Method for determining the position in a vehicle by receiving GNSS signals under calibration of a GNSS receiving antenna based on a plurality of antenna error influence maps

Abstract

Method for determining the position in a vehicle by receiving GNSS signals under calibration of a GNSS receiving antenna based on a plurality of different antenna error influence maps stored in the vehicle (1, 2), wherein the following steps are performed: a) Determine at least one signal parameter of at least one GNSS signal; b) Selecting at least two antenna error influence maps (1, 2) from the plurality of antenna error influence maps (1, 2); c) Determine at least one correction value for correcting a position determination from each of the at least two selected antenna error influence maps (1, 2) taking into account the signal parameter determined in step a); and d) Performing position determination using a Kalman filter, taking into account the correction values ​​determined in step c).

Description

State of the art

[0001] The present invention relates to a method for determining a position in a vehicle by receiving GNSS signals while calibrating a GNSS receiving antenna based on a plurality of antenna error influence maps stored in the vehicle. Furthermore, a control unit, a computer program, and a machine-readable storage medium are specified. The invention can be used in particular in GNSS-supported localization systems for autonomous or semi-autonomous driving.

[0002] A global navigation satellite system (GNSS) consists of numerous GNSS satellites orbiting the Earth and emitting electromagnetic GNSS signals. By receiving GNSS signals with an antenna, a vehicle can be positioned and navigated.

[0003] When receiving GNSS signals with the antenna, it is important to consider that group delay variations or antenna errors, depending on the azimuth and elevation angle to GNSS satellites, can affect the measurement accuracy (e.g., the accuracy of the time-of-flight measurement) of GNSS signals. In other words, the measurement accuracy of GNSS signals depends on the relative position between the antenna and the corresponding GNSS satellites.

[0004] In addition to group delay variations and the azimuth and elevation angles, the measurement accuracy of GNSS signals can also be affected by the vehicle body surface because different paint types, paint colors, contamination and / or weather conditions (e.g. wet paint in the rain, snow-covered paint in the snow) can have different reflection and attenuation properties, so that GNSS signals may be reflected by the vehicle body surface (especially in the case of vehicle bodies with silver metallic paint) and / or attenuated (especially in the case of snow-covered vehicle roofs) when received.

[0005] In the context of safety-critical automated driving functions, positioning accuracy must be in the centimeter range. Therefore, there is a desire to consider and compensate for the above-mentioned impairments when calculating navigation solutions (e.g., position, speed, time).

[0006] Based on this, the object of the present invention is to alleviate or at least partially solve the problems described with reference to the prior art. In particular, a method for determining the position in a vehicle by receiving GNSS signals and calibrating GNSS receiving antennas while taking into account vehicle characteristics, in particular vehicle body surfaces. This is particularly important for autonomous driving, because autonomous driving places particularly high demands on the accuracy of navigation solutions. Disclosure of the invention

[0007] A method for determining the position in a vehicle by receiving GNSS signals and calibrating a GNSS receiving antenna based on a plurality of different antenna error influence maps stored in the vehicle contributes to this, whereby the following steps are carried out: a) determining at least one signal parameter of at least one GNSS signal; b) selecting at least two antenna error influence maps from the plurality of antenna error influence maps; c) determining at least one correction value for correcting a position determination from each of the at least two selected antenna error influence maps, taking into account the signal parameter determined in step a); and d) Carrying out the position determination using a Kalman filter, taking into account the correction values ​​determined in step c).

[0008] As mentioned at the beginning, the measurement accuracy of GNSS signals depends on many factors. Key factors include the azimuth and elevation angles of the corresponding GNSS satellites (i.e., the relative positions between an antenna and the GNSS satellites), as well as atmospheric influences, reflections, and deflections of GNSS signals, etc. It has been shown that the measurement accuracy of GNSS signals also depends on vehicle characteristics (e.g., different paint finishes), and in particular on the vehicle's orientation in space (i.e., the azimuth and elevation angles). Depending on how a vehicle is oriented relative to the existing GNSS satellite constellation, different signal deviations occur, leading to incorrect or shifted positioning / position determinations.With the proposed method, the measurement of GNSS signals and the determination of positions using the measured GNSS signals can be carried out taking into account both azimuth and elevation angles (relative to the vehicle orientation or antenna orientation) and vehicle characteristics, thus improving positioning accuracy and the overall performance of GNSS-based navigation. Vehicle characteristics are taken into account as a function of signal parameters, with signal parameters in this context being, in particular, azimuth and elevation angles.

