Method for evaluating a digital map and evaluation system

Abstract

The invention relates to a method for evaluating a digital map (7), wherein the following steps are performed: - detecting an environment (3) of a vehicle (1) and creating a relative map of the environment based on the detected information and the relative positions of detected landmarks (A, B, G, S, V) in the environment (3); - providing a global map of the environment (3) with global positions of the landmarks (A, B, G, S, V); - generating a plurality of update hypotheses for the digital map (7) based on position deviations of the landmarks (A, B, G, S, V) relative to the global positions of at least one reference object (A, B, G, S, V, 1); - determining and evaluating an error statistic of at least some of the position deviations; and - assessing whether at least one update hypothesis of the digital map (7), in particular the entire digital map (7), is suitable for self-localization of the vehicle (1) depending on the evaluation of the error statistic.

Description

TECHNICAL FIELD

[0001] One aspect of the present application relates to a method for evaluating a digital map. Another aspect of the present application relates to an electronic evaluation system for evaluating a digital map. BACKGROUND

[0002] For the self-localization of autonomous vehicles, different designs are known. In some cases, structures and patterns in the vehicle's environment are identified by vehicle sensors and compared with corresponding entries in a digital map. Such structures are also referred to as landmarks. The premise of these designs is that the vehicle is equipped with a current and validated landmark map.

[0003] In order to keep these digital maps current, measurement data from a special measuring vehicle or from vehicles in a vehicle fleet can be summarized in the background, for example, and then deviations can be found. The deviations found can then be output as updates to the map to the vehicles.

[0004] Alternatively, it is possible for each vehicle to maintain an internal map itself, to find and evaluate deviations between the map and its measurements itself and to generate update hypotheses itself. The vehicle can use these update hypotheses for the driving task of the vehicle and for the self-localization of the vehicle. In addition, the vehicle can preferably provide these update hypotheses to the background, in particular via a wireless transmission, in particular a radio transmission.

[0005] The requirement in both alternatives that the changes in the digital map contain no systematic errors and are within the error tolerance is particularly important for long-term operation. This is an important prerequisite for the autonomous driving function of the vehicle, which depends on the position estimate of the vehicle localization system.

[0006] A method and a system for positioning a motor vehicle are known from DE 102 016 225 213 A1. The system is set up to carry out a method for positioning a motor vehicle. A landmark hypothesis is created with a computing unit of the system. A map comparison between a local occupancy map and a global occupancy map is carried out. An estimate of a global vehicle pose is carried out by means of a map matching algorithm between an environment model and the global occupancy map.

[0007] An update of a landmark map is known from DE 102 017 105 086 A1. Various classes are provided for this classification of landmark comparison processes. The landmark map can be updated in the vehicle.

[0008] In the case of known methods and systems, in accordance with the landmark assumption, although a digital map is generated and updated, this update is not further assessed or evaluated. If such an update is made, no further check is made of the update. This can lead to defects. The reason for this is that there can be situations in which these updates are disadvantageous due to inaccuracy as a result of measurement errors and / or other error influences. SUMMARY

[0009] It is the task of the present application to provide a method and an electronic evaluation system in which the creation of a digital map and the evaluation of a possible existing update status of a digital map are improved.

[0010] This task is solved by the method and the evaluation system according to the application.

[0011] One aspect of the application relates to a method for evaluating a digital map, wherein the following steps are carried out:

[0012] - detecting an environment of the vehicle, in particular using at least one detection unit of the vehicle;

[0013] - creating a relative map of the environment, in particular based on the detected information and the relative position of the detected landmarks in the environment;

[0014] - providing a global map of the environment with global positions of the landmarks;

[0015] - generating a plurality of update hypotheses for the digital map, in particular based on the position information of the landmarks in the relative map and / or based on the position information of the landmarks in the global map, wherein preferably at least when at least one position deviation, in particular at least one position deviation type, of a landmark with respect to the global position of at least one reference object occurs, an update hypothesis is generated;

[0016] - determining an error statistic of at least some of the position deviations;

[0017] - evaluating the error statistic; and

[0018] - assessing whether at least one update hypothesis, in particular the entire digital map, of the digital map is suitable for self-localization of the vehicle in accordance with the evaluation of the error statistic.

[0019] Thus, by this approach the digital map can be evaluated more precisely. Thus, in particular, the updating of the digital map is no longer carried out automatically and the updates are retained without further checking. Rather, potential updates of the digital map are checked again using the proposed method. That is, it can be provided that potential updates of the digital map have been carried out and the digital map that was updated at that time is checked again using the proposed method. In particular, this is achieved in accordance with the re-checking of the update assumptions. In particular, it is important here to determine the error statistics, to evaluate the error statistics and to assess whether at least one update assumption of the digital map is suitable for the self-localization of the vehicle, wherein this is assessed in accordance with the evaluation of the error statistics. That is, in fact the digital map can have been updated in one embodiment by one or more update assumptions in order to then check the updated digital map using the error statistics and the evaluation thereof by the above-mentioned processes. However, it can also be provided that the update assumptions have not yet been converted into an update of the digital map and thus in this embodiment the update assumptions are checked again before the digital map is updated using the update assumptions.

[0020] Thus, in particular, it is important that the update assumptions generated initially meaning that the digital map needs to be updated are additionally checked again using the error statistics, the evaluation of the error statistics and the subsequent steps by very specific scenarios.

