DE102020212771A1 - Computer-implemented method and computer program for trajectory planning for an automated system and computing system embedded in an automated system for trajectory planning and/or regulation and/or control - Google Patents

Abstract

Computer-implemented method for trajectory planning for an automated system comprising the steps in offline processing of high-definition maps (HD maps) calculation of geometric figures comprising points of sections of straight lines (g), curves (cur), spirals (sp) and polynomials (poly) and orientations (ψ) of these sections and storing the figures together with grid cell information and information links between individual ones of the figures in a raster format in an external memory (12) (V1), in online processing selecting the figures from the raster format depending on the position data of the automated system, the grid cell information and the information connections and transferring the selected figures to an internal memory (14) of the automated system (V2), in a computing system (10) of the automated system connecting the selected figures and generating lanes ( FS) (V3).

Description

The invention relates to a computer-implemented method and a computer program for trajectory planning for an automated system and a computing system embedded in an automated system for trajectory planning and/or regulation and/or control.

CN 106441319 discloses a system and method for generating lane-level navigation maps for unmanned vehicles. Grid maps with a width of 500 and a height of 750 pixels are transmitted, with each grid representing a square with a side length of 20 cm of the current street scene.

WO 2018 104 563 A2 discloses a system and method for determining a geographic location and orientation of a vehicle. High-resolution maps are generated from video recordings.

U.S. 2017 010 617 A1 discloses a system for autonomously navigating a vehicle along a road segment based on an intermediate map.

High-resolution maps, or high-definition maps, or HD maps for short, are a prerequisite for automated systems for highly automated driving. HD maps are known with centimeter-precise resolution of all road details, including, for example, 3D poses of all road geometries, traffic signs, road markings, lighting systems and pedestrian crossings. Usually, the HD maps are continuously fed to a vehicle via mobile communications via a backend connection, for example via a V2V / V2X ECU. This results in a large memory requirement for storing and using HD maps and a high bandwidth for transferring HD maps.

The object of the invention was to improve the handling of HD cards, in particular to make them more efficient.

The objects of claims 1, 9 and 10 solve this problem.

• • in einer Offline-Verarbeitung von High-Definition-Karten Berechnen von geometrischen Figuren umfassend Punkte von Abschnitten von jeweils Geraden, Kurven, Spiralen und Polynomen und Orientierungen dieser Abschnitte und Speichern der Figuren zusammen mit Gitterzelleninformationen und Informationsanbindungen zwischen einzelnen der Figuren in einem Rasterformat in einem externen Speicher,
• • in einer Online-Verarbeitung Auswählen der Figuren aus dem Rasterformat in Abhängigkeit von Positionsdaten des automatisierten Systems, der Gitterzelleninformationen und der Informationsanbindungen und Übertragen der ausgewählten Figuren in einen internen Speicher des automatisierten Systems,
• • in einem Rechensystem des automatisierten Systems Verbinden der ausgewählten Figuren und Erzeugen von Fahrspuren.

In one aspect, the invention provides a computer-implemented method for trajectory planning for an automated system. The procedure includes the steps

• • in an offline processing of high-definition maps, computing geometric figures comprising points of sections of straight lines, curves, spirals and polynomials, respectively, and orientations of these sections and storing the figures together with grid cell information and information links between individual of the figures in a raster format an external memory,
• • in online processing, selecting the figures from the grid format depending on the position data of the automated system, the grid cell information and the information links and transferring the selected figures to an internal memory of the automated system,
• • in a computing system of the automated system, connecting the selected figures and creating lanes.

• • wenigstens ein Modul, das ausgeführt ist, die Schritte des erfindungsgemäßen Verfahrens auszuführen,

oder

• • eine Schnittstelle zu einem externen Speicher, in dem geometrische Figuren umfassend Punkte von Abschnitten von jeweils Geraden, Kurven, Spiralen und Polynomen und Orientierungen dieser Abschnitte aus High-Definition-Karten extrahiert und zusammen mit Gitterzelleninformationen und Informationsanbindungen zwischen einzelnen der Figuren in einem Rasterformat gespeichert sind,
• • ein zweites Modul, das in einer Online-Verarbeitung Positionsdaten verarbeitet und das System lokalisiert,
• • einen internen Speicher,
• • ein als Kommunikationsmodul ausgebildetes drittes Modul, das in Abhängigkeit der Lokalisierung des Systems ausgewählte Figuren aus dem externen Speicher in den internen Speicher überträgt und
• • ein viertes Modul, das basierend auf den ausgewählten Figuren Fahrspuren generiert.

