroneous, the system switches to the other of the inertial measurement units.

[0019] In one embodiment, the results of the monitoring of the two inertial measurement units are included in an error detection of the object tracking unit, in which case a malfunction is only detected if the deviation of the predicted object position exceeds a threshold for one of the hypotheses but not for the other, and at the same time an error was detected via the sensor comparison of an IMU error detection unit.

[0020] In one embodiment, only one object detected as stationary or multiple objects detected as stationary are included in the error detection. A malfunction is detected for the hypothesis in which the predicted object position changes significantly, while it remains spatially constant for the other hypothesis.

[0021] The invention specifically exploits the higher sensitivity of object tracking methods with regard to errors in self-motion estimation for error detection and thus realizes detection of IMU malfunctions with very small amplitude.

[0022] Embodiments of the invention are explained in more detail below with reference to drawings.

[0023] Showing: Fig. 1 is a schematic view of a typical example of the application of an inertial measurement unit (IMU) for estimating the movement quantities of a vehicle, Fig. 2 is a schematic representation of a monitoring of IMU channels with simple redundancy, Fig. 3 is a schematic view of the solution according to the invention including the evaluation and error detection of the egomotion solutions within the object tracking, and Fig. 4A & 4B are a schematic representation to illustrate the basic idea of ​​the invention in two-dimensional space.

[0024] Corresponding parts are provided with the same reference numerals in all figures.

[0025] In a method for detecting malfunctions F in inertial measurement units with simple redundancy, two inertial measurement units 4 and 5, each with several accelerometers and gyroscopic sensors, are used, as in Figure 2 is shown.

[0026] In the figures, the two inertial measurement units 4 and 5 are also referred to as IMU A and IMU B. Each of them represents a vector ω → ib b with three angular velocity signals and one vector f → ib b with three specific force signals. To simplify the representation of these signals, the superscripts and subscripts are omitted. We can then refer to the sensor signals as ω A , ω B or f A , f B To refer to.

[0027] In Figure 2a first inertial measuring unit 4 is used as the master inertial measuring unit, wherein a second inertial measuring unit 5, the performance of which may be lower than that of the first inertial measuring unit 4, is used as the slave or backup inertial measuring unit, wherein in a monitoring unit 12 the measurements of the master inertial measuring unit 4 are used to estimate systematic error parameters in the slave inertial measuring unit 5 relative to the master inertial measuring unit 4 in models and to detect signal deviations between the signals compensated in this way, wherein an excessive signal deviation can be caused by a potential malfunction F in one of the two inertial measuring units 4 or 5. However, the detection in the monitoring unit 12 alone does not reveal which of the two inertial measuring units 4 or 5 is affected by a malfunction F. However, it is possible to reveal which specific sensor pair is affected.

[0028] To calculate the state of motion of the vehicle, the angular velocity signals ω A , ω B and specific force signals f A , f B Each of the two inertial measurement units 4, 5 is provided with a Kalman filter 13 and 14 (Vehicle Motion Observer, abbreviated VMO) as input variables and merged there with the measured variables of other vehicle sensors, e.g. wheel speeds and steering angles, whereby each of the two Kalman filters 13, 14 uses the same measured variables. The Kalman filters 13, 14 each determine a motion state BZ A, BZ B for the respective inertial measurement unit 4, 5. As can be seen from Figure 1 As can be seen, the described data fusion of signals from the inertial measurement unit 1 with the remaining vehicle sensors represents a standard method for determining the movement variables of the vehicle.

[0029] The solution proposed in [4] as state of the art provides, according to Figure 2to monitor the innovations of the two Kalman filters 13 and 14 in a monitoring unit 15 and to use them together with the information coming from the monitoring unit 12 for error detection IF in the error detection unit 16

[0030] The following formula symbols are used in the following description: t 0 time u , u ' Vehicle pose estimated from IMU and vehicle sensors Vehicle attitude angle increment Vehicle position increment C Δ q eb n rotation matrix calculated from the attitude angle increment Δ q eb n Δ p eb n H ( u ) Homogeneous transformation of earth-fixed coordinates into vehicle-fixed coordinates x Object position z Object position measurement using environmental sensors. K Kalman filter gain v innovation S Measurement error covariance atrix P -< State error covariance matrix before correction with measurement (a-priori) R Covariance matrix of the measurement κ Threshold Ψ Yaw angle of the vehicle with respect to an earth-fixed coordinate system ΔΨ Yaw angle increment ω z Yaw rate δω z Yaw rate error vx Vehicle longitudinal speed vy Vehicle lateral speed D t Time interval between two measurements l r Distance of the vehicle-fixed coordinate system from the rear axle of the vehicle.