[0021] The position deviation can be identified by the vehicle itself. For example, a detected landmark can deviate from a global position which previously existed and which is already present in the global map. This case, for example, leads to a landmark position deviation update hypothesis. The landmark position deviation update hypothesis describes a possible update requirement of the global map due to such a position deviation. Additionally or alternatively, it can also be detected in the detection process that a landmark previously did not exist in the global map and thus no global position for this landmark exists in the global map. This case, for example, leads to a landmark supplement update hypothesis. The landmark supplement update hypothesis describes a possible update requirement of the global map due to such a landmark supplement. Additionally or alternatively, there can also be a case in which a landmark which is already present in the global map is currently not detected by the detection unit. This case, for example, leads to a landmark removal update hypothesis. The landmark removal update hypothesis describes a possible update requirement of the global map due to such a no longer existing landmark which has to be removed from the global map if necessary. Additionally or alternatively, there can also be information about a landmark of another vehicle and / or another unit. In particular, when this vehicle itself, in particular, cannot utilize its detection unit to detect this information. This case, for example, leads to a landmark external information update hypothesis. Thus, various different scenarios can occur which lead to different types of update hypotheses. The mentioned scenarios should not be understood as exhaustive. The scenarios represent examples of different position deviation types. However, all scenarios are understood in the present context as follows: The scenarios represent position information and in the present context represent "position deviations" and, preferably, the update hypotheses are generated at least when at least one type of such a position deviation occurs for a landmark. In particular, with respect to a global position of at least one reference object.

[0022] Thus, in particular, in a particularly advantageous manner, the update hypotheses generated by the vehicle itself are checked or evaluated automatically. In particular, this automatic checking is in turn implemented by the vehicle itself. This checking of the update hypotheses takes place with respect to a deviation from a specified error tolerance.

[0023] By this approach of the proposed method, on the one hand, the digital map can be generated more precisely and it can be evaluated and estimated whether a potential update is indeed suitable for further use. By this, it can be achieved that error influences such as measurement tolerances and / or errors of the detection unit, etc. can be identified more improved. By this, such error influences can also be eliminated or taken into account more improved.

[0024] In an advantageous embodiment it is provided that the error statistics are evaluated in order to determine whether the positional deviation is caused by an error of the detection unit. If the positional deviation is caused by an error of the detection unit, it is ascertained that the self-localization of the vehicle is carried out using at least these updated assumptions, in particular using the digital map. This is a further advantageous procedure. The reason for this is that the occurring positional deviation can thus be analyzed more finely and in more detail. Thus, the possibility is provided to identify the possible cause of such a positional deviation. If it is identified that such an error occurs due to the detection unit, it is then also possible to take this deviation due to the system into account in future self-localizations of the vehicle. By this, it is then possible to simply implement that the digital map, which has been updated by then, is used as a basis for the self-localization. The reason for this is that it is then also known how an error is introduced by the detection unit and how this error should then be compensated for.

[0025] In a very advantageous embodiment, the evaluation procedure is carried out completely in the vehicle itself as set forth above. By this, a quick evaluation can be carried out. This is an advantageous procedure in particular when the vehicle is moving and should not only be self-localized on the basis of the digital map but also should rate the occurred or potential update of the digital map.

[0026] In an embodiment it is provided that the evaluation results of the evaluation procedure generated in the vehicle are transmitted to a background outside the vehicle. The evaluation results of a fleet of vehicles with a plurality of vehicles are deposited in the background. In the background, the evaluation results are compared with these evaluation results of the fleet of vehicles. On the basis of the comparison, a final evaluation result is generated, in particular by the background. The final evaluation result is provided to the vehicle. Thus, by this approach, information from other vehicles can also be easily and reliably incorporated into the assessment of whether the occurred or potential update of the digital map is suitable. Although the expenditure in this respect can be higher, the accuracy of the rating can be improved.

[0027] In an advantageous embodiment it is provided that, in order to generate the update assumptions, a landmark with a relative position detected by the detection unit of the vehicle is selected by the evaluation system according to a random principle. The global position of the randomly selected landmark is evaluated. That is, in this automated rating scenario, the random principle also serves as a basis. By this, an improvement can also be achieved. The reason for this is that, since not only one specific type of landmark serves as a basis for the rating scenario, an advantageous distribution with regard to the landmarks considered for the rating scenario can be achieved by the random generator. By this, the creation of the rating results is more representative.

[0028] The relative position is that position which is defined by the detection unit of the vehicle. Thus, the position is given in the coordinate system of the vehicle. Correspondingly, the global position is the position in the world coordinate system.

[0029] An evaluation of the global position is of greater importance. The reason for this is that this evaluation is also the basis for the digital map that is updated or is input into the digital map.

[0030] Preferably, the relative position of the selected landmark is disregarded in the evaluation. In an advantageous embodiment, the global position of the selected landmark is calculated from the relative positions of other landmarks that are detected by at least one detection unit in the detection process. From the calculated global position, an error term is determined, which is provided to the error statistics. This is a particularly advantageous embodiment. The reason for this is that the selected landmark is thus disregarded in the evaluation. Thus, in this embodiment, the existing global position of the selected landmark is disregarded in the evaluation. Rather, the global position is calculated in another way. For this calculation, the relative positions of other landmarks are used. In this way, an improvement can be achieved, in particular when generating the error statistics. In particular, it is thus possible to draw more precise analyses and conclusions about the importance of the global position of the selected landmark.