According to a further aspect, the invention provides a computing system that can be embedded in an automated system for trajectory planning and/or regulation and/or control. The computing system includes

• • at least one module that is designed to carry out the steps of the method according to the invention,

or

• • an interface to an external memory in which geometric figures comprehensively extracts points from sections of lines, curves, spirals and polynomials, respectively, and orientations of these sections from high-definition maps and stores them in a raster format together with grid cell information and information connections between individual figures are,
• • a second module that processes position data online and localizes the system,
• • an internal memory,
• • a third module designed as a communication module, which transfers selected figures from the external memory to the internal memory depending on the localization of the system and
• • a fourth module that generates lanes based on the selected characters.

• • aus einem externen Speicher, in dem geometrische Figuren umfassend Punkte von Abschnitten von jeweils Geraden, Kurven, Spiralen und Polynomen und Orientierungen dieser Abschnitte aus High-Definition-Karten extrahiert und zusammen mit Gitterzelleninformationen und Informationsanbindungen zwischen einzelnen der Figuren in einem Rasterformat gespeichert sind, in Abhängigkeit von Positionsdaten des automatisierten Systems, der Gitterzelleninformationen und der Informationsanbindungen einzelne der Figuren in einen internen Speicher des Computers übertragt und
• • der Computer basierend auf den einzelnen Figuren Fahrspuren generiert,

wenn das Programm auf dem Computer läuft.According to a further aspect, the invention provides a computer program for trajectory planning for an automated system. The computer program includes instructions that cause a computer

• • from an external memory in which geometric figures comprehensively extracts points of sections of straight lines, curves, spirals and polynomials, respectively, and orientations of these sections from high-definition maps and are stored in a raster format together with grid cell information and information links between individual ones of the figures, depending on position data of the automated system, the grid cell information and the information links, transmit individual ones of the figures to an internal memory of the computer and
• • the computer generates lanes based on the individual figures,

when the program is running on the computer.

Advantageous refinements of the invention result from the definitions, dependent claims, the drawing and the description of preferred exemplary embodiments.

Automated systems include automated vehicles, such as cars, trucks, buses, people movers, transportation systems and mobile platforms, especially fully automated or autonomous, and robots, such as robots that deliver mail, especially fully automated or autonomous.

A yaw angle, for example, serves as orientation for the respective sections.

According to one aspect of the invention, the external memory is implemented as a server or a cloud-based memory. An intermediate map, also known as a spare map, is stored in the external memory. The intermediate map does not contain the full information content of the HD map, but only the geometric figures extracted from the HD map.

Online processing refers to processing of data during a drive of the automated system.

The computing system can be embedded or embedded, for example, in an electronic control unit, also known as an electronic control unit, or ECU for short, of the automated system, for example in an autonomous driving ECU of an automated vehicle. For example, when used, the computing system exchanges signals with a data bus of the automated system, for example with a CAN bus or FlexRay. According to one aspect of the invention, the computing system is a system on the chip, SoC for short, comprising one or more CPUs, GPUs, FGPAs, ASICs, data buses, memory modules, communication interfaces and/or other electronic components.

The module or modules of the computing system according to one aspect of the invention are hardware modules in the form of individual electronic components, for example individual ASICS or ICs. The first module is, for example, a geometry extraction module, ie a separate computing unit in the form of an electronic component that is designed to calculate and extract geometric figures in HD maps. Accordingly, the second module is a localization module that, for example, position data of the automated system, including data from satellite navigation systems, such as NAVSTAR GPS (USA), GLONASS (Russia), Galileo (EU), Beidou (China), Indian Regional Navigation Satellite System ( India), quasi-centi-satellite system (Japan), and reads automated system odometry data and determines the position of the automated system. Correspondingly, the third module is a communication module and the fourth module is a lane generator module. According to a further aspect of the invention, the individual modules are corresponding software modules and/or the hardware modules execute software corresponding to their task or function.

