Computer-implemented method for estimating a vehicle position

Abstract

A computer-implemented method (200) for estimating vehicle position includes the following steps: providing (210) a first position estimate based on position information from at least one first source, wherein the first position estimate is assigned a first deviation value and a first statistical deviation; providing (220) a second position estimate based on position information from at least one second source; determining (230) a third deviation value and a third statistical deviation, wherein the third deviation value and the third statistical deviation are assigned to a third position estimate obtained by combining the first position estimate and the second position estimate; evaluating (240) the first position estimate and the third position estimate by means of a quality criterion that takes into account the deviation value assigned to the respective position estimate and the statistical deviation assigned to the respective position estimate; if the third position estimate is superior to the first position estimate in terms of the quality criterion, then using (250a) the third position estimate as an estimated vehicle position or as a basis for combining with position information from one or more other sources; and if the first position estimate is superior to the third position estimate in terms of the quality criterion, then using (250b) the first position estimate as an estimated vehicle position or as a basis for combining with position information from one or more other sources.

Description

Technical Field

[0001] This invention relates to a computer-implemented method for estimating vehicle position. Furthermore, the invention relates to a processing apparatus and computer program for performing this method, as well as a computer-readable (memory) medium including instructions for performing this method. Background Technology

[0002] The vehicle may have driver assistance systems (FAS) configured to influence the longitudinal and / or lateral guidance of the vehicle. For example, lane assist may be configured to keep the vehicle between lane markings. The markings may be scanned and automatically identified, for example, by means of a camera.

[0003] Many FASs (Focusing Area Systems) need to know the exact location of a vehicle. The location can be determined in the longitudinal and / or lateral directions and represented relative to a predetermined reference point. For example, the absolute geographic location can be determined using map information from a predetermined geodetic reference system such as WGS84. The relative position of a vehicle can be specified in the lateral direction relative to identified lane markings, for example.

[0004] Vehicle position determination is often subject to a range of errors and inaccuracies. For example, sensors may provide information that is affected by noise and / or distorted, or they may sometimes fail completely. Different measurement conditions or complex processing heuristics can lead to determinations of varying accuracy or reliability.

[0005] To determine the vehicle's position as accurately and reliably as possible, multiple, preferably statistically independent, location information sources can be used for statistical estimation of the vehicle's position. Different location information sources can be, for example, based on camera-based detection of lane boundaries, LiDAR-based detection of objects or obstacles, DGPS data, and / or odometry-based prediction. The multiple location information sources from different sources can then be combined to form the estimated vehicle position. Here, common sensor data fusion algorithms known to those skilled in the art, such as Kalman filtering, can be employed.

[0006] In some traditional methods of sensor data fusion, multiple location estimates based on different sources are provided, each assigned a statistical deviation in the form of covariance. Here, under the assumption of Gaussian white noise, the covariance reflects the accuracy of each location estimate. From the combination of multiple location estimates, for example using Kalman filtering, a new location estimate can usually be obtained, whose covariance is smaller than that of the individual location estimates from different sources.

[0007] A challenge associated with known methods for estimating the positions of vehicles of the aforementioned types is that the individual position estimates from different sources often do not perfectly follow a statistically normal distribution around the vehicle's true position. The corresponding average of the statistical distribution followed by each position estimate (which may deviate from a normal distribution) is typically offset by a certain distance relative to the vehicle's true position. This offset or distance is generally referred to as the deviation. For this reason, the term "deviation value" will also be used below within the scope of this specification. In simple cases, the deviation value may, for example, be a constant offset in a particular direction. However, the deviation value may also vary depending on its magnitude and direction.

[0008] Traditional (especially Kalman filter-based) sensor data fusion algorithms often fail to achieve optimal vehicle position estimates if some or all of the position estimates derived from different sources are biased. For example, combining an unbiased first position estimate with a biased second position estimate results in a combined position estimate that is also biased, and often even worse than the first position estimate.

[0009] Graph-based localization algorithms (so-called graph SLAM algorithms) are known in robotics, which address the problem of biased position data by modeling deviations as part of the position estimation. However, in applications in autonomous driving, the available information is often insufficient to accurately estimate the deviations. Furthermore, this graph-based approach suffers from the drawback of requiring significant computation time, making it unsuitable for autonomous driving applications. Summary of the Invention

[0010] The objective of this invention is to provide improved vehicle positioning. This objective is achieved through the subject matter of the independent claims. The dependent claims reflect preferred embodiments.

