CN116358514A

CN116358514A - Method, computing unit and system for providing ambient data, creating and/or perfecting a digital map

Info

Publication number: CN116358514A
Application number: CN202211650521.2A
Authority: CN
Inventors: H·米伦茨; J·罗德
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2021-12-21
Filing date: 2022-12-21
Publication date: 2023-06-30
Also published as: DE102021214767A1

Abstract

The invention relates to a method for providing surroundings data to a digital map by means of a vehicle, comprising the following steps: driving the vehicle into a predetermined region arranged in or adjacent to the map region, for which insufficient ambient data is present in particular in the digital map; determining a pose of the vehicle within the predetermined area, wherein a location position of the vehicle and an angular position of an orientation of the vehicle are determined as poses; detecting ambient data by means of at least one sensor of the vehicle; the pose of the parked vehicle and the ambient data are provided for creating and/or refining the digital map.

Description

Method, computing unit and system for providing ambient data, creating and/or perfecting a digital map

Technical field

The present invention relates to a method for creating a digital map. The invention also relates to a system for creating a digital map. The invention also relates to a computer program product.

Background

Current and future automated (AD systems) and assisted (DA systems) vehicle systems rely to a large extent on additional information from digital maps. In this way, increasingly complex automated driving tasks can be performed based on reduced vehicle sensor groups. The use of map information requires vehicle positioning (position and orientation) relative to the digital planning map used.

A method for updating an electronic land map is known from WO 2018/001670 A1, in which a sensor of a vehicle detects an object, the position of the vehicle is detected, and the land map is updated with these data.

Disclosure of Invention

It is an object of the present invention to provide an improved method for creating a digital map.

The task is solved by the following method:

a method for providing surroundings data to a digital map by means of a vehicle, the method having the following steps:

driving the vehicle into a predetermined region arranged in or adjacent to the map region, for which there is in particular insufficient ambient data in the digital map;

determining a pose of the vehicle within the predetermined area, wherein a location position of the vehicle and an angular position of an orientation of the vehicle are determined as poses;

detecting ambient data by means of at least one sensor of the vehicle;

the pose of the parked vehicle and the ambient data are provided for creating and/or refining the digital map.

In this way, a defined overhead is generated for providing mapping data in the problem area. In such problem areas, features may be missing and/or ambiguous. Advantageously, by means of the proposed method, the precision of the mapping can be significantly improved by: the measuring vehicle is dispatched into the problem area and data for the complete creation of the digital map are detected there.

This is particularly advantageous because the sensing devices of fleet vehicles are typically consumer-grade sensing devices having limited performance capabilities compared to sensing devices that measure vehicles.

As a result, high accuracy of map creation is thereby achieved by the "enhanced (angeeicherte) crowdsourcing scheme".

A method for creating and/or refining a digital map by means of ambient data is provided, wherein the pose of a vehicle and the ambient data relative to the pose of the vehicle are received from a central computing unit, wherein the central computing unit records the ambient data corresponding to the pose of the vehicle at least partially into the digital map, and wherein the refined digital map is output at least partially to the vehicle, wherein the central computing unit finds a map region in the digital map in which insufficient ambient data is present, wherein the central computing unit outputs instructions to the vehicle: driving into a predetermined area within or adjoining the map area, recording ambient data and transmitting said ambient data to the central computing unit.

A computing unit is provided, which is configured for carrying out the method.

A computer program product is provided with program code means for performing the method when the computer program product is run on a computing unit or stored on a computer-readable data carrier.

The advantageous development of the proposed method can be achieved by the measures listed in the preferred embodiments.

In one embodiment, the vehicle is first parked in a predetermined zone. The pose of the parked vehicle is then determined. Thereby improving the accuracy of determining the pose.

In a further embodiment, the pose of the vehicle is determined by means of GNSS signals, in particular by means of differential GNSS signals. Thereby further improving the accuracy of determining the pose.

In a further embodiment, the pose of the vehicle is determined by means of a plurality of measurement cycles and/or a correction method is used to refine the pose of the vehicle. Thereby further improving the accuracy of determining the pose.

In another embodiment, the pose of the vehicle is resolved to an accuracy of several centimeters and/or less than one degree.

In another embodiment, a local ambient map is detected as ambient data.

In a further embodiment, a further vehicle finds its own position relative to the vehicle, wherein the further vehicle detects further ambient data by means of at least one further sensor, wherein the further vehicle provides the relative position, in particular the relative pose, and said further ambient data relative to the vehicle for creating and/or refining the digital map. The ambient data and the pose of the further vehicle may be transmitted to a central computing unit. Thereby increasing the amount of ambient data. Furthermore, low quality ambient data of the vehicle with position determination by means of GNSS systems may be used.