The instructions of the computer program according to the invention include machine instructions, source text or object code written in assembly language, an object-oriented programming language, such as C++, or in a procedural programming language, such as C. The computer program is fully or partially on the automated system or in a remote system comprising a cloud executed.

The advantage of the subject matter according to the invention is that less data has to be transmitted in order to generate lanes and plan the trajectory. According to the invention, only selected figures and no complete or complete HD maps or excerpts from HD maps are transmitted. This reduces the information content of HD maps while maintaining the same level of accuracy in lane generation. Furthermore, the information content can be stored in internal memories with a comparatively small storage volume. According to the invention, data transmissions with a comparatively low bandwidth are thus possible. This makes communication between a backend, for example a server, for example a cloud-based server, and the automated system seamless and inexpensive. The subject of the invention enables rapid handling of information from HD maps while driving by automated systems, highly accurate generation of lanes while driving without pre-processing and data transmission with comparatively low latency times due to the lower data volume transmitted.

• • bei Geraden, Kurven, Spiralen und/oder Polynomen Punkte von jeweiligen Abschnitten in globalen und lokalen Koordinaten, Längen der jeweiligen Abschnitte, Fahrspur-Kennung, Punkt-Kennung in einem Gitter, Gitter-Kennung, Kennung von vorangehenden Punkten und/oder Kennung von nachfolgenden Punkten,
• • bei Kurven jeweils eine Krümmung und/oder bei Spiralen jeweils die Krümmung bei einem ersten Punkt der Spirale und/oder die Krümmung bei einem zweiten Punkt der Spirale und/oder
• • bei Polynomen jeweils die Polynomkoeffizienten gespeichert.

According to one aspect of the invention

• • in the case of straight lines, curves, spirals and/or polynomials, points of respective sections in global and local coordinates, lengths of the respective sections, lane identifier, point identifier in a grid, grid identifier, identifier of previous points and/or identifier of following points,
• • in the case of curves, in each case a curvature and/or in the case of spirals, in each case the curvature at a first point of the spiral and/or the curvature at a second point of the spiral and/or
• • In the case of polynomials, the polynomial coefficients are saved.

For example, the first point of the spiral is the starting point of the spiral. For example, the second point of the spiral is the endpoint of the spiral.

The lane identifier or lane ID is the identifier of the lane from the HD map, analogous to the point identifier or point or pose ID. Accordingly, the grid identifier or grid ID is the unique identifier of a grid or grid element of the grid format. The grid ID enables a faster search for connection points of the individual geometric figures. The Lane ID, Point ID and Grid ID are examples of grid cell information. The identification of preceding points and/or the identification of subsequent points are examples of the information links according to the invention. Due to the information connections, divisions or splits of road segments are known. This information connects the individual sections of the extracted geometric figures point by point in order to generate complete lanes.

This type of information storage represents a new and advantageous structure for data storage of HD maps and enables lossless compression of the information content of HD maps for trajectory planning. Furthermore, this information storage represents an efficient format for storing HD maps and enables rapid caching in the automated driving systems.

In a further refinement, the trajectory planning is based on a lane center or on lane boundaries. In the case of the middle of the lane, the lane widths of the respective sections are saved, and in the case of lane boundaries, the lane widths and identifiers of neighboring points are saved. This enables trajectory planning in two alternative ways.

According to a further aspect, discretization levels of the raster format can be selected. The selected figures are transferred in an application-specific discretization. The discretization is thus scalable. For example, automated and/or autonomous driving in rural areas does not need an HD map with high information content every half meter, whereas in urban areas such a discretization is required. For example, a discretization of 1 m distance between two grid points is selected for driving over country or on freeways. For example, for driving in urban environments, a discretization of 0.5 m distance between two grid points is chosen. Automated and/or autonomous robots that deliver parcels and other mail, for example, require highly accurate information on scales below half a meter. For such applications, for example, a discretization of 0.2 m distance between two grid points is selected. Due to the selectable discretization levels, the subject of the invention can be used for a number of road applications.

According to a further aspect of the invention, traffic signs, lane markings and/or orientation points are stored in the raster format and are transmitted as a function of the selected figures. This improves trajectory planning. According to one aspect of the invention, the traffic signs, lane markings and/or orientation points are stored as additional information links.