[0031] In contrast to the state of the art in Figure 2 are according to the invention Figure 3 the results of the calculations of the state of motion (correct vehicle pose u and incorrect vehicle pose u' ) of the vehicle and the information from the monitoring unit 12 is fed to the object tracking unit 17. There, the movement state is used in a prediction block 18 to predict the position x -< of the objects in the vehicle's surroundings. The predicted positions x -< are then corrected in a correction unit 19 for the object position with object measurements z from the environmental sensors and corrected accordingly (corrected position x +< of the objects in the vehicle environment). In case of malfunction F of an inertial measurement unit 4, 5, a large error between the predicted position x -< and measured or corrected position x +< occur. This is then used in combination with the independent IMU error detection in the monitoring unit 12 to detect and localize the IMU malfunction F (assignment to the relevant IMU) and, if necessary, to isolate the IMU error by switching to the redundant solution, i.e., the error-free inertial measurement unit 4, 5. If none of the inertial measurement units 4, 5 exhibits a malfunction F, then a good agreement between the predicted position x -< and measured or corrected position x +< are present.

[0032] The detection problem is solved by propagating the two redundant self-motion estimates to the object tracking unit 17. These are then available there as hypotheses of the vehicle's self-motion.

[0033] The object tracking unit 17 typically receives data regarding objects in the vehicle's surroundings from various sensor sources (e.g., camera, radar, lidar, and ultrasound). This data is typically fused and tracked over time to extract object information, track the temporal movement of these objects, and predict their future movement. For example, dynamic objects such as vehicles and pedestrians can be tracked. The detection of static objects, e.g., small obstacles on the roadway, is another example of tasks performed by the object tracking unit 17.

[0034] When fusing objects in the vehicle environment, the vehicle's own motion must be compensated in order to calculate the correct position and speed of these objects. For this purpose, an accurate and reliable motion estimate from the VMO 13 or 14 is used as input to the object tracking unit 17. In the object tracking unit 17, the own motion state is used to predict the position of the surrounding objects between observations via the environment sensors. An incorrect own motion state leads to the position of these objects being incorrectly predicted. For example, this can lead to static objects changing their position from one observation time to the next. All tracked static objects would be affected simultaneously. This leads to the basic idea of ​​the invention: If for an object in the vehicle environment that was previously classified as static (e.g.a wall), a movement is suddenly predicted, then there is probably a disturbance in the movement calculation.

[0035] The inventive solution now consists in calculating two ego-motion estimates in respective Kalman filters 13, 14 or VMO as hypotheses for the two redundant inertial measurement units 4 and 5. At the same time, the sensor values ​​of the inertial measurement units 4, 5 are fed to a monitoring unit 12. If both inertial measurement units 4, 5 are operating correctly, they provide similar sensor signals with differences that are below the thresholds of the monitoring unit 12, and the ego-motion states calculated from them are also identical and consistent with the objects in the vehicle's surroundings observed by the object tracking unit 17. If a malfunction F occurs in one of the IMU sensors 4, 5, this event is detected by the monitoring unit 12.Secondly, the hypotheses for the vehicle's own motion state are now different, and the faulty solution leads to a discrepancy between the predicted and measured position of objects. On the other hand, the faulty motion state calculated from the faulty IMU will be consistent with the observed objects. By combining the two methods, both the faulty IMU 4, 5 and the sensor with the malfunction F can be clearly localized, and the system can switch to the backup solution.

[0036] In one embodiment, the motion states BZ A, BZ B calculated from the two redundant inertial measuring units 4, 5 can be fed to the object tracking unit 17 for further processing.

