Methods and systems for generating and updating digital maps

Abstract

This invention relates to a method and system for generating and updating digital maps. The digital map is generated and updated by at least one road vehicle using multiple lanes along a road segment. Each road vehicle includes a perception system with at least one sensor configured to monitor the vehicle's surroundings. The method includes: acquiring positioning data and sensor data for each lane from the at least one road vehicle; forming a sub-map representation of the surroundings at each acquired longitudinal position based on the acquired sensor data, and estimating a longitudinal error for each acquired longitudinal position within each road segment; determining multiple new longitudinal positions for each road vehicle in each lane by applying the estimated longitudinal error to each corresponding acquired longitudinal position; and applying the determined multiple new longitudinal positions to the associated sensor data to generate a first layer of a map representation of the surroundings along the road segment.

Description

[0001] Cross-references to related applications

[0002] This application claims priority to European Patent Office application No. 19218793.8, filed on December 20, 2019, entitled “Method and System for Generating and Updating Digital Maps,” which has been assigned to its assignee and is expressly incorporated herein by reference. Technical Field

[0003] This disclosure relates to methods and systems for generating and updating digital maps, and vehicles navigating using such maps. Specifically, this disclosure relates to the detection source of digital maps based on longitudinal error estimation of attitude sequences. Background Technology

[0004] Over the past few years, the development of autonomous vehicles has been rapid, and many different solutions are being explored. Today, the development of both autonomous driving (AD) and advanced driver assistance systems (ADAS) (i.e., semi-autonomous driving) is taking place in several different technological areas within these fields. One such area is how to consistently and accurately position the vehicle, as this is a crucial safety aspect when the vehicle is driving in road traffic.

[0005] Therefore, maps have become a crucial component of autonomous vehicles. The question is no longer whether they are useful, but how to create and maintain maps efficiently and scalably. In the future of the automotive industry, and specifically for autonomous driving, it is foreseeable that maps will become the input for localization, planning, and decision-making tasks, rather than human interaction.

[0006] The traditional approach to simultaneously solving the mapping and localization problems is to use Simultaneous Localization and Mapping (SLAM) techniques. However, SLAM methods do not perform well in practical applications. Limitations and noise in the sensor input propagate from the mapping stage to the localization stage and vice versa, resulting in inaccurate mapping and localization. Therefore, new, accurate, and sustainable solutions are needed to meet the requirements for precise localization.

[0007] Therefore, precise location maps have so far been primarily created by survey vehicles using high-precision devices, specifically expensive high-precision Global Positioning System (GPS) sensors (such as so-called Real-Time Kinematic (RTK) GPS sensors) with centimeter-level accuracy. However, with the continuous changes in road networks, this costly process is not well-suited to the growing demand for continuously updated maps of the target environment for AD (Advanced Location and Design). Therefore, this method is no longer available when maps require rapid and continuous updates.

[0008] To address this issue, it has been suggested that fleets use consumer-grade satellite positioning devices (such as the consumer-grade GPS devices available in most cars) to crowdsource maps. However, using consumer-grade satellite positioning devices introduces uncertainty relative to the actual location of each vehicle. This, in turn, will cause any maps generated and / or updated based on data from such positioning devices to inherit their inaccuracies.

[0009] Therefore, new and improved solutions are needed that provide more accurate, robust and scalable means for generating and updating maps suitable for autonomous driving applications. Summary of the Invention

[0010] Therefore, the purpose of this disclosure is to provide a method for generating and updating digital maps, a control system for generating and updating digital maps, and a vehicle for navigation using these maps, which mitigates all or at least some of the disadvantages of currently known solutions.

[0011] This objective is achieved by providing a method for generating and updating digital maps, a control system for generating and updating digital maps, and a vehicle for navigation using these maps, as defined in the appended claims. The term "exemplary" is understood in this context to mean instance, example, or illustration.

[0012] This disclosure is based, at least in part, on the understanding that consumer-grade systems employed in most modern vehicles today (such as consumer-grade GPS, IMU, and other peripheral systems) can estimate lane departure (lane lateral deviation) and vehicle heading in a relatively accurate manner. This can be achieved, for example, using high-definition maps along with standard automotive GPS and vision-based lane trackers. However, estimation of "along the road" (i.e., longitudinal position on the road) is often difficult to estimate accurately by consumer-grade systems, and uncertainties on the order of 10 meters are not uncommon. Therefore, the inventors recognized that this longitudinal error can be estimated by using data (position data and perception data) from several channels traversed by one or more vehicles, and by running the data through an optimized algorithm. Accurate digital maps can then be generated / updated in a scalable and cost-effective manner by applying the correct offset (based on the longitudinal error) to each "snapshot" image of the surrounding environment.

[0013] According to a first aspect of this disclosure, a method is provided for generating and updating a digital map by using multiple lanes along a road segment by at least one road vehicle. Each road vehicle includes a perception system having at least one sensor configured to monitor the vehicle's surrounding environment. The method includes acquiring location data and sensor data for each lane from the at least one road vehicle. The location data includes multiple longitudinal positions of each lane within multiple road segments of the road segment, and the sensor data includes information about the surrounding environment of each road vehicle at each longitudinal position. Further, the method includes forming a sub-map representation of the surrounding environment at each acquired longitudinal position based on the acquired sensor data, and estimating a longitudinal error for each acquired longitudinal position within each road segment. This estimation is accomplished by performing digital optimization on the longitudinal error of each acquired longitudinal position within a common road segment based on a similarity level from the formed sub-map representations of common road segments. Further, the method includes determining multiple new longitudinal positions for each road vehicle in each lane by applying the estimated longitudinal error to each corresponding acquired longitudinal position, and applying the determined multiple new longitudinal positions to associated sensor data to generate a first layer of a map representation of the surrounding environment along the road segment.

