Vehicle testing method, test control box, device, and computer storage medium

Abstract

The application provides a vehicle test method, a test control box, a vehicle test device and a computer storage medium. The vehicle test method is applied to a test control box, and the test control box is used for automatic control, data collection and test evaluation of an unmanned vehicle. The vehicle test method comprises the following steps: a test sequence for the unmanned vehicle is generated; the unmanned vehicle is driven to sequentially execute each test working condition according to the test sequence; vehicle state data in the process of executing each test working condition by the unmanned vehicle is acquired; a holding sample set and an outlying sample set of each test working condition in the vehicle state data are determined; an upper limit of a confidence interval of each test working condition is determined based on the holding sample set and the outlying sample set; and a lateral acceleration declared value of each test working condition is determined according to the upper limit of the confidence interval of each test working condition. Through the above vehicle test method, the scientificity and efficiency of the steering control system certification test of the unmanned vehicle are improved.

Description

Technical Field

[0001] This application relates to the field of unmanned vehicle testing technology, and in particular to a vehicle testing method, a test control box, a vehicle testing device, and a computer storage medium. Background Technology

[0002] As the level of automotive intelligence increases, stringent certification requirements have been imposed on the Advanced Steering Function (ACSF) of autonomous vehicles. In particular, for lane-keeping assist functions, manufacturers must declare the maximum lateral acceleration limit to the certification body and confirm it through field certification testing.

[0003] However, existing testing technologies have the following shortcomings: First, existing testing equipment has fragmented functions. For example, some technologies only have CAN (Controller Area Network) communication capabilities and lack high-precision positioning access and statistical analysis capabilities for GPS (Global Positioning System) / GNSS (Global Navigation Satellite System), resulting in low testing efficiency and difficulties in data integration. Second, existing methods mostly rely on bench simulation or production line EOL (End Of Line) testing, which cannot simulate lateral acceleration threshold testing under real-world conditions and lacks traversal testing and threshold statistical analysis capabilities. Third, the determination of hold / deviation states mostly relies on manual interpretation or a single threshold, lacking an automated engine for multi-condition joint determination, resulting in poor consistency and repeatability of the determination results. Fourth, the determination of declared values ​​lacks quantitative statistical basis and does not consider the overlap (grey area) in probability distribution between hold and deviation samples, failing to guarantee the dual reliability of declared values ​​in certification testing and functional testing. Summary of the Invention

[0004] To address the aforementioned technical problems, this application proposes a vehicle testing method, a test control box, a vehicle testing device, and a computer storage medium.

[0005] To address the aforementioned technical problems, this application proposes a vehicle testing method. This method is applied to a test control box, which is used for the automated control, data acquisition, and test evaluation of an autonomous vehicle. The vehicle testing method includes: Generate a test sequence for the autonomous vehicle; The driverless vehicle is driven to perform each test condition sequentially according to the test sequence. Acquire vehicle status data during the execution of each test condition by the unmanned vehicle; Determine the hold-up sample set and the off sample set for each test condition in the vehicle state data; The upper bound of the confidence interval for each test condition is determined based on the maintained sample set and the deviated sample set. The lateral acceleration declaration value for each test condition is determined based on the upper bound of the confidence interval for each test condition.

[0006] The generation of test sequences for the autonomous vehicle includes: Obtain the test configuration file, which includes at least one of the following fields: a list of corner radii, a vehicle speed range configuration, a target lateral acceleration step value, regulatory limits for each vehicle speed range, the number of test repetitions, a test direction configuration, and corner entry control parameters; The test configuration file is analyzed to determine the target lateral acceleration calculated for each curve radius and each target vehicle speed; Test instances are generated by combining test fields whose target lateral acceleration is less than or equal to the maximum limit of the speed range in which the target vehicle speed is located. All test cases are combined into a test sequence for the autonomous vehicle.

[0007] The step of acquiring vehicle status data during the execution of various test conditions by the autonomous vehicle includes: The location data and vehicle operation data of the unmanned vehicle during the execution of each test condition are acquired, wherein the vehicle operation data includes at least a lateral acceleration signal; Using the timestamp of the location data as a reference time axis, the vehicle operation data is aligned with the location data according to the timestamp to form the vehicle status data.

[0008] The vehicle testing method further includes aligning the vehicle operation data with the location data according to the timestamps: The theoretical number of data frames for the location data is determined based on the reference time axis. Determine the number of valid data frames in the aligned vehicle operation data that correspond to the data frame timestamps of the location data; The frame drop rate of the test condition is determined based on the theoretical number of data frames and the effective number of data frames. When the frame loss rate exceeds a preset threshold, the autonomous vehicle is re-driven to perform the test condition and data is collected again.

[0009] The vehicle testing method further includes aligning the vehicle operation data with the location data according to the timestamps: The aligned lateral acceleration and yaw rate signals are processed using digital low-pass filtering. The filtering method used is zero-phase filtering.