According to a further aspect of the invention, trajectories for the automated system are determined and/or objects are predicted based on the lanes generated. This improves trajectory planning for the automated system.

According to a further aspect of the invention, the automated system is regulated and/or controlled based on the generated lanes. For example, actuators for longitudinal and/or lateral guidance of the automated system are regulated and/or controlled in such a way that the automated system drives in the lanes generated.

According to a further aspect of the invention, the high-definition maps are created internally by sensor data fusion of the automated system or are provided externally. For example, lidar, radar and/or camera data from the automated system are merged or an HD map is provided by a third party.

A further aspect of the invention is a computer-readable data carrier for trajectory planning tion for an automated system on which the computer program according to the invention is stored. The computer-readable data carrier is, for example, a random access memory, a read-only memory, a flash memory, a USB stick, an SD card, a CD, DCD or BluRay. If the automated system is not equipped with the object according to the invention when it is delivered, it can be retrofitted using the computer-readable data carrier.

Another aspect of the invention is a data carrier signal for trajectory planning for an automated system. The data carrier signal transmits the computer program according to the invention. The data carrier signal can be used to download the computer program, for example via a network, for example via the Internet, using radio technology, for example WLAN, and software-as-a-service.

• 1 eine schematische Darstellung eines Ausführungsbeispiels des erfindungsgemäßen Verfahrens und eines Ausführungsbeispiels des erfindungsgemäßen Rechensystems,
• 2 Ausführungsbeispiele geometrischer Figuren,
• 3 ein Ausführungsbeispiel von erfindungsgemäß verbundenen geometrischen Figuren aus 2,
• 4 ein Ausführungsbeispiel einer High-Definition-Karte,
• 5 ein Ausführungsbeispiel der High-Definition-Karte aus 4 nach erfindungsgemäßer Extraktion von geometrischen Figuren und
• 6 eine detailliertere schematische Darstellung der 1.

The invention is illustrated in the following exemplary embodiments. Show it:

• 1 a schematic representation of an embodiment of the method according to the invention and an embodiment of the computing system according to the invention,
• 2 Examples of geometric figures,
• 3 an exemplary embodiment of geometric figures connected according to the invention 2 ,
• 4 an embodiment of a high-definition card,
• 5 an embodiment of the high definition card 4 after extraction of geometric figures according to the invention and
• 6 a more detailed schematic of the 1 .

In the figures, the same reference symbols denote the same or functionally similar reference parts. For the sake of clarity, only the relevant reference parts are highlighted in the individual figures.

According to the overview of 1 Using the method according to the invention and the computing system 10 according to the invention, a high-definition map HD map is pre-processed offline in a method step V1 in a first module 11. The first module 11 is a geometry extraction module that extracts geometric from the HD map Figures are calculated and extracted according to the invention, ie with the relevant points, orientations and information connections. The geometric figures are stored in an external memory 12 in a raster format. The external storage 12 is a cloud storage, for example.

While the automated system is moving, i.e. online, in a method step V2 the automated system is localized with the aid of a second module 13 in the form of a localization module and, based on the localization from the external memory 12, those extracted figures are selected whose position in correspond to the grid format of the position of the automated system or that are located in an adjustable environment of the position of the automated system based on the information bindings. Here, global coordinates x, y to indicate the position of the automated system and local coordinates s, t for referencing in the HD map and the intermediate map are transformed into one another. The selected geometric figures are transferred from the external memory 12 to an internal memory 14 of the automated system via a third module 15, which represents a communication module, for example a radio module. Since only certain geometric figures and not the full information content of the HD card is transmitted, lower bandwidths for data transmission and lower latency times for data transmission are possible here and comparatively small internal memories 14.

The computing system 10 of the automated system also includes a fourth module 16 designed as a track generator which, in a method step V3, connects the transmitted geometric figures using the grid cell information and the information connections and generates the lanes FS depending on the set discretization level. The lanes FS generated according to the invention are made available to a trajectory planning in order to plan a mission of the automated system and/or to predict objects on predicted paths. According to one aspect of the invention, the trajectory planning is a software or hardware module of the computing system 10. Alternatively, the computing system 10 includes an interface to the trajectory planning.