[0031] In an advantageous embodiment, the calculated global position of the selected landmark is compared with the global position of the landmark from the global map. When a positional deviation occurs between these global positions, this positional deviation is formed as an error term and is provided. Thus, in particular, error terms are formed from the positional deviations. The error term then represents, in particular, the determined positional deviation. By this approach, it is possible to make a particularly detailed and precise evaluation of whether the digital map that is updated or the digital map that is provided for updating using the update hypotheses is suitable for the self-localization of the vehicle, not only here but also in the advantageous embodiments described above. Thus, in particular, a particularly precise and in-depth detailed analysis of the update requirements of the digital map can be achieved. In this respect, the update hypotheses that have already been generated can subsequently be improved to some extent and assessed in more detail as to whether they are indeed suitable for the self-localization of the vehicle in the digital map.

[0032] Preferably, in the evaluation of the error statistics, the average error of the overall system is given in advance. The error terms characterizing the positional deviations are averaged. The average error is compared with the average error term. On the basis of the comparison, it is assessed whether the at least one update hypothesis of the digital map, in particular the entire digital map, is suitable for the self-localization of the vehicle. The provision of such error statistics with a plurality of error terms enables still more precise information about the measured and / or calculated positional deviations. Thus, such error statistics allow an exact analysis of what deviations actually exist. By this, it is possible to more precisely and more detailed deduce whether there are errors of the detection unit and / or other error influences. By this more detailed analysis of what errors exist, it is also possible to make the assessment of whether the update hypothesis is suitable for recording or remaining in the digital map more accurate, in particular with respect to the self-localization of the vehicle which is desired and required in the case of the digital map.

[0033] In an advantageous embodiment, a map comparison between the global map and the relative map is carried out before the generation of the update hypothesis. This can also be referred to as map matching. In an advantageous embodiment, on the basis of the map comparison, the global position of the vehicle, in particular the current global position, is determined. That is, the global position in which the vehicle is located is advantageously identified.

[0034] Preferably, on the basis of the map comparison, a correlation between the relative positions of the landmarks and the global positions of the landmarks is carried out. In particular, a correlation between those relative landmarks which are detected with the detection unit in this respect is carried out.

[0035] Preferably, in an advantageous embodiment, the landmarks are removed or disregarded in the case of the map comparison. In particular, the global positions of the landmarks, in particular randomly selected, are disregarded. Then, the following procedure for the assessment of the selected landmarks can be carried out as already set out above in the advantageous embodiments.

[0036] In particular, such removed or disregarded in the assessment landmarks are also referred to as test landmarks.

[0037] In particular, on the basis of the error statistics, it is assessed whether the system, in particular the at least one detection unit, is functioning as assumed or intended.

[0038] In an advantageous embodiment, an error estimation for the overall system is carried out by the error statistics and the evaluation thereof.

[0039] In an advantageous embodiment, the average error of the entire system is 0. By means of the generated error terms mentioned above, the total actual error term can be determined, which is defined by the average error term. If this average error term is also 0 and thus corresponds to the predefined average error, there is a functioning, error-free system. If the average error term deviates from the average error, this deviation can in turn be assessed. If the deviation is within a predefined tolerance interval, the update assumption can be retained and thus also the updated digital map or the digital map to be updated by means of it, in particular for self-localization of the vehicle, can be used. If the deviation is greater than a predefined value, the update assumption is not suitable. The update assumption should then be discarded, in particular.

[0040] In another embodiment, it can be provided that, upon recognition that the landmark detected with the detection unit has no counterpart in the global map, the global position of the new landmark is calculated from the relative positions of other landmarks detected with the detection unit, in particular during the detection process. That is, if it is recognized that the currently detected landmark is newly added, the global position of the landmark can be determined from the relative positions of the other landmarks. Since such a new landmark did not exist before and thus the global position in the global map is unknown and does not exist, this approach with the calculation is advantageous.

[0041] It can also be provided that, upon recognition that the landmark detected with the detection unit has no counterpart in the global map, the global position of the new landmark is calculated from the calculated global position of the vehicle using the detection unit to detect the landmark. In an advantageous implementation, if a new landmark that did not exist before is detected, the global position of the vehicle that performed the detection is calculated. This global position is calculated, in particular, from the relative positions of other landmarks detected with the detection unit of the vehicle during the detection process. Preferably, the global position of the detected new landmark is determined from the calculated global position of the vehicle. In particular, the calculated global position of the vehicle is the update assumption. In this approach and in this calculation, in particular, at least one further landmark that was detected in the past with the detection unit of the vehicle is also taken into account. In particular, the position deviation and thus the error term of the at least one landmark detected in the past is taken into account here. Since the error of the at least one landmark detected in the past is known, it is also known that the calculated global position of the vehicle has this error. Thus, in this respect, the error term of the global position of the vehicle is also known.