[0010] The vehicle testing method further includes, after acquiring the vehicle status data during the execution of each test condition by the autonomous vehicle: Steady-state window data is extracted from several data frames after the vehicle enters a preset state in the vehicle state data. Perform multi-condition joint determination on each data frame in the steady-state window data; Data frames that meet the judgment criteria are marked as held, and data frames that do not meet the judgment criteria are marked as deviated. The set of held samples is determined based on all data frames marked as held states; The set of biased samples is determined based on all data frames marked as biased.

[0011] The step of extracting steady-state window data from several data frames after the vehicle enters a preset state in the vehicle state data includes: Obtain candidate window data after the vehicle enters a preset state from the vehicle status data; If the candidate window data simultaneously meets the following conditions, then the candidate window data is determined as steady-state window data: The standard deviation of lateral acceleration in all data frames of the candidate window data is less than a first preset difference. The maximum difference in vehicle speed among all data frames of the candidate window data is less than or equal to the second preset difference. The duration of the candidate window data is greater than or equal to a preset duration.

[0012] The determination conditions used in the multi-condition joint determination include at least one of the following: lane distance condition, system state condition, lateral acceleration condition, and lateral acceleration stability condition.

[0013] The step of performing multi-condition joint determination on each data frame in the steady-state window data includes: Based on the position information in the data frame, determine the lateral offset of the vehicle center relative to the lane center; The distance from the outermost point of the vehicle to the lane line is determined based on the lateral offset, the lane width of the test site, and the vehicle width. When the distance between the outermost point of the vehicle and the lane line is greater than a preset distance, the data frame is determined to meet condition one; When the system status signal of the data frame is an active and control signal, it is determined that the data frame satisfies condition two. When the average lateral acceleration within the first preset window where the data frame is located is less than or equal to a preset average threshold, the data frame is determined to satisfy condition three. When the standard deviation of the lateral acceleration within the second preset window where the data frame is located is less than the preset standard deviation threshold, the data frame is determined to meet condition four. Data frames that meet all the above conditions are marked as held, and other data frames are marked as off-target.

[0014] The step of determining the upper bound of the confidence interval for each test condition based on the maintained sample set and the deviated sample set includes: The optimal estimate of the threshold is calculated based on the maintained sample set and the deviated sample set; Calculate the standard deviation of the lateral acceleration of the maintained sample set; Calculate the standard error of the best estimate of the threshold based on the standard deviation of the lateral acceleration and the number of samples in the hold-up sample set; The upper bound of the confidence interval for each test condition is determined based on the standard error and the optimal estimate of the threshold.

[0015] After determining the hold-up sample set and the deviation sample set for each test condition in the vehicle state data, the vehicle testing method further includes: The width of the gray zone is determined based on the maximum lateral acceleration in the maintained sample set and the minimum lateral acceleration in the deviated sample set; When the width of the gray area is greater than a preset value, it is determined that there is a ambiguity area in the test condition, and a warning message is output.

[0016] After determining the declared lateral acceleration values ​​for each test condition, the vehicle testing method further includes: Obtain the maximum declared value of lateral acceleration for all test conditions; Based on the maximum declared value of lateral acceleration and the declared value of lateral acceleration for each test condition, calculate the functional test margin for each test condition. When the functional test margin is less than or equal to the preset value, a warning signal indicating insufficient functional test margin is output, and the declared value of the test condition is marked as unavailable.

[0017] After determining the declared lateral acceleration values ​​for each test condition, the vehicle testing method further includes: The over-limit trigger margin is calculated based on the declared lateral acceleration value, the preset over-limit amplitude value, and the best estimated threshold value for each test condition. When the over-limit trigger margin is less than or equal to a preset value, an over-limit trigger margin insufficient warning signal is output, and the declared value of the test condition is marked as unavailable.

[0018] After determining the declared lateral acceleration values ​​for each test condition, the vehicle testing method further includes: The minimum value among all the declared lateral acceleration values ​​for all test conditions is determined as the final declared value of the autonomous vehicle's lateral acceleration control capability.

[0019] To address the aforementioned technical problems, this application also proposes a test control box, which includes: The test case management module is used to generate test sequences for autonomous vehicles; The main control unit is used to drive the unmanned vehicle to execute each test condition sequentially according to the test sequence; The signal receiving module is used to acquire vehicle status data during the execution of various test conditions by the unmanned vehicle; A data processing engine is used to determine the hold-up sample set and the deviation sample set for each test condition in the vehicle state data; determine the upper bound of the confidence interval for each test condition based on the hold-up sample set and the deviation sample set; and determine the declared lateral acceleration value for each test condition based on the upper bound of the confidence interval for each test condition.

[0020] To address the aforementioned technical problems, this application also proposes a vehicle testing apparatus, which includes a memory and a processor coupled to the memory; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the vehicle testing method described above.

[0021] To address the aforementioned technical problems, this application also proposes a computer storage medium for storing program data, which, when executed by a computer, is used to implement the aforementioned vehicle testing method.