[0011] This objective is achieved through the features of the independent claim. Advantageous embodiments are described in the dependent claims.

[0012] It should be noted that additional features of a claim subordinate to an independent claim, even without the features of the independent claim or only in combination with a subset of the features of the independent claim, can constitute a separate invention independent of all combinations of features of the independent claim, and can be the subject of the independent claim, divisional application, or subsequent application. This also applies to the technical theories described in the specification, which can form an invention with features independent of the independent claim.

[0013] The first aspect of the invention relates to a computer-implemented method for estimating vehicle position. This method can be performed, for example, by means of a (data) processing device (see below for a second aspect of the invention).

[0014] One step of the method provides a first location estimate based on one or more location information from at least one first source, wherein the first location estimate is assigned a first bias value and a first statistical deviation.

[0015] For example, a statistical distribution such as a normal distribution can be assigned to the first location estimate. In particular, the first location estimate can be given in the form of such a statistical distribution.

[0016] In this case, the statistical deviation can be given relative to the statistical distribution, for example in the form of covariance, standard deviation, three times the standard deviation, etc.

[0017] For example, the deviation value can be given as the maximum expected deviation (in the sense of worst-case deviation). If the deviation value is correct, the actual deviation that occurs (i.e., the deviation that actually occurs in individual cases) lies between 0 and the given deviation value in this case.

[0018] Alternatively, the deviation value can also be given in the form of (other) statistical deviations, such as the covariance, standard deviation, or three-times standard deviation of a normal distribution. For example, a corresponding real-world model in this case could calculate the deviation once for a series of measurements, and then calculate the current error for each measurement point. The total error would then be the sum of the deviation and the current error.

[0019] For example, the deviation value can be constant at least for a certain period of time, or it can be assumed to be constant.

[0020] Another step is to provide a second location estimate based on one or more location information from at least one second source. Here, a second bias value and a second statistical deviation can be assigned to the second location estimate.

[0021] The source may include, for example, one or more elements from the following list: (environmental) sensors, particularly in the form of lidar sensors, radar sensors, and / or cameras; odometers; and receiving modules for satellite navigation systems. Here, the aforementioned sensors or odometers and / or receiving modules are preferably arranged in or at the vehicle whose position is to be estimated. In the case of an odometer, the position estimate can be provided, for example, in the form of a prediction based on odometer data. The odometer data may, for example, represent the movement the vehicle has made relative to a previously occupied point, and is determined, for example, based on signals provided by speed sensors at the vehicle's wheels.

[0022] For example, one or more sources can provide information about lane boundaries (e.g., optically), azimuth points (also called landmarks) (e.g., optically or by means of lidar), or objects or obstacles (e.g., optically or by means of lidar) as the basis for position estimation. In other words, lane boundaries and / or azimuth points and / or objects in the vehicle environment can be identified, for example, in images provided by cameras and / or lidar sensors.

[0023] Here, in particular, position estimation can be performed relative to one or more lane boundaries, and this can, for example, lead to conclusions about which lane the vehicle is traveling in among multiple lanes. That is, in order to ensure sufficient safety, for example, in autonomous or highly automated driving, it is necessary to reliably determine the vehicle's position relative to surrounding lanes. Here, within the scope of the overall safety concept, the aim is to determine the vehicle's position within the lane with very high statistical reliability. In particular, the most stable possible positioning within the lane is also desired.

[0024] As mentioned above, a source can also be configured as a receiving module that provides location information based on signals received by a satellite-supported navigation system. Therefore, vehicle position can be estimated using data from a satellite navigation system (such as GPS or DGPS).

[0025] Typically, vehicle position can be estimated, for example, relative to or using map information; that is, the location of the vehicle on the map can be estimated. Map information can be provided, for example, by a map storage device. Map data can specifically include the location or orientation of objects or azimuth points and / or the routes that vehicles can travel.

[0026] Here, map data may be based, for example, at least in part, on data detected by sensors, which are recorded during one or more survey trips of the survey vehicle. The environmental sensors used here may include, for example, receivers of global navigation satellite systems (e.g., GPS or DGPS), one or more optical cameras, one or more radar sensors, and / or one or more lidar sensors.