[0042] In another embodiment, it can also be provided that a detected landmark that also existed in the past and thus also has a global position in the global map is taken into account for the analysis or evaluation. Thus, in this embodiment, the test landmark for the assessment is not removed, as in the above-described embodiments. In another embodiment, it can be provided that, upon recognition that the landmark detected with the detection unit has no counterpart in the global map, the global position of the new landmark is calculated from the relative positions of other landmarks detected with the detection unit, in particular during the detection process. That is, if it is recognized that the currently detected landmark is newly added, the global position of the landmark can be determined from the relative positions of the other landmarks. Since such a new landmark did not exist before and thus the global position in the global map is unknown and does not exist, this approach with the calculation is advantageous.

[0041] It can also be provided that, upon recognition that the landmark detected with the detection unit has no counterpart in the global map, the global position of the new landmark is calculated from the calculated global position of the vehicle using the detection unit to detect the landmark. In an advantageous implementation, if a new landmark that did not exist before is detected, the global position of the vehicle that performed the detection is calculated. This global position is calculated, in particular, from the relative positions of other landmarks detected with the detection unit of the vehicle during the detection process. Preferably, the global position of the detected new landmark is determined from the calculated global position of the vehicle. In particular, the calculated global position of the vehicle is the update assumption. In this approach and in this calculation, in particular, at least one further landmark that was detected in the past with the detection unit of the vehicle is also taken into account. In particular, the position deviation and thus the error term of the at least one landmark detected in the past is taken into account here. Since the error of the at least one landmark detected in the past is known, it is also known that the calculated global position of the vehicle has this error. Thus, in this respect, the error term of the global position of the vehicle is also known.

[0042] In another embodiment, it can also be provided that a detected landmark that also existed in the past and thus also has a global position in the global map is taken into account for the analysis or evaluation. Thus, in this embodiment, the test landmark for the assessment is not removed, as in the above-described embodiments.

[0043] In an advantageous embodiment, the type of position deviation is analyzed. Depending on the type of position deviation identified, an error statistic is determined or created. As already set out above, it is thus possible to increase the degree of scaling with regard to error recognition and thus also with regard to the assessment of the suitability of the update hypothesis and thus the depth of the analysis.

[0044] In particular, the direction of the position deviation and / or the distance of the detection unit from the landmark is taken into account in the analysis. In this way, for example, it is possible to recognize that if the detection unit, in particular the camera, is pointing in a particular direction, for example is oriented too far to the left, then any measurements have a corresponding deviation in this respect and thus are measured too far to the left in this respect. This type of position deviation can then be identified and analyzed and taken into account in the assessment. Likewise, this can be done with regard to the distances already mentioned above.

[0045] It can also be provided that if a position deviation of a landmark has been determined and thus an error term of the deviation of the calculated global position from the position present in the global map, this error term is stored.

[0046] In principle, the proposed method and thus the evaluation process is also performed such that the detection of the environment with the at least one detection unit of the vehicle is carried out during the movement of the vehicle, in particular at least at discrete time intervals, in particular continuously. Thus, one or more landmarks are always detected continuously in time. In accordance with the scenarios described above, individual update hypotheses are thus always recreated within a certain time interval and thus at certain consecutive points in time. The assessment of at least one or more of these update hypotheses is then carried out individually. By this means, it is also possible to assess a temporally subsequent update hypothesis taking into account a temporally previously created update hypothesis. If the temporally previous update hypothesis has a certain error assessment and thus an individual error term, this error term can be taken into account in the update hypothesis to be assessed subsequently.

[0047] In particular, by this means it is possible to continuously create a plurality of error terms over time. At least some of these error terms can then be taken into account when creating an error statistic. In particular, an average error term can then be determined when such an assessment of the error statistic is made. This average error term can then be compared with an average error, in particular a prescribed average error, as already set out above.

[0048] Especially when all error terms have the same or very similar values, it can be concluded that the detection unit of the vehicle has an error. Then, especially the digital map can be used. In this respect, it can be used in such a way that the known error of the detection unit is taken into account when determining the self-localization by means of the digital map. Additionally or alternatively, it can be provided that the digital map is not updated using the updated hypotheses. The digital map can be updated using the updated hypotheses which are corrected in this respect. In an advantageous embodiment, the digital map can be corrected in this respect if it has been updated with false updated hypotheses.

[0049] Especially when the entire evaluation scenario and thus the entire evaluation process is only carried out in the vehicle having the detection unit and thus without using a background, the storage of the generated updated hypotheses can be carried out in the first layer of the digital map. Thereby, a layer can be generated which is equipped with a certain uncertainty value, which relates to the degree of updating or the suitability of use of the update. In the first layer of the digital map, it is systematically assumed that the map is up to date. New landmarks are entered into the above-mentioned second map layer with a corresponding uncertainty. If the uncertainty of the landmark and thus of the related updated hypothesis is too high, the landmark is moved and is not used in the digital map.

[0050] Especially, the comparison of the error terms is also carried out when creating the error statistics. If all or substantially all of the error terms have the same error, it can be concluded that the detection unit has a systematic error. Especially when different errors occur and these different errors of the error terms equal a predefined average error, the system is suitable and the digital map can be used for the self-localization determination of the vehicle in the updated state or in the state to be updated.

[0051] A further aspect of the application relates to a method for self-localization of a vehicle. A global position of the vehicle is determined from a digital map. The use of the digital map for the self-localization determination is evaluated according to the method according to the above-mentioned aspect or advantageous design solutions thereof.