[0022] Compared with existing technologies, the beneficial effects of this application are as follows: By generating test sequences and driving vehicles to execute various test conditions, comprehensive vehicle state data is obtained, enabling the tests to cover multiple operating scenarios required by regulations and improving the representativeness of the test data. By determining the hold-and-go sample set and the deviation sample set, quantitative classification of the hold-and-go states of the vehicle system under different lateral accelerations is achieved, eliminating subjective interference from manual interpretation and improving the accuracy and consistency of state determination. Determining the upper bound of the confidence interval based on the hold-and-go sample set provides a statistical basis for the determination of the declared values, reflecting the capability boundary of the system at a specific confidence level and avoiding deviations in declared values ​​caused by sample fluctuations. Determining the declared lateral acceleration value based on the upper bound of the confidence interval ensures that the final declared value guarantees both the reliability of over-limit triggering in certification testing and the stability within the functional testing range, achieving dual reliability assurance for certification testing and functional testing, which is conducive to improving the scientificity and efficiency of certification testing of the steering control system of autonomous vehicles. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the first embodiment of the vehicle testing method provided in this application; Figure 2 This is an overall flowchart of the test method for confirming the declared lateral acceleration value provided in this application; Figure 3 This is a flowchart illustrating the second embodiment of the vehicle testing method provided in this application; Figure 4 This is a flowchart illustrating the third embodiment of the vehicle testing method provided in this application; Figure 5 This is a flowchart illustrating the fourth embodiment of the vehicle testing method provided in this application; Figure 6 This is a flowchart illustrating the fifth embodiment of the vehicle testing method provided in this application; Figure 7 This is a flowchart of the multi-condition joint determination logic of the maintain / deviation dual-state determination engine in the data processing engine provided in this application; Figure 8 This is a flowchart illustrating the sixth embodiment of the vehicle testing method provided in this application; Figure 9 This is a flowchart illustrating the seventh embodiment of the vehicle testing method provided in this application; Figure 10 This is a flowchart illustrating the eighth embodiment of the vehicle testing method provided in this application; Figure 11 This is a flowchart illustrating the ninth embodiment of the vehicle testing method provided in this application; Figure 12 This is a schematic diagram of the structure of an embodiment of the test control box provided in this application; Figure 13 This is a schematic diagram of another embodiment of the test control box provided in this application; Figure 14 This is a schematic diagram of the structure of an embodiment of the vehicle testing device provided in this application; Figure 15 This is a schematic diagram of the structure of an embodiment of the computer storage medium provided in this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0025] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] This application provides an unmanned driving field test control box and a test method for confirming the declared value of lateral acceleration, aiming to solve the following technical problems: (1) Existing test equipment has fragmented functions and lacks a dedicated test control box that integrates GPS / GNSS signal reception, CAN bus communication, automatic test case management, data processing and analysis, and report generation.

[0027] (2) The existing test methods rely on manual interpretation for the determination of the hold / deviation state, lack an automated engine for multi-condition joint determination, and the consistency and repeatability of the determination are insufficient.

[0028] (3) The existing methods for determining the declared value lack quantitative statistical analysis basis, do not consider maintaining the overlap of probability distribution between the sample and the deviation sample, and cannot guarantee the dual reliability of the declared value in certification testing and functional testing.

[0029] (4) The existing testing process separates the regulatory compliance verification step from the data analysis step, resulting in low efficiency in the verification process.

[0030] Please refer to details. Figure 1 and Figure 2 , Figure 1 This is a flowchart illustrating the first embodiment of the vehicle testing method provided in this application. Figure 2 This is the overall flowchart of the test method for confirming the declared lateral acceleration value provided in this application.

[0031] in, Figure 2 The overall flowchart shows the complete steps from parameter preloading, test sequence generation, test execution, data acquisition, steady-state segment extraction, filtering, hold / off determination, threshold statistical analysis, declaration value calculation, regulatory compliance verification to report generation.

[0032] The vehicle testing method provided in this application is applied to a test control box, which is used for the automated control, data acquisition, and test evaluation of unmanned vehicles.

[0033] Figure 1 The vehicle testing method shown is to determine the declared value of lateral acceleration by dividing a specific sample set, statistical modeling and confidence interval calculation, thereby solving the problem that existing technologies lack quantitative statistical basis for determining the declared value.

[0034] like Figure 1 As shown, the specific steps are as follows: Step S11: Generate test sequences for autonomous vehicles.

[0035] In this embodiment, the test control box generates test sequences containing different combinations of vehicle speeds and curve radii, which can cover a variety of operating conditions required by regulations, ensuring the comprehensiveness of test data.

[0036] Step S12: Drive the autonomous vehicle to perform each test condition in sequence according to the test sequence.

[0037] In this embodiment, the test control box drives the vehicle to perform operating conditions in sequence and acquires status data, providing a raw data foundation for subsequent analysis.

[0038] Step S13: Obtain vehicle status data during the execution of each test condition by the unmanned vehicle.

[0039] In this embodiment, the test control box uses a multi-condition joint decision engine to mark the state of each data frame, and summarizes the results into a hold-up sample set and a deviated sample set. This division method can reflect the probability distribution characteristics of the system near the limit boundary.

[0040] Step S14: Determine the hold-up sample set and the off sample set for each test condition in the vehicle state data.

