Parking function automatic evaluation method and system and medium

Abstract

The invention discloses a parking function automatic evaluation method and system and a medium. The method comprises the following steps: acquiring each theme data in a parking process; analyzing the main body data to extract evaluation data required by evaluation; preprocessing the evaluation data, and generating parking initial state data, parking end state data and parking end state data according to the preprocessed evaluation data; and calculating key performance indexes of the parking function based on the parking initial state data, the parking end state data and the parking end state data. The method is used for solving the problems that manual evaluation of the parking function is high in test cost and low in efficiency, some recorded data are not accurate enough, and some KPI indexes cannot be counted.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving evaluation technology, and in particular to a parking function automatic evaluation method, system and medium. Background Art

[0002] Fully automatic parking is a lateral and longitudinal auxiliary control function in low-speed parking scenarios based on surround-view cameras, ultrasonic radars and intelligent driving controllers. It supports low-speed background silent search for parking spaces, that is, the background search is turned on by default when the conditions are met and the driver is reminded of the parking space search results. After the user confirms to start parking, the system controls the vehicle to park in and out horizontally and longitudinally according to the planned path; if a narrow parking space is detected, the driver is supported to get off in advance when parking part of the space, and automatically complete the second half of the parking after confirmation.

[0003] In order to provide users with a better driving experience, it is necessary to conduct functional tests on the actual vehicle and output the evaluation KPI indicators. By comparison, it is possible to find out whether the new version is better, how much each indicator has been improved, and where there are still major problems, and then the automatic parking function can be optimized. However, many parking tests are currently manually evaluated, through manual marking, manual measurement, and then documented. The use of manual evaluation methods has problems such as high testing costs, low efficiency, inaccurate recorded data, and the inability to count some KPI indicators. Summary of the invention

[0004] In view of the above technical problems, the present invention provides a parking function automatic evaluation method, system and medium, which are used to solve the problems of high testing cost, low efficiency, inaccurate recorded data, and inability to count some KPI indicators in manual evaluation of parking functions.

[0005] A first aspect of the present invention provides a method for automatically evaluating a parking function, comprising: Obtain data on various topics during the parking process; Parsing the data of each subject to extract the evaluation data required for evaluation; Preprocessing the evaluation data, and generating parking start state data, parking end state data, and parking exit end state data according to the preprocessed evaluation data; Based on the parking start state data, the parking end state data and the parking exit end state data, a key performance indicator of a parking function is calculated.

[0006] In a possible implementation, the acquiring of various subject data during the parking process includes: The data recording tool is used to record the perception data, fusion data, map data, positioning data and vehicle status data during the parking process.

[0007] In a possible implementation, parsing the subject data to extract the evaluation data required for evaluation includes: Identify the software version of the recorded package; Pull the message definition corresponding to the software version from the code repository; Extract the required subject data and signal data according to data requirements.

[0008] In a possible implementation, the preprocessing of the evaluation data includes: Temporal alignment of the topics; Combine all subject data into one large table; Fill missing values ​​based on business logic.

[0009] In a possible implementation, generating the docking initial state data includes: Obtain the coordinate data of the four corner points of the parking space, the parking space type data, the parking space length and width data, and the vehicle body width data; Calculate the parking space position and lateral distance at the starting state.

[0010] In a possible implementation, generating the parking end status data includes: Calculate the steering angle after parking and the wheel track after parking.

[0011] In a possible implementation, generating the parking end status data includes: Calculate the lateral distance after parking, the longitudinal distance after parking and the deflection angle after parking.

[0012] In a possible implementation, calculating the key performance indicators of the parking function includes calculating at least one of the following indicators: Parking time, average parking speed, maximum parking speed, number of parking maneuvers and number of parking re-plannings.

[0013] In a possible implementation, calculating the key performance indicators of the parking function further includes calculating at least one of the following indicators: The duration of parking, average parking speed, maximum parking speed, number of parking maneuvers and number of parking re-planning.

[0014] In a possible implementation, the automatic parking function evaluation method further includes: The total number of parking attempts and the number of completed parking attempts are counted, and the parking completion rate is calculated.

[0015] In a possible implementation, the automatic parking function evaluation method further includes: The total number of parking outs and the number of parking out completions are counted, and the parking out completion rate is calculated.

[0016] A second aspect of the present invention provides a parking function automatic evaluation system, comprising: A data acquisition module, at least used to acquire various subject data during the parking process; A data extraction module, at least used to parse the data of each subject to extract the evaluation data required for the evaluation; A data processing module, at least used to pre-process the evaluation data, and generate parking start state data, parking end state data and parking exit end state data according to the pre-processed evaluation data; The result generating module is configured to calculate a key performance indicator of a parking function based on at least the parking start state data, the parking end state data and the parking exit end state data.

[0017] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a computer, the method described in the first aspect is executed.

[0018] The present invention uses an automated method to replace manual evaluation, which can accurately obtain and analyze multi-topic data in the parking process, extract the information required for evaluation, and pre-process it efficiently, thereby generating the start and end state data of parking in and out. Based on these data, the key performance indicators (KPIs) of the parking function can be automatically calculated, including vehicle speed, parking time, parking space offset angle, number of re-planning, etc., and some KPIs that are difficult or impossible to count manually are output. Compared with the existing technology, this method greatly reduces the testing cost, improves the evaluation efficiency, ensures the accuracy and integrity of data records, and provides strong support for the optimization and quality evaluation of the parking function of autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flowchart of an embodiment of the automatic evaluation method of parking function of the present invention.

[0020] Figure 2 FIG. 4 is a flow chart of an embodiment of a method for automatically evaluating parking function according to the present invention.

[0021] Figure 3 FIG. 1 is a schematic diagram of the structure of an embodiment of the automatic parking function evaluation system of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0023] It should be understood that the terms "first", "second", and "third" etc. in the claims, specifications, and drawings of the present disclosure are used to distinguish different objects rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present disclosure indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections. It should also be understood that the terms used in this disclosure specification are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0024] Reference Figure 1 and Figure 2 , an embodiment of the present invention discloses a method for automatic evaluation of parking function, comprising the following steps.

