Method and apparatus for providing high resolution digital maps - Patents.com

Abstract

The present invention relates to a method for providing a high-resolution digital map (300), comprising the steps of: locating an engine, in particular a vehicle; providing sensor data at the located location during a measured journey of the engine; determining, at the located location, an identification indicator for at least one identified object using the provided sensor data; and adding at least one further layer (310a-310n) to the high-resolution digital map (300), wherein the at least one further additional layer (310a-310n) includes an identification indicator for the at least one identified object.

Description

[Technical field]

[0001] The present invention relates to a method for providing a high resolution digital map.The present invention further relates to an apparatus for providing a high resolution digital map.The present invention further relates to a computer program product. [Background technology]

[0002] Prior Art Automated driving is nowadays one of the major trends in the automotive industry. Over the last decade, car manufacturers have gradually increased the number and functionality of their driver assistance systems. In addition to the technical aspects of realizing fully automated vehicles, validation and approval are major challenges in the automotive industry, due on the one hand to the complexity of the interaction with the surroundings and, on the other hand, to the difficulty of proving that automated vehicles perform at least as well as a human driver from a safety point of view.

[0003] It is foreseen that only real vehicle tests for validation would not be economical. For these reasons, simulation is considered as a key element for the development and validation of fully automated vehicles. Simulation-based driving tests are therefore an important pillar of verification and validation strategies. However, for a meaningful simulation, all applied models must be validated beforehand. The functions to be protected are based on virtual models of the vehicle's surrounding sensor systems. To protect the driving functions, highly accurate sensor models (i.e. mathematical simulations of real sensors) are required, but currently there are no generally recognized quality standards for evaluating and validating such sensor models. This is because conventional maps do not provide sufficient resolution or sufficient information about the surroundings of a fully automated system. Therefore, the safety demands are enormous, especially in urban areas with high traffic density. The information contained in today's high-resolution maps is shown in so-called layers, each describing a different aspect of the surroundings. Here, there are various definitions of which layers in particular contain which information. However, often in the base layer there is information about the road network. On this basis, in other layers static objects such as e.g. traffic signs, e.g. buildings, trees etc. are described. Summary of the Invention [Problem to be solved by the invention]

[0004] Disclosure of the Invention The object of the present invention is to provide a method for better providing high-resolution digital maps for at least partially automated vehicles. [Means for solving the problem]

[0005] The above mentioned problem is solved according to a first aspect by a method for providing a high resolution digital map, the method comprising: - determining the location of the vehicle, in particular the vehicle; Providing sensor data at localized locations during a measured run of the engine; determining, at the located location, an identification indicator for at least one object identified using the provided sensor data; - adding at least one further layer to the high resolution digital map, the at least one further layer comprising identification indicators for the at least one identified object; has.

[0006] At least one additional layer is used as a so-called "confidence layer" for the validation of the sensor model. Such an additional layer based on pure sensor information would contain much more data. The thus extended high-resolution digital map can advantageously be used to improve the algorithms of the sensor model, which can thereby be validated and improved. This results in a low-cost validation of the sensor model. The creation of the high-resolution digital map is based on real sensor data of a measurement run, which must be performed in any case to test the sensor system. As a result, the extended high-resolution digital map according to the invention can be created with only a small increase in cost.

[0007] The above problem is solved according to a second aspect by an apparatus for providing a high resolution digital map, the apparatus comprising: at least one sensor device providing sensor data during the measurement journey; A location device for determining the location of the engine, in particular the vehicle; an identification device for determining an identification indicator for at least one identified object from the provided sensor data; an additional device for adding at least one additional layer to the high resolution digital map, the additional layer including identification indicators for at least one identified object; It is equipped with:

[0008] According to a third aspect, the above mentioned problem is solved by a computer program product comprising program code means for carrying out the proposed method when executed on the proposed electronic device or when stored on a computer readable data carrier.

[0009] Advantageous developments of the method are the subject of the respective dependent claims.

[0010] In an advantageous development of the method, it is provided that one additional layer is created for each sensor type, which makes it possible to create one additional layer for each sensor type (e.g. radar, lidar, camera, etc.), which further improves the means of validation of the sensor model.