[0014] This method provides a cost-effective and scalable means for creating and updating digital maps, especially high-definition maps. Furthermore, longitudinal error estimation can be used to generate and update several different layers of high-definition maps. For example, even if radar or lidar data is used to estimate the longitudinal error, subsequent error calculation can be used to accurately locate, for example, various landmarks (e.g., specific road signs at their current longitudinal position along a road section) by applying the longitudinal error estimation to image data from one or more cameras from the vehicle.

[0015] In this context, longitudinal is understood as a location “along the road”, that is, a location along the main extension of the road.

[0016] In some embodiments of this disclosure, the similarity level of the sub-map representation formed within a common road segment includes an entropy function, and wherein numerical optimization is configured to minimize the entropy function.

[0017] Furthermore, in some embodiments of this disclosure, the similarity level of the formed sub-map representation within the public road segment includes the edge size within the formed sub-map representation, and wherein digital optimization is configured to maximize the edge size.

[0018] Furthermore, according to an exemplary embodiment of this disclosure, the digital optimization is further based on a (first) condition, which defines that the sum of the longitudinal errors from multiple lanes within a common road segment is approximately zero. Therefore, the solution of the digital optimization, where the longitudinal errors from multiple lanes within the common road segment deviate from zero, is associated with an increased cost factor. Here, it is assumed that the resulting optimization problem (solved by a digital optimization algorithm) includes a cost function. In other words, the first condition defines that different lanes of one or more vehicles typically have different longitudinal errors within the same road segment, but their average value is approximately zero.

[0019] Furthermore, according to another exemplary embodiment of this disclosure, the digital optimization is further based on a (second) condition, which defines that the longitudinal error within a single segment of a corridor is (approximately / substantially) the same for all acquired longitudinal positions within that single segment. In other words, the second condition is based on the assumption that the deviation of the longitudinal error of a specific segment of a particular corridor is approximately zero.

[0020] According to a second aspect of this disclosure, a (non-transient) computer-readable storage medium is provided for storing one or more programs configured to be executed by one or more processors of a processing system, the one or more programs including instructions for performing methods according to any embodiment of the present disclosure. Similar advantages and preferred features exist for this aspect of the disclosure as for the first aspect previously discussed.

[0021] As used herein, the term "non-transient" is intended to describe computer-readable storage media (or "memory") that exclude the propagation of electromagnetic signals, but is not intended to further limit the type of physical computer-readable storage device included by the phrase computer-readable media or memory. For example, the terms "non-transient computer-readable media" or "tangible memory" are intended to include types of storage devices that do not necessarily permanently store information, including, for example, random access memory (RAM). Program instructions and data stored in a non-transient form on a tangible computer-readable storage medium can be further transmitted via communication media or signals (e.g., electrical signals, electromagnetic signals, or digital signals), which can be transmitted via communication media (such as networks and / or wireless links). Therefore, as used herein, the term "non-transient" is a limitation on the medium itself (i.e., tangible, not signaling), not on the persistence of data storage (e.g., RAM or ROM).

[0022] According to a third aspect of this disclosure, a control system is provided for generating and updating a digital map by using multiple lanes along a road segment by at least one road vehicle. Each road vehicle includes a perception system having at least one sensor configured to monitor the surrounding environment of the road vehicle. The control system includes control circuitry configured to acquire positioning data and sensor data for each lane from the at least one road vehicle. The positioning data includes multiple longitudinal positions of each lane within multiple segments of the road segment, and the sensor data includes information about the surrounding environment of each road vehicle at each longitudinal position. The control circuitry is further configured to form a sub-map representation of the surrounding environment at each acquired longitudinal position based on the acquired sensor data, and to estimate a longitudinal error for each acquired longitudinal position. The longitudinal error within each segment is estimated by performing digital optimization on the longitudinal error of each acquired longitudinal position within a common road segment based on a similarity level from the formed sub-map representations of common road segments. The control circuitry is further configured to determine multiple new longitudinal positions for each road vehicle in each lane by applying the estimated longitudinal error to each corresponding acquired longitudinal position, and to apply the determined multiple new longitudinal positions to associated sensor data to generate a first layer of a map representation of the surrounding environment along the road section. Similar advantages and preferred features exist for this aspect of the present disclosure as in the first aspect of the present disclosure previously discussed.

[0023] Furthermore, according to a fourth aspect of this disclosure, a vehicle is provided, comprising: a perception system having at least one sensor device and a vehicle control system, the at least one sensor device being used to generate sensor data including information about the vehicle's surrounding environment, the vehicle control system including an autonomous driving module. The autonomous driving module has control circuitry configured to acquire map data from the control system according to any embodiment of the third aspect. The map data includes a first layer of map representation of the surrounding environment along a road portion. The control circuitry of the autonomous driving module is further configured to: when the vehicle is traveling along the road portion, generate a signal controlling at least one of steering angle, acceleration, and braking actuation based on a comparison between the generated sensor data and the acquired map data, in order to control the vehicle.