[0027] Therefore, a digital map stored in a map memory can contain multiple layers, such as one layer based on data from a global navigation satellite system, another based on data from optical cameras, and yet another based on data from lidar. Different layers can contain features that can be identified using the corresponding sensor technologies.

[0028] The provision of the first or second position estimate may, for example, include receiving the first position estimate by a processing unit performing the method or by a (software) functional module implemented in the processing unit, wherein the position estimate can be used for further processing after receipt. For example, the first position estimate may be output to a processing device or the originally mentioned functional module by another processing unit (which may be, for example, part of a sensor module) or by another functional module of the processing unit performing the method for further processing.

[0029] The first location estimate can be a combined location estimate based on fused location information from multiple different first sources. Therefore, the first location estimate can be, for example, an intermediate or provisional result of a multi-level combination of location estimates, each based on location information from different sources. Alternatively, the first location estimate can be based on location information from a single first source.

[0030] The second location estimate can preferably be based on location information from a single second source. Alternatively, the second location estimate can also be a combined location estimate, as described above regarding the first location estimate.

[0031] Another step of the method is to determine a third deviation value and a third statistical deviation, wherein the third deviation value and the third statistical deviation are assigned to a third position estimate obtained by combining the first position estimate and the second position estimate.

[0032] For example, the third position estimate can be the result of a statistical combination, such as averaging or multiplying the two probability distributions corresponding to the first or second position estimates (followed by normalization). Preferably, the third position estimate is the result of combining the first and second position estimates using a Kalman filter. Therefore, the third deviation value and the third statistical deviation can also be obtained from a statistical combination of the first and second position estimates.

[0033] Depending on some implementation variations, it is not always necessary to actually determine the third position estimate to obtain the third deviation and the third statistical deviation. If the second position estimate provides a corresponding second deviation and a corresponding second statistical deviation, then it can be considered, for example, that the third deviation and the third statistical deviation can be determined based on the first deviation, the first statistical deviation, the second deviation, and the second statistical deviation.

[0034] Another step is to evaluate the first and third location estimates by means of a quality standard that takes into account the first or third deviation value assigned to the corresponding location estimate and the first or third statistical deviation assigned to the corresponding location estimate.

[0035] If the assessment shows that the third location estimate is superior to the first location estimate in terms of quality criteria, the third location estimate is used in further steps as an estimate of the vehicle's location or as a basis for combining it with location information from one or more other sources.

[0036] Conversely, if the evaluation shows that the first position estimate is superior to the third position estimate in terms of quality criteria, the first position estimate is used as the result of the vehicle position estimate or as the basis for combining it with position information from one or more other sources (i.e., sources different from the second source). In particular, it can be specified that, in this case, the second position estimate (and the third position estimate) will be discarded (i.e. ignored) for the purpose of estimating the vehicle position.

[0037] In the above scenario, the first or third position estimate is used as the result of the vehicle position estimate, which can be output to downstream functional modules of the vehicle for further processing, and can be used as the basis for driver assistance functions that control the vehicle, for example.

[0038] According to one implementation, the evaluation of the first position estimate and the third position estimate based on quality criteria includes: comparing the first sum of the first deviation value and the first statistical deviation with the third sum of the third deviation value and the third statistical deviation.

[0039] According to an improved embodiment, the first sum and the third sum can also be weighted sums, that is, weighted sums of each deviation and each statistical deviation using different factors.

[0040] For example, according to a variation of this implementation, if the third sum (in value) is less than the first sum, then the third position estimate can be determined to be superior to the first position estimate in terms of quality criteria. Alternatively or additionally, it can also be specified that if the first sum is less than the third sum, then the first position estimate is determined to be superior to the third position estimate in terms of quality criteria.

[0041] Furthermore, within the scope of this invention, the evaluation of the first and third position estimates based on quality standards may include comparing a third deviation value with a first statistical deviation. For example, if the third deviation value is greater than the first statistical deviation, it can be determined that the first position estimate is superior to the third position estimate in terms of quality standards.

[0042] The method according to the invention can be improved by determining the vehicle's attitude, i.e., estimating the vehicle's attitude in the manner proposed according to the invention. In addition to the vehicle's position, the attitude also includes the vehicle's orientation (e.g., orientation relative to one or more lane boundaries). Here, for example, the position can be specified in Cartesian coordinates along a right-handed coordinate system along two or three axes, and the orientation can be specified as rotation angles about these axes.