[0025] S1, obtain various subject data during the parking process.

[0026] In some embodiments, various data topics during the parking process may be recorded in real time by a data-logger tool installed in the vehicle. These topic data may include perception data, fusion data, map data, positioning data, and vehicle status data.

[0027] In some embodiments, the data recording may combine the outputs of different sensors (such as surround view cameras, ultrasonic radar, lidar, etc.) and vehicle controllers to ensure comprehensive coverage of key information involved in the parking process.

[0028] In some implementations, data may be organized in time series and recorded in a compressed package format to improve storage and transmission efficiency.

[0029] In some implementations, relevant metadata, such as recording time, number of recording topics and specific content, etc., need to be synchronized during the recording process and stored as a YAML file to provide support for subsequent data analysis and evaluation.

[0030] In some implementations, if there are breakpoints or anomalies in the recording process, incomplete data packets can be discarded through a real-time verification mechanism to ensure the integrity and accuracy of the data.

[0031] S2, parsing the data of each subject to extract the evaluation data required for evaluation.

[0032] In some implementations, the corresponding software version is automatically identified based on the meta information of the recorded data packet, and a matching message definition file (msg) is obtained from a code management platform (such as GitLab) to ensure the correctness and consistency of the parsing.

[0033] In some implementations, the data compression package is decompressed to extract the stored subject data (such as perception data, vehicle status data, etc.), and the integrity of the metadata file is checked. If necessary, the missing yaml file is generated or supplemented through the ROS2 yaml repair tool.

[0034] During the parsing process, target signal data is extracted from each subject according to the evaluation requirements, such as vehicle speed, APA (intelligent parking assist) state machine status, parking space geometry information, etc.

[0035] In some implementations, abnormal or incomplete data packets may be discarded to ensure that the extracted evaluation data is of high quality and credibility.

[0036] Finally, the parsed key data are organized in a predefined format to provide reliable input for subsequent preprocessing and evaluation.

[0037] S3, preprocessing the evaluation data, and generating parking start state data, parking end state data and parking exit end state data according to the preprocessed evaluation data.

[0038] In some embodiments, the extracted subject data are time-aligned, and the time error is allowed to be within a certain range (such as 0.1 second) through a fuzzy matching method, so that the multi-source data are synchronously integrated into a unified time series.

[0039] In some implementations, missing data values ​​are processed according to business logic, such as by interpolation or logical deduction, to ensure data integrity and continuity.

[0040] In some implementations, the start and end nodes of parking in and parking out are identified based on the state machine changes, for example, state data such as parking in start time, parking in end time and parking out end time are extracted using key frame markers of the APA state machine.

[0041] At the same time, in some implementations, key information such as the corner coordinates of the parking space, the parking space type, the initial and end positions of the vehicle, etc. can be calculated in combination with the geometric data.

[0042] Finally, the preprocessed data is organized in a structured format, and the parking start state data, parking end state data, and parking exit end state data are output respectively, laying the foundation for the subsequent performance index calculation.

[0043] S4, calculating a key performance indicator of a parking function based on the parking start state data, the parking end state data and the parking exit end state data.

[0044] In some implementations, indicators related to parking efficiency, such as parking entry time, parking exit time, average vehicle speed, and maximum vehicle speed, may be statistically analyzed based on the parking entry and exit status data.

[0045] In some implementations, the geometric position information of the start and end of parking can be used to calculate the deflection angle of the vehicle after parking, the offset distance of the vehicle after parking, and whether it exceeds the allowable range (such as the longitudinal distance exceeds the limit).

[0046] In some implementations, the stability of path planning and control during parking can be evaluated by counting the number of parking maneuvers and replanning times through vehicle status data (such as the number of gear shifts and path replanning signals).

[0047] In some implementation manners, the steering wheel angle changes at the end of parking and exiting can be directly calculated to measure the accuracy of the parking trajectory.

[0048] Finally, these key performance indicators are integrated and structured into an evaluation report format, daily parking test results are generated and automatically exported to a storage system (such as NAS) to support subsequent analysis and optimization.

[0049] Furthermore, as an implementation mode of the present invention, step S1, obtaining various subject data during the parking process, includes: recording perception data, fusion data, map data, positioning data and vehicle status data during the parking process through a data recording tool.

[0050] In some embodiments, a professional data logging tool (such as a data-logger) may be used to access the vehicle's sensor and controller interfaces to record multi-topic data generated during the parking process in real time using high-speed sampling.

[0051] In some embodiments, the perception data may obtain surrounding environment information through a surround-view camera, radar, or lidar, and the fusion data is generated by the comprehensive calculation results of the perception module and the map positioning module.

[0052] In some embodiments, map data may be obtained from parking space location and path planning information provided by a high-precision map system.

[0053] In some embodiments, the positioning data may be obtained by updating the vehicle's global position in real time via a GNSS or an onboard inertial navigation system.

[0054] In some implementations, the vehicle status data may be obtained through dynamic status information such as the vehicle speed, gear position, steering wheel angle, etc.

[0055] The data recording process can be automatically started or ended in conjunction with trigger conditions (such as the start or stop of the parking state machine). At the same time, the data integrity can be verified in real time and organized by timestamp and subject classification during storage to facilitate subsequent parsing and analysis.

[0056] Further, as an implementation mode of the present invention, step S2, parsing the various subject data to extract the evaluation data required for evaluation, includes: identifying the software version of the recording package; pulling the message definition corresponding to the software version from the code repository; and extracting the required subject data and signal data according to data requirements.

[0057] In some implementations, identifying the software version of the recorded package includes: directly reading and matching the corresponding software version information through a metadata field (such as a software version number or a build timestamp) embedded in the recorded package file.

[0058] In some implementations, if the package recording file does not include an explicit software version number, the version can be inferred by comparing the package recording time with the software release history record.