[0024] Further embodiments of this disclosure are defined in the dependent claims. It should be emphasized that, when used in this specification, the term "comprising / including" is used to specify the presence of the stated features, integers, steps, or components. It does not exclude the presence or addition of one or more other features, integers, steps, components, or combinations thereof.

[0025] These and other features and advantages of this disclosure will be further described below with reference to the embodiments described herein. Attached Figure Description

[0026] Further objects, features, and advantages of the embodiments of this disclosure will become apparent from the following detailed description with reference to the accompanying drawings, wherein:

[0027] Figure 1 This is a flowchart representation of a method for generating and updating a digital map by using multiple lanes along a road section by at least one road vehicle, according to embodiments of the present disclosure.

[0028] Figure 2 This is a perspective view of a control system for generating and updating digital maps based on information from two vehicles traveling on the same section of road, according to embodiments of the present disclosure.

[0029] Figure 3 It includes having configured to from Figure 2 The depicted control system is an autonomous driving module that acquires map data, and is shown as a side view of the vehicle's control system. Detailed Implementation

[0030] Those skilled in the art will understand that the steps, services, and functions described herein can be implemented using separate hardware circuitry, software that runs in conjunction with a programmable microprocessor or general-purpose computer, one or more application-specific integrated circuits (ASICs), and / or one or more digital signal processors (DSPs). It will also be understood that, when this disclosure is described according to a method, it can also be embodied in one or more processors and one or more memories coupled to the one or more processors, wherein the one or more memories store one or more programs that, when executed by the one or more processors, perform the steps, services, and functions disclosed herein.

[0031] In the following description of exemplary embodiments, the same reference numerals denote the same or similar components.

[0032] To create and update accurate digital maps, precisely estimating the location of the detection vehicle is crucial. This can be achieved using high-end positioning systems provided in dedicated detection vehicles, a very expensive and non-scalable process. Alternatively, sensor data from consumer-grade systems can be utilized, providing the necessary data output for effective scaling, but at the cost of reduced accuracy.

[0033] Therefore, the inventors recognized that consumer-grade systems employed in most modern vehicles today (such as consumer-grade GPS, IMU, and other peripheral systems) can estimate lane departure (lane lateral deviation) and vehicle heading in a relatively accurate manner. This can be achieved, for example, using high-definition maps (including lane markings) as well as standard automotive GPS and vision-based lane trackers. However, the estimation of "along the road" (i.e., longitudinal position on the road) is often difficult to estimate accurately by consumer-grade systems, and uncertainties on the order of 10 meters are not uncommon. Therefore, the inventors recognized that this longitudinal error can be estimated by using data (position data and perception data) from several channels traversed by one or more vehicles, and by running the data through an optimized algorithm. Then, by applying the correct offset (based on the longitudinal error) to each "snapshot" image of the surrounding environment, an accurate digital map can be generated / updated in a scalable and cost-effective manner.

[0034] Initial position estimation can be performed within the vehicle or in a system outside the vehicle (e.g., in a cloud-based service). In the latter case, the vehicle can send raw data (GNSS data, IMU data, sensor data, etc.) to an external service, where the initial position / attitude estimation is performed. This can be advantageous in enabling the use of “ADAS vehicles” to generate / update digital maps, as they are typically not equipped with high-resolution maps, which increases the scalability of the methods and systems disclosed herein.

[0035] Figure 1 This is a flowchart representation of method 100, which is used to generate and update a digital map using multiple lanes traversed by one or more road vehicles (e.g., cars, buses, trucks, etc.) along a road segment. In other words, the digital map can be updated / generated based on a single road vehicle traversing multiple lanes along the same road segment, or multiple road vehicles each traversing one or more lanes along the same road. Furthermore, each vehicle is equipped with a perception system having one or more sensors configured to monitor the vehicle's surrounding environment. In this context, the perception system is understood as a system responsible for acquiring raw sensor data from sensors (such as cameras, LiDAR, RADAR, and ultrasonic sensors) and transforming that raw data into scene understanding. Each road vehicle is further provided with a positioning system configured to at least monitor the vehicle's geographic location and preferably to monitor the vehicle's attitude (geographic coordinates and heading).

[0036] Method 100 includes acquiring positioning data and sensor data for each lane from one or more road vehicles. The term "acquire" is interpreted broadly herein and includes receiving, obtaining, collecting, acquiring, etc. Positioning data includes multiple longitudinal positions (pi) for each lane within multiple road segments of a road section.x Sensor data (S) includes information about each (reported) longitudinal position (p). x The location data includes information about the surrounding environment of each road vehicle at a given location. Additional information about the location of each vehicle may be included. In some embodiments, the location data includes multiple poses (P(p)) for each lane within multiple road segments of the road section. y p x ,Θ)), that is, the attitude sequence, where p y p represents the lateral position of a section of the road (i.e., the offset within the lateral lane). x Θ represents the longitudinal position of the road segment, and Θ represents the heading angle. In other words, the sensor data can be understood as a snapshot of the vehicle's surroundings at each reported location along the road segment.

