METHOD FOR DETERMINING A PROTECTION LEVEL BY MEANS OF A Copula-ASSISTED Bayesian FRAMEWORK WHEN A VEHICLE HAVING A GNSS-SUPPORT POSITIONING

Abstract

The invention relates to a method for determining a protection level (7) by means of a Copula-assisted Bayesian framework during driving of a vehicle having a GNSS-supported positioning system, in which a plurality of environment classes (10, 11) and a plurality of GNSS quality indicators (9) are predefined, and in which a Copula model (15, 16, 17) is provided offline for each environment class (10, 11) and each GNSS quality indicator (9) and stored in a memory, the method comprises the following steps: a) classifying the environment while driving on the basis of GNSS quality indicators (9) and ascertaining an environment category (10, 11) to which the environment belongs, b) reading a Copula model (15, 16, 17) corresponding to the ascertained environment category (10, 11) for the respective quality indicator (9) from the memory, c) extracting at least one likelihood function (1, 2, 3, 4, 12) from the read Copula model (15, 16, 17) as a function of the quality index value (13), d) ascertaining a posterior distribution (6) for the error by multiplying the at least one likelihood function (1, 2, 3, 4, 12) with a prior distribution (5) for the position error on the basis of the Bayesian theorem, and e) determining a protection level (7) from the posterior distribution (6).

Description

Background Technology

[0001] This invention relates to a method for determining protection levels using a Copula-assisted Bayesian framework while a vehicle with a GNSS-supported positioning system is in motion. Furthermore, a controller, a computer program, a machine-readable storage medium, and a positioning system are also provided. This invention is particularly applicable to GNSS-supported positioning systems for automated or autonomous driving systems.

[0002] It is known that positioning and navigation on the ground and in the air can be achieved by receiving navigation satellite signals using a Global Navigation Satellite System (GNSS). With the aid of multi-frequency and multi-constellation reception, position can be determined on the ground with centimeter-level precision. The quality of these positions is determined by meeting the requirements imposed upon them, particularly accuracy, continuity, availability, and integrity.

[0003] It can be determined that integrity plays an extremely important role, just as important as positioning accuracy, in safety-critical autonomous driving, because integrity ensures the reliability of positioning accuracy, and in safety-critical environments, failure to monitor integrity can lead to catastrophic consequences.

[0004] Initially, the term "integrity" was introduced for aerial positioning and navigation, and positional error can be described by the following parameters: - AL (Alert Limit) describes a position error tolerance that must not be exceeded. Otherwise, an alarm will be triggered. - TTA (Time to Alert) describes the maximum permissible time interval from exceeding the alarm limit until the alarm is triggered. - IR (Integrity Risk) describes the probability that the position error exceeds the alarm limit. - PE (Position Error) describes the deviation between the determined position and the actual position. - PL (Protection-Level) describes the position error that the algorithm guarantees will not be exceeded without being noticed. - FA (False Alarm) describes an event that triggers an alarm without exceeding the alarm limit. - MI (Misleading Information) describes an event where the protection level is less than the position error and both the protection level and the position error are less than the alarm limit. HMI (Hazardously Misleading Information) describes events where the protection level is less than the position error and alarm limit, and the position error exceeds the alarm limit.

[0005] The parameter "Protection Level" (PL) is the core of integrity monitoring. It can be output by the positioning system along with the location and ensures that the entire system is safe when the location error is below a specified alarm time.

[0006] Known methods for determining protection levels are typically within the framework of ABSA, GBAS, or SBAS schemes, and although developed for airborne positioning and navigation (e.g., Greer et al., 2007; Gratton et al., 2010; Zhu et al., 2018), their standardized integrity algorithms are generally defined considering the receiving conditions of an aircraft in flight, and therefore are not applicable to ground-based positioning and navigation for autopilot given the many critical environmental conditions (e.g., in urban environments) and receiving conditions (e.g., multipath reception). Furthermore, methods developed within the framework of ABSA, GBAS, or SBAS schemes typically involve single-frequency reception. In contrast, as mentioned at the beginning, autopilot requires multi-frequency and multi-constellation reception.