[0009] To prepare the proposed method, a plurality of different antenna error influence maps can be created using measurement technology under different vehicle characteristics, for example, by determining measurement errors of GNSS signals with differently painted vehicles on a rotating plate. By rotating the rotating plate, the measurement errors at different azimuth and elevation angles can be determined, and the determined measurement errors at the corresponding azimuth and elevation angles can be plotted on an antenna error influence map. Under different vehicle characteristics, the measurement errors determined at the same azimuth and elevation angle can also have different magnitudes. These differences in measurement error magnitudes are possibly caused (only) by the different vehicle characteristics, since the measurement errors at the same azimuth and elevation angle are determined under different vehicle characteristics.Thus, the antenna influence maps used here can be so different from each other that the measurement errors entered on the different antenna influence maps can have different sizes, even though these measurement errors are associated with the same azimuth and elevation angle.

[0010] To ensure accurate measurement of GNSS signals during a vehicle's journey, even taking vehicle characteristics into account, it is essential to use at least two different antenna influence maps, which may differ from each other. Preferably, the at least two different antenna influence maps each represent the maximum possible errors or a particularly extreme filter in a specific situation (for specific azimuth and elevation angles). Thus, a real measured value of a GNSS signal can be corrected with the measurement errors derived from the at least two different antenna error influence maps for the same azimuth and elevation angle of the GNSS satellite transmitting the GNSS signal.

[0011] The measured values ​​of GNSS signals include, for example, the travel time of GNSS signals from the corresponding GNSS satellites to the GNSS receiving antenna. Based on the measured travel times and the speed of light, the distances between the GNSS satellites and the GNSS receiving antennas can also be calculated. Accordingly, the antenna error influence maps proposed here can include the travel time and / or distance measurement errors associated with the azimuth and elevation angles.

[0012] The vehicle characteristics relevant to the measurement accuracy of GNSS signals include, in particular, the vehicle body surface, which, with different paint types, different paint colors, and / or different levels of contamination, can reflect and / or attenuate the GNSS signals to be received differently. This means that the GNSS signals may be received with an extended or shortened propagation time, and this can lead to propagation time and distance measurement errors. Furthermore, the vehicle body surface can exhibit different reflection and / or attenuation properties under different weather conditions due to the moisture / snow adhering to it. For example, in rainy weather, a vehicle body surface with wet paint can influence the GNSS signals more strongly due to reflection than a vehicle body surface with dry paint.In situations where snow has fallen on the vehicle, the vehicle body surface with the snow-covered paint and the snow-covered vehicle roof may affect the GNNS signals more than the vehicle body surface and the vehicle roof without snow due to attenuation.

[0013] In principle, the travel time and / or distance measurement errors caused by azimuth and elevation angles and vehicle characteristics are so-called "systematic errors" (i.e., non-random errors) and can occur regularly under certain circumstances according to a deterministic pattern. It is therefore possible to experimentally detect such travel time and / or distance measurement errors in advance and, while the vehicle is in motion, to correct the currently measured travel times and / or distances with the previously detected travel time and / or distance measurement errors.

[0014] It can be provided that the previously recorded travel time and / or distance measurement errors are displayed in the form of an antenna error influence map. Such an antenna error influence map can be stored in a control unit. This allows the measured travel times and / or the calculated distances or the determined positions / position determinations to be corrected using the travel time and / or distance measurement errors read from the antenna error influence map while the vehicle is moving. The control unit can be a control unit of a GNSS localization system with which the vehicle is equipped.

[0015] According to step a), at least one signal parameter is determined from at least one GNSS signal. The signal parameter is, for example, the azimuth and elevation angle of the GNSS satellite transmitting this GNSS signal. For example, from the ephemeris contained in the received GNSS signal, it is possible to determine which GNSS satellite the received GNSS signal is coming from and how / where this GNSS satellite is currently positioned in the sky. Furthermore, using the vehicle's current heading determined by the sensors (e.g., IMU sensors) and / or the camera, it is possible to determine the azimuth and elevation angle from the vehicle's GNSS receiving antenna to the GNSS satellite transmitting the received GNSS signal.