[0059] In some implementations, the software version may also be indirectly confirmed by recording the format or field definition of certain specific subject data in the package and comparing it with the message definition of a known version using a unique identifier (such as a message field name or structure difference).

[0060] In some implementations, the history records of version management tools (such as Git, SVN) may be queried, and verification may be performed based on the environment configuration files or compilation records when the package was recorded to ensure the accuracy and consistency of the identification results.

[0061] In some implementations, if the version information still cannot be determined, a default version may be used or the user may be prompted for manual confirmation to ensure the reliability of the parsing process.

[0062] In some embodiments, pulling the message definition corresponding to the software version from the code repository includes: using the software version number as a key identifier, locating the corresponding version branch or tag from the version control system (such as GitLab, GitHub, SVN, etc.), and ensuring that the obtained message definition is consistent with the software environment when the package is recorded.

[0063] In some implementations, the pulling process may use an API interface or a command line tool (such as Git clone, Gitfetch) to download a message definition file (such as a msg or IDL file) to a local parsing environment.

[0064] In some implementations, if the warehouse adopts a distributed storage structure, it can automatically switch to the correct remote warehouse according to the server address or warehouse path in the package recording metadata.

[0065] In some implementations, if the version control system supports an archiving function, the packaged file of a specific version may also be directly downloaded for decompression and extraction.

[0066] In some implementations, to ensure the efficiency and reliability of the pulling process, a version cache mechanism may be combined to preferentially load message definitions from the local cache for commonly used versions without having to re-download from the remote repository each time.

[0067] In some implementations, if the pull fails, the user may be prompted to check the network connection or repository permissions to ensure that the acquired message definition is accurate.

[0068] In some embodiments, extracting required subject data and signal data according to data requirements includes: identifying all recorded subject data and their corresponding timestamps and field information by parsing recording package metadata (such as a yaml file).

[0069] In some implementations, the required topic data and specific signal data may be screened out based on a predefined requirement configuration file (including a target topic name and a signal field list).

[0070] In some implementations, the time series information of the subject data may be combined to obtain target data in a specific time period or under event triggering by using index matching or dynamic query.

[0071] In some implementations, if the recording package format is a compressed file (such as rosbag), it can be first decompressed into an intermediate format (such as a db3 file) in order to directly read the target content.

[0072] In some implementations, for complex requirements, nested structures or array fields of the subject data may be parsed to extract specific signal values ​​therein.

[0073] In some embodiments, if certain topics are missing or the data is abnormal (such as a time series break), the missing signal value can be derived and calculated from the related topics according to the backup logic.

[0074] Finally, the extracted subject data and signal data are formatted and stored as required, providing clear and structured input for subsequent analysis.

[0075] Furthermore, as an implementation mode of the present invention, in step S3, the preprocessing of the evaluation data includes: time alignment of each subject; merging all subject data into a large table; and filling in missing values ​​according to business logic.

[0076] In some implementations, time alignment of each topic includes: sorting all topic data by timestamp and unifying the time reference, such as using the timestamp of a certain topic as the main reference or selecting the earliest time point as the global reference.

[0077] In some implementations, a fuzzy matching method is used for timestamps of different topics, and close time points are regarded as alignment points within a set time difference tolerance range (such as 0.1 seconds).

[0078] In some embodiments, if the subject data timestamps are inconsistent, the misaligned data can be filled in by linear interpolation, nearest neighbor method, or spline interpolation.

[0079] In some implementations, for topics with high-frequency data, sampling and frequency reduction can be performed based on the time points of low-frequency topics to reduce data redundancy.

[0080] In some implementations, for subjects with missing or abnormal timestamps, global time series generation rules can be used or time points can be derived from other synchronized subjects to complete supplementary alignment.

[0081] Ultimately, all aligned data are stored in a sequence format with a unified timeline, ensuring that multi-subject data can be accurately matched on the same time basis, laying the foundation for subsequent data processing and analysis.

[0082] In some embodiments, merging all subject data into a large table includes: after completing time alignment, horizontally splicing the data of each subject according to the timestamp alignment rules, using the timestamp as the primary key, to ensure that multiple subject data at the same time point are merged into one row.

[0083] In some implementations, for incompletely aligned subject data, placeholders may be filled with default values ​​(such as NaN) or interpolated to maintain the integrity of the table.

[0084] In some implementations, when merging, a source identifier or field prefix may be added to each subject data to avoid field name conflicts.

[0085] In some implementations, if the amount of subject data is large, batch loading or block merging methods can be used to reduce memory consumption.

[0086] In some implementations, derived fields that need to be calculated (such as ratios or differences of multi-subject data) can be completed synchronously during the merging process, thereby simplifying subsequent processing steps.

[0087] Finally, the merged results are stored in a structured table format (such as DataFrame or CSV), which supports efficient retrieval and storage, forming a comprehensive data table that is convenient for subsequent analysis.

[0088] In some embodiments, filling missing values ​​according to business logic includes: selecting a suitable filling method according to the data type and context logic. For example, for time series data, forward filling (filling the missing point with the last valid value), backward filling (using the next valid value), linear interpolation, or polynomial interpolation can be used to fill missing values.

[0089] In some implementations, for categorical data, the mode of the field or a default value defined by business rules may be used to fill in the gaps.

[0090] In some embodiments, if the values ​​of certain fields have a logical association (such as the relationship between speed and acceleration), the missing values ​​can be derived through calculations of the related fields.

[0091] In some implementations, for some key indicators, if business rules allow, random values ​​within a threshold range or business-defined specific values ​​may be set for filling to ensure the integrity and feasibility of subsequent analysis.

[0092] In some embodiments, obviously abnormal or uninferable missing data may also be marked as special values ​​(such as -1 or NaN) for screening during analysis.

[0093] The final filling logic needs to be based on specific business needs and data distribution characteristics to ensure that the filled data is meaningful and reliable for result analysis.

[0094] Further, as an implementation mode of the present invention, in step S3, generating parking starting state data includes: obtaining the coordinate data of the four corner points of the parking space, the parking space type data, the parking space length and width data, and the vehicle body width data; and calculating the parking space position and lateral distance in the starting state.

