Method for producing high-precision map based on virtual information track model

Abstract

Embodiments of the present disclosure provide a method for making a high-precision map based on a virtual information track model, which includes extracting road network information from original map data or surveying data and generating a reference path according to the road network information; establishing a plurality of Frenet local coordinate systems with the reference path as a reference, each local coordinate system including at least an arc length axis along the reference path, a lateral offset axis perpendicular to the reference path, and an elevation axis perpendicular to the road surface; dividing the reference path into virtual information track segments, each local coordinate system including a plurality of virtual information track segments; generating a globally unique identification code for each virtual information track segment, which includes encrypting at least part of the geographic information of each virtual information track segment; and binding a static attribute parameter associated with the arc length axis for each virtual information track segment. The map made by the method can meet the requirements of L3 or above level autonomous driving, and at the same time meet the requirements of geographic information secrecy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to a high-precision map making method based on a virtual information track model. BACKGROUND

[0002] With the rapid development of automatic driving technology, high-level automatic driving (L3 and above) relies more and more on high-precision maps. According to the national standard “Automotive Driving Automation Classification” (GB / T 40429-2021), L3 (conditional automatic driving), L4 (high-level automatic driving) and L5 (full automatic driving) systems need to obtain real-time road geometry and physical parameters with centimeter-level precision to support path planning, vehicle control and dynamic decision-making. Currently, intelligent networked basic maps can be divided into “road level”, “lane level” and “driving control level” according to the precision level. Among them, “lane level” maps carry high-precision road information and can meet the general needs of L3 / L4 automatic driving, with a relative positioning accuracy of facilities and targets of up to 10 centimeters; and “driving control level” maps further integrate vehicle networking dynamic data (such as roadside unit exchange information and vehicle coordination data) to form real-time driving information on the “digital twin lane” to solve the problem of L5 level over-the-horizon machine driving and coordinated control in dense vehicle environment.

[0003] However, since the publicly used navigation geographic information must meet the relevant legal regulations of surveying and mapping geographic information secrecy, the real geographic coordinates of the high-precision map must be nonlinearly transformed by using the state-recognized geographic information secrecy processing technology before being publicly used, and the real space position must be hidden, and the matching of the vehicle real-time position and the secrecy-processed map must be completed by integrating the terminal plug-in at the vehicle end. This scheme involves data processing and plug-in integration, and the implementation cycle is long, and the precision can only meet the needs of L3 automatic driving, so the contradiction between the production and use of high-precision maps and the needs of the development of automatic driving technology is highlighted. At present, there is no high-precision map solution on the market that can meet the needs of L3 and above automatic driving functions and meet the requirements of geographic information secrecy in China. Therefore, a high-precision map making method that can meet the needs of L3 and above automatic driving and meet the requirements of geographic information secrecy in China is needed. SUMMARY

[0004] Based on the defects of the prior art, the present application aims to provide a high-precision map making method based on a virtual information track model.

[0005] The embodiment of the present disclosure provides a high-precision map production method based on a virtual information track model, which comprises extracting road network information from original map data or surveying data, and generating a reference path according to the road network information; taking the reference path as a reference, a plurality of Frenet local coordinate systems are established, each Frenet local coordinate system at least comprising an arc length axis along the reference path, a lateral offset axis perpendicular to the reference path and an elevation axis perpendicular to the road surface; the reference path is divided into virtual information track segments along the arc length axis, and each Frenet local coordinate system comprises a plurality of virtual information track segments; a globally unique identification code is generated for each virtual information track segment, which comprises at least one part of geographical information of each virtual information track segment being encrypted; and a static attribute parameter associated with the arc length axis is bound for each virtual information track segment.

[0006] According to the high-precision map production method of the embodiment of the present disclosure, the method further comprises online verification of a vehicle, determination of a virtual information track segment where the vehicle is located; centimeter-level positioning of information detected by a plurality of sensors of the vehicle in the Frenet coordinate system; calibration of positioning deviation by using a road side information code and / or a handshake area; and reception of road information from the vehicle, updating of a dynamic attribute parameter of the virtual information track segment where the vehicle is located according to the road information, the dynamic attribute parameter being associated with the arc length axis.

[0007] Optionally, according to the high-precision map production method of the embodiment of the present disclosure, the information detected by the plurality of sensors comprises a vehicle heading angle, which is compared with a tangential direction of the arc length axis, and is used to determine deviation of the vehicle in a traveling direction.

[0008] Optionally, according to the high-precision map production method of the embodiment of the present disclosure, the encryption processing of at least one part of road geographical information of each virtual information track segment comprises irreversible encryption processing of the at least one part of geographical information. For example, the encryption processing can be implemented by using an SM3 algorithm.

[0009] Optionally, according to the high-precision map production method of the embodiment of the present disclosure, the plurality of Frenet local coordinate systems and the globally unique identification code of each virtual information track segment do not comprise real geographical coordinates, and the globally unique identification code of each virtual information track segment corresponds to a position of the each virtual information track segment in the Frenet local coordinate system.

[0010] Optionally, the high-precision map making method according to the embodiments of the present disclosure, wherein the static attribute parameters bound for each virtual information track segment include one or more of curvature, slope, lane width and / or lane width range, road facilities, traffic sign information, and driving restriction information corresponding to the mileage position of the arc length axis.

[0011] Optionally, the high-precision map making method according to the embodiments of the present disclosure, wherein the dynamic attribute parameters associated with the arc length axis include one or more of road surface friction coefficient, traffic flow, visibility, road event, and weather condition.