[0037] In one embodiment, the error state from the IMU monitoring unit 12, which was determined by comparing corresponding IMU sensors of the two inertial measuring units 4, 5, can be fed to the object tracking unit 17, so that only in the case of a malfunction F do both hypotheses regarding the motion state have to be checked.

[0038] In one embodiment, the proper movement of the vehicle 22 is determined by a vehicle pose u _ = Δ q eb n Δp b n described, where Δ q eb n the position increment (angular increment) of the vehicle 22 relative to a local earth-fixed (tangential) coordinate system 20 with the axes xn< and yn< and Δ p b n is the position increment of the vehicle 22 in the earth-fixed coordinate system 20. Now denotes C Δ q eb n the rotation matrix, which can be calculated from the position increment, then the entire transformation (rotation and translation) of an object from earth-fixed to vehicle-fixed coordinates can be described by the homogeneous matrix (Homogeneous transformation of earth-fixed coordinates into vehicle-fixed coordinates) from [8, 10], which is well known from practical applications. H u _ = C − 1 Δ q eb n − C − 1 Δ q eb n Δ p b n 0 1 be described.

[0039] In one embodiment, predicted values ​​for the position of objects in the vehicle environment can be calculated from the two states for the vehicle's own motion and previous object positions. Δ p o n the position of an object in earth-fixed coordinates, its position is Δ p o b in vehicle-fixed coordinates by the equation Δ p o b 1 = H u _ ⋅ Δ p o n 1 given. Here, Δ p o b the position of the object as it can be measured by a vehicle-mounted environmental sensor (camera, lidar, radar) relative to the vehicle 22.

[0040] In one embodiment, the object position x _ = Δ p o n 1 of the object in earth-fixed coordinates represents the state of the object for object tracking and z _ = Δ p o b 1 is the measured value provided by the environment sensor (object position measurement by the environment sensor), which must be fused in a Kalman filter 13, 14 to localize the object. For simplification, it is assumed here that the object has already been classified as static. In this case, x be constant and the fusion task in the Kalman filter 13, 14 consists in a correction (innovation) of the estimated object state x (the estimated position of the object) using the measurement z . Now refers to z k a sequence of measurements of the object at discrete times k = 0, 1, 2, ... and are u k the corresponding vehicle poses and x _ ^ k − the estimation of the object position before the correction (a-priori), then the correction is based on an improved (a-posteriori) position estimate x _ ^ k + about the equation x _ ^ k + = x _ ^ k − + K ⋅ z _ k − H u _ k x ^ _ k − The difference between the measurement and the predicted measurement value determined from the vehicle's own motion and a priori position estimation is considered as innovation v k is called [2, 3, 1]: ν _ k = z _ k − H u _ k x _ ^ k − According to the Kalman filter theory, the statistical distribution of innovation can be described by its covariance matrix S k = H u _ k P k − H T u _ k + R k calculate where P k − the covariance matrix of the a priori state estimation error and R k is the covariance matrix of the measurement error. The latter contains basic statistical assumptions about the accuracy (error distribution) of the measurement error and the egomotion estimation. For each component v ik the innovation vector v k then the estimate applies ν ik < κ s iik , κ = 3 … 4

[0041] This means that the difference between the measurement and the predicted measurement should be smaller than a multiple of its value, determined by the diagonal element s iik the covariance given, standard deviation. (If κ = 3 then on average 99.73% of the innovations should be in the range described by equation (9). κ = 4 it would be 99.994%). If this is not the case, there is either an error in the measurement (outlier) or an error in the prediction (e.g. due to incorrectly calculated proper motion). Now let u _ k A and u _ k B two hypotheses for the proper motion based on the respective inertial measurement units A and B, 4 and 5 and x ^ _ k − the a-priori estimated position of an object classified as static in the vehicle environment. If a malfunction F in IMU A, 4 and the resulting error in u _ k A that for the elements of the assigned innovation ν ik A > κ s iik applies, but for the elements of the hypothesis u _ k B assigned innovation applies ν ik B < κ s iik then hypothesis A can be clearly detected as false and there is an error in the self-motion A. In this way, an error previously detected by the IMU monitoring of the monitoring unit 12 can then be clearly assigned to one of the two IMUs 4, 5.