[0095] In some embodiments, the coordinate data of the four corner points of the parking space are obtained from the perception module or the fusion data, and the exact positions of these coordinates in the global or local coordinate system are confirmed in combination with the vehicle positioning information.

[0096] In some embodiments, the parking space type data may be directly extracted through a parking state machine or a parking space recognition algorithm (eg, based on parking space line detection or space perception) to distinguish between vertical parking spaces, horizontal parking spaces, or special types of parking spaces.

[0097] Exemplarily, the parking space type data includes two dimensions: "parking space type_vertical or horizontal" and "parking space type_line or space". "Parking space type_vertical or horizontal" is determined by the type field in the APA state machine at the beginning of parking, where type 1 indicates that the parking space is vertical, and type 2 indicates that the parking space is horizontal, which is used to describe the geometric arrangement direction of the parking space. "Parking space type_line or space" is determined by the source field in the APA state machine at the beginning of parking, where source 1 indicates a line parking space (based on parking space line recognition), 2 indicates a space parking space (based on spatial measurement judgment), 3 indicates a fused parking space (comprehensive recognition of line and space information), and 4 indicates a virtual fused parking space (inferred comprehensive result). This classification method clearly distinguishes the geometric layout of parking spaces from the source of perception, providing basic support for the adjustment and evaluation of parking strategies.

[0098] In some embodiments, the length and width data of the parking space are directly obtained by calculating the side lengths of four corner points or from the parking space attribute field of the perception module.

[0099] Exemplarily, the parking space length and width data include "parking space length" and "parking space width". The calculation of "parking space length" and "parking space width" is based on the length and width values ​​in the APA state machine at the beginning of parking, representing the initial length and width of the parking space, respectively. To ensure that the definition of the length and width of the parking space is consistent with the actual physical meaning, if the width (width) of the parking space is greater than the length (length), the values ​​of the two are swapped to ensure that the "length" is always the larger side of the parking space and the "width" is the smaller side. This processing method avoids confusion caused by differences in data recording direction or coordinate system, making the description of the geometric dimensions of the parking space more intuitive and unified, and providing standardized input data for subsequent parking path planning and key performance indicator statistics.

[0100] In some implementations, the vehicle body width data may be read from vehicle status data, which is usually a static parameter.

[0101] In some implementations, the parking space position in the initial state is determined by determining the geometric center of the parking space or the distribution positions of the four corner points (eg, the left or right side).

[0102] Exemplarily, the parking space position in the initial state includes "initial state_parking space position", which is defined as determining the left and right position of the parking space relative to the vehicle according to the y coordinate value of the corner point of parking space No. 0 at the beginning of parking. This is based on the vehicle's own coordinate system, where the y axis usually represents the offset in the width direction of the vehicle: if the y value of the corner point of parking space No. 0 is greater than 0, it means that the parking space is located on the left side of the vehicle; if the y value is less than 0, the parking space is located on the right side of the vehicle.

[0103] In some embodiments, the lateral distance can be calculated based on the horizontal distance between the coordinates of the corner point of the parking space and the center line of the vehicle, and the results can be graded (such as too close, near end, middle end, far end) in combination with business logic.

[0104] Exemplarily, the lateral distance includes “starting state_lateral distance”, which is calculated as follows: Starting state_lateral distance = |y value of the 0th corner point of the parking space at the beginning of parking| - width of the vehicle body / 2; Among them, the y value of the parking space 0 corner point represents the lateral offset of the parking space relative to the center line of the vehicle. Taking the absolute value can uniformly handle the situation where the parking space is on the left or right side of the vehicle. Subtracting half of the body width is to correct the influence of the actual space occupied by the vehicle, so as to accurately reflect the net lateral distance from the center of the vehicle to the parking space. This calculation provides important reference data for evaluating the rationality of the initial parking state and path planning.

[0105] The horizontal distance also includes "Starting state_Horizontal distance_Grading", which is determined as follows: If the parking space type data is a line parking space or a fused parking space: less than 0.3 meters is too close; 0.3-1 meters is near; 1-2 meters is mid-end; 2-3 meters is far; more than 3 meters is too far; If the parking space type data is spatial parking space: less than 0.5 meters is too close; 0.5-1 meters is near; 1-1.8 meters is far; more than 1.8 meters is too far.

[0106] Further, as an implementation of the present invention, in step S3, generating the parking end state data includes: calculating the post-parking deflection angle and the post-parking wheel track.

[0107] In some embodiments, calculating the deflection angle after parking includes: determining the two long sides of the parking space by using the coordinates of the four corner points of the parking space at the end of parking, and selecting the longer one as the reference side. Calculating the two direction vectors of the long side, and selecting the direction vector with an acute angle with the current direction of the vehicle as the parking space reference vector. Then, according to the direction vector of the vehicle after parking (provided by the vehicle posture or positioning module), the angle between the vehicle direction and the reference parking space direction is calculated using the vector angle formula to obtain the deflection angle after parking.

[0108] In some implementations, the rotation direction (clockwise or counterclockwise) may be further determined based on the relative relationship between the parking space coordinate system and the vehicle coordinate system.

[0109] The calculation of the deflection angle after parking can be achieved based on geometric algorithms, positioning data or sensor fusion data to ensure the accuracy of the calculation of the deflection angle, providing an important basis for the evaluation of parking accuracy and completion status.

[0110] In some implementations, calculating the wheelbase after parking includes: using different wheelbase calculation methods according to the parking space type data (line parking space or space parking space).

[0111] In some embodiments, the parking space type data is a line parking space or a fused parking space. The straight line equation of the parking space boundary can be calculated through the coordinates of the two corner points of the parking space boundary. Combined with the specific position coordinates of the right front wheel, right rear wheel, left front wheel, and left rear wheel of the vehicle, the vertical distance from each wheel to the parking space boundary is calculated to obtain four wheelbase values.