[0012] Optionally, the high-precision map making method according to the embodiments of the present disclosure, wherein each handshake area corresponds to one of the plurality of Frenet local coordinate systems, has an encrypted arc length axis reference point, on-line accumulates deviation data of the arc length axis and the lateral offset axis when the vehicle travels to the handshake area, and resets the Frenet local coordinate system in which the vehicle is currently located according to the received calibration parameters.

[0013] The reconstructed coordinate system of the handshake area needs to perform strict encryption processing. Preferably, the arc length axis reference point coordinates can be encrypted by using the SM4 algorithm. The geographical position of the arc length axis reference point can be saved in the globally unique identification code of the virtual information track segment where the reference point is located. Because all subsequent node segment roads (including all road nodes between handshake areas) are determined relative to the space position of the original point coordinate system, the principles of handshake area reconstruction standard and handshake area setting are very important.

[0014] Optionally, the high-precision map making method according to the embodiments of the present disclosure, wherein the road side information code is arranged at least on a road where a navigation satellite system positioning signal is lost or a road with high curvature, and the vehicle can correct its deviation on the arc length axis and the lateral offset axis in real time by reading the road side information code. The road with high curvature is, for example, an overpass, a roundabout, a continuous turning road section, a mountainous road section, etc. The navigation satellite system can be selected from a Beidou navigation satellite system, a global navigation satellite system (GNSS), etc.

[0015] Optionally, the high-precision map making method according to the embodiments of the present disclosure, wherein dividing the reference path along the arc length axis into virtual information track segments includes dividing the virtual information track segments according to the collection nodes of the original map data.

[0016] Compared with the prior art, the embodiments of the present disclosure have the following beneficial effects:

[0017] First, the map data is highly confidential. The entire high-precision map production and application process does not involve the storage of real geographical positions at all. Core data such as virtual information tracks (VIT) are encrypted. Since the Frenet coordinate system is a local parameterized coordinate system based on a reference path, the coordinate axes have no real directional property, and the virtual information track segment coding is the "core index" of the map data, which is designed in accordance with the principles of no real geographical coordinates and "irreversible encryption" coding. Therefore, even if the map data is leaked, it cannot be associated with a specific road through coding, that is, the real geographical position cannot be obtained through coding.

[0018] Second, the map is highly compatible with autonomous driving. The Frenet coordinate system is naturally compatible with the "motion control logic" of autonomous driving, that is, no real coordinates are needed, only "local relative parameters" are needed. The vehicle realizes accurate positioning in the Frenet coordinate system through multi-sensor fusion between multiple sensors (for example, at least two of an inertial measurement unit (IMU), a visual camera, and a laser radar), without relying on real coordinates. All parameters of the virtual information track segment (such as curvature, slope, lane width, dynamic friction coefficient, etc.) are stored in the form of "Frenet coordinate system bound attributes", and multiple types of information are fused. After the vehicle is matched to the current segment through the virtual information track coding, the parameters can be directly read and transmitted to the control module without any "real coordinate calculation" step.

[0019] Third, the embodiments of the present disclosure can achieve centimeter-level precision and stable reliability. The roadside information code is mainly used for vehicle repositioning in special areas such as navigation satellite system positioning lockout environment, and the handshake area is used to correct the error accumulation problem of autonomous driving in continuous long sections. Through the strict adherence of the Frenet coordinate system to the road form, combined with the calibration mechanism of the roadside information code and the handshake area, the vehicle's full-section cumulative positioning error can be controlled within a few centimeters, which is much better than the traditional scheme of ±15 cm. Even in long distances and complex scenarios, the calculation of map parameters (curvature, slope, etc.) is still accurate and has no cumulative error.

[0020] Fourth, the embodiments of the present disclosure can support navigation for various complex road scenarios (intersections, roundabouts, interchanges, etc.) through flexible segmentation. Alternatively, the segmentation of the virtual information track can be established according to the segmentation nodes determined by the map vendor during static map construction, such as the nodes established every second during static data collection. Alternatively, the virtual information track is adaptively segmented according to the complexity of the road environment, so that shorter segmentation lengths are used in areas with complex road geometry, and longer segmentation lengths are used in areas with simple geometry, to balance modeling accuracy and data efficiency.

[0021] The automatic driving vehicle can directly obtain accurate path guidance and control reference from the map in these scenarios, without being limited by the problem of reduced accuracy of traditional maps in special scenarios.

[0022] Fifth, the virtual information track segment coding mode adopts a joint coding mode of region segment identification, additional binding of static information (such as road curvature, slope), and dynamic information (such as traffic control). The coding design can optionally follow the principle of "fixed field and algorithm generation". It can be directly generated based on the existing collected map data without human intervention or complex encryption process, and can be directly embedded in the existing coding generation link. The calibration logic (handshake area calibration, dynamic calibration, etc.) of the virtual information track is completely based on the "local relative parameters" of the Frenet coordinate system, and does not involve real coordinate conversion. It can reuse the "geometric accuracy verification" and "parameter adaptation" links in the existing map production process, and the process is simple and has no additional complexity.

[0023] Implementing any device or method of the present disclosure does not necessarily require all the advantages described above to be achieved simultaneously. Other features and advantages of the present disclosure will be set forth in the description that follows, and will in part be apparent from the description, or will be learned by practice of the present disclosure. The objectives and advantages of the embodiments of the present disclosure can be achieved and obtained by the structures indicated in the specification, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present disclosure, but not limit the present disclosure.