[0016] According to step b), at least two antenna error influence maps are selected from the plurality of antenna error influence maps. It is possible to select only two extreme antenna error influence maps that exhibit the greatest differences from one another. For example, one extreme antenna error influence map can be selected so that it was previously created using a silver metallic vehicle, while another extreme antenna error influence map can be selected so that it was previously created using a matte black vehicle. It has been found that precisely this combination of antenna error influence maps (silver vehicle and matte black vehicle) captures particularly extreme error influences. Thus, the currently measured GNSS signals can be corrected using only these two extreme antenna error influence maps. This is particularly advantageous because it saves computing resources (e.g.Computation time and memory requirements can be reduced to a minimum. It is also conceivable to select more than two antenna error influence maps, allowing the measured GNSS signals to be corrected using more than two selected antenna error influence maps. This is particularly advantageous for further improving the accuracy of the overall navigation solution.

[0017] According to step c), at least one correction value for correcting a position determination is determined from each of the at least two selected antenna error influence maps, taking into account the signal parameter determined in step a). At this point, the signal parameter serves as a bridge connecting a real measured value with a previously recorded measurement error as a correction value. This correction value can, for example, be the propagation time error of a GNSS signal or the distance measurement error calculated from the propagation time error using the speed of light between the corresponding GNSS satellite and the GNSS receiving antenna, while the signal parameter determined in step a) is the azimuth and elevation angle of the corresponding GNSS satellite.For this purpose, the propagation time error and / or distance measurement error associated with the azimuth and elevation angles determined in step a) can be read from the respective antenna influence maps selected in step b). The read propagation time error and / or distance measurement error are used to determine the position in step d).

[0018] Preferably, a correction value is determined for each GNSS signal considered. Correction values ​​determined from an antenna error influence map are also referred to here as a "set" of correction values. Preferably, for each GNSS signal used for position determination, correction values ​​are determined using each of the at least two antenna error influence maps depending on the respective signal parameter (in particular depending on the respective azimuth and elevation angles relative to the vehicle). For a set of GNSS signals used for position determination, preferably several sets of correction values ​​(based on the individual antenna error influence maps) exist after step c).

[0019] According to step d), the position determination is carried out using a Kalman filter, comprising merging data that are corrected and / or estimated using the correction values ​​determined in step c).

[0020] The conventional Kalman filter essentially performs two tasks, or rather, two different determinations. The Kalman filter makes predictions of expected measured values ​​in a next (future) time interval. The Kalman filter fuses predicted measured values ​​with the actual measured values ​​to determine the position to be output to the vehicle.

[0021] Two different approaches are proposed here for using the correction values ​​determined in step c) for position determination.

[0022] According to a first possibility, in step d) correction values ​​from different antenna error influence maps are used together in the Kalman filter for position determination.

[0023] According to a second possibility, in step d) first positions are first determined separately for each antenna error influence map selected in step b) using a Kalman filter and then first positions are merged to form a position to be output.

[0024] In the first approach, the various antenna error influence maps are considered together in the Kalman filter for position determination. In principle, correction values ​​(each derived from the most extreme antenna error influence maps) are adjusted, and then the corrected measured values ​​are merged with the predicted measured values ​​to produce the output position.

[0025] In the second option, for example, a corresponding first position is first estimated using each of the correction values ​​determined in step c), and then the estimated first positions are merged to form the output position. The fusion of the first positions to form the output position can also occur outside the Kalman filter, for example, by using the output position as an average position.

[0026] With the described method, the measurement errors caused by the vehicle's body surface can be compensated purely at the software level without any hardware effort, without having to replace the existing GNSS receiving antennas (e.g., vehicle antennas) with, for example, high-precision geodetic antennas. Compensating for such measurement errors is particularly advantageous for autonomous driving, where positioning accuracy in the centimeter range is required.

[0027] It is further preferred if the first positions are weighted using probability values, and a position to be output is averaged from the weighted first positions.

[0028] The initial positions and the output position have the same timestamp. The initial positions are only estimated internally. The output position is the final determined position and is sent from the GNSS control unit to the vehicle. The initial positions are only the estimated possible positions. An output position can be fused, averaged, and / or weighted from several initial positions, with each initial position being estimated using a correction value determined from a selected antenna error influence map.

[0029] The estimated first positions are preferably weighted using probability values, and the position to be output is averaged from the weighted first positions.

[0030] Alternatively, if the estimated first positions do not differ significantly from each other, the position to be output can be directly averaged from the first positions (i.e., without weighting).

[0031] It is preferred if the initial positions are weighted using likelihood values, and a position to be output is averaged from the weighted initial positions. The following substeps can be performed: - Estimating first positions, - Determination of residuals using real measured values, - Determining likelihood values ​​using residuals and measurement covariances, and - Determining a position to be output which is averaged from the first positions, whereby the first positions are weighted using the likelihood values.