In one embodiment, the vehicle or the further vehicle is an autonomously driven vehicle, in particular a robotic taxi. The autonomous vehicle enters a predetermined area and detects ambient data, in particular when no person is present in the vehicle. Thus, robotic taxis may be used to obtain accurate maps when not carrying passengers.

An embodiment of the proposed method provides for generating a comprehensive list of recognized regions. The integration of the regions can be performed, for example, by combining dual or multiple detections. Advantageously, a minimum number of problem areas into which the measuring vehicle is subsequently dispatched is thereby defined. Advantageously, the outlay and costs for measuring the vehicle can thereby be kept low.

An embodiment of the method provides for the creation of a defined high-precision local map at Yun Zhongchuang. Advantageously, the high performance capabilities of the cloud are thereby used to create a local map with defined high accuracy.

An embodiment of the method provides for creating a local map with defined high accuracy at least in part on the measuring vehicle. In this way, a map with a defined high accuracy may already be created in part at the survey vehicle, which then transmits the high accuracy map onto the cloud for further mapping.

A further advantageous development of the method provides for at least one region which does not have sufficient data material for creating the digital map to be automatically identified. This may occur, for example, due to failure of optimization steps, failure of data correlation, etc. throughout the mapping framework. In this way the overhead for finding the problem area can be kept low.

An embodiment of the method provides that the position of the parked measuring vehicle is determined with high accuracy and corresponding data are generated. The following facts are advantageously used here: the measuring vehicle has a high-precision GNSS sensor with which a positioning accuracy in the region of low unit-order cm can be achieved. This advantageously enables such high positioning accuracy to be achieved in combination with long measurement durations.

In one embodiment of the method, the measuring vehicle is parked in such a way that it detects at least one sensor-wise region which can also be detected by the fleet vehicle. In this way, an overlap between the ambient data detected by the measuring vehicle and the ambient data detected by the fleet vehicle is derived, which can be efficiently processed.

In one embodiment of the method, the position of the parked measuring vehicle is reconstructed from the fleet data if the measuring vehicle cannot be located by means of GNSS. In this case, the measuring vehicle must be stationary in order to be detected by the fleet vehicle in a sensor-wise manner. This makes it possible to achieve at least one advantageous temporary effect as long as the vehicle is measured to be parked.

In one embodiment of the method, the determined ambient data of the at least one measuring vehicle is stored in a digital map and is not modified and marked as ambient data of the measuring vehicle. The high-precision GNSS position determined with high precision by the high overhead of the measuring vehicle is therefore marked as a high-precision position of the measuring vehicle. Measuring the ambient data of the vehicle may have a higher priority than the ambient data of the fleet vehicle. For example, the ambient data of the measuring vehicle may be prohibited from being changed or overwritten by the ambient data of the fleet vehicle. Such a high precision reference point in the form of a measured vehicle pose would only potentially be degraded by fleet vehicle data.

Drawings

The invention is described in detail below in connection with additional features and advantages according to two figures. All of the features described or illustrated herein constitute subject matter of the present invention by itself or in any combination, irrespective of their generalization in the various embodiments or their back-reference and irrespective of their presentation or illustration in the specification or drawings.

Features and advantages of the disclosed methods derive from the disclosed features and advantages of the system in a similar manner, and vice versa.

The drawings show:

fig. 1: a diagram of a data collection process for the method;

fig. 2: a principle flow of one embodiment of the proposed method;

fig. 3: a measurement vehicle and a central computing unit.

Detailed Description

The digital HD map may distinguish the three main layers of map information.

a) Planning layer: information is included such as precise lane geometry, lane topology and connectivity information between lanes (e.g., turn information), information about speed limits, information about drivability or utilization in different areas, etc. This layer is used to keep the vehicle in the lane and assist in maneuver planning.

b) Positioning layer: containing detectable objects of different sensor types. The current sensor information is compared to the layer to estimate the vehicle pose.

c) Dynamic layer: including dynamically altered information such as road status, traffic, weather, information about parking, etc. The layer is mainly used for the comfort function of an automated vehicle.

In addition to the improved accuracy compared to SD maps, HD maps should also generally be very real-time, i.e. have a high degree of accuracy in time and space.

The positioning layer contains geometric information for determining the position of the vehicle, while the planning layer contains semantic information required by the vehicle in order to be able to plan and implement driving maneuvers. Typical contents of the planning map are, for example:

-lane geometry, lane edges, centerline

Traffic lights, stop lines, traffic signs, rules of advance

Since the localization map or localization layer is essentially a geometrically correct representation of the surroundings of the vehicle, which is built up from the perception of a plurality of sensors or vehicles or drives, the localization map or localization layer can serve as a basis for creating a planning map. The steps performed in constructing the positioning map can basically be regarded as preconditions for deriving the planning map.