[0025] Figure 1 is a schematic diagram of a coordinate system for constructing a virtual information track model according to an embodiment of the present disclosure;

[0026] Figure 2 is a method flowchart for constructing a virtual information track model based on a coordinate system and generating an offline map according to an embodiment of the present disclosure;

[0027] Figure 3 is a method flowchart for online updating of a map according to an embodiment of the present disclosure;

[0028] Figure 4A is a schematic diagram of the distribution of handshake areas on a physical road according to an embodiment of the present disclosure;

[0029] Figure 4B is a schematic diagram of a virtual information track map and the distribution of handshake areas thereon corresponding to Figure 4A DETAILED DESCRIPTION ​

[0030] In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure clearer, the following will be combined with the drawings of the embodiments of the present disclosure to make a clear and complete description of the technical solutions of the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all the embodiments. The various different embodiments can be combined with each other to constitute other embodiments which are not shown in the following description. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present disclosure without creative labor are within the scope of protection of the present disclosure.

[0031] Unless otherwise defined, technical terms or scientific terms used herein should be understood as having the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs. The terms "first", "second" and similar terms used in the description and claims of the present disclosure do not necessarily mean any order, number or importance, but are only used to distinguish different components. Similarly, the terms "one" or "a" and the like do not necessarily mean the number limit. The terms "include" or "contain" and the like mean that the elements or objects before the terms cover the elements or objects listed after the terms and their equivalents, without excluding other elements or objects. The terms "connected" or "connected" and the like are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to represent relative positional relationships, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0032] Figure 1 is a schematic diagram of a coordinate system for constructing a virtual information track model according to an embodiment of the present disclosure. As shown in Figure 1 The Frenet coordinate system of the virtual information track model is a parameterized coordinate system established along the reference path. The virtual information track model is a virtual digital road information model that integrates road static parameters and real-time dynamic parameters in a local coordinate system without real geographical position.

[0033] The coordinate system can take the road entrance as the coordinate starting point of the carrier, the forward direction arc length mileage of the reference path of the road as the s axis, the transverse right perpendicular to the s axis as the d axis, and the elevation axis perpendicular to the road surface as the z axis (not shown in the figure), thereby forming a Frenet geometric space.

[0034] Each segment is taken as a "road segment unit" to mark lane width, curvature, slope, road surface mechanics parameters and other information with centimeter-level accuracy. The coordinate system thus constructed is independent of the real geographical position and only serves the path tracking and control of autonomous driving.

[0035] The reference path can be extracted from existing collected map data. The reference path preserves relative geometry and can not contain any absolute geographic position information. The reference path can preferably be a center line of a lane, or a center line of a certain lane, a lane boundary line, etc.

[0036] The main parameters of the Frenet coordinate system are described below:

[0037] The s-axis (arc length axis): the s-axis is established along the cumulative arc length direction of the reference path. Optionally, the unit of the s-axis can be meters (m). The coordinate origin O (s=0) is set at the starting point of the road, and the arc length accumulates incrementally in the forward direction of the vehicle. For example, the point on the reference path 35.2 meters away from the origin O has an s coordinate of 35.2 m.

[0038] The d-axis (lateral axis): the d-axis is established perpendicular to the s-axis, and the d-axis represents the lateral offset of the lane. Optionally, the unit of d can be centimeters (cm), decimeters (dm), etc., and d=0 is defined on the reference path, for example, the right side of the reference path can be agreed to be the positive direction and the left side to be the negative direction. For example, when the vehicle is located 15 centimeters to the right of the center line of the lane, the d value is +15 cm. The d-axis is used to describe the lane width range and the lateral position of the vehicle relative to the reference path.

[0039] The z-axis (elevation axis): the z-axis is the height axis perpendicular to the road surface. Optionally, the unit of the z-axis can be set according to the actual road segment needs, such as centimeters (cm), decimeters (dm), etc. z=0 is set at the starting point of the road, which is used to describe the vertical height corresponding to the s-axis position, which is the integral result of the slope along the s-axis. On a slope road section, the value of the z-axis can be used to calculate the road surface slope i.

[0040] Based on the above coordinate system, the following parameters can be further obtained:

[0041] The θ-axis (heading angle) is the deviation angle of the actual driving direction of the vehicle relative to the tangent direction of the s-axis. The unit of the θ-axis is degrees (°). When the vehicle drives towards the s-axis direction, θ=0°, and the clockwise deflection can be set as a positive angle and the counterclockwise deflection can be set as a negative angle. θ is used to correct the heading pose of the vehicle in real time, and cooperates with the positioning of the s-axis and the d-axis to improve the path tracking accuracy.

[0042] Road curvature refers to the degree of bending of the road in the horizontal plane, usually represented by "1 / R", where R is the radius of the road curve. The greater the curvature, the higher the degree of road bending, and the greater the steering angle the vehicle needs to make during driving. The road curvature k(s) is a static description of the bending degree of the reference path along the s-axis, which is the rate of change of the tangent heading angle of the s-axis. The road curvature provides an indirect calibration target for the heading angle by generating a tangent heading angle (static reference) of the s-axis.