[0032] It is preferred if the weighting of the first positions is filtered using a PT1 filter with a time constant. The time constant can be at least 30 seconds.

[0033] It is preferred if the time constant is optimized using machine learning.

[0034] It is preferred if the at least one signal parameter comprises the reception direction of a GNSS signal.

[0035] It is preferred if the vehicle characteristics include at least the vehicle body surface.

[0036] It is preferred if the antenna error influence maps were created using the following steps: i) providing a vehicle with a GNSS receiving antenna and a specific vehicle body surface on a rotatable plate, ii) Measuring GNSS signals with the GNSS receiving antenna while rotating the turntable, iii) Determining measurement errors based on the measured GNSS signals iv) creating an antenna error influence map showing the measurement errors determined in step iii) as a function of different reception directions, and v) Repeating steps i) to iv) to create a plurality of antenna error influence maps for vehicles with different vehicle body surfaces.

[0037] It is also preferred if, in step v), the antenna error influence maps were created taking into account different weather conditions, wherein the weather conditions are simulated with a snowy paint, a wet paint or a dry paint.

[0038] Since the reception of GNSS signals can be influenced by the relative positions between the GNSS receiving antenna and the respective GNSS satellites, according to step i), a vehicle with a GNSS receiving antenna on a rotatable plate is provided.

[0039] As the plate rotates, the vehicle rotates with it. Theodolite measurements can be used to determine the different orientations of the vehicle and the different true positions and reception directions of the GNSS receiving antenna as the vehicle rotates.

[0040] According to step ii), GNSS signals are measured using the GNSS receiving antenna. In particular, the propagation time of the GNSS signals between the GNSS receiving antenna and the respective GNSS satellites is measured.

[0041] According to step iii), measurement errors between the measured values ​​and the true values ​​are determined. These measurement errors include, for example, the travel-time measurement errors or the distance measurement errors calculated using the travel-time measurement errors. The true values ​​can be determined using known methods, such as those based on the ephemerides of the respective GNSS satellites.

[0042] According to step iv), the measurement errors recorded during rotation in different reception directions are plotted on an antenna error influence map according to the corresponding reception directions. On the antenna error influence map, the different reception directions can be represented by the corresponding azimuth and elevation angles.

[0043] As mentioned above, vehicle body surfaces with different coating types and / or colors can have different effects on the reception of GNSS signals with the vehicle antenna. Furthermore, GNSS signal reception can be affected by the coating being covered with little or more snow, moisture, and / or dirt.

[0044] Therefore, according to step v), different antenna influence maps can be created with different vehicles, each belonging to the same vehicle model but with different body surfaces. The body surfaces can differ from each other in at least one of the following aspects: coating type, color, snow, wetness, and dirt.

[0045] Preferably, the body surfaces are selected so that the reflection and / or attenuation properties of the body surfaces differ as much as possible from each other in order to capture as many extreme measurement errors of GNSS signals as possible through the measurements, which can then be further processed in the form of extreme antenna influence maps.

[0046] The recorded extreme measurement errors can represent the limits that the actual measurement errors during travel should not exceed. Therefore, the actual measurement errors of GNSS signals during travel usually lie between the recorded extreme measurement errors, so the actual measurement errors or the actual measured values ​​can be compensated or corrected using mathematical methods such as interpolation, fusion, averaging, and / or weighting.

[0047] For example, in the simplest case, an antenna error influence map can be created with a silver, metallic, wet vehicle, which may contain extreme measurement errors due to extreme reflection (compared to other colors). In contrast, another antenna error influence map can be created with a black, matte, dry vehicle, which may contain extreme measurement errors due to extreme attenuation (compared to other colors). While driving, the two extreme antenna error influence maps can be used to correct the current measured values.

[0048] It is preferred if a control device is configured to carry out the described method.

[0049] It is also preferred if a computer program is used to carry out a method described here. In other words, this particularly relates to a computer program (product) comprising instructions that, when executed by a computer, cause the computer to carry out a method described here.

[0050] It is also preferred if a machine-readable storage medium is used on which the computer program proposed here is stored. The machine-readable storage medium is usually a computer-readable data carrier.

[0051] The invention and the technical environment are described below with reference to Fig. 1. It should be noted that the Fig. 1 is of a schematic nature. It should also be noted that the Fig. 1, the features shown do not necessarily have to be used in the combination shown; rather, partial features can be extracted and combined with other explanations of the description. Anything different should only apply if this is explicitly stated here as a mandatory combination of features.

[0052] Fig. Figure 1 schematically shows two exemplary antenna error influence maps 1, 2, each created with a differently painted vehicle.