[0041] In this embodiment of the application, the maintained sample set refers to the set of lateral acceleration data obtained in multiple repeated tests under specific test conditions that satisfy the multi-condition joint judgment logic and are marked as maintained; the deviation sample set refers to the set of data marked as deviation under the same test conditions.

[0042] Step S15: Determine the upper bound of the confidence interval for each test condition based on the maintained sample set and the deviated sample set.

[0043] In the embodiments of this application, the upper bound of the confidence interval refers to the upper limit value with a specific confidence level (such as 95%) calculated based on the statistical characteristics of the sample set.

[0044] The test control box calculates the upper bound of the confidence interval based on the hold sample set because the hold sample represents the area where the system can stably maintain its lane-keeping capability, and its statistical upper bound can quantify the boundary of the system's lane-keeping capability in most cases (such as 95% probability). By using the upper bound of the confidence interval as the calculation benchmark, the subjectivity and inconsistency caused by relying solely on a single threshold or manual interpretation can be avoided.

[0045] Step S16: Determine the declared lateral acceleration value for each test condition based on the upper bound of the confidence interval for each test condition.

[0046] In this embodiment of the application, the declared lateral acceleration value refers to the maximum lateral acceleration limit that is ultimately submitted to the certification body for the declaration of regulatory compliance.

[0047] The test control box determines the declared value based on the upper bound of the confidence interval, ensuring that the declared value is based on a strict statistical foundation and thus guaranteeing its dual reliability in certification testing and functional testing.

[0048] In one specific implementation method, during a test at a speed range of 60-100 km / h and a curve radius of 250 m, a driving robot controls the vehicle to enter the curve at a speed of 90 km / h, repeating the test 5 times. The data processing engine filters and judges the steady-state data of each test. If the results are "hold" 3 times and "deviation" 2 times, the average lateral acceleration of the 3 "hold" data is included in the "hold" sample set, and the average of the 2 "deviation" data is included in the "deviation" sample set.

[0049] After the test control box summarizes all operating condition samples across all vehicle speed ranges, it calculates the 95th percentile and standard deviation of the sample set, and then calculates the upper bound of the 95% confidence interval. The final declared value can be set as the upper bound of this confidence interval minus a margin of 0.3 m / s².

[0050] Another embodiment involves adjustments under different road surface adhesion coefficients. When the adhesion coefficient of the test site is low, the target lateral acceleration step value and vehicle speed range configuration can be adjusted, but the statistical logic of the sample and the deviation sample remains unchanged. The declared value is still determined by the upper bound of the confidence interval to adapt to the system performance evaluation under different environments.

[0051] The vehicle testing method of this application generates test sequences and drives the vehicle to execute various test conditions, acquiring comprehensive vehicle state data. This ensures that the tests cover multiple operating scenarios required by regulations, improving the representativeness of the test data. By determining the hold-and-go sample set and the deviation sample set, the method achieves quantitative classification of the hold-and-go states of the vehicle system under different lateral accelerations, eliminating subjective interference from manual interpretation and improving the accuracy and consistency of state determination. Determining the upper bound of the confidence interval based on the hold-and-go sample set provides a statistical basis for the determination of the declared values, reflecting the system's capability boundary at a specific confidence level and avoiding deviations in declared values ​​due to sample fluctuations. Determining the declared lateral acceleration value based on the upper bound of the confidence interval ensures that the final declared value guarantees both the reliability of over-limit triggering in certification testing and the stability within the functional testing range. This achieves dual reliability assurance for certification and functional testing, which is beneficial to improving the scientific rigor and efficiency of certification testing for autonomous vehicle steering control systems.

[0052] This application's embodiments are aimed at... Figure 1 The vehicle testing method step S11 shown provides a specific implementation plan, which automates the construction of the test case matrix through specific configuration file field parsing, physical formula calculation, and limit filtering logic. Please refer to [link / reference] for details. Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the vehicle testing method provided in this application.

[0053] like Figure 3 As shown, the specific steps are as follows: Step S21: Obtain the test configuration file, which includes at least one of the following fields: curve radius list, speed range configuration, target lateral acceleration step value, regulatory limits for each speed range, number of test repetitions, test direction configuration, and curve entry control parameters.

[0054] In this embodiment of the application, the test configuration file refers to a data file that stores test parameters and constraints in a structured data format (such as JSON).

[0055] Among them, the curve radius list refers to a set of preset standard curve curvature radius values; the vehicle speed range configuration refers to the configuration item that divides the vehicle speed into several ranges and specifies the specific test target vehicle speed and corresponding regulatory limit parameters in each range; the target lateral acceleration step value refers to the difference between two adjacent target lateral acceleration test points, which is used to control the precision of the test traversal; the curve entry control parameters refer to the dynamic control parameters such as the aiming distance and lateral gain adjustment ratio when the driving robot performs steering operations in the curve entry transition section.

[0056] The test control box obtains a test configuration file containing fields such as a list of curve radii, vehicle speed range configuration, and regulatory limits, thereby achieving standardized and modular input of test parameters. This makes the generation of test sequences no longer dependent on hard coding, but driven by the configuration file.