[0011] In a further advantageous development of the method, it is provided that a discrimination index is determined for each sensor type, which supports an optimal diversification of the discrimination index for each additional layer of the high-resolution digital map.

[0012] In another advantageous development of the method, it is provided that the discrimination indicator is at least one of the following: the distance at which the object is identified, the probability that the object is identified, etc. For example, this makes it possible to determine a discrimination indicator for a concrete static object.

[0013] In another advantageous development of the method, it is provided that the identification indicators associated with certain weather conditions or data relating to the weather conditions are stored in a high-resolution digital map, which advantageously provides different means for providing identification indicators for different weather conditions.

[0014] In a further advantageous development of the method, it is provided that the discrimination feature is determined during or after the measurement run has been carried out, which advantageously provides for different means for determining the discrimination feature.

[0015] In a further advantageous development of the method, it is provided that the identification indicator is determined in the vehicle or in the cloud, which also advantageously provides for different means for determining the identification indicator.

[0016] In another advantageous development of the method, it is provided that during the localization using the provided sensor data, the positioning data is generated using a SLAM algorithm. Advantageously, different types of localization using the provided sensor data are performed in the two ways described above.

[0017] In another advantageous development of the method, it is provided that an estimated trajectory is determined, which can be provided, for example, by means of a SLAM algorithm known per se, so that the sensor model can also be checked by means of the SLAM algorithm.

[0018] In another advantageous development of the method, it is provided that the sensor data is determined for at least one of the following sensor types: radar, lidar, camera, ultrasound. The proposed method is therefore advantageously suitable for different sensor types.

[0019] In the following, the invention together with further characteristics and advantages is explained in detail on the basis of two drawings, in which all the characteristics described or shown, by themselves or in any combination, constitute the subject matter of the invention, independently of the generalization in the claims or their references and independently of their expression or depiction in the description or drawings.

[0020] The disclosed features and advantages of the method may likewise be derived from the disclosed features and advantages of the apparatus, and vice versa. [Brief description of the drawings]

[0021] [Figure 1] FIG. 2 illustrates the basic flow of an embodiment of the proposed method. [Diagram 2] FIG. 1 is a block diagram of a proposed electronic system for implementing the proposed method. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Description of the embodiments The core idea of ​​the present invention is to extend the high-definition digital map (HD Map) in particular to allow both a qualitative and a quantitative evaluation of the sensor model with the high-definition digital map, which plays an important role in the field of autonomous driving.

[0023] According to the invention it is proposed to provide an improved high-resolution digital map 300 for at least partially automated vehicles. In particular, an extension of such a high-resolution digital map 300 comprises further additional layers 310a-310n, which have so-called sensor-specific confidence or discrimination indicators, so-called Key Performance Indicators (KPIs in English).

[0024] Such discriminatory indicators represent key performance indicators, based on which progress or satisfaction with important goals or critical success factors can be measured. These are defined from the requirements for the sensor system or also from further processing stages, such as object detection. In order to use such key performance indicators as a basis for comparison with sensor data generated from a simulation, it is necessary to specify a tolerance within which the results of the sensor model are considered acceptable, since it is expected that the real sensor data (from a real measurement run) and the synthetic sensor data (from a simulated run with the sensor model) will differ to some extent. For this purpose, measurement runs are performed with real sensors and thresholds for the quality of the localization are calculated based on the resulting measurement data and marked in the high-resolution digital map 300.

[0025] In the following, the term "automated vehicle" is used synonymously with the terms "fully automated vehicle", "autonomous vehicle" and "semi-autonomous vehicle".

[0026] The steps of generating at least one additional layer 310a-310n are explained in detail below. Firstly, the current position of the engine is determined, in particular in the form of an automated vehicle. For this purpose, a positioning device can preferably be used to provide GNSS or GPS data that are used for the positioning of a measurement vehicle (not shown). By means of the measurement vehicle, within the scope of the measurement journey, a sensory detection of the surrounding data by a sensor device can be carried out, where as sensor device for example radar sensors and / or lidar sensors and / or cameras and / or ultrasonic sensors can be used. In this way, the position of the engine can be determined precisely; for this purpose any suitable positioning method can be used.

