Method for determining vehicle position based on landmarks

Abstract

The invention relates to a method (1) for determining a vehicle position based on landmarks, wherein in a method part (VA) a plurality of position hypotheses (PH) for the vehicle position are formed based on forming associations (A) between sensor landmark objects (SLM) measured by a sensor and map landmark objects (KLM) stored in a digital map, the most probable vehicle position is determined as a positioning result (LE) based on probabilistic filtering of the position hypotheses (PH), and a guaranteed position range (PL) is determined in which a predetermined upper error limit is not exceeded. According to the invention, the method part (VA) is executed multiple times with different probabilistic filtering designs, wherein the positioning result (LE1, LE2, LE3) with the smallest guaranteed position range (PL1, PL2, PL3) is selected as the vehicle position (eFP) if the guaranteed position ranges (PL1, PL2, PL3) completely overlap in pairs.

Description

Technical Field

[0001] The invention relates to a method for determining a vehicle position based on landmarks according to the features of the preamble of claim 1 . Background Art

[0002] As described in WO 2019 / 243031 A1, a method for determining a vehicle position is disclosed by the prior art. A plurality of position hypotheses are determined by comparing landmark objects detected by a vehicle sensor system with landmark objects stored in a map, in particular a map portion. A position hypothesis with an integrity value is determined by analyzing all position hypotheses and filtering out all erroneous information using a probability analysis. By filtering according to a predetermined limit value, a position hypothesis with sufficient position accuracy for determining the vehicle position in a predetermined manner is determined. Summary of the invention

[0003] The object of the present invention is to specify a method for determining the position of a vehicle based on landmarks which is improved compared to the prior art.

[0004] This object is achieved according to the invention by a method for determining a vehicle position based on landmarks having the features of claim 1 .

[0005] Advantageous embodiments of the invention are the subject matter of the dependent claims.

[0006] In a method for determining a vehicle's position based on landmarks, in one method part, multiple position hypotheses of the vehicle's position are formed based on associations between sensor landmark objects measured by sensors in use, in particular by means of an environmental detection sensor system of the vehicle, and map landmark objects stored in a digital map, and also in this method part, based on probabilistic filtering of the position hypotheses, the most likely vehicle position is determined as a positioning result, and a guaranteed position range (also called a protection limit) that does not exceed a predetermined upper error limit is determined. The protection limit, i.e., the guaranteed position range, is therefore a position range for which a certain maximum error rate is guaranteed, i.e., within this position range it is guaranteed that a certain, i.e., predetermined upper error limit is not exceeded. The certain, in particular guaranteed maximum error rate is the upper error limit guaranteed for the guaranteed position range (protection limit), for example, 10 per hour. -3 An error means that the error rate of the positioning result within the guaranteed position range (protection limit) is lower than the guaranteed maximum error rate.

[0007] According to the invention, the method part, in which the most probable vehicle position is determined as a positioning result based on probabilistic filtering of position hypotheses and a guaranteed position range which does not exceed a predetermined upper error limit, at least a part of this method part is performed multiple times, for example at least twice, for example at least three times, for example exactly three times, with different probabilistic filtering designs, in particular with different designs of probabilistic feature association filters. As a result, a plurality of positioning results with different accuracies and different guaranteed position ranges are thereby obtained. If the guaranteed position ranges completely overlap in pairs, i.e., in particular, if for each possible pairing of the guaranteed position ranges determined in different embodiments of this method part, they completely overlap, the positioning result with the smallest guaranteed position range is selected as the vehicle position, i.e., as the vehicle position to be searched for and to be determined by means of the method.

[0008] The different designs of probability filtering relate in particular to feature association filters, in particular only to feature association filters. In the feature association filter, position hypotheses are filtered and impossible position hypotheses are discarded. Therefore, through the different designs of the feature association filter, the probability of the position hypothesis is evaluated differently in different embodiments of the method part or at least the above-mentioned part of the method part, so that the position hypothesis is evaluated as possible or impossible accordingly, that is, each position hypothesis can be evaluated as possible by one of the feature association filter designs, and evaluated as impossible and discarded by another design. This leads to different numbers of position hypotheses and / or different position hypotheses in different embodiments of the above-mentioned method part or at least the above-mentioned part of the method part, in particular, on the one hand, especially when the feature association filter is strictly designed, it leads to a small number of but very safe / reliable position hypotheses, and on the other hand, especially when the feature association filter design is not too strict, it leads to a large number of but less safe position hypotheses.