[0052] A further aspect of the application relates to an electronic evaluation system for evaluating a digital map, especially with regard to the use of the digital map for self-localization of a vehicle. The evaluation system has at least one detection unit of a vehicle. The evaluation system also has at least one evaluation unit. The evaluation system is designed to carry out the method according to the above-mentioned aspect or advantageous embodiments thereof. Especially, the method is carried out with the evaluation system. Especially, the evaluation system is designed to evaluate updated hypotheses for a digital map. Especially, a method for evaluating at least one updated hypothesis for a digital map is also provided.

[0053] Even in embodiments in which the update data and thus the update hypotheses are transmitted by the vehicle to the backend, these update hypotheses are evaluated internally in the vehicle.

[0054] The overall advantage of performing the evaluation process by the vehicle is also that the backend can take these evaluation results into account when comparing them with the fleet data or can already discard these evaluation results directly if the update data and thus the update hypotheses are outside the error tolerance.

[0055] In the context of the present application, the term "hypothesis" describes a possible required update of the digital map. This update can be, for example, a change in the position of a landmark, the addition of a previously unknown landmark and the deletion of a landmark that no longer exists. Now, with the proposed method, it is possible that the landmark has been determined as well as possible on the one hand and that it is now additionally checked whether the landmark is indeed determined directly and sufficiently precisely. In previous methods, although the position of the landmark can be determined, it cannot be concluded whether systematic measurement errors have distorted the result since no further check is performed. It is precisely the present application that now makes this possible. With the proposed method, in which conclusions can also be drawn about the functioning of the system by means of the error statistics, the suitability analysis of the digital map can be improved since the deviations are again assessed in more detail, since a multi-stage process is performed. It is particularly advantageous that for the creation of the error statistics, not only one update hypothesis, but a plurality of update hypotheses are taken into account. Since a plurality of error terms are drawn from this, an error statistics can be provided that has a more precise information content. In particular, at least two, in particular at least five, in particular at least ten such error terms are taken into account.

[0056] In the above-described method, it is preferably also provided that if it is determined in the evaluation of the error statistics that the system and thus a plurality of update hypotheses have a deviation of the value x on average and thus there is an average error term x for the position of a landmark, a further newly detected landmark and first of all the as good as possible position of the new landmark is determined on the basis of the current measurement data and thus the previous error terms. If an error term greater than x occurs here, the update hypothesis is discarded. Despite the position of the new landmark having been estimated as well as possible, there is a corresponding probability of an incorrect calibration of the detection unit or other errors. If, on the other hand, the error term of the newly detected landmark is less than x, the update hypothesis about the new landmark is accepted. The new landmark is then considered a global landmark and used with the error term of x when the vehicle next drives past the newly detected landmark.

[0057] Therefore, the detected landmarks are usually assigned an update hypothesis with a separate error term or are characterized thereby. In particular, an average error is predefined. The average error x can be predefined individually. The average error can also be changed individually. For example, a dynamic change can also be provided.

[0058] The average error can depend on the vehicle and / or the detection unit. Additionally or alternatively, the average error can also depend on the traffic route on which the vehicle is located. For example, in this respect, the average error can be different on a motorway than on a country road or other small traffic routes.

[0059] In principle, the landmarks can be specific objects in the environment of the vehicle. The objects can be stationary objects. For example, this can be traffic control elements such as traffic lights, traffic signs, etc. However, the landmarks can also be, for example, specific buildings or specific trees. The landmarks can also be non-stationary objects. For example, this can be a mobile traffic sign, etc.

[0060] Preferably, the landmarks are divided into at least two different categories. The categories can be predefined. For example, the categories can be the specification of the landmarks in the environment. Additionally or alternatively, the classification can also depend on the time of day. For example, a distinction can be made here between day and night. For example, the classification can include that the estimation of the landmarks and thus the position assessment of the detection unit can be better possible during the day than at night. Thereby, the error term can also be weighted in accordance with the classification of the landmarks. This is a further advantage in order to improve the accuracy of the analysis and thus the evaluation process.

[0061] In particular, it can also be provided that a quality measure of the system is determined in accordance with the evaluation process. Therefore, such a quality measure is determined for a plurality of update hypotheses. In an advantageous embodiment, the quality measure can also be assigned to the digital map. In this respect, a quality measure of the aforementioned classification can also be carried out. Therefore, a quality measure can also be assigned to the respective category division.

[0062] In the case of the detection of the landmarks with the detection unit, the landmarks can also be detected several times as the vehicle drives by and this can also be done, for example, from different perspectives. In particular, the detected landmarks are entered in the vehicle coordinate system so that the aforementioned relative positions are derived. The detected landmarks are preferably calculated with one another by means of an estimation method, for example, a graph-based optimization system from a preferably SLAM algorithm.

[0063] The individual measurements of the landmarks and thus the individual detections of the landmarks can be weighted differently, for example, depending on the angle of view and / or the spacing or distance of the detection unit. In this regard, it is also possible to include odometry or GNSS information and / or other sensor information in order to determine the position of the vehicle. In particular, the global position of the vehicle is determined here.