[0037] In some embodiments, location data is defined / acquired in the form of P = (x, y, Θ) at each location reported by each vehicle. Furthermore, the x and y (longitudinal and lateral) positions along the road segment can be projected as a trajectory. Therefore, there is a cumulative length s at each point, i.e., x(s), y(s), Θ(s). In this implementation, the "longitudinal position" along the road segment corresponds to the "arc length," i.e., the vehicle's "position" on the estimated trajectory. In other words, it is assumed (due to the relatively high accuracy of the lateral and heading estimates) that the estimated trajectory has a precise shape / form, but there is a time / "arc length" error for each reported location. Therefore, numerical optimization can be configured to utilize the longitudinal error δ. x Resample the trajectory: x(s+δ) x ), y(s+δ x ), Θ(s+δ x This ensures that the sensor data of the surrounding environment of all passages in a specific road segment are at the "maximum similarity level".

[0038] Method 100 may further include the step of defining multiple road segments (not shown) for the road portion. The road segments may be selected / defined such that it can be assumed that the reported longitudinal error is constant over that road segment. In other words, the road segments are selected to be small enough to assume that the vehicle's positioning system has the same error for every report of each longitudinal position within the same road segment. The length of each road segment may vary depending on various factors, such as road curvature, proximity to tree lines, proximity to tall buildings or other obstacle structures, the presence of tunnels, the quality of GNSS, whether the vehicle is an AD or ADAS vehicle, etc. For example, if the road portion is a curved road portion (containing numerous curves), conventional positioning algorithms are better at estimating the longitudinal position of a vehicle on the road than when the vehicle is traveling on a straight road.

[0039] Furthermore, method 100 forms a sub-map representation of the environment surrounding 102 at each longitudinal location of acquisition 101 based on the sensor data acquired 101. In other words, the sub-map representation is formed from each snapshot of the surrounding environment 102. More specifically, sensor detections can form a map quantized in space, such as a raster map. A raster map, or occupied raster map, divides the world into grates, where each raster consists of a cell block, each cell either occupied by an object with a certain probability or unoccupied. This probability can be approximated by dividing the number of detections returned from the object by the largest possible number of detections returned. After post-processing the raw data collected from the sensors, features of interest can be mapped to corresponding grates, where each grates can be examined independently of other grates.

[0040] Next, method 100 includes estimating 103 the longitudinal error of each acquired longitudinal position within each road segment by performing numerical optimization on the longitudinal error of each acquired longitudinal position within the common road segment based on the similarity level of the formed sub-map representations from the common road segment. More specifically, the numerical optimization may be performed such that, for each road segment, each longitudinal error is applied to each of the multiple sub-map representations (under one or more predefined constraints). Then, compared with other longitudinal error values, some longitudinal error values ​​will thus generate a “better match” for the sub-map representation within each road segment, and the longitudinal error value associated with each longitudinal position that yields the “best match” for each sub-map representation is selected as the correct error value.

[0041] The term "similarity level" is interpreted broadly in this context and can, for example, be the level of correlation between two or more submaps, entropy value, edge size in grouped submap images, pixel level, shape level, or any other suitable means of providing a quantitative measure of the matching level between two or more images. More specifically, longitudinal error estimation and optimization can be understood as the process of moving two overlapping images of the same object from different reference points relative to each other along a common axis and relative to the common reference point until the contents of the images overlap each other as well as possible. The offset introduced relative to the common reference point is the longitudinal error in this example.

[0042] Therefore, in some embodiments, the similarity level of the sub-map representation formed within a common road segment includes an entropy function, and numerical optimization is configured to minimize or reduce the entropy function. More specifically, entropy is used as a quantification value to measure the degree of image overlap after "pushing" images onto each other to generate a "sharp" image of the area surrounding the road segment. Entropy can be understood as the "sharpness" level of the combined images, where low values ​​represent sharp and clean combined images, and high values ​​represent noisy or blurry combined images.

[0043] Furthermore, in some embodiments, the similarity level of the sub-map representations formed within the common road segment includes the edge size within the formed sub-map representations, and wherein digital optimization is configured to maximize the edge size. In other words, optimization aligns the individual images (each formed sub-map representation) so that the combined image has the clearest and "sharpest" edges (of objects or content) possible.

[0044] Next, method 100 includes determining 104 multiple new positions for each road vehicle in each lane by applying the estimated longitudinal error to each corresponding acquired longitudinal position. In other words, the previously acquired longitudinal position 101 is corrected based on the estimated longitudinal error 103. The determined multiple new longitudinal positions are then applied to associated sensor data to generate 105 a first layer of a map representation of the surrounding environment along the road section. These two steps do not need to be performed as two separate steps, but can be performed simultaneously. For example, in some embodiments, the estimated longitudinal error is (directly) applied to the "associated sensor data" (which includes any metadata), thereby generating a layer of map representation without any different "determination 104" of the new longitudinal positions. However, even in this embodiment where the error is applied to the associated sensor data, it is considered to include the step of "determining multiple new longitudinal positions".

[0045] Even though the above embodiments focus on longitudinal error, i.e. δ x In some embodiments, the method also additionally or alternatively includes determining the lateral error δ in a similar manner. y and / or heading error Θ x .

[0046] Optionally, executable instructions for performing these functions are included in a non-transient computer-readable storage medium or other computer program product configured to be executed by one or more processors.