[0009] The method is particularly performed in a vehicle, particularly by a vehicle, particularly at least one vehicle unit provided for this purpose and designed and configured accordingly. The digital map or at least one map portion of a digital map respectively used for the method is particularly stored in the vehicle, particularly in a unit provided for performing the method or in another unit of the vehicle connected to the unit, particularly for data transmission.

[0010] The method is particularly advantageous for vehicles with partially automatic, especially highly automatic or autonomous driving systems. The requirement of such driving systems is usually that the position determination indicates the integrity of the current output vehicle position. This is represented by the calculation and output of a guaranteed position range, i.e., a protection limit, which indicates that the maximum deviation from the actual position can be guaranteed just at the residual error rate defined for this purpose, but this is generally only applicable from a certain threshold. This threshold is also called an alert limit. Therefore, the alert limit is the threshold required for guaranteeing the position range (protection limit) in a specific application such as unmanned driving, especially partially automatic, especially highly automatic or autonomous driving. The different safety goals of the system design of such a driving system usually have different requirements for integrity. This requires the output of multiple guaranteed position ranges (protection limits) with different thresholds and acceptable residual error rates.

[0011] In the solution according to the invention, a plurality of so-called shadow levels of positioning filters, especially those designed as probability filters, such as Kalman filters, are set up, and upstream thereof are provided with error association filters, especially feature association filters, of different degrees of strict design. The vehicle positions are all estimated independently of each other, wherein the stricter shadow levels have a more complete state but obtain lower accuracy. Therefore, the guaranteed position range (protection limit) becomes safer again, i.e., fewer false positive statements occur, and the guaranteed position range (protection limit) is less than the threshold value (alarm limit), although in fact it is not the case, but the guaranteed position range (protection limit) also becomes larger, and thus the condition that the guaranteed position range (protection limit) is less than the threshold value (alarm limit) is rarely met. Because more information arrives, the less strict shadow level provides a more accurate position. Through the solution of the invention, this is optimally utilized, i.e., as the final vehicle position, i.e., as the vehicle position to be searched and determined by the method, the most accurate vehicle positions are output respectively, but still within all guaranteed position ranges (protection limits).

[0012] Through the solution of the invention and the determination of a plurality of guaranteed position ranges (protection limits) achieved thereby, a plurality of guaranteed position ranges (protection limits) can also be output, which optimally serve the various safety objectives, such as road accuracy for activating highly automated driving functions, for which a relatively large threshold value (large warning limit) with a low acceptable error rate is required, and in parallel therewith lane accuracy for automatic lane change, for which a very small threshold value (very small warning limit) with a higher acceptable error rate is required. The availability of the entire system, i.e., in particular the availability of the partially automated, in particular highly automated or autonomous driving system of the vehicle, is thereby increased. In addition, the safety requirements no longer have an influence on the accuracy of the vehicle position, since the solution of the invention advantageously also provides a shadow layer operating with the greatest possible feature density. However, this shadow layer is also monitored by the safer layer, thereby intercepting larger errors.

[0013] In one possible embodiment of the method, an error message is sent if there is at least one pair of guaranteed position ranges that only partially overlap, i.e. if the guaranteed position ranges do not all completely overlap in pairs. This avoids, in particular, functional failures of a partially automatic, in particular highly automatic or autonomous driving system of the vehicle and avoids possible dangers, in particular caused by incorrect or too inaccurate positioning. In this case, for example, the partially automatic, in particular highly automatic or autonomous driving system of the vehicle or at least one corresponding driving function of the driving system is deactivated and, for example, the vehicle driver is requested to take over the vehicle control and / or, for example, park the vehicle, in particular in a safe position.

[0014] In one possible embodiment, the selected vehicle position is assigned to all larger guaranteed position ranges. Thus, these guaranteed position ranges together with the selected vehicle position can be used for a plurality of different safety objectives as described above and can be optimized to serve these safety objectives respectively.

[0015] Advantageously, the method part, at least the above-mentioned multiple implementations of the above-mentioned parts of the method part, are parallel in time, particularly advantageously start at the same time. Advantageously, the method part, at least the above-mentioned multiple implementations of at least the above-mentioned parts of the method part, are carried out in parallel branches of the method, in particular branches that run in parallel in time. As a result, the individual positioning results and the guaranteed position range are advantageously provided at the same time for further processing. In particular, compared with the implementation in succession, a much faster processing time is obtained thereby. This is extremely important for continuously positioning moving vehicles as accurately as possible. Therefore, by means of fast processing, higher vehicle speeds can also be achieved, for example, in semi-automatic, in particular highly automatic or autonomous driving operations.