[0043] According to a second aspect of the invention, a processing apparatus is provided, wherein the processing apparatus is designed to perform the method according to the first aspect of the invention. Features or advantages of the method may be applied accordingly to the processing apparatus, and vice versa.

[0044] The processing device may be, for example, part of a vehicle control system that includes one or more processors (e.g., CPU and / or GPU), on which computational operations required to execute the method are performed.

[0045] For example, the vehicle whose location should be estimated may have processing equipment according to the second aspect.

[0046] Here, the vehicle preferably includes a drive motor and is a motor vehicle, particularly a road motor vehicle. For example, the motor vehicle can be controlled in the longitudinal direction by influencing the drive motor or braking device.

[0047] The vehicle may be equipped with one or more driver assistance systems that utilize the estimated vehicle position. The vehicle is preferably configured for at least partial automated driving, including highly automated driving or even autonomous driving.

[0048] For example, the vehicle's driving functions can be controlled based on the estimated position. Driving functions can specifically enable longitudinal and / or lateral control of the vehicle, such as in the form of speed assist or lane keeping assist. Here, the estimated position can be a safety-related parameter of the vehicle and can be measured, for example, in the longitudinal and / or lateral directions of the vehicle.

[0049] The third aspect relates to a computer program including instructions that, when the processing device runs the computer program, cause the processing device to perform the method according to the first aspect.

[0050] The fourth aspect of the invention relates to a computer-readable (storage) medium including instructions that, when run by a processing device, cause the processing device to perform the method according to the first aspect.

[0051] It should be understood that the processing apparatus mentioned above in conjunction with the third and fourth aspects of the present invention can be, in particular, a processing apparatus according to the second aspect of the present invention.

[0052] For example, the processing device may include one or more programmable microcomputers or microcontrollers, and the method may be implemented in the form of a computer program product having program code media. The computer program product may also be stored on a computer-readable data carrier.

[0053] According to another aspect, a system for estimating vehicle position is proposed. The system includes: a first source of position information and at least one second source of position information, wherein these sources may in particular be one or more environmental sensors of the vehicle; and a processing device according to a second aspect of the invention, which is data-technically connected to the sources, wherein the processing device is configured to receive position information from the sources and, based on the position information, execute a method according to a first aspect of the invention.

[0054] According to some implementations, location information from multiple different sources is combined to determine the estimated position of a vehicle. Here, a decision is made regarding whether to use specific additional location information provided by the sources for combination, based on the expected deviation and expected statistical deviation of the combined position estimate (compared to the result without considering other location information). The present invention is based on the idea that improved vehicle position estimation can be achieved efficiently, not only in terms of statistical deviation but also in terms of deviation value, through simultaneous optimization of the combined position estimate as defined by quality criteria. Attached Figure Description

[0055] The present invention will now be described with reference to the accompanying drawings and embodiments. Wherein:

[0056] Figure 1 An exemplary and schematic illustration shows a vehicle having a device for determining the vehicle's position; and

[0057] Figure 2 A schematic flowchart of a method for determining vehicle location is shown. Detailed Implementation

[0058] Figure 1 An example scenario is illustrated schematically, in which the position of a motor vehicle 105 traveling in the right lane of a two-lane road is estimated using information from multiple different sources 115. The invention will be explained exemplarily below with reference to this example scenario, wherein reference is also made to... Figure 2 It shows a schematic flowchart of the method 200 according to the present invention.

[0059] Vehicle 105 is equipped with a system 110 for estimating the position of vehicle 105. The device 110 includes multiple sources 115 for providing position information, wherein these sources specifically include a first sensor 1151 and a second sensor 1152. In this embodiment, the first sensor 1151 is a lidar sensor, and the second sensor 1152 is a camera.

[0060] Apart from Figure 1In addition to source 115 explicitly shown, one or more other sources may be additionally provided, which together form a complete set of information sources that can be used to determine the position of vehicle 105. Therefore, for example, a receiving device for receiving signals from a satellite-based navigation system (e.g., DGPS) and / or an odometer for providing odometer data may be provided as other sources among all sources 115. Furthermore, for example, one or more radar sensors may be provided as other sources.

[0061] In this embodiment, system 110 also includes a map memory 125 for providing map data, which can be used for vehicle position estimation. For example, vehicle position estimation can be performed by referring to map data from map memory 125.