[0112] In some embodiments, the parking space type data is a spatial parking space. The perception module can obtain the geometric relationship between the vehicle and the surrounding obstacles when parking is completed, select the nearest neighboring vehicle as a reference, and calculate the wheelbase based on the geometric position relationship between the boundary coordinates of the reference vehicle and the wheel position of the parked vehicle.

[0113] In some embodiments, if no obstacle is detected, the calculation may be performed by falling back to a derivation formula based on the parking space corner point and the vehicle body width.

[0114] Exemplarily, the wheel tracks after parking include the right front wheel track for parking, the right rear wheel track for parking, the left rear wheel track for parking, and the left front wheel track for parking.

[0115] In some implementations, if the parking space type data is a line parking space or a fused parking space, then: Parking right front wheelbase: first find the x and y coordinates of the two corner points on the right side of the parking space, and then use the formula to calculate the equation of the line where these two corner points are located. Then substitute when x is equal to the wheelbase, and get the value of y on the straight line equation. Finally, parking right front wheelbase = |y|-body width / 2; Parking right rear wheel width: first find the x and y coordinates of the two corner points on the right side of the parking space, and then use the formula to calculate the equation of the line where these two corner points are located. Then substitute when x equals 0 to get the value of y on the straight line equation. Finally, parking right rear wheel width = |y|-body width of the vehicle / 2; Parking left rear wheel width: first find the x and y coordinates of the two left corner points of the parking space, and then use the formula to calculate the equation of the line where these two corner points are located. Then substitute when x equals 0 to get the value of y on the straight line equation. Finally, parking left rear wheel width = |y|-body width of the vehicle / 2; Parking left front wheelbase: first find the x and y coordinates of the two left corners of the parking space, and then use the formula to calculate the equation of the line where these two corners are located. Then substitute when x is equal to the wheelbase, and get the value of y on the line equation. Finally, the parking left front wheelbase = |y|-body width / 2.

[0116] In some implementations, the parking space type data is a vertical space parking space, then: First, the perceived dynamic obstacles at the end of parking are selected, and the obstacles are selected as the entire vehicle, and the perceived position point is the data of the geometric center; Then calculate the wheelbase on the right side, filter the adjacent vehicles on the right side by the obstacle coordinates, and then select the nearest vehicle as the reference vehicle by calculating the distance; Then, according to the xy coordinates, yaw, length and width data of the right reference vehicle, the right front wheel width and the right rear wheel width are calculated. If no obstacle vehicle is sensed, the right wheel width is empty. Finally, the left wheelbase is calculated. By using the obstacle coordinates, the adjacent vehicles on the left are screened, and then the nearest vehicle is selected as the reference vehicle by calculating the distance. Then the xy coordinates, yaw, length and width data of the left reference vehicle are obtained and put into the formula to calculate the left front wheelbase and left rear wheelbase for parking. If no obstacle vehicle is sensed, the left wheelbase is empty.

[0117] It should be noted that in some embodiments, if the parking space is a horizontal space parking space, the wheelbase after parking is not calculated. This is because the parking direction of the horizontal space parking space is parallel to the long side of the parking space. At this time, the lateral distance between the four wheels of the vehicle and the parking space boundary is not a key indicator for evaluating parking accuracy, and the longitudinal position and deflection angle of the vehicle can better reflect the completion quality of parking. Therefore, the wheelbase calculation has no direct meaning for the horizontal space parking space, and this step can be omitted to simplify the data processing process.

[0118] Furthermore, as an embodiment of the present invention, the calculation of the key performance indicators of the parking function includes calculating at least one of the following indicators: parking time, average parking speed, maximum parking speed, number of parking maneuvers, and number of parking re-plannings.

[0119] In some embodiments, the parking duration can be calculated by the difference between the start time and the end time of the parking process. The parking start time is determined by the first frame time when the APA state machine enters the smart parking state, and the parking end time is recorded by the last frame time when the state machine exits the smart parking state. The time difference between the two is the parking duration, which is usually expressed in seconds.

[0120] In some implementations, if a state switch or abnormal exit occurs during the docking process, it is necessary to accurately identify the complete docking process based on the state machine log to ensure the accuracy of the calculation results.

[0121] The calculation of parking time can reflect parking efficiency and is one of the important performance indicators for evaluating parking functions.

[0122] In some embodiments, the average parking speed is obtained by accumulating the driving distance during the parking process and dividing it by the parking time. The driving distance can be calculated based on the accumulated output of the wheel speed sensor, and the parking time is determined by the difference between the parking start time and the parking end time, and the result is usually expressed in kilometers per hour (kph). The maximum parking speed is the maximum value of all instantaneous vehicle speed values ​​extracted from the wheel speed sensor data during the parking process, which directly reflects the peak speed of the vehicle during the parking process. These two indicators measure the overall speed level and speed peak of the parking process respectively, and can comprehensively reflect the parking efficiency and the vehicle's motion characteristics.

[0123] In some embodiments, the number of parking maneuvers can be achieved by analyzing the changes in gears during the parking process. Specifically, during the parking period, the gear signal of the vehicle is monitored, and the number of times the vehicle switches from the forward gear (D gear) to the reverse gear (R gear) or from the reverse gear to the forward gear is counted. Each switch is counted as a maneuvering action. This indicator can reflect the frequency of path adjustment during the parking process and is an important reference for evaluating the accuracy of parking path planning and control.

[0124] In addition, in some embodiments, in order to ensure the accuracy of the calculation, it is necessary to combine the parking state machine data to filter the gear switching in the non-parking state. For example, if the gear switching is caused by environmental interference outside the parking, it will not be counted in the number of times the parking garage is turned on.

[0125] In some embodiments, calculating the number of parking re-planning times can be achieved by monitoring the re-planning signal of the path planning module during the parking process. When the original path is unavailable due to obstacles, dynamically changing environment or other constraints detected during the parking process, the path re-planning signal will be triggered. The number of re-planning times can be obtained by counting the number of times the signal is triggered during the entire parking process. To ensure the accuracy of the calculation, it is necessary to combine the parking state machine to confirm that the re-planning signal is only triggered in the parking state. This indicator reflects the complexity and environmental adaptability of the parking process, and is an important basis for evaluating the robustness of the parking algorithm and the efficiency of path planning.