[0043] As described above, the road surface slope (i) is the static change relationship between the z-axis (elevation axis) and the s-axis (arc length axis) in the Frenet coordinate system, defined as the "derivative of elevation to s-axis arc length" (i=dz / ds). For example, a road surface slope i=3% means that it rises 3m for every 100m forward. The static slope can directly determine the distribution rule of the elevation.

[0044] Using the virtual information track model constructed based on the Frenet coordinate system can make all positioning and road parameters closely related to the reference path, directly correspond to the driving path of the vehicle, and more easily obtain the lateral deviation information of the vehicle, without the need for complex coordinate conversion or curve calculation at the vehicle end, greatly improving the utilization efficiency of the automatic driving system for map data.

[0045] Figure 2 A flowchart of a method for constructing a virtual information track model based on a coordinate system and generating an offline map according to an embodiment of the present disclosure is shown.

[0046] In step S101, first, map information including road network data is collected from static maps or surveying data, and a reference path is generated from the road network data. Specifically, the original map information is collected from the original static electronic map data conforming to the national standard or the surveying data obtained by the surveying department, for example, the CGCS2000 / WGS84 compliant electronic map. The reference path is generated according to the road network data in the map. As described above, the reference path can preferably be the lane centerline. Alternatively, in this step, the real geographical position in the original map data or surveying data can be encrypted, for example, using SM4 symmetric encryption, so that the real geographical position can be hidden in the subsequent map making process.

[0047] In step S102, parameterized modeling is performed based on the Frenet coordinate system, the road network data is converted to the Frenet local coordinate system, the virtual information track under the local coordinate system is generated, and the virtual information track segments are established.

[0048] The Frenet coordinate system is a parameterized coordinate system established along the reference path, which can refer to the description of the Frenet coordinate system above. According to the parameter definition of the Frenet coordinate system, the road network data is converted to a plurality of continuous Frenet local coordinate systems, thereby constructing a virtual information track. Specifically, taking the reference path as the reference, the coordinate origin of each local coordinate system is established, that is, s = 0, d = 0, and z = 0. The position of the coordinate origin on the road can be determined along the arc length on the s axis, and the specific manner of determining the coordinate origin will be described below in the description of the handshake area. The reference point coordinates (origin) of each local coordinate system are stored in encrypted form, for example, an irreversible encryption algorithm can be used, and the SM4 algorithm is preferred. The core advantage of the SM4 encryption processing of the reference point coordinates is that SM4 is a symmetric encryption algorithm independently designed by China (belonging to the national encryption algorithm system), and the encryption processing of the reference point coordinates is simple and efficient, low in power consumption, and the encrypted data is "tamper-proof and anti-peeping", strong in data security, and strong in compliance and interoperability.

[0049] The various encryption processes mentioned in the present disclosure can preferably use national encryption algorithms (including SM2, SM3, SM4, etc.), which are not limited to the specific examples mentioned in the present disclosure and can be selected as needed. For example, the unique code is first encrypted using a hash algorithm and then encrypted using a key, or the geographic data is separately encrypted using SM4-GCM and attached with an authentication tag, etc. The national encryption algorithm is a commercial encryption algorithm approved by the National Cryptography Administration and meets the legal requirements of the country, which can avoid compliance risks (the handshake area reference point coordinates are core data supporting high-level autonomous driving safety, and encryption using the national encryption algorithm is a "legal requirement" that can avoid project risks due to non-compliance).

[0050] In step S103, after the virtual information track is established, each local coordinate system is divided into a plurality of segments along the s axis as virtual information track segments. The virtual information track segments can use the rules of the node segments in the collected original map (for example, using the collection nodes determined every second when collecting the map data as the nodes of the virtual information track segments), or can be segmented according to fixed arc length intervals (for example, segmented according to an arc length of 5 meters or 10 meters, etc.), or different segmentation principles can be used according to the complexity of the road. When the road is segmented according to the complexity of the road, the arc length of each segment unit can be different according to the road characteristics, for example, if the road curvature changes sharply (such as sharp turns, roundabouts), the segment length can be shortened as needed to improve the fitting accuracy.

[0051] For example, starting from the origin, the first segment covers s ∈ [0, Δs1], the second segment covers s ∈ [Δs1, Δs2], and so on, so that the entire road is divided into a plurality of continuous arc length segments, each of which corresponds to a certain mileage interval on the road. According to the above local coordinate system established with the road origin as the starting point, the local coordinate system is further segmented to generate a plurality of virtual information track segments.

[0052] The information of each virtual information track segment is encoded. The information can include lane information, road facility information, etc. In order to be able to distinguish each virtual information track segment, a global code is generated for each segment according to a unified road coding system, while each code does not contain any explicit geographical position, but is only associated with the Frenet origin coordinates of the local coordinate system or encrypted data. Specifically, a globally unique identification code (VIT-ID) is generated for each virtual information track segment. The code uses a multi-field combination design. For example, the VIT-ID can include VIT identification, administrative division code, road identification, road segment serial number, lane identification, driving direction, etc. For the virtual information track segment corresponding to the handshake area, the VIT-ID includes the encrypted reference point coordinates of the handshake area. The VIT-ID can also include a check code, which can be used to verify the integrity of the encoded data and prevent tampering. Part or all of this information can be encrypted by encoding as necessary. Since the Frenet local coordinate system replaces the original geographical position, i.e. does not include the original geographical position, the VIT-ID cannot restore the original geographical position. For example, part of the geographical information can be encrypted using a random salt, multi-round iterative hash algorithm. Alternatively, important information (e.g. information related to road information and s-axis) can be irreversibly mapped using an encryption algorithm, and two pieces of information can also be associated with encryption. Through this key-driven irreversible mapping, a third party cannot calculate the actual geographical position of the road or splice a readable map from the VIT-ID data, fully protecting the confidentiality of the map data. The continuous segments within each road can also be identified by the segment index in the VIT-ID.