[0042] A simple example in two-dimensional space will now illustrate this procedure using Figure 4 illustrate. Figure 4a shows the relationship between an earth-fixed coordinate system 20 and a vehicle-fixed coordinate system 21 with the axes xb< and yb< in two-dimensional space for a vehicle 22. The yaw angle Ψ describes the rotation between the two systems. The corresponding rotation matrix is C Ψ = cos Ψ − sinΨ sinΨ cosΨ

[0043] In Figure 4b In the example shown, a vehicle 22 travels at a constant speed vxstraight ahead towards an object directly in front of the vehicle 22 at the object position 25, which is defined by the vector x _ = x o 0 1 described. At the time t 0 corresponding to the discrete time k = 0 the vehicle 22 has a yaw angle increment ΔΨ 0 = 0 and the position increment Δ p b 0 n = 0 0

[0044] The movement of the vehicle 22 is described by the following three equations: Δ Ψ ˙ = ω z Δ x ˙ = v x cosΔΨ − v y sinΔΨ Δ y ˙ = v x sinΔΨ + v y cosΔΨ

[0045] The rotation speed ω z is also referred to as yaw rate in this context and is measured via the inertial measurement unit 4, 5. The speed vx is the body-fixed longitudinal speed of the vehicle 22, which can be calculated in the VMO 13, 14 by fusion of the wheel speed sensors with the IMU data. A common simplification of the calculation for the example is that the lateral speed of the vehicle 22 is derived from the equation v y ≈ l r ω z calculated. l r the distance of the coordinate origin of the vehicle-fixed coordinate system 21 from the rear axle of the vehicle 22. This is referred to as purely kinematic driving. When the vehicle 22 is driving straight ahead, ω z = 0. It follows that ΔΨ = 0 and an integration of the equations of motion up to the time t 1 = t 0 + Δ t according to discrete time k = 1 results in the new pose 23 u 1 of the vehicle 22 ΔΨ 1 = 0 , Δ p b 1 n = v x Δ t 0

[0046] Consequently, the measured position of the object in vehicle coordinates is z _ 1 = H u _ 1 x _ o = 1 0 − v x Δ t 0 1 0 0 0 1 ⋅ x o 0 1 = x o − v x Δ t 0 1

[0047] Join now at the time t 0 a constant bias error δω z in the yaw rate sensor of the IMU, then the sensor instead of the real yaw rate ω z = 0 the faulty yaw rate ω z = 0 + δω z measured. Integration of equation (15) then leads to t 1 on the wrong yaw angle ΔΨ 1 = δω z ⋅ Δ t

[0048] This error propagates into the position calculation and leads to an incorrectly calculated pose 24 of the vehicle 22. For demonstration purposes, we assume that the yaw angle error is small and for trigonometric functions the angle approximations sinΔΨ ≈ ΔΨ , cosΔΨ ≈ 1 With this approximation, the erroneous position caused by the yaw rate disturbance can be calculated analytically by integrating equations (16) and (17). One obtains Δ p b 1 n = v x Δ t − 1 2 l r ΔΨ 1 2 1 2 v x Δ t + l r ΔΨ 1

[0049] With the rotation matrix obtained from the small angle approximation C ΔΨ 1 = 1 − ΔΨ 1 ΔΨ 1 1 can then the matrix H ( u 1 ) can be calculated. H u _ 1 = 1 ΔΨ 1 − 1 + 1 2 ΔΨ 1 ⋅ v x Δ t − 1 2 l r ΔΨ 1 − ΔΨ 1 1 1 2 v x Δ t − l r ΔΨ 1 − 1 2 l r ΔΨ 1 3 0 0 1

[0050] The final result for the prediction of the measurement from the vehicle's own motion is z ^ 1 − = H u _ 1 ⋅ x _ 1 − = x o − 1 + 1 2 ΔΨ 1 2 ⋅ v x Δ t − 1 2 l r ΔΨ 1 2 1 2 v x Δ t − l r − x o ⋅ ΔΨ 1 − 1 2 l r ΔΨ 1 3 1