[0007] Although the known ARAIM scheme (Blanch et al., 2007) focuses on multi-frequency and multi-constellation reception to provide information about failed GNSS satellites and is more robust than the ABSA, GBAS, or SBAS schemes in terms of lower ionospheric delay (due to multi-frequency reception) and higher measurement redundancy (due to multi-constellation reception), it appears difficult to implement in methods for determining protection levels for ground-based positioning and navigation. In this context, the following challenges are particularly significant:

[0008] Using non-ionospheric (IF) measurements is known to increase the degree of uncorrelated error between frequencies, such as thermal noise, multipath effects, or some degree of distortion. In a typical road environment, this can lead to large positional uncertainties.

[0009] Furthermore, GNSS receivers in aviation typically perform pseudorange phase smoothing for over 100 seconds to reduce noise and multipath effects. This approach cannot be used in automotive applications because carrier phase tracking may not be reliably maintained for extended periods due to environmental conditions.

[0010] Furthermore, for most planned applications in the automotive industry, more stringent alarm limits and alarm times are expected than in aviation (typically alarm limits are in the range of 0.5 to 10 meters, and alarm times are in the range of 1 second), without necessarily requiring less stringent integrity risks (IR). This means that typical orders of magnitude for ARAIM schemes could lead to excessively large positional errors.

[0011] Therefore, it is promising to develop a method for determining protection levels in the context of ground-based positioning and navigation. This method should be applicable not only to multi-frequency and multi-constellation reception but also to determining robust protection levels under critical environmental conditions. This is particularly important for autonomous driving, where, in addition to positioning accuracy, there are exceptionally high requirements for the security, integrity, and correctness of positioning information. Summary of the Invention

[0012] To this end, a method is provided to help determine the protection level using a Copula-assisted Bayesian framework while a vehicle with a GNSS-supported positioning system is in motion, wherein multiple environmental categories and multiple GNSS quality indices are predefined, and a Copula model is provided offline for each environmental category and each GNSS quality index and stored in memory. The method includes the following steps: a) Classify the environment during driving based on GNSS quality indicators and determine the environmental category to which the environment belongs. b) Retrieve the Copula model corresponding to the desired environment category and the corresponding quality index from memory. c) Extract at least one likelihood function from the read Copula model based on the quality index value. d) Obtain the posterior distribution of the error by multiplying at least one likelihood function with the prior distribution for the position error based on Bayes' theorem, and e) Determine the protection level from the posterior distribution.

[0013] The described method is particularly applicable to autonomous driving. Here, autonomous driving specifically refers to vehicles, mobile robots, and unmanned transport systems that move largely autonomously using GNSS receivers and based on the Global Navigation Satellite System (GNSS). It is especially advantageous if the autonomous vehicle is equipped with a positioning system for performing the described method. This positioning system can be GNSS-based or GNSS and INS-based, which identifies GNSS satellites within its field of view, receives GNSS signals, and performs positioning based on the received GNSS signals.

[0014] The described method can be configured to be executed repeatedly on a periodic or periodic basis. A period here specifically refers to the time interval between when the currently determined position is valid or until the currently determined position is updated. This could mean that the described method is executed once each time the position is updated, so the currently determined protection level can be output along with the currently determined position. It could also mean that within a period, all positioning-related parameters, such as the factor of precision (DOP) or the number of available GNSS signals in the field of view, are updated once.

[0015] Although the steps are given here in a specific order using letters a) through e), this order does not always need to be followed. For example, individual steps may be repeated independently and / or partially omitted in cases of repetition. Steps may be performed at least partially overlapping in time.