[0047] Figure 2This is a perspective view of a control system 20 used to generate and update a digital map 23 via multiple lanes along a road section 25 using at least one road vehicle 24a, 24b. In the illustrated example, two road vehicles 24a, 24b are depicted, each traversing at least one lane on the same road section 25. However, as those skilled in the art will recognize, the concepts disclosed herein are similarly applicable to the same road section 25 with multiple lanes traversed by a single road vehicle 24a, 24b. Each road vehicle is equipped with a sensing device having one or more sensors configured to monitor the surrounding environment of the road vehicle 24a, 24b. Furthermore, each vehicle is preferably equipped with a positioning system for monitoring the vehicle's geographic location, i.e., a system capable of outputting the longitudinal coordinates, lateral coordinates, and direction of travel of the vehicle 24a, 24b. The positioning system may, for example, be a consumer-grade Global Navigation Satellite System (GNSS), such as a GPS unit. Naturally, other regional GNSS solutions (such as GLONASS, BeiDou, Galileo, etc.) are also possible.

[0048] The control system 20 may be located in a remote server communicating with vehicles 24a, 24b, as a so-called cloud solution. The control system may include control circuitry 21 (also referred to as one or more control units, one or more processors, etc.) and memory 22. Memory 22 may store one or more programs configured to be executed by the control circuitry 21, the one or more programs including instructions for performing methods according to any embodiment disclosed herein.

[0049] Furthermore, vehicle 1 can connect to an external network (e.g., for transmitting sensor data) via, for example, a wireless link. The same wireless link, or some other wireless link, can be used to communicate with other vehicles near vehicles 24a, 24b, or with local infrastructure components. Cellular communication technologies can be used for long-distance communication (such as to external networks and external servers 20), and if the cellular communication technology used has low latency, it can also be used for vehicle-to-vehicle, vehicle-to-vehicle (V2V), and / or vehicle-to-infrastructure (V2X) communication. Examples of cellular wireless technologies are GSM, GPRS, EDGE, LTE, 5G, 5G NR, etc., and also include future cellular solutions. However, in some solutions, medium- to short-range communication technologies are used, such as wireless local area networks (LANs) (e.g., solutions based on IEEE 802.11). ETSI is investigating cellular standards for vehicle communication, and 5G, for example, is considered a suitable solution due to its low latency, high bandwidth, and efficient processing capabilities of the communication channel. Therefore, vehicles 24a, 24b are equipped with suitable interfaces and devices (not shown) for communicating with control system 20, and vice versa.

[0050] Next, control circuit 21 is configured to acquire positioning data and sensor data for each lane from road vehicles 24a and 24b. The positioning data includes multiple longitudinal positions (pi) of each lane within multiple road segments of road section 25. x In this context, longitudinal direction is understood as the coordinate along the main extension of road section 25, i.e., along the direction of travel or x-axis in the figure. Sensor data (S) includes information about each (reported) longitudinal position (p x Information about the surrounding environment of each road vehicle 24a, 24b at location 24a, 24b. Vertical arrow 27 is used to indicate the reported location data of road vehicles 24a, 24b. The location data may include additional information about the location of each vehicle 24a, 24b. In some embodiments, the location data includes multiple orientations p(p) of each lane within multiple segments of road section 25. y p x ,Θ)(i.e., attitude sequence), where p y p indicates the lateral position on road section 25. x Θ represents the longitudinal position on road section 25, and Θ represents the heading angle. In other words, the sensor data can be understood as a snapshot of the vehicle's surroundings at each reported location along road section 25.

[0051] exist Figure 2 The diagram illustrates three road segments #m, #(m+1), and #(m+2) of road section 25. At a specific road segment (e.g., #m), assume there exists N... m A sequence of poses Where k is the sample index. The attitude sequence can be understood as positioning data, including at least the vehicle's longitudinal position at each sample index. Figure 2 In this study, only two poses are illustrated for each channel and each segment; however, as will be apparent to those skilled in the art, the positioning data may include one or more longitudinal positions (or complete poses) within each segment.

[0052] The sensor data includes information about the surrounding environment of each road vehicle 24a, 24b at each longitudinal position 27. In some embodiments, the sensor data is acquired / obtained from an active sensor system configured to transmit and receive signals reflected from objects within the surrounding environment of the road vehicles 24a, 24b. The active sensor system can be at least one of a lidar, radar, and active sonar sensor system. One advantage of using active sensor data, and especially radar data, is that the subsequently generated map representation will take the form of a radar map, which is particularly suitable for autonomous driving applications. For example, radar-based navigation works reliably at night and in low visibility conditions. Furthermore, radar maps can provide centimeter-level accuracy with less data compared to camera-based maps. However, it can be difficult to acquire radar maps based on radar data from typical "survey" vehicles because radar data is relatively noisy. Therefore, multiple channels or vehicles are often required.

[0053] Furthermore, control circuit 21 is configured to form / generate / build a sub-map representation of the surrounding environment at each acquired longitudinal position based on the acquired sensor data. Therefore, in an exemplary embodiment where the sensor data includes radar data, control circuit 21 is configured to form / generate / build a radar map of the surrounding environment at each reported longitudinal position.

[0054] Then, control circuit 21 is configured to estimate the longitudinal error (δ) of each acquired longitudinal position within each road segment by performing digital optimization on the longitudinal error of each acquired longitudinal position within the same road segment based on the similarity level of the formed sub-map representation from the "same" road segment. x ).