[0062] A first sensor 1151 and a second sensor 1152, along with one or more optional sources from all sources 115, are configured to collect and provide information about the environment of the vehicle 105. In this embodiment, a camera 1152 and a lidar sensor 1151 are designed to detect or scan the environment of the vehicle 105. The (optical) camera 1152 can identify lane boundaries 3 drawn on the road surface. The lidar sensor 1151 can, in particular, identify objects or obstacles, such as guardrails 4.

[0063] In addition, system 110 includes processing device 120, which is connected in data technology to source 115 and map storage 125.

[0064] The processing device 125 is software-configured to receive information from the source 115 and estimate the position of the vehicle 105 based on this (referencing map information from the map memory 125 if necessary). Here, a corresponding software component, hereinafter referred to as a "matcher", may be assigned at the logical or data technology processing level, for example, to each of the entire sources 115.

[0065] Each matcher (potentially) assists in the localization of vehicle 105 by providing a location estimate based on location information from the corresponding source. Here, each matcher also specifies a corresponding statistical deviation (e.g., covariance or standard deviation) and a corresponding (expected) bias value using the location estimate. The bias value can be constant and, for example, given as a constant empirical value or a worst-case estimate for the corresponding source. Alternatively, the bias value can also vary over time, for example, based on the quality of the location data currently provided by the corresponding source.

[0066] Another software component, also known as a "combiner," is configured to combine the position estimates provided by the matchers into an estimated vehicle position. Here, the combiner determines whether to use the position estimates provided by the matchers for combination. This determination is made based on the expected deviation and expected statistical deviation of the combined position estimates (compared to the results of position estimates without considering the relevant matchers). Therefore, simultaneous optimization of the combined position estimates by quality criteria can achieve improved vehicle position estimation in an efficient manner, not only in terms of statistical deviation but also in terms of deviation.

[0067] Therefore, based on the above, the processing device 120 is specifically configured to perform... Figure 2 The following steps 210-250a, b are schematically illustrated in the diagram:

[0068] – Provide 210 a first location estimate based on one or more location information from at least one first source, wherein the first location estimate is assigned a first bias value and a first statistical deviation;

[0069] – Provides 220 second location estimates based on one or more location information from at least one second source;

[0070] – Determine the third deviation value and the third statistical deviation, wherein the third deviation value and the third statistical deviation are assigned to the third position estimate obtained by the combination of the first position estimate and the second position estimate;

[0071] – The first and third position estimates of 240 are evaluated by means of a quality standard that takes into account the first or third deviation value assigned to the corresponding position estimate and the first or third statistical deviation assigned to the corresponding position estimate.

[0072] – If the third location estimate is superior to the first location estimate in terms of quality criteria, then the 250a third location estimate is used as the vehicle location estimate or as the basis for combination with location information from one or more other sources; and

[0073] – If the first position estimate is superior to the third position estimate in terms of quality criteria, the 250b first position estimate is used as the estimated vehicle position or as the basis for combination with position information from one or more other sources.

[0074] For example, the evaluation 240 of the first and third position estimates may include comparing a first sum of a first deviation value and a first statistical deviation with a third sum of a third deviation value and a third statistical deviation. Here, for example, if the third sum (in value) is less than the first sum, the combiner may determine that the third position estimate is superior to the first position estimate in terms of quality criteria. Alternatively or additionally, if the first sum is less than the third sum, it may also be determined that the first position estimate is superior to the third position estimate in terms of quality criteria.

[0075] In other words, the quality criterion used by the combiner to determine whether the second position estimate should be used in the combined position estimate can be whether it causes or will cause the sum of the deviation value and statistical deviation (in value) to decrease or increase. If it increases, the combiner ignores the second position estimate for the purpose of estimating the vehicle position and outputs the first position estimate as the vehicle position estimate, or uses it as the basis for combining with position information from other sources. Figure 2 Step 250b in the process. In the case of reduction, the combiner outputs a third position estimate as the vehicle position estimate or uses it as the basis for combination with position information from other sources, i.e., the second position estimate is considered in the vehicle position estimate. Figure 2 Step 250a).