[0064] By this method, it is achieved that, in particular, an erroneous global position of a landmark entered in the digital map can lead to the fact that, when the vehicle next drives past the landmark, the relative measurements using the detection unit no longer match what is to be expected from the global map. This can lead to the fact that the global position of the vehicle is erroneously estimated and thus the self-localization is erroneously estimated. If this self-localization of the vehicle is erroneously estimated, in turn, all new landmarks are also globally localized erroneously. A cycle is started thereby, which is precisely interrupted by the proposed method or advantageous design solutions thereof. This is achieved in such a way that the system, in particular, constantly performs a self-diagnosis. That is, the system again internally evaluates the own update assumptions of the system. That is, a check process in at least two stages is performed. Certain update assumptions are not accepted as correct, but are again internally evaluated by specific methods as set out. Ultimately, this prevents that erroneous global positions of landmarks are entered into the digital map or remain permanently therein and are used without knowledge of the error.

[0065] In particular, it can be achieved that, in one embodiment, it is continuously or periodically estimated during the driving of the vehicle by means of the landmarks entered in the digital map how precisely the localization of the landmarks is currently taking place. Here, in particular, the extent to which the estimated position of a landmark deviates from its entered global position in the map is determined. Preferably, the landmarks used here are selected from the digital map in such a way that these landmarks can be well detected and the position accuracy of these landmarks is known to be high. If a new landmark is detected which is not contained in the digital map, the accuracy of the reference landmarks around the new landmark can be used to estimate the accuracy of the position of the new landmark, as this has been set out above. This information can be used to generate an update assumption for the landmark.

[0066] The cause of the error in the determination of the landmark position can be, for example, an erroneous sensor calibration, a sensor malfunction, a map comparison performed incorrectly and generally system errors. In this case, the sensor is the detection unit of the vehicle.

[0067] In this case, it is initially not attempted to draw conclusions about the respective specific cause of the error, since the error image is usually dependent on the sensor means used or the detection unit. Rather, the present invention primarily relates to a system which, independently of the detection unit used, is able to estimate the error in the generation of an update assumption. Then, a detailed diagnosis can be performed.

[0068] The present invention also includes combinations of features of the described embodiments. Attached Figure Description

[0069] Embodiments of the invention are described below. For this purpose:

[0070] Figure 1 A simplified diagram of a traffic scene with vehicles and landmarks is shown; and

[0071] Figure 2 A schematic diagram of a flowchart of a method according to an embodiment of the present invention is shown. Detailed Implementation

[0072] The embodiments described below are preferred embodiments of the present invention. In these embodiments, the components described are features of the invention that are to be considered independently of each other, and these features also extend the invention independently of each other, and thus can be considered as components of the invention individually or in combinations other than those shown. Furthermore, the described embodiments can also be supplemented by other features among the features already described in the present invention.

[0073] In these accompanying drawings, elements with the same function are equipped with the same reference numerals.

[0074] exist Figure 1 In the simplified schematic diagram, vehicle 1 is shown. Vehicle 1 can be a motor vehicle, such as a passenger car or a freight car. Vehicle 1 is traveling in the direction of arrow P on traffic route 2. Traffic route 2 can be, for example, a public road such as a national highway, federal highway, or expressway. Traffic route 2 can be long-distance or in a village or town. The examples of traffic routes 2 and their directions should not be construed as exhaustive, but rather as exemplary. Multiple objects are arranged in the environment 3 of vehicle 1. These objects are, in particular, landmarks. For example, these are streetlights S, traffic lights A, buildings G, trees B, and traffic signs V. For example, traffic sign V can be a movable or mobile traffic sign. In particular, other landmarks mentioned herein are statically arranged in environment 3. The types, numbers, and locations of the exemplary landmarks mentioned should not be construed as exhaustive. The static nature or mobility of these landmarks should also be construed as exemplary.

[0075] Vehicle 1 has at least one detection unit 4. This detection unit may be, for example, a camera. The detection unit 4 is arranged such that it can detect the environment 3 of vehicle 1. It is also specified that vehicle 1 has an evaluation unit 5. In the illustrated embodiment, the detection unit 4 and the evaluation unit 5 are components of an evaluation system 6. The electronic evaluation system 6 is also designed to evaluate a digital map 7. Figure 1A digital map 7 is shown in exemplary storage in the vehicle 1. However, the digital map can also be stored outside the vehicle 1. The digital map can then be correspondingly provided to the vehicle 1. The evaluation system 6 is designed to evaluate the digital map, in particular with regard to the use for the self-localization of the vehicle 1 with the digital map 7. In particular, the evaluation system 6 is designed to evaluate update assumptions, in particular with regard to the use for the self-localization of the vehicle 1 with the digital map 7. The corresponding method is carried out in particular with the evaluation system 6.

[0076] In Figure 2 A flowchart of an embodiment of the method is shown in a schematic diagram. First, the environment 3 of the vehicle 1 is detected during the moving driving of the vehicle 1 with the at least one detection unit 4. Here, at least some of the landmarks S, A, G, B and V are detected simultaneously or in succession at certain points in time with the detection unit 4. A relative map of the environment 3 is created on the basis of the detected information. In particular, in this respect, a graph-based positioning and mapping system is preferably run according to step S1. The relative positions of the detected landmarks S, A, G, B, V are entered in the relative map.

[0077] Then, a global digital map 7 of the environment 3 with global positions of the landmarks is provided. The global positions can be the positions of the landmarks S, A, G, B, V.