Because of potential sensor limitations (e.g., sensor field of view, measurement accuracy, etc.), mapping may not be performed in all areas to be mapped. And the precision of the mapping in certain areas may be significantly lower than in other areas, for example, even if it could in principle be realized.

A wide variety of environmental sensors (e.g., video, radar, laser radar, etc.) may be used to sensor ambient characteristics during travel and compare with characteristics of the locating layer of the digital map. Whereby the current vehicle pose (i.e., the position and orientation of the vehicle) can be estimated. For example, three-dimensional point clouds or semantic landmarks (road markings, traffic signs, light poles, etc.) can be used as environmental data and positioning features.

The accuracy requirements for positioning are typically very high (positioning errors relative to the map are up to 10 cm), and thus the accuracy requirements for the map are correspondingly high. A measuring vehicle with an expensive sensor device can be used to create an HD map in order to be able to achieve the required map accuracy.

Maps from mapping techniques created with crowd sourcing data may be used to generate a map of the surrounding environment from sensor observations of a single survey vehicle or large fleet of vehicles. Can be applied to the fields of robots, logistics, automobiles, aerospace, consumer goods and the like.

Automated vehicles rely heavily on map-based information to assist the automated vehicles primarily in path planning, perception, and condition understanding. These maps are often referred to as HD maps (as opposed to SD maps for traditional navigation tasks).

A method capable of performing digital map creation with improved performance is provided.

The proposed method advantageously provides data in the form of "reference points" for challenging areas in the digital mapping process.

An exemplary implementation of the proposed method is described in more detail below with reference to fig. 1.

The starting point of the proposed method is the map creation process 1.

In an optional step 10, a special area of the digital map is identified, for example, by a central computing unit or a computing unit of the vehicle. Such areas may be, for example, intersections that are high-demand and/or particularly wide roads with fewer structures at the road edges and/or areas with sparse surrounding features and/or areas with poor GNSS reception. For identification, the corresponding comparison values of the intersections that are required to be high and/or particularly wide roads with fewer structures at the road edges and/or regions with sparse surrounding features and/or regions with poor GNSS reception are stored in a data memory that can be accessed by a central computing unit or a computing unit of the vehicle. For example, such special areas have only inadequate and/or nonsensical data, in particular ambient data, with which a complete or sufficiently precise creation of a digital map cannot be achieved.

In a further optional step 20, the regions identified in optional step 10 are integrated, i.e. for example the regions (for example ambiguous or ambiguous) are cleaned off, so that the regions can be reduced in number, so that the overhead for subsequently activating the measuring vehicle can be kept as low as possible. As a result, there is an integrated list of special areas, in which the map quality can be considered low.

In step 30, the vehicle with the appropriate sensor and/or dGPS module, hereinafter referred to as a measuring vehicle, is driven into the previously identified area. For example, the central computing unit may transmit the location coordinates of the identified region to the vehicle. For this purpose, for example, so-called robotic taxis that do not carry passengers may be dispatched. Additional vehicle fleets may additionally or alternatively be dispatched from the central computing unit, thereby generating a combination of advantages from crowd-sourced and map-based vehicle fleets. The vehicles of the map-built vehicle fleet are hereinafter referred to as "survey vehicles".

The mentioned measuring vehicles are equipped with environmental sensor devices that are more valuable than fleet vehicles, and may also have additional sensor devices (e.g., lidar, radar, cameras, etc.), for example. In this way, the measuring vehicle is able to detect and provide more valuable, i.e. accurate, surrounding data of the vehicle surroundings than crowd-sourced fleet vehicles, for mapping purposes, for example by wireless transmission to a central computing unit. The central computing unit may be a mapping cloud. For example, a "mapping cloud" may be a high-performance capable electronic computer system that is used solely or for mapping purposes, among other purposes, to create and/or refine a digital map, particularly for autonomous vehicles.

In step 30, at least one measuring vehicle is driven into the region or regions integrated in step 20 and parked there, for example at least one location where a signal for position determination can be received by the vehicle. The signal for position determination may be, for example, a GNSS signal. Additionally or alternatively, the at least one measuring vehicle is driven into such an area: the overlap between the sensor field of view of the measuring vehicle and the sensor field of view of the fleet vehicle in these areas is available.