[0057] In one specific implementation, the test configuration file is stored in JSON format, in which the list of curve radii includes six standard values: 100m, 125m, 200m, 250m, 500m, and 600m; the vehicle speed range configuration includes five ranges: 0-60km / h, 60-100km / h, etc., and each range specifies a specific target vehicle speed such as 80km / h, 90km / h, etc., as well as the corresponding regulatory limit such as 3.0m / s²; the target lateral acceleration step value is set to 0.1m / s².

[0058] In another embodiment, the configuration file includes parameters such as a pre-steering start distance of 15m and a lateral gain increase of 20%. When generating test sequences, the system not only records the curve and vehicle speed parameters, but also binds these curve entry control parameters to the corresponding test instances to ensure that the driving robot can smoothly enter the curve according to the predetermined dynamic strategy when performing the test, thereby reducing test data noise caused by curve entry impact.

[0059] Step S22: Parse the test configuration file to determine the target lateral acceleration calculated for each curve radius and each target vehicle speed.

[0060] In this embodiment of the application, when the test control box parses the configuration file, it uses the physical formula a_y=V² / (3.6²×R) to transform the abstract vehicle speed V and the curve radius R into the target lateral acceleration a_y in the sense of vehicle dynamics. This step establishes a direct relationship between geometric kinematic parameters and the force state of the vehicle.

[0061] Step S23: Combine test fields that contain the target lateral acceleration less than or equal to the maximum limit of the speed range in which the target vehicle speed is located to generate a test instance.

[0062] In this embodiment of the application, when generating test instances, the test control box introduces the regulatory limits of each vehicle speed range as filtering conditions, and only retains combinations in which the target lateral acceleration is less than or equal to the maximum limit of the range. This design ensures that the generated test cases are physically feasible and comply with the regulatory framework, avoids generating invalid test conditions or test conditions that exceed the system's capabilities, thereby improving the effectiveness and execution efficiency of the test sequence.

[0063] In one specific implementation, after parsing the configuration, the test control box calculates a target lateral acceleration of 2.50 m / s² for a vehicle speed of 90 km / h and a curve radius of 250 m within the 60-100 km / h range. This value is less than the limit of 3.0 m / s² for this range, therefore a test instance is generated. If the calculated target lateral acceleration is 3.5 m / s², it is filtered out because it exceeds the limit of 3.0 m / s², and no test instance is generated.

[0064] Step S24: Combine all test instances into a test sequence for the autonomous vehicle.

[0065] In this embodiment, the test control box combines all qualified test instances to form a complete test sequence covering a multi-dimensional parameter space.

[0066] The vehicle testing method of this application achieves structured management and flexible configuration of test parameters by acquiring a test configuration file containing a list of curve radii, vehicle speed range configuration, target lateral acceleration step value, regulatory limits for each vehicle speed range, number of test repetitions, test direction configuration, and curve entry control parameters. This allows the generation of test sequences to adapt to the needs of different vehicle models and test scenarios. By parsing the configuration file and calculating the target lateral acceleration using physical formulas, automatic conversion from geometric kinematic parameters to vehicle dynamics parameters is achieved, ensuring the scientific accuracy of the test condition settings. By combining test fields where the target lateral acceleration is less than or equal to the maximum limit of each vehicle speed range to generate test instances, invalid or out-of-bounds test combinations are filtered out, making the generated test sequences more compact and efficient, and improving the utilization rate of test resources. Combining all test instances into test sequences enables automated orchestration and execution of test tasks, reducing the workload and error probability of manual test case configuration, and contributing to improving the overall efficiency and standardization of autonomous vehicle field certification testing.

[0067] This application's embodiments are aimed at... Figure 1 The vehicle testing method step S13 shown provides a specific implementation plan, clarifying the dual-source characteristics of data acquisition (location data and vehicle operation data) and the alignment mechanism for data fusion. Please refer to [link / reference] for details. Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the vehicle testing method provided in this application.

[0068] like Figure 4 As shown, the specific steps are as follows: Step S31: Acquire the position data and vehicle operation data of the unmanned vehicle during each test condition, wherein the vehicle operation data includes at least the lateral acceleration signal.

[0069] In this embodiment, location data refers to the vehicle's latitude and longitude, speed, and high-precision timestamp information obtained through the GPS / GNSS signal receiving module; vehicle operation data refers to the vehicle's dynamic parameters obtained through the CAN bus communication module, including lateral acceleration, yaw rate, vehicle speed, and system status signals.

[0070] Step S32: Using the timestamp of the location data as the base time axis, align the vehicle operation data with the location data according to the timestamp to form vehicle status data.

[0071] In this embodiment of the application, time alignment refers to the process of mapping data from different sampling clock sources to a unified time coordinate system.

[0072] Verifying the declared lateral acceleration values ​​of autonomous vehicles requires high-precision spatial location information and real-time dynamic response data. Location data is typically provided by a GNSS module at a fixed frequency (e.g., 100Hz) with a high-precision UTC (Coordinated Universal Time) timestamp. Vehicle operation data, while transmitted via the CAN bus, can also be sampled at 100Hz, but both have independent clock sources, and direct acquisition would lead to time asynchrony. Therefore, a unified time reference must be established.