[0076] According to Figure 1 In an example scenario, a first position estimate can be provided, for instance, based on lane boundary 3 detected (optically) by camera 1152. Empirically, this camera-based lane recognition provides very accurate and effective information for the localization of vehicle 105, meaning the camera-based first position estimate can have a relatively low bias value. Here, different lane boundaries 3 can have more or less "reliable" lane boundaries. For example, based on the current lane boundary 3 (i.e., ... Figure 1 Position estimation of the dashed line 3 immediately to the left of vehicle 105 and the solid line 3 immediately to the right of vehicle 105 typically results in relatively low deviation values. Conversely, considering more distant lanes (e.g., lane boundaries very far to the left 3) produces deviations that generally remain essentially constant across multiple control cycles of the driver assistance system of vehicle 105.

[0077] Therefore, according to Figure 1 In the example, a camera-based matcher provides a first position estimate along with a first deviation value, corresponding to the expected deviation due to imperfect positioning of lane boundary 3. Furthermore, the camera-based matcher specifies an expected first statistical deviation of the first position estimate, for example, in the form of a first covariance. Here, the effect of, for example, the random motion of vehicle 105 can be approximately described by white noise, which may be the main contributor to the statistical deviation of the first position estimate.

[0078] For example, a second position estimate can be provided based on the identification of obstacles via lidar sensor 1151. Figure 1 In this example, guardrail 4 is marked as an exemplary obstacle adjacent to the right and left sides of the road. Therefore, the lidar-based matcher can output a second position estimate along with a second deviation value and a second statistical deviation. Here, the second deviation value can be relatively large. For example, a lateral offset of a few decimeters relative to the actual vehicle position can be expected, such as up to 30 cm.

[0079] The availability of the second position estimate based on the detection of guardrail 4 can change in a nearly unpredictable way along the driving route, for example, with the presence or absence of guardrail 4 (i.e., at the start or end). If the second position estimate is combined with the first position estimate to form a third position estimate, the third position estimate changes with... Figure 1 In the upper region of the illustrated road segment, the start of guardrail 4 may, for example, suddenly veer to the right, causing automatic lateral guidance to cause vehicle 105 to suddenly turn left. As soon as guardrail 4 ends again, a reverse movement to the right may occur, and so on. Therefore, the availability of changes in the second position estimate based on the guardrail will, in principle, lead to sudden changes in the estimate of the lateral vehicle position, which, due to the automatic lateral guidance of vehicle 105, translates into, for example, an undesirable, uncomfortable serpentine driving pattern for the vehicle's passengers.

[0080] One solution to this problem lies in combining the (total) position estimate without using position estimates with relatively large bias values ​​(e.g., lidar-based position estimates in this case). This is because combining a second position estimate with a large bias with a first position estimate with a smaller bias results in a third position estimate with a relatively large bias. This result is typically worse than using only the first position estimate.

[0081] However, combining a first position estimate with a large statistical deviation and a small bias value with a second position estimate with a small statistical deviation and a moderate bias value is likely to improve the vehicle position estimation. For this reason, in this embodiment, the combiner determines, based on the aforementioned quality criteria, whether the second position estimate (based on LiDAR) provided to it should be used in combination with the first position estimate (based on camera), or whether the first position estimate is superior to the expected result of the combination (i.e., the third position estimate) in terms of quality criteria.

[0082] In the first numerical example, the first location estimate is assigned (0.5m). 2 The variance. In this example, the first statistical deviation is defined as the square root of the variance, i.e., the (simple) standard deviation, and is therefore 0.5m. The first bias value is 0.1m.

[0083] The second position measurement has a range of (0.1m). 2 The variance is given. Therefore, the second statistical deviation (defined here as the simple standard deviation) is 0.1m. The second bias is 0.2m.

[0084] In this case, the combiner can determine, for example, that the combination of the first position estimate and the second position estimate will produce a third position estimate, which, compared to the first position estimate, has an increased third deviation value but also a significantly reduced third statistical deviation, thereby reducing the sum of the deviation value and the statistical deviation through combination.

[0085] Specifically, in this numerical example, the variance of the third position estimate can be calculated, for example, based on the known mathematical principles of the Kalman filter, as follows:

[0086]

[0087] Therefore, the third statistical deviation (defined as the simple standard deviation of the third location estimate) is approximately 0.098m.