• • ein Punkt eines Abschnittes s einer Geraden g mit Koordinaten x, y und Gierwinkel ψ,
• • ein Punkt eines Abschnittes s einer Spiralen sp mit Koordinaten x, y und Gierwinkel ψ,
• • ein Punkt eines Abschnittes s einer Kurve cur mit Koordinaten x, y und Gierwinkel ψ,
• • ein Punkt eines Abschnittes s eine s Polynoms poly mit Koordinaten x, y und Gierwinkel ψ.

2 shows exemplary geometric figures in global coordinates x, y and in local coordinates s, t. Shown are for example:

• • a point of a section s of a straight line g with coordinates x, y and yaw angle ψ,
• • a point of a section s of a spiral sp with coordinates x, y and yaw angle ψ,
• • a point of a section s of a curve cur with coordinates x, y and yaw angle ψ,
• • a point of a section s of a s polynomial poly with coordinates x, y and yaw angle ψ.

• • Gerade g: [x, y, ψ, s, Länge des Abschnitts, Spur-ID, Punkt-ID, Grid-ID, Vorgänger-ID, Nachgänger-ID, weitere Eigenschaften],
• • Spirale sp: [x, y, ψ, s, Länge des Abschnitts, Spur-ID, Punkt-ID, Grid-ID, Vorgänger-ID, Nachgänger-ID, weitere Eigenschaften, Start der Krümmung, Ende der Krümmung],
• • Kurve cur: [x, y, ψ, s, Länge des Abschnitts, Spur-ID, Punkt-ID, Grid-ID, Vorgänger-ID, Nachgänger-ID, weitere Eigenschaften, Krümmung],
• • Polynom poly: [x, y, ψ, s, Länge des Abschnitts, Spur-ID, Punkt-ID, Grid-ID, Vorgänger-ID, Nachgänger-ID, weitere Eigenschaften, a_i].

According to the invention, the individual geometric figures are stored with the following information and displayed in the further course of the method:

• • Line g: [x, y, ψ, s, length of segment, track ID, point ID, grid ID, predecessor ID, successor ID, other properties],
• • Spiral sp: [x, y, ψ, s, length of section, track ID, point ID, grid ID, predecessor ID, successor ID, other properties, start of curvature, end of curvature],
• • Curve cur: [x, y, ψ, s, length of section, track ID, point ID, grid ID, predecessor ID, successor ID, other properties, curvature],
• • Polynomial poly: [x, y, ψ, s, length of section, track ID, point ID, grid ID, predecessor ID, successor ID, other properties, a_i].

where a_i are the polynomial coefficients of the polynomial P(x)=a_0 + a_1 *x + a_2*x^2 + ... + a_n*x^n, n>=0.

3 shows an example of how individual geometric figures are connected at the corresponding points of the geometric figures. The respective points define a change in a property of the lane FS.

4 represents a normal HD map. In the prior art, the entire information content of this HD map is sent to a planning module for trajectory planning and environment prediction. However, this information content is redundant since the majority of the points belonging to a track 1, track 2 and track 3 element have the same mathematical properties.

According to the invention, only the points at which the properties of the geometric figures change are sent to the internal memory 14 of the automated system, as shown in FIG 5 is shown. Point ID: 1 has the property of a straight line g, Point ID: 2 has the property of a curve cur and Point ID: 3 has the property of a straight line. Each point is represented with the type of particular geometric figure of which it denotes a section, the length of the particular section and the grid ID. This information is using the first module 11 in the form of the geometry extraction module from the HD map 4 extracted.

Grid cells in the grid format are selected from the external memory 12 based on the location data. If the automated system is in the vicinity of the point PointID:1, in one aspect a radius for an environment around the respective location point being selectable, for example depending on the speed of the automated system, grid cells around the point point- ID: 1 selected around application specific. For example, driving on a freeway requires a larger number of grid cells, for example 100x200, compared to driving in an urban area, for which a number of for example 100x100 is sufficient. According to one aspect of the invention, the number of grid cells of the grid format to be selected is another configurable parameter of the subject matter of the invention.