[0126] In some implementations, if multiple types of replanning are involved (such as local adjustment or global path recalculation), the parking replanning times indicator can be further subdivided and counted separately.

[0127] Furthermore, as an implementation mode of the present invention, the automatic parking function evaluation method further includes: counting the total number of parking attempts and the number of completed parking attempts, and calculating the parking completion rate.

[0128] In some embodiments, calculating the parking completion rate includes: identifying all records of entering the parking process through the parking state machine log or the APA module output data, thereby counting the total number of parking times. Then, for each parking, judging whether the parking is successfully completed according to the final state of the state machine at the end of the parking, for example, if the state machine is "completed", it is counted as a successful parking, and if it is exited due to "abnormal environment", "manual takeover" or "other reasons for failure", it is counted as unfinished. After counting the number of completions, the parking completion rate can be calculated by the formula, that is, parking completion rate = (number of parking completions / total number of parkings) × 100%, and the result is usually expressed as a percentage.

[0129] In some embodiments, to ensure the accuracy of statistics, the independence of each parking can be verified by filtering abnormal data or repeated records and combining timestamps. This indicator directly reflects the parking success rate and reliability of automatic parking.

[0130] Further, as an embodiment of the present invention, generating the parking end state data includes: calculating the lateral distance after parking, the longitudinal distance after parking, and the deflection angle after parking.

[0131] In some embodiments, calculating the lateral distance after parking includes calculating according to a relative relationship between the position of the vehicle and the parking space or a reference object at the end of parking.

[0132] In some embodiments, for linear parking spaces or fused parking spaces, the lateral distance between the center of the parked vehicle and the boundary of the parking space can be obtained by subtracting half of the vehicle body width (vehicle body width / 2) from the absolute value of the y coordinate of the corner point No. 0 of the parking space.

[0133] In some embodiments, for a spatial parking space, it is necessary to filter the geometric information of a reference object (such as a neighboring vehicle or obstacle) at the end of parking through a perception module, and combine the coordinates of the vehicle's wheel position to calculate the shortest distance between it and the boundary of the reference object as the lateral distance.

[0134] In some embodiments, if the reference object cannot be sensed during the unparking process, the parking space data may be used for alternative calculations.

[0135] Exemplarily, the lateral distance after parking includes "end state_lateral distance", then: If the parking space type data is a line parking space or a fused parking space, then: Ending state_lateral distance = |y value of the 0th corner point of the parking space at the end of parking| - width of the own vehicle body / 2; If the parking space type data is a space parking space and the reference car can be sensed, then: First, the perceived dynamic obstacles are selected when the time point is the start of parking, and the obstacles are selected when the obstacle is the entire vehicle and the perceived position point is the geometric center; If the vehicle is parked vertically on the right, the nearest vehicle on the right is used as the reference vehicle. The obstacle coordinates are used to filter the vehicles on the right, and then the nearest vehicle is selected as the reference vehicle by calculating the distance. If the vehicle is parked vertically on the left, the nearest vehicle on the left is used as the reference vehicle. The obstacle coordinates are used to filter the vehicles on the left, and then the nearest vehicle is selected as the reference vehicle by calculating the distance. If it is horizontal parking, the nearest car in front is used as the reference car. The obstacle coordinates are used to screen the adjacent vehicles in front, and then the nearest car is selected as the reference car by calculating the distance. After finding the reference car, check the information of the reference car at the end of parking through the data: x, y, yaw, length, width, etc. By substituting into the formula, calculate the x and y of the four side points of the reference car at the end of parking. Then calculate the lateral distance from the car to the four side points, and take the lateral distance of the nearest point as the end state_lateral distance; If the parking space type data is a space parking space, but the reference car cannot be sensed, then: Use the parking space calculation method instead (end state_lateral distance = |y value of the 0 corner point at the end of parking| - width of the vehicle body / 2).

[0136] In some embodiments, calculating the lateral distance after parking includes: determining the lateral distance between the vehicle and a parking space boundary or an adjacent obstacle according to a coordinate relationship between the position of the vehicle and the parking space or a reference object at the end of parking.

[0137] In some embodiments, for a linear parking space or a fused parking space, the calculation method is to subtract half of the vehicle body width (vehicle body width / 2) from the absolute value of the y coordinate of the corner point No. 0 of the parking space at the end of parking, so as to obtain the net lateral distance from the center of the vehicle to the boundary of the parking space.

[0138] In some embodiments, for a spatial parking space, the perception module needs to obtain the geometric center and boundary coordinates of a reference object (such as an adjacent vehicle) at the end of parking, and calculate the shortest lateral distance from the vehicle to the reference object in combination with the position of the vehicle's wheel position.

[0139] In some embodiments, if the reference object cannot be sensed, the parking space boundary data may be used to perform an alternative calculation.

[0140] Exemplarily, the lateral distance after parking includes "end state_longitudinal distance", then: If the parking space type data is a vertical parking space, then the absolute value of the x value of the midpoint between the 0th corner point and the 3rd corner point of the original parking space after the parking is completed is calculated; If the parking space type data is a horizontal parking space, the absolute value of (the x value of the 0th corner point of the parking space at the end of parking out - the x value of the 0th corner point of the parking space at the start of parking out) is calculated.

[0141] Exemplarily, the lateral distance after parking also includes "end status_longitudinal distance_whether exceeded limit", wherein the longitudinal distance of the horizontal parking space cannot exceed 10 meters, and the longitudinal distance of the vertical parking space cannot exceed 5 meters. If exceeded, it is exceeded.

[0142] In some embodiments, the deflection angle after parking includes: calculating according to the angle between the direction of the vehicle and the direction of the parking space when parking is completed, and determining the reference direction vector of the parking space through the coordinates of the corner points of the parking space when parking is completed.