[0016] In one possible implementation of the method, the branches of the method each include a feature association filter, wherein the feature association filters of the branches are designed in different ways, and wherein the safe position hypothesis is determined with the help of the feature association filters of each branch. Therefore, different filter designs involve the feature association filters of different branches. This also leads to multiple positioning results with different precisions and different guaranteed position ranges.

[0017] In a possible embodiment of the method, in the corresponding branch, the safe position assumption of the branch and in particular the vehicle's own motion are processed by means of a probabilistic filter, in particular a Kalman filter, thereby determining respective positioning results and respective guaranteed position ranges that do not exceed a predetermined upper error limit. Since different safe position assumptions determined by means of feature correlation filters of different designs and thus different are adopted here, different positioning results and different guaranteed position ranges that do not exceed a predetermined upper error limit are thereby determined.

[0018] In one possible embodiment of the method, the sensor landmark objects that have been passed by the vehicle are also used to form an association between the sensor landmark objects detected by the sensor and the map landmark objects stored in the digital map. The sensor landmark objects detected by the sensor are advantageously used to form an association between the sensor landmark objects detected by the sensor and the map landmark objects stored in the digital map over a predetermined period and / or over a predetermined route length. The association is therefore advantageously carried out via a recorded history of the sensor landmark objects, also referred to as a signature. This advantageously results in a higher accuracy of the vehicle position. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The embodiments of the present invention are explained in more detail below with reference to the accompanying drawings.

[0020] in:

[0021] Figure 1 Schematically illustrates an embodiment of a method for determining a vehicle position based on landmarks, and

[0022] Figure 2 A further improved specific embodiment of the method for determining the position of a vehicle based on landmarks is schematically shown. DETAILED DESCRIPTION

[0023] In all the figures, parts that correspond to one another are provided with the same reference numerals.

[0024] Figure 1 and Figure 2 Two specific embodiments of a method for determining the position of a vehicle 1 based on landmarks are shown, wherein: Figure 1An embodiment known from the prior art and described in particular in WO 2019 / 243031 A1 of the applicant is shown. The content of WO 2019 / 243031 A1, in particular the content of its drawings and the associated description of the drawings, i.e. in particular the description of an embodiment of a method for determining the position of a vehicle 1 based on landmarks Figure 1-1 2 and the description of the drawings on pages 4-14 are incorporated by reference.

[0025] As further described below, Figure 2 A significantly improved specific embodiment of a method for determining the position of a vehicle 1 based on landmarks is shown.

[0026] In both embodiments of the method, a plurality of position hypotheses PH for the vehicle position are formed based on the formation of associations A between sensor landmark objects SLM detected by sensors, in particular by means of an environment detection sensor system of the vehicle 1, and map landmark objects KLM stored in a digital map. Based on the probabilistic filtering of the position hypotheses PH, the most probable vehicle position is determined as a positioning result LE, and a guaranteed position range PL is determined, which does not exceed a predetermined upper error limit, also referred to as a protection limit.

[0027] This is Figure 1 . As described, an association A is formed between a sensor landmark object SLM detected by a sensor and a map landmark object KLM stored in a digital map, so as to determine a plurality of position hypotheses PH for the vehicle position. The filtering of these position hypotheses PH, in particular the probabilistic filtering or at least a part of the filtering of these position hypotheses PH, in particular the probabilistic filtering, is performed with the aid of a feature association filter MAF, which is in particular designed as a probabilistic filter. In particular, in the feature association filter MAF, in particular the erroneous association A between the sensor landmark object SLM detected by the sensor and the map landmark object KLM stored in the digital map is filtered out, so as to comply with a predetermined upper error limit. In this way, a safe position hypothesis SPH is determined. These safe position hypotheses SPH and the vehicle 1 own motion EM determined in particular with the aid of a corresponding sensor system, in particular an odometer sensor system, are processed in a probabilistic filter PF, for example a Kalman filter. The result is a probability distribution WV of the vehicle position in the digital map. Then, a guaranteed position range PL can be determined from this.

[0028] Therefore, the guaranteed position range PL of the vehicle 1 is determined with a guaranteed error rate. The guaranteed position range PL is a position range in which the positioning error rate is guaranteed not to exceed a predetermined maximum allowable error rate, i.e., a predetermined upper error limit. As described above, the guaranteed position range PL is also called a protection limit.