[0051] This result is best illustrated by a numerical example: v x = 30 m s l r = 1 , 5 m δω z = 1 ∘ s x o = 80 m Δ t = 0 , 3 s

[0052] For the case of the yaw rate without malfunction F, the exact prediction of the measurement z ^ 1 − = 71 m 0 1

[0053] For the case of the yaw rate with malfunction F one obtains z ^ 1 − = 70 , 9999 m − 0 , 4032 m 1

[0054] This means that with the wrong proper motion, the position of object 25 in vehicle coordinates is determined with respect to the y -component is predicted to be shifted by more than 0.4 m to the right, while the measured object position 26 is still exactly in front of the vehicle 22 in the case of malfunction F. If the corresponding standard deviation is now, for example, s iik = 0.1 m , then the corresponding hypothesis can be κ = 3 as false.

[0055] In one embodiment, the two hypotheses regarding the predicted positions of the objects can be compared with the object positions 25, 26 measured by the environmental sensors. If the deviation of the measured object position 25, 26 from the predicted object position 25, 26 exceeds a threshold value for one of the hypotheses, but not for the other hypothesis, then an error in the respective inertial measurement unit 4, 5 can be inferred.

[0056] In one embodiment, the results of monitoring the two IMU sensors 4, 5 can be included in the error detection of the object tracking unit 17. In this case, a malfunction F is only detected if the deviation of the predicted object position 25, 26 exceeds a threshold for one of the hypotheses but not for the other, and at the same time, an error was detected via the sensor comparison of the IMU error detection unit 16.

[0057] In one embodiment, only one object recognized as stationary or several objects recognized as stationary can be included in the error detection and a malfunction F can be detected for the hypothesis in which the predicted object position 25, 26 changes significantly, while it remains spatially constant for the other hypothesis.

[0058] The invention enables a fault-tolerant determination of the motion variables of a vehicle 22 in the event of an IMU malfunction F (fail operational) with simple redundancy. This does not impose any specific requirements on geometric arrangement and sensor quality. Rather, the invention is also applicable to cost-effective, heterogeneous IMU hardware, some of which is already present in the vehicle 22.

[0059] The method according to the invention allows the signals of the master inertial measuring unit 4 to be protected against sensor errors as a measure to ensure the functional reliability (fail-operational) of the estimation of the self-motion of a vehicle, taking into account increased safety requirements in automatic driving.

[0060] The focus is on the use of heterogeneous (not identical) inertial measurement units (IMU) 4, 5, which are installed in different locations in the vehicle 22.

[0061] The inventive solution achieves fault tolerance (fail operational) in the calculation of vehicle movement variables. It draws on the systems and methods for object detection and fusion already available in driver assistance systems and automated vehicles. Bibliography

[0062] [1] Berman, Z. : Outliers Rejection in Kalman Filtering-Some New Observations. In: IEEE / ION Position, Location and Navigation Symposium - PLANS 2014, S. 1008-1013 9 [2] Grewal, M. S. ; Adrews, A. P.: Kalman Filtering. New York : John Wiley, 2001 9 [3] Groves, P. D.: Principles of GNSS, Inertial and Multisensor integrated Navigation systems. Artech House, 2013 9 [4] Kalkkuhl, J. ; Bergmann, M. ; Engelhardt, T. ; Bleimund, F. : DE 10 2021 004 103 A1 : Method and arrangement for monitoring and detecting sensor errors in inertial measuring systems 3, 17 [5] Klier, W. ; Reim, A. ; Stapel, D. : Robust Estimation of Vehicle Sideslip Angle - An Approach w / o Vehicle and Tire Models. In: SAE Technical Paper Series (2008), Nr. 2008-01-0582 1 [6] Mercep, L. ; Pollach, M. : US10553044B2 : Self-diagnosis of faults with a secondary system in an autonomous driving system. 4 [7] Mercep, L. ; Pollach, M. : US11145146B2 : Self-Diagnosis of Faults in an Autonomous Driving System 4[8] Rabe, C.: Detection of Moving Objects by Spatio-Temporal Motion Analysis, Christian-Albrechts-University Kiel, Diss., 2011 7 [9] Wendel, J.: Integrated Navigation Systems. Oldenbourg, 2007