[0073] The timestamp of the location data is used as the reference timeline because GNSS timestamps have a globally unified standard and extremely high accuracy, making them suitable as a global time reference for the entire testing process. After data acquisition, the data processing engine receives location data frames from the GNSS module and vehicle operation data frames from the CAN module. For each location data frame, the engine searches the CAN data stream for the vehicle operation data frame with the closest timestamp and a deviation within the allowable tolerance (e.g., ±0.01 seconds). If it exceeds the tolerance, it is marked as out of sync or discarded, thus ensuring that each set of location data can accurately match the vehicle's dynamic state at the corresponding moment.

[0074] This alignment operation eliminates data misalignment caused by device clock drift or transmission delay, enabling subsequent multi-condition joint determination based on spatial location (such as lane line distance) and dynamic state (such as lateral acceleration) to be based on the vehicle state at the same moment, thus ensuring the physical authenticity of the determination logic.

[0075] In one specific implementation, within the test control box, the GNSS module outputs a location data frame containing longitude, latitude, speed, and a UTC timestamp every 10 milliseconds; the CAN module outputs a CAN data frame containing lateral acceleration, vehicle speed, and system status every 10 milliseconds. The data processing engine uses the timestamp of the GNSS data frame as an index to retrieve CAN data frames in the CAN data buffer whose absolute timestamp difference is less than 0.01 seconds. If the GNSS data timestamp is T at a certain moment, and the most recent CAN data timestamp is T-0.008 seconds, then the CAN data is bound to the GNSS data to form a set of vehicle status data; if the most recent CAN data timestamp is T-0.012 seconds, then the pairing is discarded because it exceeds the tolerance, and the frame loss rate is recorded.

[0076] Another embodiment involves interpolation alignment at different sampling rates. When the CAN bus sampling frequency is slightly lower than that of GNSS, linear interpolation can be used to calculate the estimated lateral acceleration at time T based on the two closest CAN data points before and after time T at the GNSS timestamp T, thereby achieving more precise time alignment and ensuring data consistency.

[0077] The vehicle testing method of this application achieves complete acquisition of vehicle dynamics and spatial position information by obtaining vehicle operation data and position data including lateral acceleration signals, providing the necessary data foundation for subsequent multi-dimensional state determination. By using the timestamp of the position data as a reference time axis and aligning the vehicle operation data with it, time deviations caused by different clock sources between different communication buses are eliminated. This ensures strict synchronization of position information and dynamics data in the time dimension, which is beneficial to improving the accuracy of the multi-condition joint determination engine, ensuring the consistency and reliability of test data, and thus improving the data quality of the declaration value confirmation process.

[0078] This application's embodiments are aimed at... Figure 4 The vehicle testing method shown in step S32 provides a specific data quality control scheme, introducing a frame count statistics mechanism based on time alignment to quantify data integrity, and establishing an automatic retest closed-loop logic when data quality is unqualified. Please refer to [link / reference] for details. Figure 5 , Figure 5 This is a flowchart illustrating the fourth embodiment of the vehicle testing method provided in this application.

[0079] like Figure 5 As shown, the specific steps are as follows: Step S41: Determine the theoretical number of data frames for the location data based on the reference time axis.

[0080] In this embodiment, the reference time axis refers to a time coordinate system with the timestamp of the location data as a unified reference; the theoretical data frame number refers to the total number of expected data frames calculated according to the sampling frequency of the reference time axis during the test duration, and the calculation formula is the test duration multiplied by the sampling frequency.

[0081] Step S42: Determine the number of valid data frames in the aligned vehicle operation data that correspond to the data frame timestamps of the location data.

[0082] In this embodiment of the application, the number of valid data frames refers to the number of vehicle operation data frames that successfully match the corresponding location data timestamp and whose time deviation is within the allowable tolerance range during the time alignment process.

[0083] Autonomous vehicle testing demands extremely high data continuity and integrity; any missing data can lead to errors in subsequent steady-state segment extraction or deviations in state determination. During the data alignment phase, transient congestion in CAN bus communication or brief loss of GNSS signals can cause some CAN data frames to fail to precisely match the GNSS timestamp. Therefore, the data processing engine needs to count the amount of lost data. Specifically, using the GNSS timestamp sequence as a benchmark, the engine iterates through each time point in the sequence, searching the CAN data buffer for the CAN data frame with the closest timestamp. If the absolute value of the difference between the found CAN data frame's timestamp and the GNSS timestamp is less than or equal to the tolerance (e.g., 0.01 seconds), it is counted as a valid data frame; if no match is found or the difference exceeds the tolerance, it is counted as lost. The number of valid data frames is obtained by counting the number of valid matches across all time points.

[0084] Step S43: Determine the frame drop rate for the test condition based on the theoretical number of data frames and the effective number of data frames.

[0085] In this embodiment of the application, the frame drop rate refers to the percentage of the difference between the theoretical number of data frames and the effective number of data frames relative to the theoretical number of data frames, and is used to measure the completeness of data sampling.