[0055] More specifically, numerical optimization can be based on one or more predefined assumptions. For example, as mentioned above, at a certain road segment (e.g., #m), it is assumed that there exists N. m A sequence of poses Where k is the sample index. Longitudinal error δ x (i; m) is defined as the error of the i-th attitude path (i.e., the attitude path of the first vehicle 24a) along road segment #m. Furthermore, compared with the longitudinal error, other errors (such as incorrect lane assignment by the positioning algorithm, lateral error, and heading error) can be assumed to be negligible.

[0056] Furthermore, it is assumed that the longitudinal error of multiple lanes within each road segment (passed by one or more vehicles 24a, 24b) is approximately zero, i.e. Furthermore, the road segment is prioritized, allowing us to assume δ x(i;m) is constant above this road segment. In other words, the road segment is chosen to be small enough to assume that the GNSS reports of vehicles 24a and 24b for each longitudinal position within the same road segment have the same error. The length of each road segment can vary depending on a variety of factors, such as road curvature, proximity to tree lines, proximity to tall buildings or other obstacle structures, the presence of tunnels, the quality of the GNSS, whether the vehicle is an AD or ADAS vehicle, etc. For example, if the road section is a curved section (containing many curves), conventional positioning algorithms are better at estimating the longitudinal position of vehicles on the road than when the vehicle is traveling on a straight road.

[0057] Furthermore, in order to eliminate the longitudinal error δ from the associated sensor data generated from subsequent maps... x Cloud processing can be applied. Therefore, the goal of the error estimation step is to estimate...

[0058] In some embodiments, in order to estimate road segment #m The longitudinal errors of all attitude sequences can be used to form / build a radar grid map. Furthermore, the "focal point" of the map (the similarity level between offset sub-maps) is measured by applying the function F:

[0059]

[0060] In some embodiments, the function F may correspond to entropy or some other equivalent function that provides a quantified value of the similarity level. Furthermore, λ is an adjustable parameter that specifies predefined constraints. The degree of enforcement. For example, when N m When the value is large, λ can be set to a large value.

[0061] Furthermore, the aforementioned formulaic optimization problem can be extended to optimize across multiple road segments. Additionally, the extended optimization problem can enforce another predefined constraint δ. x (i;m)≈δ x (i;m+1)≈δ x (i; m-1). That is, for a given attitude sequence (i.e., for locating data reports originating from the same channel), the variation in longitudinal error over adjacent segments should be small (e.g., below a predetermined threshold).

[0062] In addition, radar grid maps could be considered. The similarity level of the formed sub-map representation within a common road segment includes the edge size within the formed sub-map representation, rather than the measurement entropy. Therefore, in some embodiments, the similarity level of the formed sub-map representation includes the edge size within the formed sub-map representation, and wherein numerical optimization is configured to maximize the edge size. In this context, a radar map can be understood as radar road markings, or a radar map can be understood as a perception of the vehicle's surroundings from the perspective of a radar device.

[0063] Furthermore, the positioning system and any associated positioning algorithm for each vehicle 24a, 24b can be configured to generate an estimate of the uncertainty in its longitudinal position. Therefore, in some embodiments, the control circuitry 21 is configured to include these uncertainties in the optimization problem. For example, positioning data associated with higher uncertainties may be weighted differently from positioning data associated with lower uncertainties.

[0064] Next, once the longitudinal error of each acquired longitudinal position within each road segment has been estimated, the control circuit 21 determines multiple new longitudinal positions for each road vehicle 24a, 24b in each lane. Multiple new longitudinal positions are formed by applying the estimated longitudinal error to each corresponding acquired longitudinal position.

[0065] Furthermore, control circuit 21 is configured to apply the determined multiple new longitudinal positions to the associated sensor data to generate a first layer of a map representation of the surrounding environment along the road section. A map layer can be understood as a representation of a geographic region based on a specific dataset. For example, one layer could be a map representation of a geographic region based on radar sensor data, another layer could be a map representation of a geographic region based on lidar sensor data, and yet another layer could be a map representation of a geographic region based on image data from one or more cameras.

[0066] Figure 3 This is a side view of a vehicle including a vehicle control system 10, which is configured to... Figure 2 The depicted control system is an autonomous driving module that acquires map data. Vehicle 1 further includes a perception system 6 and a positioning system 5. In this context, the perception system 6 is understood to be responsible for acquiring raw sensor data from sensors 6a, 6b, 6c (such as cameras, LiDAR and RADAR, ultrasonic sensors) and converting this raw data into scene understanding. The positioning system 5 is configured to monitor the vehicle's geographic location and heading, and may be in the form of a Global Navigation Satellite System (GNSS) (e.g., GPS). The vehicle may further include at least one of an inertial measurement unit (IMU) and wheel speed sensors to assist the positioning system.

[0067] The automatic (semi-automatic or fully automatic) drive module includes one or more processors 11, a memory 12, a sensor interface 13, and a communication interface 14. The processor 11 may also be referred to as a control loop 11 or a control circuit 11. The control circuit 11 is configured to implement the functions of the automatic driving module, and more specifically, executes instructions stored in the memory 12 to perform methods for controlling at least one of steering angle, acceleration, and braking actuation to maneuver the vehicle 1, as described below. In other words, the memory 12 of the control device 10 may include one or more (non-transient) computer-readable storage media for storing computer-executable instructions, such as instructions that, when executed by one or more computer processors 11, cause the computer processors 11 to perform the techniques described below. The memory 12 may optionally include high-speed random access memory (such as DRAM, SRAM, DDRRAM, or other random access solid-state storage devices); and may optionally include non-volatile memory (such as one or more disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid-state storage devices).