[0088] The third deviation value can be calculated based on the mathematical principles of the Kalman filter, for example, as follows:

[0089]

[0090] The third statistical deviation (defined as the simple standard deviation of the third position estimate) and the third deviation value (“third”) are slightly less than 0.3m, which is less than the sum of the first standard deviation and the first deviation value (“first”), which is 0.6m. Therefore, the combiner determines to output the third position estimate as a result of the vehicle position estimate, or, if necessary, to use it as the basis for combining with one or more other (e.g., DGPS-based) position estimates to estimate the vehicle position (…). Figure 2 Step 250a).

[0091] Depending on the application, the evaluation 240 of the first and third position estimates based on quality standards may also include comparing the third deviation value with the first statistical deviation. For example, if the third deviation value is greater than the first statistical deviation, it can be determined that the first position estimate is superior to the third position estimate in terms of quality standards.

[0092] Therefore, in other words, the quality criterion used by the combiner to determine whether the second position estimate should be used in the combined position estimate can also consider whether this would result in a third deviation value greater than the statistical deviation of the first position estimate. If this is the case, the combiner ignores the second position estimate and outputs the first position estimate as the estimated vehicle position, or uses it as the basis for combining with position information from other sources to estimate the vehicle position. Figure 2(Step 250b in the original text). If this is not the case, the combiner can output a third position estimate as an estimate of the vehicle's position, or, if necessary, use it as the basis for combining with position information from other sources (…). Figure 2 Step 250a).

[0093] This will be illustrated with a second numerical example. Here, the variance of the first location estimate is (0.1m). 2 Therefore, the first statistical deviation given as the (simple) standard deviation is 0.1m. The corresponding expected first deviation value is 0.1m. The first position estimate can be, for example, purely based on the camera, or has already been based on a combination of camera-based and lidar-based position estimates.

[0094] The DGPS-based matcher additionally provides the combiner with a variance of (0.1m). 2 The second location estimate has a deviation of 0.3m from the expected value. In other words, the second statistical deviation is 0.1m, and the second bias is 0.3m.

[0095] In this second numerical example, the variance of the third location estimate is calculated as follows:

[0096]

[0097] Therefore, the third statistical deviation (defined as the simple standard deviation of the third location estimate) is approximately 0.07m.

[0098] The worst-case deviation of the third position estimate (i.e., the third deviation value) can be calculated as follows:

[0099]

[0100] The sum of the third statistical deviation (defined as the simple standard deviation of the third position estimate) and the third deviation value (“third”) is approximately 0.27m in this case, which is greater than the sum of the first standard deviation and the first deviation value (“first”), which is 0.2m. If a quality criterion is used (i.e., if the first sum is less than the third sum, then the first position estimate is superior to the third position estimate in terms of quality), then the combiner will therefore continue to operate with the first position estimate and will not use the second position estimate for combination.

[0101] Furthermore, the combiner may have discarded the second position estimate based on DGPS and continued to operate with the first position estimate for the following reasons: the expected deviation (0.2m) of the combined (third) position estimate will be greater than the statistical deviation (0.1m) of the first position estimate, so the vehicle 105 may be "pulled" in the wrong direction due to the combined position estimate.

[0102] For example, it is feasible to combine the (partial) quality standard (regarding the comparison of the third deviation value with the first statistical deviation) with one or more other (partial) quality standards (e.g., the aforementioned quality standard regarding the sum of the deviation value and the statistical deviation) to form a more comprehensive quality standard, such that the second position estimate is used for the combination only if, for example, the sum of the deviation value and the statistical deviation is thereby reduced on the one hand, and the (third) deviation value obtained on the other hand is not greater than the first statistical deviation.

[0103] It is important to note, particularly in the second numerical example above, that if the corresponding (first / second / third) position estimates are based, for example, on three or four times the standard deviation as the corresponding (first / second / third) statistical deviations instead of simple standard deviations, and then the sum of the first statistical deviation and the first deviation value is compared with the sum of the third statistical deviation and the third deviation value on this basis, the combiner may produce different results. That is, in terms of this quality criterion, the first and third position estimates will be approximately the same when based on three times the standard deviation. When based on four times the standard deviation, the third position estimate will be better than the first position estimate. This clearly demonstrates that, in the second numerical example, the first position estimate is preferred when primarily concerned with driving comfort. If rare exceptions are also considered within the scope of the safety concept, the combined (third) position estimate should be used.