[0053] In step S104, static attributes and dynamic attributes related to the s-axis position can be bound for each VIT segment. The static attributes mainly include physical attributes of the road and traffic control information, etc. The known information can be obtained from the original static map or surveying data mentioned above and converted into information associated with at least the s-axis coordinates in the offline stage. The static attributes can be pre-configured offline and bound in the form of additional attributes at the s-axis corresponding mileage position of the local coordinate. The physical attribute information can include curvature, slope, lane width, etc. For example, each segment unit associated with the s-axis position can be attached with the following parameters: curvature k(s) (e.g. k=0 for straight line segments, curvature function or average value such as 1 / 50m -1), slope i (e.g., i value is calculated according to i = ΔZ / Δs), lane width range (e.g., the lane d to which the segment belongs belongs to [-175, 175] cm). Traffic control information includes, for example, signal light position (e.g., there is a traffic signal light at a certain segment s = 50 m, the red light duration is 25 seconds, the green light duration is 35 seconds), road speed limit or turning restriction (e.g., left turn is prohibited in the interval s ∈ [1200, 1300] m), etc. These static elements are stored in association with the specific s-axis position by means of key-value pairs or structured data, etc., so that the autonomous vehicle can directly look up the relevant road information according to the current position. These information can be attached to the offline map as static attributes, or can be updated in real time as dynamic attributes through real-time collection and calculation (see the description of dynamic road attributes below). The parameters of the static attributes are associated with the s-axis arc length, which can be directly called by the autonomous driving system.

[0054] Dynamic attributes can also include physical properties and traffic control information of the road. Dynamic attributes can also include other information related to driving, such as visibility, traffic flow, road friction coefficient, etc. The information collected by online vehicles or road management information collection systems on the road can be collected to the control system (e.g., can be called "virtual information track" network) in the cloud, and the control system generates or updates the dynamic attributes according to the collected information, and sends the dynamic attributes to the traveling vehicle according to the vehicle position, to update the dynamic attributes of the road in real time for the vehicle. The parameters of the dynamic attributes are associated with the s-axis arc length, which can also be directly called by the autonomous driving system.

[0055] Figure 3 A flow chart of a method for online real-time updating of a map and vehicle positioning according to an embodiment of the present disclosure is shown. The method realizes accurate positioning of vehicles and VIT s / d axes, cross-segment calibration, and dynamic information fusion.

[0056] At step S201, the autonomous vehicle performs VIT matching and key verification with the cloud. After the autonomous vehicle is powered on or enters a certain area, it requests a VIT index list near the current location from the control system of the cloud. The autonomous vehicle preliminarily filters one or more VIT segments that may be located according to the starting point coordinates of the VIT segments, for example, based on a navigation satellite system positioning to filter candidate VIT segments. Then, the vehicle verifies the ID of the candidate VIT segment using a preloaded local key, for example, sends the local key and the VIT-ID to the cloud for authorization verification. After verification, the cloud confirms the VIT segment corresponding to the autonomous vehicle according to the starting point coordinates of the preliminarily filtered VIT segment. The target VIT segment corresponding to the vehicle returns the coordinate information in the Frenet coordinate system and / or the offline map information, such as the s-axis range, the d-axis lane parameters, and the static attributes bound to the segment. The above data do not contain any real geographical location. Through the above process, the vehicle can safely obtain the required high-precision map information, and only authorized vehicles can interpret the complete VIT content, further improving the security of data use.

[0057] At step S202, the vehicle calculates the positioning in the Frenet coordinate, and calibrates the positioning deviation through the roadside information code and / or the handshake area. The calibration of the positioning deviation is mainly used to calibrate the coordinate deviation of the s-axis, and also calibrates other parameters on the coordinate. After the vehicle obtains the map information of the corresponding VIT segment, it calculates the coordinate of the vehicle relative to the reference path in the Frenet coordinate system by using the multi-sensor fusion and pose estimation technology of the vehicle itself, to realize the centimeter-level positioning.

[0058] Optionally, when positioning the s-axis, the relative position of the vehicle in the local coordinate system can be determined by measuring the incremental arc length of the vehicle through the wheel odometer.

[0059] In order to eliminate the cumulative error, the visual sensor can be combined to identify the roadside information code (for example, two-dimensional code, RFID tag or visual signboard, ground magnetic sign, etc.) pre-deployed along the road, which contains the information of the current accurate positioning value (for example, s value, etc.). The vehicle reads the information of the roadside information code through the visual camera (front / side camera), radio frequency card reader or magnetic sensor device, and accurately calibrates the position of the s-axis, and the accuracy can even reach ±1 cm. For example, when the s position calculated by the odometer is 50.15 m, and the true s reference obtained by reading the roadside information code is 50.00 m, the vehicle immediately corrects its own s to 50.00 m, thereby eliminating the drift error of the odometer. The roadside information code can also carry the identification of the current VIT segment for the vehicle to identify, so that the vehicle can quickly confirm the segment to which it belongs without frequent requests to the cloud, improve the positioning efficiency, and reduce the communication delay.