[0143] In some implementations, for a linear parking space, the reference direction is the boundary line direction of the parking space.

[0144] In some embodiments, for a horizontal space parking space, the reference direction is the x-axis direction of the vehicle when parking begins.

[0145] In some embodiments, for a vertical space parking space, the reference direction is the y-axis direction of the vehicle when parking begins.

[0146] In some embodiments, the actual driving direction of the vehicle at the end of the parking is calculated by the posture data or direction vector of the vehicle. The angle between the vehicle direction vector and the parking space reference direction vector is calculated by the vector angle formula, and the size of the angle is the deflection angle after the parking. In addition, the rotation direction of the angle (clockwise or counterclockwise) can be determined by the cross product of the vectors. This deflection angle can reflect the deviation of the vehicle posture after the parking, and is an important indicator for evaluating the parking accuracy.

[0147] Furthermore, as an embodiment of the present invention, the calculation of the key performance indicators of the parking function also includes calculating at least one of the following indicators: parking time, average parking speed, maximum parking speed, number of parking maneuvers and number of parking re-plannings.

[0148] In some embodiments, the calculation of the parking out duration can be completed by analyzing the start time and end time of the parking out process. The parking out start time is marked by the first frame time when the APA state machine enters the parking out mode of the smart parking, and the parking out end time is determined by the last frame time when the state machine exits the parking out mode. The parking out duration is calculated by the difference between these two time points, usually expressed in seconds.

[0149] In some implementations, if the unparking process involves a pause or abnormal state, the accuracy of the time point needs to be verified through the state machine log to eliminate the interference of non-continuous unparking behavior. The unparking time is an important indicator for evaluating the efficiency of the vehicle leaving the parking space, which can directly reflect the smoothness and time consumption of the unparking operation.

[0150] In some embodiments, the average parking speed is calculated by calculating the ratio of the total distance traveled by the vehicle during the parking process to the parking time, wherein the total distance is calculated based on the accumulation of wheel speed sensor data, and the parking time is determined by the difference between the parking start time and the parking end time, and the result is usually expressed in kilometers per hour (kph). The maximum parking speed is obtained by monitoring all instantaneous vehicle speed data during the parking process and extracting the maximum value as the maximum vehicle speed.

[0151] In some implementations, to ensure data accuracy, it is necessary to filter out non-continuous parking actions (such as pause or non-parking state), and only count the vehicle speed data in the parking state based on the timestamp or state machine mark. These two indicators reflect the efficiency and dynamic characteristics of the parking process, and are important parameters for measuring the parking performance.

[0152] In some embodiments, calculating the number of parking maneuvers can be achieved by counting the number of times the vehicle gear is switched during the parking process. The gear signal of the vehicle is monitored during the parking period. When the gear is switched from the forward gear (D gear) to the reverse gear (R gear) or from the reverse gear to the forward gear, each switch is counted as a maneuvering action. The parking state is marked by the APA state machine to ensure that the counted gear switches only occur during the parking stage. To avoid repeated counting or misjudgment, the timestamp and parking path data can be combined to filter the gear switches caused by non-parking operations (such as environmental interference or non-parking state). Finally, the number of maneuvers can be used as a frequency indicator for parking path adjustment, reflecting the complexity and accuracy of path planning and control during the parking process.

[0153] In some embodiments, the number of re-planning times for parking out can be calculated by monitoring the re-planning signal of the path planning module during the parking out process. When the original plan is unavailable due to the detection of obstacles, environmental changes or blocked paths during the parking out process, the path re-planning signal is triggered. Each triggering of the re-planning signal is counted as one re-planning. In combination with the APA state machine, it is ensured that the statistical re-planning signal only occurs in the parking out state, and the signal interference caused by non-parking out operations is excluded.

[0154] In some implementations, different types of replanning (such as global replanning or local adjustment) can be further classified and counted separately. The final number of replanning can reflect the environmental complexity of the parking process and the adaptability and stability of the path planning algorithm, and is an important indicator for evaluating the parking performance.

[0155] Furthermore, as an embodiment of the present invention, the automatic parking function evaluation method further includes: counting the total number of parking out and the number of parking out completions, and calculating the parking out completion rate. The implementation of the parking out completion rate is similar to the implementation of the parking in completion rate, and will not be repeated here.

[0156] Reference Figure 3The embodiment of the present invention also discloses a parking function automatic evaluation system, including a data acquisition module 1, a data extraction module 2, a data processing module 3 and a result generation module 4.

[0157] In some embodiments, the data acquisition module 1 is used to collect multi-topic data during the parking process in real time through an integrated data recording tool (such as a data-logger), including perception data, fusion data, map data, positioning data, and vehicle status data.

[0158] In some embodiments, the data acquisition module 1 can be connected to a controller interface (such as a CAN bus or a ROS interface) through a sensor interface (such as a camera, a radar, a GNSS module, etc.) of the vehicle to receive relevant data streams.

[0159] In some embodiments, the data acquisition module 1 can be configured with trigger conditions, such as the start or end event of the parking state machine, to dynamically turn on or off the data recording function. The recorded data is organized into compressed files based on timestamps, and metadata files (such as subject type, recording time, and device status) are saved for subsequent analysis.

[0160] In some embodiments, the data acquisition module 1 can also verify data integrity in real time, discard incomplete or abnormal data packets, ensure the consistency and accuracy of the acquired data, and provide high-quality input for subsequent processing.

[0161] In some embodiments, the data extraction module 2 is used to parse the recording package file generated by the data acquisition module 1, identify its metadata (such as software version, recording theme list) and pull the corresponding message definition file (msg or IDL) from the code repository according to the software version. The data extraction module 2 decompresses the recording package file and extracts the required theme data, such as perception information, vehicle status, path planning data, etc.

[0162] In some embodiments, the data extraction module 2 may extract the signal data required for the evaluation (such as vehicle speed, APA state machine status, parking space geometry information, etc.) from the subject data according to the target field list defined in the configuration file.