[0029] Furthermore, a probability distribution WV of the vehicle position in the digital map is determined. The determination of the probability distribution WV and the guaranteed position range PL is based, as described above, on the formation of an association A between a sensor landmark object SLM detected by the sensor and a map landmark object KLM stored in the digital map, on the establishment of a plurality of position hypotheses PH based on the association A as potential vehicle positions, and on the probabilistic filtering of the position hypotheses PH, which at this point are hidden, i.e. filtered out and thus discarded, and thus only the remaining safe position hypotheses SPH are further processed. The probabilistic filtering is based on a probabilistic assessment of whether the individual position hypotheses PH are based more on a correct association A or more on an erroneous association, and on the filtering out of position hypotheses PH that are based more on erroneous associations.

[0030] The probability evaluation is in turn based on a probability analysis with the aid of a feature association filter MAF (also called a feature association filter, FeatureAssociation Filter). In this case, the position hypothesis PH is advantageously evaluated using a distribution model of misleading information and a distribution model of correct information, respectively. Here, the degree of agreement between the position hypothesis PH and the distribution model of correct information and the distribution model of misleading information is calculated. The evaluation is based in particular on the idea that the position hypothesis PH distributed in a spatially adjacent cumulative manner is based on a correct association A between a sensor landmark object SLM detected by a sensor and a map landmark object KLM stored in a digital map, i.e., based on correct information, and the position hypothesis PH that is not applicable to this is based on an incorrect association A between a sensor landmark object SLM detected by a sensor and a map landmark object KLM stored in a digital map, i.e., based on misleading information.

[0031] The result is a numerical value, also called integrity value, which indicates whether the respective position hypothesis PH is based more on correct information or more on misleading information. The numerical value thus indicates how likely it is that the respective position hypothesis PH is based on a correct association A rather than an erroneous association. By filtering out the position hypotheses PH according to this numerical value, the position hypotheses PH that are based more on erroneous associations than on the correct association A are concealed, i.e., filtered out and thus discarded.

[0032] according to Figure 1 The approach of the first embodiment of the method shown is problematic, that is, in filtering, on the one hand, not all erroneous hypotheses, that is, not all position hypotheses PH based on erroneous associations are concealed, that is, not all erroneous hypotheses are filtered out, and on the other hand, a portion of the correct position hypotheses PH, that is, position hypotheses PH based on the correct association A are concealed, that is, filtered out. The reason for this is that, based on probability analysis, erroneous hypotheses and correct position hypotheses PH can only be distinguished according to probability. Therefore, different degrees of concealment, that is, filtering, of erroneous hypotheses and correct position hypotheses PH are obtained for different filtering designs.

[0033] Therefore, strict filtering leads to a strong suppression of wrong hypotheses, that is, the false positive rate (False Positive Rate) is reduced, but strict filtering also leads to the suppression of the correct position hypothesis PH. This suppression of the correct position hypothesis PH results in low positioning accuracy, which in turn results in a larger guaranteed position range PL (protection limit). The strict design of the filter and especially the feature association filter MAF therefore leads to a low guaranteed error rate and low positioning accuracy, especially because the guaranteed position range PL (protection limit) is wide.

[0034] In contrast, a less stringent design of the filtering, in particular the feature association filter MAF, leads to a higher guaranteed error rate (i.e. more errors have to be considered and therefore more errors may occur in the localization) and a higher localization accuracy, especially because the guaranteed position range PL (protection limit) is narrower.

[0035] Therefore, the size of the guaranteed position range PL (protection limit) and the guaranteed maximum error rate are determined by the design of the filter, especially the design of the feature association filter MAF. Figure 2 This fact is illustrated by the different probability distributions WV1, WV2, WV3 and the corresponding different guaranteed position ranges PL1, PL2, PL3 (protection limits). Figure 2 The method sequence of the second embodiment of the improved method is shown, from which these different probability distributions WV1, WV2, WV3 and guaranteed position ranges PL1, PL2, PL3 (protection limits) are derived.

[0036] If the maximum error rate is guaranteed to be high, it can be guaranteed that the vehicle 1 is within a narrow range. The frame around the vehicle 1, i.e. the frame around its vehicle position, shows the respective guaranteed position ranges PL1, PL2, PL3 (protection limits). The narrower this frame, i.e. the narrower the guaranteed position ranges PL1, PL2, PL3, the more accurate the positioning results LE1, LE2, LE3. The curve shows the respective probability distributions WV1, WV2, WV3 of the position hypothesis PH, in particular the safety position hypothesis SPH1, SPH2, SPH3, in particular the probability histogram. The maximum value of the probability distributions WV1, WV2, WV3 is the most likely vehicle position. The maximum value or the entire probability distribution WV1, WV2, WV3 represents the positioning results LE1, LE2, LE3.