[0060] The roadside information code is described in detail below.

[0061] The roadside information code is a passive / low-power physical coding device (in the form of a two-dimensional code, an anti-interference bar code, a radio frequency identification tag (RFID), a laser reflection code, a magnetic marker, etc.) deployed at the edge of the road (such as the ground, guardrails, kerbs, etc.). The roadside information code can be set at fixed intervals, or set according to the complexity of the road (for example, areas with large changes in curvature such as curves and roundabouts can be set with encryption), or set according to the "on-demand deployment" scheme (as described below). The roadside information code only serves the positioning calibration in the Frenet coordinate system.

[0062] The roadside information code can include local parameters corresponding to virtual information track segments, and in the design, the principle of "not revealing real geographical information, only serving Frenet positioning" is strictly followed. All fields come from preset parameters in the offline construction phase, and there is no sensitive information such as geographical position, real road name or milepost number, etc. Even if the coding plate is maliciously obtained, the real position of the road cannot be reversely deduced. All data of the roadside information code is pre-prepared in the offline construction phase, and the vehicle does not need to be connected when reading, only real-time collection and analysis by local sensors, which can ensure that the positioning delay is less than 50ms, and can also avoid the risk of data transmission.

[0063] The roadside information code can contain core necessary information and other extended information as needed. The core necessary information of the roadside information code includes the basic fields to ensure positioning calibration and matching, which is static data pre-prepared in the code offline, and it can only contain local static parameters of a single segment and cannot be modified. For example, the roadside information code includes VIT segment identification and s-axis reference value (with high precision), which are completely aligned with the VIT model and strictly consistent with the parameters of the VIT segment constructed offline. In addition, the roadside information code can also include coding plate numbers continuously numbered in the same road, with each number corresponding to a roadside information code of a road. The roadside information code can also have built-in verification information, which can verify the data integrity locally after being read by the vehicle (such as matching the verification code to confirm that the code has not been tampered with), to ensure calibration accuracy.

[0064] The "on-demand deployment" scheme of the roadside information code described above is only set in high-risk areas of positioning, such as navigation satellite system positioning lock loss or high-curvature road segments (such as interchanges), continuous tunnel groups, etc. Special scenarios are set, so that the following effects can be achieved:

[0065] (1) Construction cost is greatly reduced: the total deployment quantity can be reduced by more than 80% (for example, for a 100-kilometer road, only 20-50 information codes are needed), significantly reducing equipment procurement and maintenance costs;

[0066] (2) Stronger scene focus: Prioritize coverage of high-risk areas for positioning, such as GNSS positioning lockout, visual limitations, and high-curvature turns, to improve resource utilization efficiency.

[0067] (3) Improved maintenance convenience: Information codes are mainly distributed in well-defined sections that can be centrally maintained, facilitating periodic inspection and replacement.

[0068] Roadside information codes are offline-prepared "local coordinate calibration anchors." Their core function is to assist autonomous vehicles in accurately calibrating s-axis (arc length axis along the reference path) positioning in the Frenet coordinate system, eliminating the cumulative errors of sensors such as wheel odometry and IMU, while avoiding the disclosure of any real geographical information. Roadside information codes can be used primarily in situations where vehicles are in areas without network coverage (such as tunnels) or when cloud interaction delays are too high (such as during peak traffic congestion). They can independently complete s-axis calibration without relying on the cloud, ensuring uninterrupted positioning.

[0069] When positioning the d-axis, sensors such as vehicle-mounted cameras and laser radars can be used to identify lane lines or road edges, measure the lateral distance of the vehicle from the reference path, and calculate the d-axis offset, with an accuracy of ±2 cm. Combined with the lane width data of the section in the map, the vehicle's lateral position can be monitored to ensure it remains within a safe range.

[0070] When positioning the θ-axis, high-precision IMU can be used to measure the vehicle's heading and compare it with the tangent direction of the s-axis in the map to calculate the deviation θ between the vehicle's heading angle and the s-axis in real time, and adjust the vehicle's driving posture accordingly. The typical accuracy can reach ±0.1°. By integrating s, d, and θ positioning, the vehicle can achieve real-time centimeter-level positioning in the Frenet coordinate system.

[0071] In this method, to prevent small errors from gradually accumulating after long-distance driving, the disclosure also provides a periodic cloud calibration mechanism - "handshake zone." Figure 4A is a schematic diagram of the distribution of handshake zones on physical roads according to an embodiment of the disclosure; Figure 4B is a schematic diagram of the corresponding virtual information track map and its handshake zone distribution Figure 4A Handshake zones are used for coordinate resetting and position calibration. As shown in the figure, handshake zones are arranged along the s-axis, and the coordinates will be reset when the vehicle passes through the handshake zone. The interval of the handshake zone corresponds to the local coordinate system.

[0072] The following principles can be followed when setting handshake zones:

[0073] (1) Based on the dynamic adaptation of road complexity, complex scene encryption configuration handshake area. Specifically, the cumulative error rate of different road scenes is different, so the interval needs to be differentiated to ensure that the error does not exceed the threshold when the calibration is completed. On the regular road, such as the city curvature of the low road section, the cumulative rate of vehicle heading angle drift and d-axis offset is slow, even if a larger interval is set, the calibration can be completed before the error exceeds the threshold, for example, set the handshake area with 10km interval, the length of the s-axis of the corresponding local coordinate system is 10km. The handshake area can be set at the VIT connection place of different types of roads, different administrative areas, etc., to synchronize the calibration of the coordinate system deviation across regions. For complex roads, such as continuous turns, interchanges, roundabouts, etc., the error accumulation will be faster, so smaller interval handshake areas need to be set.