[0027] After detection of the individual sensor data, a respective discrimination index can be calculated for at least one object identified from the sensor data. For this, the current position of the vehicle is used to be able to convert the calculated values ​​into the high-resolution digital map 300. The calculation of the discrimination index can be performed for one, for several or for all sensor types used by the automated measurement vehicle. Here, it may be envisaged that for each sensor type used, a specific additional layer 310a-310n is added to the high-resolution digital map 300.

[0028] Here, for example, static objects (e.g. traffic lights, traffic signs, buildings, etc.) can be identified by the detected sensor data. For example, identification indicators can be indicated at what distance (e.g. 50 m) the static object is identified by the sensor device and / or with what probability the static object is identified by the sensor device. Alternatively or additionally, the detected sensor data can be used to determine a predicted trajectory using a SLAM algorithm known per se. In this way, additional layers 310a-310n are created, which provide sensor-specific characteristics as an extension of the high-resolution digital map 300.

[0029] Optionally, it may be envisaged that the above-mentioned identification indicators may be generated in certain weather conditions (e.g., sunshine with clear visibility, night, fog, rain, etc.) or that data relating to the weather conditions may be stored in the high-resolution digital map 300 as additional information in additional layers 310a-310n.

[0030] The improved high-resolution digital map 300 can be advantageously used as a basis for validating the sensor model, thereby verifying that the developed sensor model is modeled sufficiently accurately and with the correct characteristics. In this case, from the actual measurement run, the observed scenario is extracted, so that this scenario can be imitated in the simulation. This scenario can then be resimulated in the virtual environment by the developed sensor model. The calculation of the discrimination index in the simulation can carry out a comparison between the discrimination index determined by the actual sensor data and the discrimination index determined by the synthetically based sensor data.

[0031] As a result, the performance mismatch between the actual sensor system and the sensor model is not only visualized but also quantified. Furthermore, the illustrated methodology not only enables validation of the sensor model but also allows the use of the validated location-related model in simulations.

[0032] FIG. 1 shows, in a highly schematic manner, the basic flow of the proposed method.

[0033] In step 100, the engine, particularly the vehicle, is located.

[0034] In step 110, sensor data is provided at located locations during a measurement run of the engine.

[0035] In step 120, an identification indicator is determined for at least one object identified at the located location using the provided sensor data.

[0036] In step 130, at least one additional layer 310a-310n is added to the high resolution digital map 300, the at least one further additional layer 310a-310n including an identification indicator for the at least one identified object.

[0037] As a result, an augmented high-resolution digital map 300 for an engine, e.g., an at least partially automated vehicle, is provided that can be used to efficiently simulate or validate sensor models in a subsequent simulation process. In this manner, costs of validating sensor models can be advantageously minimized.

[0038] In the following, the proposed flow is considered in detail by means of an example, for which the so-called "Simultaneous Localization and Mapping" (SLAM) algorithm is provided as an example to explain the above-mentioned methodology.

[0039] The SLAM algorithm is advantageous in that it can be used with all (e.g., three) sensor types (e.g., radar, lidar, camera). The goal of the SLAM algorithm is to generate a consistent high-resolution digital map 300 of the surroundings and at the same time estimate the vehicle's position within this high-resolution digital map 300. As a result, the measurement run results in both a high-resolution digital map of the surroundings and a predicted trajectory.

[0040] To evaluate a SLAM algorithm, for example, the absolute error of the orbit estimation with respect to the GNSS data can be calculated as a confidence metric. Since it is assumed that such deviations vary from sensor type to sensor type, the results of all sensor types are stored and added as new layers 310a-310n of the high-resolution digital map 300. This allows the current evaluation results of the system to remain within the high-resolution digital map 300.

[0041] This has the advantage that the comparison can be carried out efficiently without recalculating the results, for example by storing the calculated confidence indicators of the last system state in relation to the position for comparison after a resimulation of a measurement run, a subsequent check or a system update.

[0042] FIG. 2 shows a basic block diagram of a proposed electronic system 200 for obtaining a high-resolution digital map 300 .