After the vehicle is measured to be parked, in step 40, the vehicle pose of the vehicle (the position of the vehicle and the angular position of the longitudinal axis are measured) is initialized by the computing unit of the vehicle. The determination of the orientation of the vehicle pose is performed, for example, by means of a plurality of GNSS antennas.

The position of the antenna relative to the vehicle coordinate system, in particular relative to the longitudinal axis of the vehicle, is known or can be determined separately from one another. The global orientation of the vehicle may then be determined based on the position of the antennas relative to the vehicle coordinate system and their absolute positions. For example, the GNSS signals and correction data can be processed over a plurality of measurement cycles in order to achieve a cm accuracy or sub-degree accuracy (subsegdgenauigkeit) of the vehicle position.

After the vehicle pose initialization is completed, ambient data is recorded by the vehicle with at least one sensor. The vehicle pose is a reference point for the ambient data measured later.

In step 50, the ambient data is used with the vehicle pose to create and/or refine a digital map with the ambient data. A high-precision local digital map of the surroundings of the measuring vehicle can thus be created from the measurement of the measuring vehicle, and thus a high-precision local digital map of the identified region can be created. For example, the ambient data may be used by the computing unit of the vehicle to create a high-precision digital map. In another embodiment, the vehicle pose of the measuring vehicle is output from the measuring vehicle to the central computing unit together with the detected ambient data. The central computing unit uses these high-precision vehicle poses and the ambient data detected with respect to the vehicle poses to create and/or refine a high-precision digital map at least in the identified regions. This can preferably be performed by means of a central computing unit (mapping cloud), or alternatively or additionally also at least partly by a computing unit measuring the vehicle.

The high-precision local map created in the measuring vehicle can then be transmitted to a central computing unit (map cloud) together with the high-precision pose. In the central computing unit, these ambient data and/or local maps can be used for aligning them with the imported crowd-sourced data and optimizing the region of the digital map, wherein the reference point is not changed here.

In the event that there is neither GNSS reception nor possibility of parking in the area reached for the measuring vehicle, the measuring vehicle detects the ambient data in a sensor-wise manner and creates a local map using the data, which map is then further transmitted to the map cloud for further processing.

The advantage of the "mobile reference point" produced by means of the proposed method compared to the data of the fleet vehicles is that the superordinate scheme for creating the digital map is still crowd-sourced at all times, and that the additionally required fleet of vehicles with measured vehicles can be smaller in size (e.g. using only available robotic taxis) compared to a dedicated mapping fleet.

Advantageously, the proposed method can be used for any kind of digital map or layer besides planning a map or planning a layer, for example for creating a positioning map or positioning layer.

If the measuring vehicle is parked in a position which can be detected by a sensor of a further vehicle, in particular of a fleet vehicle, the further computing unit of the further fleet vehicle can measure its own position and/or its own pose with respect to the measuring vehicle with the sensor. Thereby, the further vehicle is able to more accurately determine its own position and/or pose. These positions or poses can then be used in the mapping process. For example, such a scheme may be of interest for areas with poor GNSS reception. In this way, the measuring vehicle can be temporarily used as long as it is a parked and identifiable object for the fleet vehicle.

The proposed method can advantageously be implemented as software, for example running in a distributed manner in a central computing unit (drawing cloud) or in part in the cloud. In this way, a simple adaptability of the method is supported.

Fig. 2 shows a schematic flow of the proposed method for creating a digital map in a highly schematic way.

In an optional step 100, at least one region with insufficient data material for creating a digital map, in particular a high-precision digital map, is identified.

In step 110, data material for creating a digital map of the region identified in step a) is provided by means of the measurement of the at least one measuring vehicle. For this purpose, data material representing the ambient data can be transmitted to the central computing unit.

In step 120, a defined high-precision local map is created from the data of the at least one measuring vehicle by means of a computing unit or a central computing unit of the measuring vehicle.

In step 130, the created local map is used to create a digital map.

Positioning accuracy of 2 to 5 cm can be achieved by means of Differential Global Positioning System (DGPS). The reference station is used to: the position of the measuring station at a given time is determined and the deviations occurring there are transmitted to the position actually to be measured, the so-called Rover (roller). This may be done over the radio or internet. The distance between the base station and the rover, referred to as the baseline, should be no greater than 10 km. Correction of the position with the reference data can also be done in the later course of the measurement, which is called post-processing. This is advantageous when accurate positions are desired but these are not needed in real time. Another high performance correction method is Real Time Kinematic (RTK). RTKs reduce the measurement accuracy to less than one centimeter in level. This is achieved by combining two optimizations: the RTK uses not only the range determination by carrier phase but also the correction signal from the reference station. Provided that the receiver not only outputs the location through the interface but also has access to the original data.