[0068] More specifically, the control circuit 11 is configured to control the system according to any of the embodiments discussed above (e.g., refer to...). Figure 2 20) Acquire map data. The map data includes a first layer of map representation of the surrounding environment along the road section on which the vehicle 1 travels. The map data may be acquired directly from an external control system or via a subsystem of the vehicle 1 (e.g., the vehicle 1's positioning system 5).

[0069] Furthermore, the control circuit 11 is configured to generate signals based on a comparison between generated sensor data and acquired map data when the vehicle 1 is traveling along a road section, to control at least one of the steering angle, acceleration, and braking actuation, thereby maneuvering the vehicle 1. Naturally, the control circuit can be configured to acquire or read sensor data generated by one or more sensors 6a, 6b, 6c of the perception system. Therefore, the vehicle 1 is capable of automatic or semi-automatic navigation based on a radar map, and real-time radar data is provided. This can be achieved through the methods and control systems discussed above (e.g., referencing...). Figure 2 This is achieved by generating a high-precision radar map using 20).

[0070] This disclosure has been presented above with reference to specific embodiments. However, other embodiments besides those described above are also possible and are within the scope of this disclosure. Method steps, performed by hardware or software, different from the method steps described above, may be provided within the scope of this disclosure. Thus, according to exemplary embodiments, a non-transient computer-readable storage medium is provided for storing one or more programs configured to be executed by one or more processors of a control system, the one or more programs including instructions for performing the methods according to any of the embodiments discussed above. Alternatively, according to another exemplary embodiment, a cloud computing system may be configured to perform any of the methods presented herein. The cloud computing system may include distributed cloud computing resources that collectively perform the methods presented herein under the control of one or more computer program products.

[0071] In general, computer-accessible media can include any tangible or non-transient storage medium or memory medium, such as electrical, magnetic, or optical media, for example, a disk or CD / DVD-ROM coupled to a computer system via a bus. The terms “tangible” and “non-transient” as used herein are intended to describe computer-readable storage media (or “memory”) excluding those that propagate electromagnetic signals, but are not intended to further limit the type of physical computer-readable storage device included by the phrase computer-readable medium or memory. For example, the terms “non-transient computer-readable medium” or “tangible memory” are intended to include types of storage devices that do not necessarily permanently store information, including, for example, random access memory (RAM). Program instructions and data stored in a non-transient form on a tangible computer-accessible storage medium can be further transmitted via a transmission medium or signal (such as an electrical signal, electromagnetic signal, or digital signal), which can be transmitted via a communication medium (such as a network and / or wireless link).

[0072] Processors 11, 21 (associated with control systems 10, 20) may be or include any number of hardware components for performing data or signal processing or for executing computer code stored in memories 12, 22. Systems 10, 20 have associated memories 12, which may be one or more devices for storing data and / or computer code to perform or implement the various processes described herein. Memory may include volatile or non-volatile memory. Memory 12, 22 may include database components, object code components, script components, or any other type of information structure for supporting the various activities described herein. According to exemplary embodiments, the systems and methods described herein may utilize any distributed or local storage device. According to exemplary embodiments, memories 12, 22 may be communicatively connected to the corresponding processors 21, 22 (e.g., via circuitry or any other wired, wireless, or network connection) and include computer code for performing one or more processes described herein.

[0073] It should be understood that the sensor interface 13 of the vehicle control system may also provide the possibility of directly acquiring sensor data or acquiring sensor data via a dedicated sensor control circuit 6 in the vehicle 1. The communication / antenna interface 14 further provides the possibility of sending output to a remote location 2 (e.g., a remote operator or control center) via antenna 8. Furthermore, some sensors in the vehicle may communicate with the control system 10 using local network settings (such as CAN bus, I2C, Ethernet, fiber optics, etc.). The communication interface 14 may be configured to communicate with other control functions of the vehicle and therefore may also be considered a control interface; however, a separate control interface (not shown) may also be provided. Local communication within the vehicle may also be wireless, using protocols such as WiFi, LoRa, Zigbee, Bluetooth, or similar mid-range / short-range technologies.

[0074] It should be noted that the word "comprising" does not exclude the presence of other elements or steps besides those listed, and the word "a" preceding an element does not exclude the presence of a plurality of such elements. It should be further noted that no reference numerals in the drawings limit the scope of the claims, and that this disclosure can be implemented at least in part by both hardware and software, and that some "apparatus" or "units" can be represented by the same hardware item.

[0075] Although the figures may illustrate a specific order of method steps, the order of steps may differ from the depicted order. Furthermore, two or more steps may be performed simultaneously or partially simultaneously. For example, the steps of determining multiple new (longitudinal) positions and applying multiple new (longitudinal) positions need not be two separate steps. In some embodiments, the estimated error is directly applied to the "correlated sensor data," thus there is no separate step of "determining multiple new (longitudinal) positions." This variation will depend on the choice of software and hardware systems and the designer's choices. All these variations are within the scope of this disclosure. Similarly, software implementations may utilize standard programming techniques with rule-based logic and other logic to implement various connection steps, processing steps, comparison steps, and decision steps. The embodiments mentioned and described above are given by way of example only and should not be limited to this disclosure. Other solutions, uses, objectives, and functions claimed within the scope of this disclosure in the patent embodiments described below will be apparent to those skilled in the art.