[0086] Step S44: When the frame loss rate exceeds the preset threshold, restart the autonomous vehicle to perform the test condition and collect data again.

[0087] In this embodiment of the application, the preset threshold refers to the critical value for determining whether the data quality is qualified, such as 1%.

[0088] The test control box compares the number of valid data frames with the theoretical number of data frames to calculate the frame drop rate. If the frame drop rate exceeds a preset threshold, it indicates that the data from this test has serious defects, and direct use will lead to unreliable statistical results. In this case, the test control box triggers a retest mechanism, outputs an exception prompt through the report generation module, and the test case management module adds this condition back to the end of the test sequence, waiting to be re-executed. This mechanism ensures that the data entering the subsequent statistical analysis stage is of high quality and high completeness.

[0089] In one specific implementation, during a 10-second test, the GNSS module outputs at a frequency of 100Hz, with a theoretical data frame count of 1000 frames. During the alignment process, the data processing engine found that three GNSS time points failed to find matching CAN data frames, while the remaining 997 time points successfully matched. At this point, the effective data frame count is 997, and the frame loss rate is (1000-997) / 1000. 100% = 0.3%. If the preset threshold is 1%, then 0.3% is less than 1%, the data is qualified, and it will proceed to the next step of processing.

[0090] Another embodiment involves a scenario where the frame loss rate exceeds the limit. If 20 consecutive frames of data are lost during a test due to CAN bus interference, the frame loss rate is 2%, exceeding the preset threshold of 1%. The data processing engine determines that the data quality for this condition is unqualified, sends a command to the driving robot control system via the communication interface, requesting that the test condition be re-executed, and marks the collected data as abnormal and archives it, excluding it from the final declaration value calculation.

[0091] The vehicle testing method in this application calculates the frame loss rate by determining the theoretical number of data frames based on a reference timeline and counting the number of valid data frames. This achieves quantitative monitoring of the integrity of data acquisition, enabling objective evaluation of the quality of test data. When the frame loss rate exceeds a preset threshold, the autonomous vehicle is re-driven to execute the test condition and re-acquire data. This helps to eliminate data defects caused by communication interference or signal loss, thereby further improving the reliability and consistency of test data. This ensures that subsequent threshold statistics and declaration value calculations are based on high-quality data, avoiding adverse effects of erroneous data on certification results.

[0092] Furthermore, the embodiments of this application are aimed at Figure 4 or Figure 5 In the data preprocessing of the vehicle testing method shown, in addition to aligning the dual-source data, this application can also perform specific types of filtering on key dynamic signals, namely, digital low-pass filtering using zero-phase filtering technology.

[0093] Specifically, the lateral acceleration signal refers to the acceleration value of the vehicle along the lateral axis, which is usually obtained by the vehicle's inertial measurement unit via the CAN bus; the yaw rate signal refers to the angular velocity of the vehicle rotating around the vertical axis, which is also obtained by the inertial measurement unit.

[0094] During vehicle testing, lateral acceleration and yaw rate signals are highly susceptible to interference from road surface irregularities, vehicle vibrations, and sensor noise. These high-frequency noises can interfere with subsequent determinations of the vehicle's steady-state driving condition. Therefore, a low-pass filter is needed to remove high-frequency interference and extract low-frequency signals that reflect the vehicle's true motion trend.

[0095] This application employs a fourth-order Butterworth low-pass filter with a cutoff frequency of 0.5Hz. Because the lateral acceleration changes in lane-keeping assist systems are typically gradual, a 0.5Hz cutoff frequency effectively preserves the steady-state acceleration characteristics of the vehicle during cornering while filtering out high-frequency jitter. However, traditional digital filters introduce phase delay during the filtering process, meaning the peak value of the filtered signal lags behind the original signal. In the testing method, subsequent steps require extracting the steady-state window and calculating the moving average lateral jerk (Jerk). The phase delay causes the calculated jerk value to mismatch with the actual moment, thus affecting the accurate determination of whether to maintain or deviate from the target state.

[0096] The test control box employs zero-phase filtering. Zero-phase filtering cancels out phase delays through two filtering operations (forward and reverse). The filtered signal not only meets the amplitude-frequency characteristics requirements but also has a time-domain waveform that is strictly synchronized with the original signal, without phase lag. This means that each sampling point of the lateral acceleration signal after zero-phase filtering precisely corresponds to the vehicle's actual physical state at that moment, without any time deviation caused by algorithm processing.

[0097] Zero-phase filtering is a special digital filtering method that eliminates the phase delay caused by traditional causal filters by first performing forward filtering and then reverse filtering on the signal, thus ensuring that the filtered signal waveform is completely aligned with the original signal on the time axis.

[0098] In one specific implementation, pre-calculated fourth-order Butterworth low-pass filter coefficients are loaded into the embedded processor. For the acquired lateral acceleration sequence, it is first forward-filtered into the input filter to obtain the forward-filtered result; then, the time sequence of this forward-filtered result is reversed, and it is input into the same filter again for reverse filtering; finally, the result of the second filtering is reversed again to obtain the final zero-phase filtered output. This process can be implemented using existing digital signal processing library functions, such as the `filtfilt` function in MATLAB or the `scipy.signal.filtfilt` function in Python.