[0163] In some embodiments, the data extraction module 2 can automatically repair or complete missing metadata files (such as yaml files) during extraction, check the integrity of the data packet, and discard abnormal or incomplete records. The extracted subject data is aligned and formatted and stored in time to ensure that structured high-quality input is provided for subsequent data processing modules.

[0164] In some embodiments, the data processing module 3 is used to time-align the structured data generated by the data extraction module 2, and synchronously integrate the multi-topic data according to a unified time reference through fuzzy matching.

[0165] In some implementations, the data processing module 3 is further used to merge the time-aligned data into a large table to ensure that the data of all subjects can be organized in a unified format.

[0166] In some embodiments, the data processing module 3 is further used to fill in missing values ​​according to business logic, such as processing data gaps by interpolation, using default values, or inferring calculations.

[0167] In some embodiments, the data processing module 3 is further used to generate parking start state data, parking end state data and parking exit end state data based on the state machine information of the parking process, which includes calculating key parameters of the vehicle in these states (such as parking space position, lateral distance, longitudinal distance, deflection angle, etc.). The processed data is stored in a structured form to provide high-quality basic data support for the result generation module.

[0168] In some embodiments, the result generation module 4 is used to calculate the key performance indicators (KPIs) of the parking function based on the parking start state data, parking end state data and parking exit state data generated by the data processing module 3. These KPIs include parking entry and exit time, average vehicle speed, maximum vehicle speed, number of parking maneuvers, number of re-planning, deflection angle and wheelbase after parking entry and exit, etc.

[0169] In some embodiments, the result generation module 4 is further used to calculate the indicators one by one through logical rules and formulas, and output the calculation results in a structured form as an evaluation report, supporting multiple formats (such as Excel or JSON).

[0170] In some implementations, the result generation module 4 is also used to count the total number of docking in and out, the number of completions, and calculate overall indicators such as completion rate. The final generated results are automatically stored in a specified path (such as NAS storage), supporting real-time query and long-term archiving, so as to facilitate system performance evaluation and version optimization comparison.

[0171] In some implementations, the result generation module 4 may also generate a visual chart based on the configuration options to intuitively display the parking performance data.

[0172] The embodiment of the present invention also discloses a readable storage medium.

[0173] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the automatic parking function evaluation method described in any one of the above embodiments.

[0174] It can be understood that computer-readable storage media may include: any entity or device capable of carrying a computer program, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc. A computer program includes a computer program code. The computer program code may be in source code form, object code form, an executable file, or some intermediate form, etc. A computer-readable storage medium may include: any entity or device capable of carrying a computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc.

[0175] In certain embodiments of the present invention, the electronic device may include a controller or a processor, and the controller is a single-chip microcomputer chip that integrates a processor, a memory, a communication module, etc. The processor may refer to a processor included in the controller. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.

[0176] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0177] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0178] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A parking function automatic evaluation method, characterized in that: include: Obtain data on various topics during the parking process; Parsing the data of each subject to extract the evaluation data required for evaluation; Preprocessing the evaluation data, and generating parking start state data, parking end state data, and parking exit end state data according to the preprocessed evaluation data; Based on the parking start state data, the parking end state data and the parking exit end state data, a key performance indicator of a parking function is calculated.

2. The automatic parking function evaluation method according to claim 1, characterized in that: The acquisition of various subject data during the parking process includes: The data recording tool is used to record the perception data, fusion data, map data, positioning data and vehicle status data during the parking process.

3. The automatic parking function evaluation method according to claim 2, characterized in that: The step of parsing the subject data to extract the evaluation data required for the evaluation includes: Identify the software version of the recorded package; Pull the message definition corresponding to the software version from the code repository; Extract the required subject data and signal data according to data requirements.

4. The automatic parking function evaluation method according to claim 1, characterized in that: The preprocessing of the evaluation data includes: Temporal alignment of the topics; Combine all subject data into one large table; Fill missing values ​​based on business logic.

5. The automatic parking function evaluation method according to claim 1, characterized in that: The generating of the docking initial state data comprises: Obtain the coordinate data of the four corner points of the parking space, the parking space type data, the parking space length and width data, and the vehicle body width data; Calculate the parking space position and lateral distance at the starting state.

6. The automatic parking function evaluation method according to claim 1, characterized in that: The generating of the parking end status data comprises: Calculate the steering angle after parking and the wheel track after parking.

7. The automatic parking function evaluation method according to claim 1, characterized in that: The generating of the parking end status data comprises: Calculate the lateral distance after parking, the longitudinal distance after parking and the deflection angle after parking.

8. The automatic parking function evaluation method according to claim 1, characterized in that: The calculating of the key performance indicators of the parking function includes calculating at least one of the following indicators: Parking time, average parking speed, maximum parking speed, number of parking maneuvers and number of parking re-plannings.

9. The automatic parking function evaluation method according to claim 8, characterized in that: The calculating of the key performance indicators of the parking function further comprises calculating at least one of the following indicators: The duration of parking, average parking speed, maximum parking speed, number of parking maneuvers and number of parking re-planning.

10. The automatic parking function evaluation method according to claim 1, characterized in that: Also includes: The total number of parking attempts and the number of completed parking attempts are counted, and the parking completion rate is calculated.

11. The automatic parking function evaluation method according to claim 1, characterized in that: Also includes: The total number of parking outs and the number of parking out completions are counted, and the parking out completion rate is calculated.

12. A parking function automatic evaluation system, characterized in that: include: A data acquisition module, at least used to acquire various subject data during the parking process; A data extraction module, at least used to parse the data of each subject to extract the evaluation data required for the evaluation; A data processing module, at least used to pre-process the evaluation data, and generate parking start state data, parking end state data and parking exit end state data according to the pre-processed evaluation data; The result generating module is configured to calculate a key performance indicator of a parking function based on at least the parking start state data, the parking end state data and the parking exit end state data.

13. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is run by a computer, the automatic parking function evaluation method according to any one of claims 1 to 11 is executed.

Images