[0037] The upper frame, namely the first guaranteed position range PL1, is narrow, but the guaranteed maximum error rate therefor is high. In contrast, the lower frame, namely the third guaranteed position range PL3, is wide, but the guaranteed maximum error rate therefor is low.

[0038] That is, the more accurately the vehicle position eFP is determined and the narrower the frame indicating the guaranteed position range PL1 , PL2 , PL3 is, the less reliable the determined vehicle position eFP is, since the guaranteed maximum error rate is higher.

[0039] This relationship is used to improve the Figure 2 The second embodiment of the method shown by way of example is as follows: Figure 1 A first known embodiment of the method shown, in particular Figure 1 The method part denoted by the reference symbol VT is executed multiple times in particular in parallel with different filter designs, in each case Figure 2 In the example shown, it is executed three times.

[0040] It can therefore be provided that the association A between the sensor landmark objects SLM detected by the sensor and the map landmark objects KLM stored in the digital map is formed only once in order to thereby determine a plurality of position hypotheses PH for the position of the vehicle. These determined position hypotheses PH are then subjected to probabilistic filtering in the manner described above, but in a second improved embodiment in a plurality of, in particular parallel and in particular temporally parallel branches Z1, Z2, Z3 of the method with different filter designs.

[0041] In the respective branches Z1, Z2, Z3 of the method, the position hypotheses PH are in particular probabilistically filtered or at least part of the probabilistic filtering is performed with the aid of the respective feature correlation filter MAF, which is in particular designed as a probabilistic filter, wherein in particular the feature correlation filter MAF is designed in different ways. As already mentioned above, in the respective feature correlation filter MAF, in particular, false correlations A between sensor landmark objects SLM detected by the sensor and map landmark objects KLM stored in the digital map are filtered out. In this way, different safe position hypotheses SPH1, SPH2, SPH3 are determined in the different branches Z1, Z2, Z3 of the method. These safe position hypotheses SPH1, SPH2, SPH3 and advantageously the vehicle 1 own motion EM determined in particular by the corresponding sensor system, in particular the odometer sensor system, are processed in a probabilistic filter PF, such as a Kalman filter. The probabilistic filter PF is advantageously designed in the same way in all branches Z1, Z2, Z3 of the method. The result is therefore a probability distribution WV1, WV2, WV3 in the digital map of the vehicle position for each branch Z1, Z2, Z3 of the method, from which the respective guaranteed position range PL1, PL2, PL3 can then be determined.

[0042] As a result, multiple Figure 2In the example shown, there are three positioning results LE1, LE2, LE3 with different accuracy and different guaranteed position ranges PL1, PL2, PL3, that is, different guaranteed maximum error rates. From the set of these positioning results LE1, LE2, LE3, the most accurate positioning result LE1 is selected, here the positioning result LE1 of the upper branch Z1, whose guaranteed position range PL1 is within all the wider guaranteed position ranges PL2, PL3 and is output as the searched and determined vehicle position eFP.

[0043] Therefore, combined with Figure 1 The explained first embodiment of the known method is carried out a plurality of times in a second improved embodiment as method part VA of the method, in particular in a plurality of parallel branches Z1, Z2, Z3, in which Figure 2 In the example shown, it is executed three times, and thus is executed three times in three parallel branches Z1, Z2, and Z3. The result is therefore a plurality of positioning results LE1, LE2, and LE3 in the form of probability distributions WV1, WV2, and WV3, in the example shown, and a plurality of guaranteed position areas PL1, PL2, and PL3, in the example shown, respectively belonging to the positioning results and presented in the form of respective boxes around the respective vehicle positions, in the form of respective boxes.

[0044] exist Figure 2 In the example shown, the most accurate positioning result LE1 is obtained in the upper branch Z1 of the method, because the frame representing the guaranteed position range PL1 is narrower than in the other branches Z2, Z3, and the probability distribution WV1 shows the smallest dispersion. However, the guaranteed maximum error rate is higher for this than in the other two branches Z2, Z3, i.e., the positioning result LE1 is the least reliable.

[0045] The least accurate positioning result LE3 is obtained in the lowermost branch Z3, because the box representing the guaranteed position range PL3 is wider than in the other two branches Z1 and Z2, and the probability distribution WV3 shows the maximum dispersion. However, the guaranteed maximum error rate is lower than in the other two branches Z1 and Z2, that is, the most reliable positioning result LE3 is obtained in the lower branch Z3.