[0016] The basic concept of this invention is based on providing a Copula model offline in advance, using training data obtained through test measurements and / or simulation detection, and storing it in memory. The Copula model describes the statistical correlation between rank-ordered datasets. Copula is a multivariate cumulative distribution function for which the marginal probabilities of each variable are uniformly distributed over the interval [0, 1] (Schmidt, T. (2007). Coping with copulas. Copulas-From theory to application in finance, 3-34). The offline-provided Copula model can be read from memory according to step b) and used to generate a likelihood function according to step c). Therefore, the protection level can be determined online according to steps d) and e), using a posterior distribution obtained by multiplying the likelihood function with the prior distribution based on Bayes' theorem. In step d), the error can be a deviation of all parameters output by a GNSS-supported positioning system, such as position error, velocity error, and / or orientation error. In step e), the protection level can be determined from the posterior distribution based on the target integrity risk.

[0017] The advantage of this invention is that the protection level is not dependent on variance.

[0018] When providing a Copula model, it can be configured to model the Copula model using training data, taking into account different environmental conditions and GNSS quality metrics. Here, environmental conditions can be categorized into different environmental classes, allowing at least one Copula model to be provided for each environmental class. Furthermore, multiple GNSS quality metrics can be predefined, enabling a Copula model to be provided for each environmental class and each GNSS quality metric.

[0019] GNSS quality indices characterize the quality of position estimation for a positioning algorithm and can be given as a number between zero and a positive real number (see [reference]). Figure 2 Here, zero represents very poor quality, while a positive real number represents very good quality. GNSS quality metrics involve the key signals or parameters of the GNSS that can be measured or calculated online.

[0020] GNSS quality metrics can be expressed as the factor of accuracy (DOP). DOP is a measure of the quality of available GNSS signals under line-of-sight conditions and describes how well the GNSS satellites transmitting these available signals are positioned relative to each other at a given location for positioning. GNSS quality metrics can also be expressed as the number of available GNSS signals in the field of view. Since at least four GNSS signals are typically required for positioning, at least five for integrity monitoring, and at least six for identifying faulty GNSS satellites, the number of available GNSS signals is equally important for positioning quality.

[0021] As mentioned at the beginning, ground positioning is strongly influenced by environmental conditions. It is possible to classify environmental conditions not only during offline modeling of the Copula model, but also when determining the protection level online according to step a).

[0022] Environmental conditions can be categorized into different types based on multipath effects caused by obstacles during GNSS signal propagation: open environments, urban environments, and critical environments. Open environments include scenarios such as driving in the wilderness or on highways, where there are virtually no obstacles. Urban environments include scenarios such as driving deep into city streets, where numerous buildings obstruct GNSS signal propagation and / or reflect GNSS signals, where multipath reception must be considered. Critical environments include scenarios such as driving on roads with multiple lanes (where multiple lanes overlap).

[0023] When operating online, not only the DOP (Depth of Optimum) but also the number of available GNSS signals is typically output along with the determined location. In other words, GNSS quality metrics can be measured and / or calculated online. Alternatively, the DOP and the number of available GNSS signals in the field of view can be (pre-calculated) based on the location, time, and motion pattern to provide a Copula model. Because each GNSS system has multiple GNSS satellites (e.g., Galileo has 28 satellites, GPS has 24), these satellites are evenly distributed in the sky or orbits and move according to a specific motion pattern. This means that at a specific location and at a specific time interval, only specific GNSS signals from specific GNSS satellites can be received. Therefore, these motion patterns are publicly available and can be downloaded (e.g., from the International GNSS Service (IGS)).

[0024] Preferably, when a Copula model is provided through the following steps: i) Provide training data through test measurements and / or simulated detection, ii) Classify the training data into environment categories. iii) Provide experience copulas for each environmental category and each quality indicator, and iv) Match the parsed Copula model to the empirical Copula.

[0025] Training data can come from measurements of real-world locations and / or simulated test data. Simulated test data can be generated, for example, by a GNSS signal generator capable of taking multipath propagation into account, i.e., different environmental categories.

[0026] Preferably, in step c), at least one likelihood function is extracted from the Copula model through the following sub-steps: 1) Read the conditional distribution of position error from the Copula model, and 2) Detect at least one likelihood function from the conditional distribution based on different GNSS quality indices.

[0027] Furthermore, preferably, in sub-step 1), the conditional distribution is provided by training data obtained through test measurements and / or simulation detection.