[0078] In this respect, it can be the case that at least one of the landmarks did not exist before and thus represents a new landmark. This can be, for example, a traffic sign V that has just been erected. Also, in another embodiment, it can be the case that all detected landmarks have existed before, i.e. in particular also at the last time the vehicle 1 drove through, and are not new landmarks. Then, all of the landmarks have a global position that exists in the global digital map 7. In another embodiment, it can also be provided that all detected landmarks have existed before, however, for example, one of the landmarks has changed its position in the meantime. This can be the case, for example, for a traffic sign V, which is mobile and thus can change its position.

[0079] Then, in a further step, when a positional deviation of the landmarks with respect to the global positions of the at least one reference object occurs, a plurality of update hypotheses for the digital map can be generated on the basis of the position information of the landmarks in the relative map and / or on the basis of the position information of the landmarks in the global map. That is, in this algorithm, positional deviations of the global positions of the detected and / or newly added landmarks and / or of the landmarks whose positions have changed are searched for or determined. In particular when such a positional deviation occurs for a detected landmark, an update hypothesis is first assumed in this respect. This means that on the basis of this assessment, the system has come to the preliminary result that the digital map should be updated on the basis of this update hypothesis.

[0080] Then, in the proposed method, this preliminary result is additionally checked again. The above-mentioned reference object can be, for example, a landmark or a vehicle 1.

[0081] Then, in the method, an error statistic is determined on the basis of the assumed update hypothesis. Here, a plurality of update hypotheses are preferably taken into account, which are thus incorporated into the determination of the error statistic. In a further step, the error statistic is evaluated. On the basis of the evaluation of the error statistic, it is assessed whether the at least one update hypothesis of the digital map, in particular the entire digital map, is suitable for the self-localization of the vehicle 1. This also encompasses assessing the degree of suitability, if necessary. Thus, it is not only possible to distinguish between complete suitability or complete unsuitability in a digital manner. Rather, in an advantageous embodiment, it is also possible to determine a certain degree of suitability. Thus, the degree of suitability can also be a percentage between 0% and 100%.

[0082] In an embodiment according to Figure 2 In an advantageous embodiment, after step S1, the random selection of a landmark within the field of view of the vehicle 1 is carried out according to step S2. Thus, for the generation of an update hypothesis, the landmark S, A, G, B, V with a relative position detected by the detection unit 4 is selected by the evaluation system 6 according to the principle of randomness. Then, the global position of this randomly selected landmark is evaluated.

[0083] In an advantageous embodiment, according to step S3, this landmark, which is referred to as test landmark, is temporarily removed from the digital map. The associated global position of the test landmark is stored as a reference position. In a further step S4, the global position of the randomly selected landmark, which is referred to as test landmark, is recalculated on the basis of the detected relative positions of the other landmarks detected by the detection unit 4. This is also carried out by the evaluation system 6.

[0084] Next, in a further step S5 in the illustrated embodiment, an error term between the estimated or calculated global position of the temporarily removed landmark and a stored reference position of the particular landmark is determined, the reference position also being a global position. The error term is provided to an error statistic. This is achieved, in particular, for a plurality of landmarks. According to step S6, the error terms are stored in a database 8. The database 8 can preferably be arranged in the vehicle 1.

[0085] According to step S7, the error terms forming the error statistic are evaluated. Here, a statistical evaluation is performed, in particular, on the basis of all previously estimated or calculated error terms. Here, in an advantageous embodiment, according to step S8, a predefined average error of the overall system is compared with the average error term. The error terms on which the error statistic is based characterize position deviations. The error terms are averaged, for example arithmetically averaged, whereby an average error term is derived. From the comparison between the average error of the overall system and the determined average error term, it is assessed: whether the at least one update assumption of the digital map, in particular the entire digital map, is suitable for the self-localization of the vehicle 1; and / or the degree of suitability of the at least one update assumption. If the average error term is greater than the average error, in one scenario, the update assumption is rated as unsuitable for the digital map and thus for the self-localization of the vehicle 1 based on the digital map. If the average error term is less than or equal to the average error, the basic suitability of the update assumption can be rated. Preferably, the entire evaluation process is carried out entirely in the vehicle 1 itself.

[0086] In an alternative, it can be provided that, after step S8, the result of the evaluation with regard to the update assumption is transmitted to a backend 9 outside the vehicle. The evaluation result is taken into account in the backend 9. In addition, evaluation results of at least one further vehicle of a vehicle fleet are also taken into account, additionally. Next, a final evaluation result is determined.

[0087] With regard to Figure 2 , a scenario is exemplarily set forth in which the randomly selected test landmark already exists a priori and is present in the digital map with a global position.

[0088] In another scenario, it is also possible that the landmark, which is randomly selected and already existed a priori, is not temporarily ignored when calculating the global position of the selected landmark, but is also taken into account subsequently.

[0089] In another scenario, it is also possible that the randomly selected landmark is new and did not exist previously. This means that the landmark does not yet have a global position in the global digital map. In this scenario, it can be provided that further detected landmarks with relative positions are taken into account in order to calculate the global position of the vehicle 1. Next, the global position of the new landmark is determined from the calculated global position of the vehicle 1.