For example, the internal memory 14, which is, for example, a local intermediate memory or cache memory, only the in 5 information shown. The fourth module 16 of the computing system 10, which is designed as a lane generator module, for example, uses this information from the internal memory 14, see also 6 , to generate the lanes FS, which are like those in 4 lane shown.

Reference List

HD map

High definition map

10

computing system

11

first module

12

external storage

13

second module

14

Internal memory

15

third module

16

fourth module

G

Just

cur

Curve

sp

spiral

poly

polynomial

x, y

global coordinates

s, t

local coordinates

ψ

yaw angle

track i

Lane identifier, i = 1, 2, 3

Point ID: i

Point identifier, i = 1, 2, 3

Grid ID

grid identifier

FS

lane

V1-V3

process steps

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• CN106441319 [0002]
• WO 2018104563 A2 [0003]
• US2017010617A1[0004]

Claims (10)

Computer-implemented method for trajectory planning for an automated system, comprising the steps • in an offline processing of high-definition maps (HD map) calculation of geometric figures comprising points of sections of straight lines (g), curves (cur), spirals (sp) and polynomials (poly) and orientations (ψ) of these sections and storing the figures together with grid cell information and information connections between individual figures in a grid format in an external memory (12) (V1), • in online processing, selecting the figures from the grid format depending on the position data of the automated system, the grid cell information and the information connections and transferring the selected figures to an internal memory (14) of the automated system (V2), • in a computing system (10) of the automated system, connecting the selected figures and generating lanes (FS) (V3). procedure after claim 1 , where • for straight lines (g), curves (cur), spirals (sp) and/or polynomials (poly) points of respective sections in global (x, y) and local coordinates (s, t), lengths of the respective sections, Lane identifier (Lane 1, Lane 2, Lane 3), Point identifier (Point ID: 1, Point ID: 2, Point ID: 3) in a grid, Grid identifier (Grid ID), Identifier of preceding points and/or identifiers of subsequent points are stored, • in the case of curves, a curvature and/or in the case of spirals, the curvature at a first point of the spiral and/or the curvature at a second point of the spiral are stored and/or • in the case of polynomials, the polynomial coefficients are stored in each case. procedure after claim 2 , wherein the trajectory planning is based on a lane center or on lane boundaries, wherein the lane widths of the respective sections are stored in the case of the lane center and the lane widths and identifiers of neighboring points are stored in the case of lane boundaries. Procedure according to one of Claims 1 until 3 , where discretization levels of the raster format can be selected and the selected figures are transmitted in an application-specific discretization. Procedure according to one of Claims 1 until 4 , wherein traffic signs, lane markings and/or landmarks are stored in the raster format and are transmitted as a function of the selected figures. Procedure according to one of Claims 1 until 5 , wherein trajectories for the automated system are determined and/or objects are predicted based on the generated lanes (FS). Procedure according to one of Claims 1 until 6 , wherein the automated system is regulated and/or controlled based on the generated lanes (FS). Procedure according to one of Claims 1 until 7 , where the high-definition maps are created internally by sensor data fusion of the automated system or provided externally. Computing system (10) embeddable in an automated system for trajectory planning and / or regulation and / or control comprising • at least one module that is executed, the steps of the method according to one of Claims 1 until 8th to execute, or • an interface to an external memory (12), in which geometric figures comprising points of sections of straight lines (g), curves (cur), spirals (sp) and polynomials (poly) and orientations of these sections from high- definition maps are extracted and stored in a raster format together with grid cell information and information links between individual figures, • a second module (13), which processes position data in online processing and localizes the system, • an internal memory (14), • a third module (15) designed as a communication module, which transfers selected figures from the external memory (12) to the internal memory (14) depending on the localization of the system, and • a fourth module (16), which calculates lanes based on the selected figures (FS) generated. Computer program for trajectory planning for an automated system comprising commands that cause a computer • from an external memory (12) in which geometric figures comprising points of sections of straight lines (g), curves (cur), spirals (sp) and Polynomials (poly) and orientations of these sections are extracted from high-definition maps and stored in a raster format along with grid cell information and information links between individual ones of the figures, depending on automated system position data, the grid cell information and information links individual figures in an internal memory (14) of the computer and • the computer generates lanes (FS) based on each character when the program is running on the computer.

Images