[0028] Furthermore, preferably, in step d), the prior distribution is based on the training data and is defined when using a parameterized distribution.

[0029] Particularly preferably, in step d), the posterior distribution for the position error is obtained for each period.

[0030] Compared to known methods for determining protection levels based on the variance of estimates from the parameter estimation process or on GNSS residuals from RAIM, the described method does not rely on the variance of the estimates but is based on GNSS quality indices. Quality indices are key signals or parameters of the system that can be measured and / or calculated online and characterize the quality of the position estimation status for the positioning algorithm.

[0031] To this end, Bayes' theorem is used to obtain the posterior distribution for position errors for each period by multiplying multiple likelihood functions by a prior. The likelihood functions are obtained from the conditional probability distribution of position errors (from training data) at different GNSS quality indices. One of the main advantages of this invention, particularly compared to known methods, is the use of Copula to generate the likelihood functions. This allows for the acquisition of reliable likelihood functions, thus directly deriving the posterior distribution. By directly deriving the posterior distribution, robust protection levels can be determined, thereby reducing the probability of HMI problems.

[0032] Preferably, the controller for the GNSS receiver is configured to perform the described method.

[0033] Furthermore, preferably, a computer program is used to perform the methods described herein. In other words, this particularly relates to a computer program (product) that includes instructions, which, when executed by a computer, cause the computer to perform the methods described herein.

[0034] More preferably, a machine-readable storage medium is used, on which the computer program described herein is stored. Typically, a machine-readable storage medium is a computer-readable data carrier.

[0035] Particularly preferably, the positioning system for the vehicle is configured to perform the methods described herein. Attached Figure Description

[0036] The solutions and their technical background presented herein will be described in more detail below with reference to the accompanying drawings. It should be noted that the invention should not be limited to the embodiments shown. In particular, unless otherwise expressly stated, certain aspects of the facts illustrated in the drawings may be extracted and combined with other components and / or insights from other drawings and / or this specification. The following is schematically illustrated:

[0037] Figure 1 For example function plots of the likelihood function, prior distribution, and posterior distribution,

[0038] Figure 2 Extract the likelihood function from the Copula model based on the quality index values.

[0039] Figure 3: Used to provide a block diagram of the Copula model, and

[0040] Figure 4 : A block diagram describing the method. Detailed Implementation

[0041] Figure 1 and Figure 2 The mathematical functions used in the described method and their associated data are shown and illustrated herein to facilitate a better understanding of the invention at the data level.

[0042] Figure 1 A function graph is shown, in which the likelihood functions P(Qi1|dx)1, P(Qi2|dx)2, P(Qi3|dx)3, P(Qi4|dx)4, prior distribution 5 and posterior distribution 6 are compared with each other and distributed on the dx axis 8 corresponding to the position error.

[0043] To obtain the protection level 7 during online operation, the corresponding likelihood functions 1, 2, 3, and 4 are first multiplied by the prior distribution 5 based on Bayes' theorem, thus obtaining the posterior distribution 6, whose two boundaries are as follows: Figure 1 As shown, this corresponds to protection level 7, which is yet to be determined.

[0044] Each likelihood function 1, 2, 3, and 4 is associated with a quality index (not shown). For example... Figure 2 As shown, during online driving, for example, the likelihood function 12 being extracted / already extracted can be extracted from the corresponding Copula model (not shown) based on the quality index value 13. Figure 2 It can also be seen that the scale of quality index 9 ranges from zero to thirty-five.

[0045] Figure 3 The flow of the method proposed herein for providing Copula models 15, 16, and 17 is illustrated schematically. The order of sub-steps i), ii), iii), and iv) shown with respect to blocks 210, 220, 230, and 240 is merely exemplary. In block 210, training data is provided through test measurements and / or simulated detection. In block 220, the provided training data is classified into different environment categories 10 and 11. In block 230, an empirical Copula is provided for each category 10 and 11, between a quality metric 9 and a position error. In block 240, the resolved Copula model is matched to the empirical Copula to provide Copula models 15, 16, and 17.