[0053] In Fig. 1 shows that the relative positions between the vehicle or the vehicle antenna and the respective GNSS satellites are represented by a circle with angles between 0 degrees and 360 degrees. The center of the circle, for example, indicates that a GNSS satellite is directly above the vehicle or the vehicle antenna. The intersection of the diameter and the circumference at 90 degrees indicates that a GNSS satellite is directly above the horizontal on the right side of the vehicle. The intersection of the diameter and the circumference at 0 degrees indicates that a GNSS satellite is directly above the horizontal in the direction of travel from the vehicle.

[0054] In Fig. 1 further shows that the two antenna error influence maps 1, 2 can exhibit different distance measurement errors 3 at the same locations. For example, the first antenna error influence map 1 shows that the distance measurement errors 3 in the lower left area between 180 and 270 degrees are much larger (almost over 1 meter) than the distance measurement errors 3 in the same area of ​​the second antenna error measurement map 2. These differences are possibly caused by the different paint finishes. During travel, the two antenna influence maps 1, 2 can be used to correct real measurement values ​​by reading the affected distance measurement errors from the two antenna error influence maps 1, 2 according to the relative positions between the vehicle antenna and the GNSS satellites. The relative positions can be described in various known forms, such as with the Fig.1, for example, on the vehicle coordinate system or the azimuth and elevation angles on an astronomical coordinate system or the reception directions. By using these two antenna error influence maps 1, 2, the measurement of GNSS signals and the positioning with the measured GNSS signals can be carried out taking into account both relative positions and vehicle characteristics, especially paintwork, so that the positioning accuracy and the overall performance of GNSS-based navigation can be improved.

Claims

[1] Method for determining the position in a vehicle by receiving GNSS signals while calibrating a GNSS receiving antenna based on a plurality of different antenna error influence maps (1, 2) stored in the vehicle, wherein the following steps are carried out: a) determining at least one signal parameter of at least one GNSS signal; b) selecting at least two antenna error influence maps (1, 2) from the plurality of antenna error influence maps (1, 2); c) determining at least one correction value for correcting a position determination from each of the at least two selected antenna error influence maps (1, 2) taking into account the signal parameter determined in step a); and d) Carrying out the position determination using a Kalman filter, taking into account the correction values ​​determined in step c). [2] Method according to claim 1, wherein in step d) correction values ​​from different antenna error influence maps (1, 2) are used together in the Kalman filter for position determination. [3] Method according to claim 1, wherein in step d) first positions are first determined separately from each other using a Kalman filter for each antenna error influence map (1, 2) selected in step b) and then first positions are merged to form a position to be output. [4] Method according to claim 3, wherein the first positions are weighted using probability values, and the position to be output is averaged from the weighted first positions. [5] Method according to claim 4, wherein the first positions are weighted using likelihood values, and the position to be output is averaged from the weighted first positions. [6] Method according to claim 3 or 4, wherein the weighting of the first positions is filtered by means of a PT1 filter with a time constant. [7] The method of claim 6, wherein the time constant is optimized using machine learning. [8] Method according to one of the preceding claims, wherein the at least one signal parameter determined in step a) and taken into account in step c) comprises the reception direction of a GNSS signal. [9] A method according to any one of the preceding claims, wherein the vehicle characteristics comprise at least the vehicle body surface. [10] Method according to one of the preceding claims, wherein the antenna error influence maps (1, 2) were created with the following steps: i) providing a vehicle with a GNSS receiving antenna and a specific vehicle body surface on a rotatable plate, ii) Measuring GNSS signals with the GNSS receiving antenna while rotating the turntable, iii) Determining measurement errors (3) based on the measured GNSS signals, iv) creating an antenna error influence map (1, 2) which represents the measurement errors (3) determined in step iii) as a function of different reception directions, and v) repeating steps i) to iv) to create a plurality of antenna error influence maps (1, 2) for vehicles with different vehicle body surfaces. [11] Method according to claim 10, wherein in step v) the antenna error influence maps (1, 2) were created taking into account different weather conditions, the weather conditions being simulated with a snowy paint, a wet paint or a dry paint. [12] Method according to one of the preceding claims, wherein in step b) two antenna error influence maps (1, 2) are selected which have the greatest differences from each other. [13] Control device which is arranged to carry out a method according to one of the preceding claims 1 to 9. [14] Computer program for carrying out a method according to one of the preceding claims 1 to 9. [15] A machine-readable storage medium on which the computer program according to claim 14 is stored.

Images