[0090] In addition, scenarios can also be evaluated in which a landmark present previously is no longer present or cannot be detected in the current detection process.

[0091] In particular, in all possible scenarios, the error statistics are evaluated to determine whether the position deviations are caused by errors of the detection unit 4. If these position deviations are caused by errors of the detection unit, it is confirmed that the self-localization of the vehicle 1 is carried out using at least these update hypotheses, in particular using the digital map 7. This is identified, in particular, by all error terms having the same error. It is then identified that the error lies at the detection unit. The known errors are then taken into account in the self-localization, respectively.

[0092] Preferably, a map comparison between the global map and the relative map is carried out before the generation of the update hypotheses. In particular, on the basis of the map comparison, the current global position of the vehicle 1 is determined.

[0093] In a further advantageous embodiment, the position deviation type is analyzed and the error statistics are determined depending on the position deviation type. In particular, the direction of the position deviation and / or the distance of the detection unit 4 from the landmarks S, A, G, B, V are taken into account in the analysis.

[0094] List of reference signs

[0095] 1 vehicle

[0096] 2 traffic route

[0097] 3 environment

[0098] 4 detection unit

[0099] 5 evaluation unit

[0100] 6 evaluation system

[0101] 7 digital map

[0102] 8 database

[0103] 9 background

[0104] A traffic light

[0105] B tree

[0106] G building

[0107] P arrow

[0108] S street lamp

[0109] V traffic sign

[0110] S1 step

[0111] S2 step

[0112] S3 step

[0113] S4 step

[0114] S5 step

[0115] S6 step

[0116] S7 step

[0117] S8 step.

Claims

1. A method for evaluating a digital map, wherein, The following steps are performed: - detecting the environment of the vehicle using at least one detection unit of the vehicle; - creating a relative map of the environment based on the detected information and the relative positions of the detected landmarks in the environment; - providing a global map of the environment with global positions of the landmarks; - generating a plurality of update hypotheses for the digital map based on the position information of the landmarks in the relative map and / or based on the position information of the landmarks in the global map, wherein an update hypothesis is generated when a position deviation of a landmark relative to the global position of at least one reference object occurs; - determining error statistics of at least some of the position deviations; - evaluating the error statistics; and - assessing whether at least one update hypothesis of the digital map is suitable for self-localization of the vehicle depending on the evaluation of the error statistics, wherein the error statistics are evaluated to determine whether the position deviations are caused by errors of the detection unit, wherein, if the position deviations are caused by errors of the detection unit, it is assessed whether at least one update hypothesis of the digital map is suitable for self-localization of the vehicle taking into account a correction of the errors of the detection unit, wherein, for generating an update hypothesis, a landmark with a relative position detected using the detection unit is selected by an evaluation system according to a random principle, wherein the global position of the randomly selected landmark is evaluated, wherein the relative position of the selected landmark is ignored in the evaluation and the global position of the selected landmark is calculated from the relative positions of other landmarks detected in the detection process and / or by the detection unit in the detection process, wherein an error term is determined from the calculated global position, which is provided to the error statistics, wherein the calculated global position of the selected landmark is compared to the global positions of the landmark from the global map, wherein, when a position deviation occurs between these global positions, the position deviation is formed as an error term.

2. The method according to claim 1, wherein whether the entire digital map is suitable for self-localization of the vehicle is assessed depending on the evaluation of the error statistics.

3. The method according to claim 1 or 2, wherein the evaluation process is performed entirely in the vehicle itself.

4. The method according to claim 1 or 2, wherein transmit evaluation results of an evaluation process generated in the vehicle to a backend outside the vehicle, wherein the evaluation results of a fleet of vehicles with a plurality of vehicles are deposited in the background, wherein the evaluation results are compared to the evaluation results of the fleet in the background and a final evaluation result is generated from the comparison, which is provided to the vehicles.

5. The method according to claim 1 or 2, wherein In evaluating the error statistics, the average error of the entire system is given in advance, wherein the error terms characterizing the position deviations are averaged and the average error is compared to an average error term and it is assessed whether at least one update hypothesis of the digital map is suitable for self-localization of the vehicle depending on the comparison.

6. The method according to claim 1 or 2, wherein Before generating the update hypothesis, a map comparison between the global map and the relative map is performed in order to determine the global position of the vehicle therefrom.

7. The method according to claim 6, wherein According to the map comparison, a correlation between the relative position of a landmark and the global position of a landmark is performed.

8. The method according to claim 1 or 2, wherein Upon recognizing that the landmark detected with the detection unit has no counterpart in the global map, the global position of a new landmark is calculated from the relative position of other landmarks detected with the detection unit and / or from the calculated global position of the vehicle.

9. The method according to claim 1 or 2, wherein The position deviation type is analyzed and the error statistics are determined therefrom.

10. The method according to claim 9, wherein The direction of the position deviation and / or the distance of the detection unit from the landmark are considered in the analysis as position deviation types.

11. A method for self-positioning of a vehicle, wherein, The global position of the vehicle is determined from a digital map, wherein the suitability of using the digital map for the self-localization is evaluated according to the method of any one of claims 1 to 10.

12. An evaluation system for evaluating a digital map using a detection unit of a vehicle and using an evaluation unit, wherein The evaluation system is designed to perform the method according to any one of claims 1 to 10.

Images