[0043] At least one sensor device 220a-220n is visible providing sensor data recorded during the measurement journey. A location determination device 210 for determining the location of the vehicle is functionally connected to the sensor devices 220a-220n. An identification device 230 is functionally connected to the location determination device 210 for determining an identification indicator, as described in detail above, for at least one object identified using the provided sensor data. An addition device 240 is functionally connected to the identification device 230 for adding at least one additional layer 310a-310n to the high-resolution digital map 300 having a number of layers 310a-310n containing specific information such as, for example, roadway boundaries, traffic signs, buildings, etc. At least one additional layer 310a-310n comprises an identification indicator for at least one identified object.

[0044] Advantageously, the proposed system 200 may be located in an automated measurement vehicle or in the cloud. In the first case, the additional layer 310a-310n is determined immediately upon detection of the sensor data. In the second case, the additional layer 310a-310n can be determined during the measurement run or after the end of the measurement run.

[0045] In a particular application, the high-resolution digital map 300 extended according to the invention can be used to validate a sensor model. In this way, it can be proven that the sensor model performs localization using the high-resolution digital map 300 as well as a real sensor device using real sensor data. In this way, the sensor device is efficiently tested by the simulation method. This can then actually contribute to the effective validation of automated vehicles using the simulation process.

[0046] Advantageously, the proposed method is implemented, for example, as software executed on the system 200. In this way, easy adaptability of the method is supported.

[0047] Those skilled in the art will modify and / or combine the features of the present invention in any suitable manner without departing from the essence of the present invention.

Claims

1. A method for providing a high resolution digital map (300), comprising: - performing a location determination of the engine; - detecting, by means of a sensor device, ambient data at the location of the located engine during a measurement run of the engine; - determining an identification indicator for at least one object identified using the ambient data at the located engine location; - adding at least one additional layer (310a-310n) to the high resolution digital map (300), said at least one additional layer (310a-310n) comprising the discrimination indicators determined for the at least one identified object according to the type of the sensor device used for the detection of the ambient data; having The discrimination index represents a key performance indicator that is a sensor-specific confidence index or discrimination index of the sensor device. method.

2. The method of claim 1, further comprising creating one additional layer (310a-310n) for each sensor device type.

3. The method according to claim 1 or 2, wherein the discrimination index is determined for each type of sensor device.

4. 4. The method of claim 1, wherein the identification metric is at least one of: a distance at which the object is identified; a probability that the object is identified.

5. 5. The method according to any one of claims 1 to 4, further comprising storing said identifying indicators associated with a given weather condition or data relating to the weather condition in said high resolution digital map (300).

6. 6. The method according to claim 1, wherein the discrimination indicator is determined during or after the measurement run has been performed.

7. The method of claim 1 , wherein the identification indicator is determined at the institution or in the cloud.

8. 8. The method of claim 1, wherein during the step of determining an identification indicator for at least one object using the surrounding data, a SLAM algorithm is used to generate positioning data.

9. The method of claim 8 , further comprising determining an estimated trajectory.

10. The method according to any one of the preceding claims, wherein the ambient data is determined for at least one of the following sensor device types: radar, lidar, camera, ultrasonic.

11. An electronic system (200) for producing a high resolution digital map (300), comprising: A location determination device (210) for determining the location of the engine; at least one sensor device (220a to 220n) for detecting ambient data at the localized engine position during a measurement run of the engine; an identification device (230) for determining an identification indicator for at least one object identified using the detected ambient data; an adding device (240) for adding to the high resolution digital map (300) at least one additional layer (310a-310n) comprising said identification indicators for said at least one identified object, said identification indicators being according to the type of said at least one sensor device (220a-220n) used for the detection of said ambient data; It is equipped with the discrimination index represents a key performance indicator that is a sensor-specific confidence index or discrimination index of the sensor device (220a-220n); 1. An electronic system (200).

12. A computer program comprising a program code for implementing the method according to any one of claims 1 to 10, when the computer program is executed on an electronic system (200) for producing a high resolution digital map (300).

13. 13. A computer readable data carrier on which a computer program according to claim 12 is stored.

Images