[0074] (2) The setting of the handshake area needs to make the whole road section without calibration blind area, to realize the positioning calibration coverage of the whole life cycle of the road, and to avoid the out-of-control of the cumulative error.

[0075] (3) The handshake area and the roadside information code are complementary, forming a system that combines "low-frequency global calibration" and "high-frequency local correction". The roadside information code can be used to ensure uninterrupted positioning and is responsible for high-frequency local correction in high-risk areas. The handshake area is responsible for low-frequency global calibration in the interval, solving the problem of large error accumulation, and covering the whole scene calibration demand through the cooperation of the two.

[0076] (4) The handshake area must have clear physical identification, and its material is suitable for long-term stable work outdoors, which is convenient for vehicle sensors to quickly identify and trigger the reconstruction process, avoiding calibration omission.

[0077] According to the above principle, the nodes of the handshake area correspond to the local coordinate system and are arranged at intervals along the s axis. For example, a handshake area is arranged at the junction of adjacent local coordinates. The information of the handshake area does not contain real coordinates, and only the encrypted reference identifier of the location is stored in the cloud. At the same time, an obvious identifiable marker can be deployed on the corresponding roadside. When the vehicle approaches the handshake area, the vehicle-mounted system can upload the accumulated positioning error information in the current VIT segment to the cloud, for example, reporting "s-axis deviation Δs = +0.15m, d-axis deviation Δd = -0.08m". The cloud calculates the calibration parameters (for example, the s value, d value, and z value should be calibrated to 0) according to the pre-saved encrypted s reference point of the handshake area, and sends these parameters to the vehicle. When the vehicle passes the handshake area identifier, it triggers the execution of calibration. The vehicle resets its own coordinates and adjusts the θ axis direction to be consistent with the tangent direction of the new road segment reference path. Through the "resetting" effect of the handshake area, the accumulated error caused by long-distance cross-segment driving can be completely eliminated, and the positioning accuracy at the junction of road types (for example, the junction of urban trunk roads and expressways), administrative regions, or complex road segments (for example, on and off the interchange, and in and out of the roundabout) can be ensured. In the scheme of the present disclosure, the combination of high-frequency local calibration (road test information code) and low-frequency global calibration (handshake area) can ensure that the positioning accuracy of the entire road segment is always maintained within ±10cm. This method takes into account the advantages of high-frequency and low-frequency calibration.

[0078] The s value of the roadside information code is the Frenet arc length, and the encrypted control point of the handshake area is a "fuzzy reference without real coordinates" (only the cloud can calculate, and the real geographical position is not sent to the vehicle), which fully meets the national geographic information security requirements and does not leak sensitive information.

[0079] Optionally, in order to prompt the approach to the handshake area and avoid the jump of coordinates during the handshake area correction, at least one roadside information code can be deployed near the handshake area (for example, it can be arranged about 100m in front). Before the vehicle enters the handshake area, the deviation of the s axis can be corrected through the roadside information code (for example, s = 999.85m is corrected to 999.98m), and then the global calibration is completed in the handshake area to further improve the reconstruction accuracy.

[0080] In step S203, the cloud updates the static and / or dynamic attributes according to the VIT segment and the road information obtained from the vehicle / road side unit (RSU, Road Side Unit) and sends them to the vehicle. When the road side unit uploads data to the cloud, it can also be verified by the above-mentioned key and fixed code matching method.

[0081] As described above, the static attributes can be curvature, slope, lane width, signal light position, turning restriction, etc. The static attributes are relatively stable and generally do not need to be frequently updated. When the cloud detects that the static attributes change, the cloud will update the static attributes stored by it in real time. The dynamic attributes can include road, traffic and environmental information that changes over time, etc. For example, the RSU collects dynamic information related to the s-axis in real time, fuses to the VIT segment through the navigation satellite system / IMU / odometry / visual / radar, etc., and uploads the measurement results. The cloud can fuse, verify and conditionally update the dynamic information. When the information validity is verified, the dynamic attributes of the corresponding VIT segment are updated, and can be issued to all vehicles within the segment through the vehicle wireless communication network. The dynamic information received by the cloud can include, for example, roadside unit information, road positioning information, navigation satellite system positioning geographic position information, vehicle dynamic information, road scene information, traffic control information, weather information, etc. Effective information can be extracted from them, and the dynamic attributes of the VIT segment are fused, generated and / or updated, so as to realize advanced automatic driving through information decision. For example, the dynamic information can be the road friction coefficient μ (such as detecting that a certain road section is wet and slippery due to rain, then the μ value of the segment is reduced), traffic flow, visibility (such as global visibility of 800m) and other event information such as accidents, construction, congestion, etc. The initial value of the dynamic attribute can be set to a default value in the offline stage, and is corrected by online data in actual application. All dynamic attributes are managed with s-axis mile segments as indexes, and the latest dynamic parameters can be directly obtained according to the VIT segment to which the current position of the vehicle belongs during vehicle driving, so as to realize timely response to environmental changes.