[0046] Figure 4A block diagram illustrating the described method for determining protection levels using a Copula-assisted Bayesian framework is shown schematically and exemplary, wherein multiple environmental categories and multiple GNSS quality indices are given, and wherein Copula models 15, 16, and 17 are provided offline for each environmental category and each GNSS quality index and stored in memory (not shown). Method steps a), b), c), d), and e) are performed periodically during operation (i.e., online). The order of steps a), b), c), d), and e) shown with respect to blocks 110, arrows 120, blocks 130, 140, and 150 is merely exemplary. In block 110, the environment during operation is classified based on GNSS quality index 9, and the environmental category to which the environment belongs is determined, wherein online detection data 14, such as GNSS signal data or sensor data, is used. At arrow 120, the Copula model 15, 16, or 17 corresponding to the determined environmental category is read from memory. In block 130, at least one likelihood function 1, 2, 3, 4, 12 is extracted from the read Copula model 15, 16, or 17 based on the quality index value 13. In both blocks 140, a posterior distribution 6 for positional error is obtained, wherein, firstly in the first block 140, a prior distribution 5 is defined based on the training data and using a parameterized distribution; subsequently, in the second block 140, at least one likelihood function 1, 2, 3, 4, 12 is multiplied by the prior distribution 5 for positional error based on Bayes' theorem. In block 150, a protection level 7 is determined from the posterior distribution 6 based on the target integrity risk.

Claims

1. A method for determining a protection level (7) using a Copula-supported Bayesian framework while a vehicle is in motion with a GNSS-supported positioning system, wherein multiple environmental categories (10, 11) and multiple GNSS quality indicators (9) are predefined, and wherein a Copula model (15, 16, 17) is provided offline for each environmental category (10, 11) and each GNSS quality indicator (9) and stored in memory, the method comprising the steps of: a) Classify the environment during the driving based on the GNSS quality index (9) and determine the environmental category (10, 11) to which the environment belongs. b) Read from the memory the Copula model (15, 16, 17) corresponding to the desired environment category (10, 11) and the corresponding quality index (9). c) Extract at least one likelihood function (1, 2, 3, 4, 12) from the read Copula model (15, 16, 17) based on the quality index value (13). d) Obtain the posterior distribution (6) for the error by multiplying the at least one likelihood function (1, 2, 3, 4, 12) with the prior distribution (5) for the position error based on Bayes' theorem, and e) Determine the protection level (7) from the posterior distribution (6).

2. The method of claim 1, wherein the Copula model (15, 16, 17) is provided through the following steps: i) Provide training data through test measurements and / or simulated detection, ii) Classify the training data into the environment categories (10, 11). iii) Provide empirical copulas for each environmental category (10, 11) and each quality indicator (9), and iv) Match the parsed Copula model (15, 16, 17) to the empirical Copula.

3. The method according to claim 1 or 2, wherein in step c), the at least one likelihood function is extracted from the Copula model (15, 16, 17) by the following sub-steps: 1) Read the conditional distribution of position errors from the Copula model (15, 16, 17), and 2) Detect at least one likelihood function (1, 2, 3, 4) from the conditional distribution according to different GNSS quality indices (9).

4. The method of claim 3, wherein in sub-step 1), the conditional distribution is provided by the training data obtained through test measurements and / or simulation detection.

5. The method according to any one of the preceding claims, wherein in step d), the prior distribution (5) is based on the training data and is defined using a parameterized distribution.

6. The method according to any one of the preceding claims, wherein in step d), the posterior distribution (6) for the error is obtained for each period.

7. The method according to any one of the preceding claims, wherein the error relates to all parameters output by the positioning system supported by the GNSS.

8. The method according to any one of the preceding claims, wherein the error is a position error, a velocity error, and / or an orientation error.

9. A controller configured to perform the method according to any one of the preceding claims.

10. A computer program for performing the method according to any one of claims 1 to 8.

11. A machine-readable storage medium on which a computer program according to claim 10 is stored.

12. A positioning system for a vehicle, the positioning system being configured to perform the method according to any one of claims 1 to 8.

Images