[0046] Now check, for each possible pair of frames (i.e. each pair of frames of the guaranteed position ranges PL1, PL2, PL3), i.e. for all possible pairings of the guaranteed position ranges PL1, PL2, PL3 of the branches Z1, Z2, Z3, whether one of the frames of each pair is completely within the other frame of the corresponding pair, i.e. whether one of the guaranteed position ranges PL1, PL2, PL3 of each pair is completely within the other guaranteed position ranges PL1, PL2, PL3 of the corresponding pair. It is therefore checked whether these frames, i.e. the guaranteed position ranges PL1, PL2, PL3, respectively, completely overlap in pairs.

[0047] If they completely overlap in pairs, the positioning result LE1 belonging to the narrowest frame, that is, the narrowest guaranteed position range PL1, is selected as the final positioning result LE1 and output. In the example shown here, the final positioning result is the guaranteed position range PL1 of the upper branch Z1, and therefore the positioning result LE1 thereof. All frames, that is, all guaranteed position ranges PL1, PL2, PL3 are also output.

[0048] Therefore, the most accurate positioning result LE1 is output, which is the positioning result of the upper branch Z1 and the guaranteed position ranges PL1, PL2, PL3 suitable for different maximum error rates, especially the position ranges PL1, PL2, PL3 of all branches Z1, Z2, Z3.

[0049] As a result, the most accurate positioning result LE1 is selected from these branches, in the example shown, the three branches Z1, Z2, and Z3, in the example shown, the curve of the upper branch Z1, that is, the probability distribution WV1, and the less accurate positioning results LE2 and LE3 of the other two branches Z2 and Z3 are replaced by the selected most accurate positioning result LE1. Through this replacement, for the guaranteed position range PL3 with the most reliable positioning, that is, the lowest maximum error rate, in the example shown here, the frame of the lower branch Z3, the accurate positioning result LE1, which is the upper branch Z1, is obtained.

[0050] Figure 1 The first embodiment of the method shown for providing a positioning result LE either with high accuracy and low integrity, i.e. with low reliability and a low maximum error rate, or with low accuracy and high integrity is modified by the described solution, i.e. by an improved second embodiment of the method, such that high accuracy and high integrity of the positioning results LE1, LE2, LE3 are obtained.

[0051] If one of the frames only partially overlaps another of the frames, ie if at least one pair of guaranteed position ranges PL1 , PL2 , PL3 only partially overlaps, an error is advantageously assumed and an error message is then advantageously output.

[0052] Therefore, in the method for determining the position of the vehicle 1 based on landmarks, Figure 2 In the second embodiment shown, in particular according to the present invention, in a method part VA, multiple position hypotheses PH for the vehicle position are formed based on the association A between sensor landmark objects SLM detected by a sensor, in particular an environmental detection sensor system of the vehicle 1, and map landmark objects KLM stored in a digital map, and if necessary, in the method part VA, the most likely vehicle position is determined as a positioning result LE and a guaranteed position range PL that does not exceed a predetermined error upper limit based on probabilistic filtering of the position hypotheses PH.

[0053] The method part VA is executed multiple times, for example at least twice, for example at least three times, for example exactly three times, in particular in parallel and in particular in time parallel, in particular in parallel branches Z1, Z2, Z3 of the method with different probability filter designs. That is, in this improved second embodiment of the method, in multiple, in particular in parallel and in particular in time parallel branches Z1, Z2, Z3, for example in at least two or at least three or exactly three branches Z1, Z2, Z3, a plurality of position hypotheses PH for the vehicle position are respectively formed based on the formation of associations A between sensor landmark objects SLM detected by the sensor in use, in particular by means of the environment detection sensor system of the vehicle 1, and map landmark objects KLM stored in the digital map, and based on the probability filtering of the position hypotheses PH, the most likely vehicle position is determined as a positioning result LE1, LE2, LE3, and a guaranteed position range PL1, PL2, PL3 is determined which does not exceed a predetermined upper error limit, wherein the branches Z1, Z2, Z3 have different probability filter designs from one another.

[0054] As a result, a plurality of positioning results LE1, LE2, LE3 with different accuracies and different guaranteed position ranges PL1, PL2, PL3 result from this. If the guaranteed position ranges PL1, PL2, PL3 completely overlap in pairs, i.e., in particular, if any possible pairing of the guaranteed position ranges PL1, PL2, PL3 determined in different embodiments of the method part VA completely overlaps, the positioning result LE1 with the smallest guaranteed position range PL1, here the positioning result of the upper branch Z1, is selected as the vehicle position eFP, i.e., as the vehicle position eFP that is searched for and to be determined by means of the method.