[0104] Furthermore, in a variant implementation, it's possible to specifically prevent abrupt shifts in bias between two values. For example, a vehicle might travel to a certain point in time based on DGPS and then suddenly obtain a position estimate based on LiDAR. Here, for example, both the LiDAR-based and DGPS-based position estimates might contribute a bias of 0.3m. If the combiner determines to use LiDAR information in addition to DGPS-based information, this will reduce the total error in the expected value of the combined position estimate. However, it's possible that the DGPS-based position estimate has a bias of exactly 30cm to the left at the relevant time point, while the LiDAR-based position estimate has a bias of 30cm to the right. In this case, the combiner can discard the use of LiDAR information to avoid a sudden lateral shift in the position estimate. However, the new DGPS position information will still be fused, as this will not cause the bias to suddenly jump from left to right. Additionally, in sensor fusion, new position information based on lane boundary detection with a bias value of, for example, only 0.1m will be considered, as this can reduce the bias in the combined position estimate.

[0105] Therefore, according to the above implementation variation, the combiner can store information about which matcher has contributed to the deviation, and in particular, can determine which other matchers will be used for the combined position estimation later. Each time a position estimate with a lower deviation (than the current position estimate) is combined, the total deviation decreases. Each time a position estimate with a higher deviation is added, the total deviation increases, and the combiner remembers the matchers whose position estimates have caused the deviation to increase. If the sum of the deviation value and statistical deviation of the combined position estimates will decrease as a result, the matcher that previously caused the deviation increase is allowed to continue increasing the deviation. Other matchers are only allowed to increase the deviation if the sum of the deviation value and statistical deviation of the combined position estimates will decrease significantly as a result. This strategy of the combiner prevents the combined position estimate from oscillating back and forth and leading to a "drunk" driving style (serpentine curve).

Claims

1. A computer-implemented method (200) for estimating vehicle position, comprising the following steps: – Provide (210) a first location estimate based on one or more location information from at least one first source, wherein the first location estimate is assigned a first deviation value and a first statistical deviation; – Provide (220) a second location estimate based on one or more location information from at least one second source; – Determine (230) a third deviation value and a third statistical deviation, wherein the third deviation value and the third statistical deviation are assigned to a third position estimate obtained by a combination of the first position estimate and the second position estimate; – The first location estimate and the third location estimate are evaluated by means of a quality standard, which takes into account a first or third deviation value assigned to the corresponding location estimate and a first or third statistical deviation assigned to the corresponding location estimate; – If the third location estimate is superior to the first location estimate in terms of the quality criteria, then the third location estimate (250a) is used as the estimated vehicle location or as the basis for combining it with location information from one or more other sources; and – If the first position estimate is superior to the third position estimate in terms of the quality criteria, then the first position estimate (250b) is used as the estimated position of the vehicle or as the basis for combination with position information from one or more other sources.

2. The method (200) of claim 1, wherein the source comprises at least one element from the following list: - sensor; – Odometer; – The receiving module of a satellite navigation system.

3. The method (200) according to claim 1, wherein the evaluation (240) comprises: Compare the first sum of the first deviation value and the first statistical deviation with the third sum of the third deviation value and the third statistical deviation.

4. The method (200) of claim 3, wherein if the third sum is less than the first sum, then the third position estimate is determined to be superior to the first position estimate in terms of the quality criterion.

5. The method (200) according to claim 3 or 4, wherein if the first sum is less than the third sum, then the first position estimate is determined to be superior to the third position estimate in terms of the quality criterion.

6. The method (200) according to any one of claims 1 to 4, wherein the evaluation (240) comprises: Compare the third deviation value with the first statistical deviation.

7. The method (200) of claim 6, wherein if the third deviation value is greater than the first statistical deviation, the first position estimate is determined to be superior to the third position estimate in terms of the quality standard.

8. The method (200) according to claim 2, wherein the sensor is in the form of a radar sensor and / or a camera.

9. The method (200) according to claim 2, wherein the sensor is in the form of a lidar sensor.

10. A processing apparatus (120) wherein the processing apparatus (120) is designed to perform the method (200) according to any one of claims 1 to 9.

11. A computer program product comprising instructions which, when run by a processing apparatus (120) according to claim 10, cause the processing apparatus to perform the method (200) according to any one of claims 1 to 9.

12. A computer-readable storage medium comprising instructions which, when operated by a processing apparatus (120) according to claim 10, cause the processing apparatus to perform the method (200) according to any one of claims 1 to 9.

Images