[10] Woo, A.; Fidan, B.; Melek, WW: Localization for Autonomous Driving. In: Zekavat, SA (ed.); Buehrer, RM (ed.): Handbook of Position Localization. IEEE Press Wiley, Chapter 29 7

[11] Brian K. Rupnik; Brian T. Whitehead: US 2020 / 159224 A1: SYSTEMS AND METHODS OF ADJUSTING POSITION INFORMATION

Claims

1. Method for detecting malfunctions (F) in inertial measuring units (4, 5) used in a vehicle (22) for measuring angular velocities (ωA, ωB) and specific forces (fA,fB), two inertial measuring units (4, 5) being used that each have a plurality of sensors, comprising accelerometers and gyroscopic sensors, a first inertial measuring unit (4) being used as the master inertial measuring unit, a second inertial measuring unit (5), of which the performance capabilities may be lower than those of the first inertial measuring unit (4), being used as the slave inertial measuring unit, measurements of the master inertial measuring unit (4) being used as reference values in order to compensate for measurements of the slave inertial measuring unit (5) by estimating error model parameters compared with the master inertial measuring unit (4), in order to detect a malfunction (F) on the basis of two corresponding sensor signals of the two inertial measuring units (4, 5) by comparison, characterized in that of the two inertial measuring units (4, 5), the inertial measuring unit of which the motion estimation, calculated in a particular downstream Kalman filter (13, 14), leads in an object tracking unit (17) to an incorrect prediction of the calculated positions of objects in the surroundings of the vehicle (22) is detected as faulty.

2. Method according to claim 1, characterized in that an error state is transmitted from a monitoring unit (12), to which sensor values of the inertial measuring units (4, 5) are supplied, to the object tracking unit (17), so that in the event of a malfunction (F) a plurality of hypotheses regarding the movement state (BZ A, BZ B) can be tested, while in normal operation only one movement state (BZ A, BZ B) is used.

3. Method according to either claim 1 or claim 2, characterized in that a proper motion of the vehicle (22) is described by a vehicle pose (u, u') and transmitted to the object tracking unit (17).

4. Method according to claim 3, characterized in that in the object tracking unit (17), a new object position (25, 26) of an object or new object positions (25, 26) of objects in vehicle-fixed coordinates are predicted from the calculated hypotheses about the vehicle pose (u, u') and a previously calculated object position (25, 26) of this object or a plurality of previously calculated object positions (25, 26) of the objects.

5. Method according to any of claims 1 to 4, characterized in that the object position (25, 26) of one or more objects measured by an environment sensor system of the vehicle (22) is compared with the object positions (25, 26) predicted via proper-motion hypotheses of the vehicle (22) determined by means of the Kalman filters (13, 14), proper-motion hypotheses of which the associated deviation between the predicted and measured object position (25, 26) exceeds a threshold value being detected as false.

6. Method according to any of claims 1 to 5, characterized in that in the event of a malfunction (F) in one of the inertial measuring units (4, 5), if the malfunction (F) was detected by comparison with a redundant sensor of the other of the inertial measuring units (4, 5) and the proper-motion hypothesis calculated with a faulty sensor value was rejected as faulty, a switch is made to the other of the inertial measuring units (4, 5).

7. Method according to any of claims 1 to 6, characterized in that the results of the monitoring of the two inertial measuring units (4, 5) are included in an error detection of the object tracking unit (17), in this case a malfunction (F) being detected only if the deviation of the predicted object position (25, 26) exceeds a threshold for one of the hypotheses but not for the other, and at the same time an error was detected via the sensor comparison of an IMU error detection unit (16).

8. Method according to any of claims 1 to 7, characterized in that only one object recognized as stationary or a plurality of objects recognized as stationary are included in the error detection and a malfunction (F) is detected for the hypothesis in which the predicted object position (25, 26) changes significantly, while it remains spatially constant for the other hypothesis.