[0055] The selected vehicle position eFP is advantageously assigned to all larger guaranteed position ranges PL2 , PL3 .

[0056] If at least one pair of guaranteed position ranges PL1 , PL2 , PL3 overlaps only partially, ie if the guaranteed position ranges PL1 , PL2 , PL3 do not all completely overlap in pairs, an error message is advantageously output.

[0057] Now combine Figure 2 The method is explained in more detail again, in particular with respect to the exemplary embodiment of the method shown here with three branches Z1 , Z2 , Z3 .

[0058] Based on the formation of an association A between a sensor landmark object SLM detected by means of a sensor, in particular by means of an environment detection sensor system of the vehicle 1, and a map landmark object KLM stored in a digital map, a plurality of position hypotheses PH for the vehicle position are formed. This can be carried out, for example, in parallel in a plurality of branches Z1, Z2, Z3 of the method, in this case in three branches Z1, Z2, Z3, where this then leads to the same position hypothesis PH. Therefore, this can alternatively also be carried out together, as in the example shown here, and the position hypothesis PH thus determined is then used in a plurality of, in particular parallel branches Z1, Z2, Z3 of the method, in the example shown here in three branches Z1, Z2, Z3. That is, either the entire method part VA of the method is executed respectively with different probability filter designs up to the end of the plurality of branches Z1, Z2, Z3, or at least a part VT of the method part VA with the method steps shown in these branches Z1, Z2, Z3 is executed respectively with different probability filter designs after the position hypothesis PH has been determined up to the end of the branches Z1, Z2, Z3.

[0059] Then, in each branch Z1, Z2, Z3, the determined position hypotheses PH are subjected to probability filtering in the above-described manner with different probability filter designs. In the individual branches Z1, Z2, Z3 of the method, in particular probability filtering or at least a part of in particular probability filtering of these position hypotheses PH is performed with the aid of respective feature association filters MAF1, MAF2, MAF3, which are in particular designed as probability filters, wherein in particular these feature association filters MAF1, MAF2, MAF3 are designed in different ways. However, these feature association filters can work together here, i.e., they do not have to be completely independent of each other, for example, but can share one or more initial processing steps.

[0060] In the individual feature association filters MAF1 , MAF2 , MAF3 , as described above, in particular erroneous associations A between sensor landmark objects SLM detected by the sensor and map landmark objects KLM stored in the digital map are filtered out. In this way, different safety position hypotheses SPH1, SPH2, SPH3 are determined in different branches Z1, Z2, Z3 of the method. In the example shown, multiple safety position hypotheses SPH1 are determined in the upper branch Z1, especially more than in the other two branches Z2, Z3, but less secure than in the other two branches Z2, Z3; few safety position hypotheses SPH3 are determined in the lower branch Z3, especially fewer than in the other two branches Z1, Z2, but highly secure, especially more secure than in the other two branches Z1, Z2; a certain number of safety position hypotheses SPH2 are determined in the middle branch Z2, which are less than in the upper branch Z1 but more than in the lower branch Z2, wherein the safety position hypothesis SPH2 of the middle branch Z2 is more secure than the safety position hypothesis SPH1 of the upper branch Z1, but less secure than the safety position hypothesis SPH3 of the lower branch Z3.

[0061] These safe position hypotheses SPH1, SPH2, SPH3 and advantageously the own movement EM of the vehicle 1 determined in particular by means of a corresponding sensor system, in particular an odometer sensor system, are processed in a probability filter PF, for example a Kalman filter. The probability filter PF is advantageously designed in the same way in all branches Z1, Z2, Z3 of the method. The result is therefore a unique positioning result LE1, LE2, LE3 for each branch Z1, Z2, Z3 of the method, i.e. a probability distribution WV1, WV2, WV3 of the vehicle position in the digital map and a respective guaranteed position range PL1, PL2, PL3.

[0062] Due to the different designs of probability filtering, especially feature association filters MAF1, MAF2, MAF3, and the different safe position hypotheses SPH1, SPH2, SPH3 obtained therefrom, there are different positioning results LE1, LE2, LE3, that is, different probability distributions WV1, WV2, WV3 of the vehicle position in the digital map and thus different accuracies regarding the vehicle position, as well as different guaranteed position ranges PL1, PL2, PL3. In the example shown, in the upper branch Z1, the probabilities of the vehicle positions are very close to each other, thus providing high accuracy regarding the vehicle position, but low integrity, in the middle branch Z2, the probabilities of the vehicle positions are more dispersed, so that the accuracy regarding the vehicle position here is lower, but the integrity is slightly higher, and in the lower branch Z3, the probabilities of the vehicle positions are more dispersed, so that the accuracy regarding the vehicle position here is the lowest, but the integrity is the highest. The guaranteed position ranges PL1, PL2, PL3 are narrow in the upper branch Z1, wider in the middle branch Z2, and wider still in the lower branch Z3.