Claims

1. A method for generating and updating a digital map by using multiple lanes along a road segment by at least one road vehicle, each road vehicle including a perception system having at least one sensor configured to monitor the vehicle's surrounding environment, the method comprising: Location data and sensor data for each lane are acquired from the at least one road vehicle. The positioning data includes multiple longitudinal positions of each lane within multiple road segments of the road section, and the sensor data includes information about the surrounding environment of each road vehicle at each longitudinal position. Based on the acquired sensor data, a sub-map representation of the surrounding environment is formed at each acquired longitudinal location; The longitudinal error of each acquired longitudinal position within each road segment is estimated by performing digital optimization on the longitudinal error of each acquired longitudinal position within the public road segment based on the similarity level of the sub-map representation formed from the public road segment. By applying the estimated longitudinal error to each corresponding acquired longitudinal position, multiple new longitudinal positions for each road vehicle in each lane are determined; and The determined new longitudinal positions are applied to the associated sensor data to generate a first layer of a map representation of the surrounding environment along the road section.

2. The method according to claim 1, wherein, The similarity level represented by the sub-map formed within the public road segment includes an entropy function, and the numerical optimization is configured to minimize the entropy function.

3. The method according to claim 1 or 2, wherein, The similarity level of the sub-map representation formed within the public road segment includes the edge size within the formed sub-map representation, and wherein the numerical optimization is configured to maximize the edge size.

4. The method according to claim 1 or 2, wherein, The digital optimization is further based on the following condition: the sum of the longitudinal errors of the plurality of channels within the common road segment is defined to be approximately zero, such that the solution of the digital optimization that deviates from zero in the longitudinal errors of the plurality of channels within the common road segment is associated with an increased cost factor.

5. The method according to claim 1 or 2, wherein, The digital optimization is further based on the following condition: the longitudinal error within a single segment of a channel is defined to be the same for all acquired longitudinal positions within that single segment of the channel.

6. The method according to claim 1 or 2, wherein, The sensor data includes radar sensor data acquired from one or more radar sensors of the plurality of road vehicles, and wherein the sub-map representation is a radar map representation formed based on the radar sensor data.

7. The method according to claim 1 or 2, wherein, The location data includes multiple longitudinal positions for each vehicle, determined by the consumer-grade positioning system for each road vehicle.

8. The method according to claim 1 or 2, wherein, The digital optimization is further based on the following condition: the variation of the longitudinal error of adjacent road segments of a channel is approximately zero, such that the solution of the digital optimization that deviates from zero in the variation of the longitudinal error of adjacent road segments of a channel is associated with an increased cost factor.

9. The method according to claim 1 or 2, further comprising: Define the plurality of road segments of the road portion.

10. The method according to claim 1 or 2, wherein, The sensor data further includes image sensor data acquired from one or more image sensors of each road vehicle, and the method further includes: The multiple new longitudinal positions are applied to the associated image sensor data to generate a second layer of map representation of the surrounding environment along the road section.

11. A computer-readable storage medium storing one or more programs configured to be executed by one or more processors of a processing system, the one or more programs including instructions for performing the method according to any one of the preceding claims.

12. A control system for generating and updating a digital map by using multiple lanes along a road segment by at least one road vehicle, each road vehicle including a perception system having at least one sensor configured to monitor the surrounding environment of the road vehicle, the control system including control circuitry configured to: Location data and sensor data for each lane are acquired from the at least one road vehicle. in, The positioning data includes multiple longitudinal positions of each lane within multiple road segments of the road section, and the sensor data includes information about the surrounding environment of each road vehicle at each longitudinal position. Based on the acquired sensor data, a sub-map representation of the surrounding environment is formed at each acquired longitudinal location; The longitudinal error of each acquired longitudinal position within each road segment is estimated by performing digital optimization on the longitudinal error of each acquired longitudinal position within the public road segment based on the similarity level of the sub-map representation formed from the public road segment. By applying the estimated longitudinal error to each corresponding acquired longitudinal position, multiple new longitudinal positions for each road vehicle in each lane are determined; as well as The determined new longitudinal positions are applied to the associated sensor data to generate a first layer of a map representation of the surrounding environment along the road section.

13. The control system according to claim 12, wherein, The sensor data further includes image sensor data acquired from one or more image sensors of each road vehicle, and wherein the control circuit is further configured to: The multiple new longitudinal positions are applied to the associated image sensor data to generate a second layer of map representation of the surrounding environment along the road section.

14. A vehicle comprising: A perception system having at least one sensor device for generating sensor data including information about the vehicle’s surrounding environment; The vehicle control system includes an autonomous driving module, which is configured to: Map data acquired from the control system according to claim 12 or 13, wherein, The map data includes a first layer representing the map of the surrounding environment along the road section; As the vehicle travels along the road section, signals are generated based on a comparison between generated sensor data and acquired map data to control at least one of steering angle, acceleration, and braking actuation in order to manipulate the vehicle.

Images