In conjunction with the open source software RTKLIB, GPS resolution with centimeter accuracy has been enabled using conventional low cost single frequency receivers. The required reference data can be generated by itself by adding a second receiver at the precisely determined location, i.e. completely similar to the case of differential GPS.

Fig. 3 schematically shows a vehicle 1 with a computing unit 5, which is connected to a sensor 6, a positioning system 8 and a transmitting/receiving unit 7. The sensor 6 may be configured as a lidar sensor and/or a radar sensor and/or a video sensor. The at least one sensor is configured to detect ambient data in a predefined area in the vicinity of the vehicle. Video sensors identify objects such as lanes, buildings, bridges, traffic lights, trees, traffic signs, other vehicles, as well as bicycles and pedestrians. The stereo camera in the roof attachment is able to detect the surroundings up to 50 meters in three-dimensional view. The radar sensor recognizes objects up to a distance of 250 meters. The sensors determine the position of the objects and also measure the speed of movement of the objects relative to the own vehicle. These sensors thus provide important information about road traffic to the auxiliary system. Lidar sensors provide similar ambient data as radar sensors. The object and its relative position to the vehicle can thus be detected by means of these sensors.

The transmitting/receiving unit may be, for example, a mobile phone or a WLAN transmitting/receiving unit. Furthermore, a central computing unit 4, which is configured as a server unit, for example, in the cloud, is provided. The positioning system may be, for example, a GNSS positioning system, by means of which the pose of the vehicle can be determined. The computing unit 5 has a data memory in which the data of the sensors can be stored. Furthermore, data of digital maps, in particular road maps and ambient maps, can be stored in the data memory.

The central computing unit has a data memory in which data of a digital map, i.e. a road map and a surrounding map, are stored. Furthermore, a further vehicle 3 is provided, which likewise has a computing unit, sensors, a positioning system and a transmitting/receiving unit. The other vehicle's sensor and/or positioning system may have a lower quality than the vehicle's 2 sensor and/or positioning system. The vehicle 2 is thus a measuring vehicle and the other vehicle is a vehicle of a fleet of vehicles.

Furthermore, a computer program product with program code means for performing the method is shown in fig. 3.

Claims

1. A method for providing surroundings data to a digital map by means of a vehicle, the method having the following steps:

driving the vehicle into a predetermined region arranged within or adjacent to a map region for which insufficient ambient data is present in particular in the digital map;

detecting ambient data by means of at least one sensor of the vehicle;

providing pose of parked vehicles and the ambient data for creating and/or refining the digital map.

2. The method of claim 1, wherein the vehicle is first parked in the predetermined area, wherein a pose of the parked vehicle is subsequently determined.

3. The method according to any of the preceding claims, wherein the pose of the vehicle is determined by means of GNSS signals.

4. Method according to any of the preceding claims, wherein the pose of the vehicle is determined by means of a plurality of measurement cycles, and/or wherein a correction method is used for refining the pose of the vehicle.

5. The method according to any of the preceding claims, wherein the pose of the vehicle is solved to an accuracy of several centimeters and/or less than one degree.

6. The method according to any of the preceding claims, wherein a local ambient map is detected as the ambient data.

7. Method according to any of the preceding claims, wherein a further vehicle finds itself in relation to the vehicle, wherein the further vehicle detects further ambient data by means of at least one further sensor, wherein the further vehicle provides the relative position, in particular the relative pose, and the further ambient data in relation to the vehicle for creating and/or refining the digital map.

8. The method according to any of the preceding claims, wherein the vehicle or the further vehicle is an autonomously driven vehicle, in particular a robotic taxi, wherein the autonomous vehicle drives into the predetermined area and detects ambient data, in particular when no person is in the vehicle.

9. Method for creating and/or refining a digital map by means of ambient data, wherein the pose of a vehicle and the ambient data relative to the pose of the vehicle are received from a central computing unit, wherein the central computing unit records the ambient data corresponding to the pose of the vehicle at least partially into the digital map, and wherein the refined digital map is output at least partially to the vehicle, wherein the central computing unit finds a map region in the digital map in which insufficient ambient data is present, wherein the central computing unit outputs instructions to the vehicle: driving into a predetermined area within or adjoining the map area, recording ambient data and transmitting the ambient data to the central computing unit.

10. A computing unit (4, 5) configured for implementing the method according to any one of the preceding claims.

11. A system for creating a digital map, the system being configured for implementing the method according to any one of the preceding claims 1 and 9.

12. A computer program product (11) having program code means for performing the method according to any of claims 1 to 9 when the computer program product is run on a computing unit (4, 5).