[0063] These positioning results LE1, LE2, LE3 are now provided to the reference position selector RPS, which selects the positioning result LE1 with the smallest guaranteed position range PL1, in the example shown, the positioning result LE1 of the upper branch Z1 as the vehicle position eFP, i.e., as the vehicle position eFP searched for and to be determined by the method, when the guaranteed position ranges PL1, PL2, PL3 completely overlap in pairs as in the example shown. This is because this is the vehicle position best determined within all guaranteed position ranges PL1, PL2, PL3 of the branches Z1, Z2, Z3, and the selected vehicle position eFP also corresponds to all larger guaranteed position ranges PL2, PL3, i.e., also from the guaranteed position ranges PL2, PL3 from other branches Z2, Z3.

Claims

1. A method for determining the position of a vehicle (1) based on landmarks, wherein: In a Methods section (VA): - forming a plurality of position hypotheses (PH) for the position of the vehicle based on forming associations (A) between sensor landmark objects (SLM) measured by the sensor and map landmark objects (KLM) stored in the digital map, - based on the probability filtering of the position hypothesis (PH), determining the most likely vehicle position as the positioning result (LE), and determining a guaranteed position range (PL) within which a predetermined upper error limit is not exceeded, Characterized in that the method part (VA) is executed multiple times with different probability filtering designs, wherein, if the guaranteed position ranges (PL1, PL2, PL3) completely overlap in pairs, the positioning result (LE1, LE2, LE3) with the smallest guaranteed position range (PL1, PL2, PL3) is selected as the vehicle position (eFP).

2. The method according to claim 1, characterized in that Different probability filter designs involve feature association filters (MAF1, MAF2, MAF3) in which the position hypotheses (PH) are filtered and unlikely position hypotheses (PH) are discarded.

3. The method according to claim 1 or 2, characterized in that: If there is at least one pair of guaranteed position ranges (PL1, PL2, PL3) that only partially overlap, an error report is sent.

4. The method according to claim 1 or 2, characterized in that: The selected vehicle position (eFP) is assigned to all larger guaranteed position ranges (PL2, PL3).

5. The method according to claim 1 or 2, characterized in that: This method part (VA) is performed at least twice.

6. The method according to claim 5, characterized in that This method part (VA) is performed at least three times.

7. The method according to claim 6, characterized in that This method part (VA) is executed exactly three times.

8. The method according to claim 1 or 2, characterized in that: This method part (VA) is executed multiple times in parallel in time.

9. The method according to claim 8, characterized in that The method part (VA) is executed multiple times in parallel in a plurality of parallel branches (Z1, Z2, Z3) of the method.

10. The method according to claim 9, characterized in that Each branch (Z1, Z2, Z3) respectively includes a feature association filter (MAF1, MAF2, MAF3), wherein the feature association filters (MAF1, MAF2, MAF3) of the branches (Z1, Z2, Z3) are designed in different ways, and wherein the safe position hypothesis (SPH1, SPH2, SPH3) is determined with the help of the feature association filters (MAF1, MAF2, MAF3) of the respective branches (Z1, Z2, Z3).

11. The method according to claim 10, characterized in that In each branch (Z1, Z2, Z3), the safe position hypothesis (SPH1, SPH2, SPH3) of the branch (Z1, Z2, Z3) and the own movement (EM) of the vehicle (1) are processed with the help of a probability filter (PH), thereby determining the respective positioning results (LE1, LE2, LE3) and the respective guaranteed position ranges (PL1, PL2, PL3) which do not exceed a predetermined error upper limit.

12. The method according to claim 11, characterized in that The probability filter (PH) is a Kalman filter.

13. The method according to claim 1 or 2, characterized in that: The sensor landmark objects (SLM) that have been passed by the vehicle (1) are also used to form an association (A) between the sensor landmark objects (SLM) detected by the sensor and the map landmark objects (KLM) stored in the digital map.

14. The method according to claim 1 or 2, characterized in that: The sensor landmark objects (SLM) detected by the sensor are used to form an association (A) between the sensor landmark objects (SLM) detected by the sensor and map landmark objects (KLM) stored in the digital map within a predetermined time interval and / or within a predetermined route length.

Images