[0082] As can be seen from the above embodiments, the method of the present disclosure can effectively construct and apply high-precision maps in actual road environments. In the offline stage, the original map is encrypted and parameterized to generate a virtual information track that meets the security requirements; in the online stage, the dynamic attributes are updated by the information collected by the RSU. The vehicle realizes accurate positioning and dynamic information acquisition with the help of the VIT. The vehicle positioning error can be kept at a few centimeters, and timely support information can be obtained from the map layer when encountering complex road conditions such as overpasses, signal lights, rain, etc. Therefore, the method of the present disclosure can be applied to advanced automatic driving.

[0083] The virtual information track of the present disclosure is a new type of map expression mode for automatic driving control and path planning. The system replaces the traditional latitude and longitude sandbox or absolute geographic coordinate system with a defined Frenet local coordinate system, realizing direct abstraction of road geometry and physical properties. In the VIT model, the map is no longer centered on "human-readable" geographic locations (latitude, longitude, road name), but is converted into a "control-available geometric tensor space", where each coordinate point carries a set of parameters that can be directly used by the automatic driving control system. This not only facilitates the automatic driving system to directly read the control required data, but also eliminates the dependence on readable geographic locations at the architecture level, thereby having natural security and anti-leakage characteristics.

[0084] Without reducing the function of the vehicle, the irreversible / key-controlled mapping is used to encrypt the road coding and geometric meta-information and publish the VIT, avoiding any receiving end (if without the key) to obtain the geographic location or generate the map for human identification. The combination of VIT segmentation and encryption method is different from the conventional encryption or simple cutting of the public, and it is an encryption strategy for the available data of the automatic driving function.

[0085] The local coordinate system collects the local geometry, dynamics and properties (such as friction, roughness, slope) of the road on the virtual information track segment node, supports centimeter-level precision, and facilitates the controller to directly read the required information, reducing the real-time geometry solving overhead.

[0086] Embodiments of the present application can be modified and changed in various ways without departing from the spirit and scope of the present application. Therefore, it should be understood that the scope of protection of the present application should not be limited to the above exemplary embodiments, but should cover the full scope defined by the claims and their equivalents.

Claims

1. A method for producing high-precision maps based on virtual information track models, characterized in that, include Extract road network information from raw map data or surveying data, and generate reference routes based on the road network information; Based on the reference path, multiple Frenet local coordinate systems are established. Each Frenet local coordinate system includes at least an arc length axis along the reference path, a lateral offset axis perpendicular to the reference path, and an elevation axis perpendicular to the road surface. The reference path is divided into virtual information track segments along the arc length axis, and each Frenet local coordinate system includes multiple virtual information track segments; A globally unique identifier code is generated for each virtual information track segment. This globally unique identifier code includes encryption of at least a portion of the geographic information of each virtual information track segment. Each virtual information track segment is bound to a static attribute parameter associated with the arc length axis.

2. The method for producing high-precision maps based on virtual information track models as described in claim 1, characterized in that, Also includes The vehicle is verified online to determine the virtual information track segment in which it is located; Centimeter-level positioning is performed in the Frenet coordinate system based on information detected by multiple sensors of the vehicle. The positioning deviation is calibrated using the roadside information code and / or handshake area; as well as The vehicle receives road information and updates the dynamic attribute parameters of the virtual information track segment in which the vehicle is located based on the road information. The dynamic attribute parameters are associated with the arc length axis.

3. The method for producing high-precision maps based on virtual information track models as described in claim 2, characterized in that, The information detected by the multiple sensors includes the vehicle's heading angle, which is compared with the tangential direction of the arc length axis to determine the deviation of the vehicle in the direction of travel.

4. The method for producing high-precision maps based on virtual information track models as described in claim 1, characterized in that, The encryption process for at least a portion of the road geographic information of each virtual information track segment includes irreversible encryption of the at least a portion of the geographic information.

5. The method for producing high-precision maps based on virtual information track models as described in claim 1, characterized in that, The multiple Frenet local coordinate systems and the globally unique identifier of each virtual information track segment do not include real geographic coordinates. The globally unique identifier of each virtual information track segment corresponds to the position of each virtual information track segment in the Frenet local coordinate system.

6. The method for producing high-precision maps based on virtual information track models as described in claim 1, characterized in that, The static attribute parameters bound to each virtual information track segment include one or more of the following: curvature, slope, lane width and / or lane width range, road facilities, traffic sign information, and driving restriction information, corresponding to the mileage position of the arc length axis.

7. The method for producing high-precision maps based on virtual information track models as described in claim 2, characterized in that, The dynamic attribute parameters associated with the arc length axis include one or more of the following: road surface friction coefficient, traffic flow, visibility, road events, and weather conditions.

8. The method for producing high-precision maps based on virtual information track models as described in claim 2, characterized in that, Each handshake area corresponds to one of the multiple Frenet local coordinate systems and has an encrypted arc length axis reference point. When the vehicle travels to the handshake area, it uploads the currently accumulated deviation data of the arc length axis and lateral offset axis online, and resets the Frenet local coordinate system in which the vehicle is currently located according to the received calibration parameters.

9. The method for producing a high-precision map based on a virtual information track model according to claim 2, characterized in that, The roadside information code is at least set on roads where the navigation satellite system positioning signal is lost or on roads with high curvature. The vehicle can correct its deviation on the arc length axis and lateral offset axis in real time by reading the roadside information code.

10. The method for producing a high-precision map based on a virtual information track model according to claim 1, characterized in that, Dividing the reference path into virtual information track segments along the arc length axis includes dividing the virtual information track segments according to the collection nodes of the original map data.

Images