[0099] Another embodiment involves parameter adjustment at different sampling rates. When the sensor sampling frequency is 200Hz, in order to keep the normalized cutoff frequency corresponding to the 0.5Hz cutoff frequency consistent, the filter coefficients need to be recalculated or scaled according to the new sampling rate to ensure the consistency of the filtering effect, but still follow the principle of zero-phase filtering.

[0100] By applying digital low-pass filtering to the aligned lateral acceleration and yaw rate signals, sensor noise and high-frequency vibration interference are effectively filtered out, extracting low-frequency signals that reflect the vehicle's true motion trend. This improves the accuracy of subsequent steady-state segment extraction and state determination. Zero-phase filtering eliminates the phase delay introduced by traditional digital filters, ensuring that the filtered signal waveform is strictly synchronized with the original signal on the time axis. This allows subsequent calculations of lateral acceleration and steady-state window determination to be based on precise time correspondence, thereby improving the consistency and reliability of the multi-condition joint determination engine and ensuring data quality during the declaration value confirmation process.

[0101] This application's embodiments are aimed at... Figure 1 The vehicle testing method step S14 shown proposes specific data processing steps to determine the maintained sample set and the deviated sample set, and introduces automated processing logic for steady-state window extraction and multi-condition joint judgment.

[0102] Please refer to details. Figure 6 and Figure 7 , Figure 6 This is a flowchart illustrating the fifth embodiment of the vehicle testing method provided in this application. Figure 7 This is a flowchart of the multi-condition joint decision logic of the maintain / deviation dual-state decision engine in the data processing engine provided in this application.

[0103] like Figure 6 As shown, the specific steps are as follows: Step S51: Extract steady-state window data from several data frames after the vehicle enters the preset state in the vehicle state data.

[0104] In this embodiment, steady-state window data refers to a sequence of continuous data frames that meet specific filtered lateral acceleration standard deviation and vehicle speed fluctuation limits and last for no less than 5 seconds, starting from when the vehicle enters the stable turning phase after entering the curve; preset state refers to the moment when the vehicle completes the curve transition and enters the stable driving phase of the curve.

[0105] Since the raw data acquisition includes multiple stages such as corner entry transition, steady-state driving, and corner exit transition, directly analyzing the entire dataset would introduce non-steady-state noise, leading to distorted state determination. Therefore, the test control box extracts steady-state window data from several data frames after the vehicle enters the preset state, thus isolating dynamic interference during the transition phase.

[0106] The process of extracting the steady-state window involves applying a fourth-order Butterworth low-pass filter with a cutoff frequency of 0.5Hz to eliminate high-frequency noise, and selecting continuous data segments with a standard deviation of lateral acceleration less than 0.1m / s² and vehicle speed fluctuations not exceeding ±1km / h after filtering. This ensures that the data involved in the judgment are in the plateau period of stable vehicle dynamic response.

[0107] Specifically, because the dynamic response of a vehicle fluctuates after entering a corner during actual track testing, directly capturing data may include noise from the transition phase, leading to distortion in subsequent state assessments. Therefore, by acquiring candidate window data and applying multi-dimensional filtering conditions, it is possible to accurately identify the period when the vehicle truly enters steady-state driving.

[0108] Specifically, the feature that the standard deviation of lateral acceleration is less than the first preset difference utilizes the statistical concept of standard deviation to quantify the dispersion of the filtered lateral acceleration signal. When the standard deviation is less than 0.1 m / s², it indicates that the vehicle's lateral force is in a highly stable state, without violent left-right swaying or control oscillations. The feature that the maximum speed difference is less than or equal to the second preset difference limits longitudinal speed fluctuations by calculating the difference between the highest and lowest speeds within the candidate window. When the difference does not exceed 2 km / h, it indicates that the vehicle maintains a constant driving speed, eliminating the interference of centrifugal force changes caused by acceleration or deceleration on lateral stability. The feature that the duration is greater than or equal to the preset duration introduces a time dimension constraint.

[0109] A system is considered to be in a steady state only if both of the aforementioned kinetic parameters are consistently satisfied for at least 5 seconds. This ensures that the window data has a sufficient time span to represent the system's sustained performance under specific operating conditions, rather than a brief coincidence or transient stability.

[0110] These three conditions work together to form a rigorous selection logic: only when the lateral force is stable, the longitudinal velocity is constant, and this condition persists for a sufficiently long period of time is the system considered to be in an analyzable steady state. This design avoids the arbitrariness of manual judgment and achieves automated and objective extraction of the steady-state window.

[0111] In one specific implementation, during a test at a speed range of 60-100 km / h and a curve radius of 250m, the vehicle entered the curve at 90 km / h. The data processing engine began scanning the candidate window data two seconds after entering the curve. The initial candidate window consisted of 10 consecutive seconds of data. Calculations revealed that the standard deviation of the filtered lateral acceleration within this window was 0.08 m / s² (less than the first preset difference of 0.1 m / s²), the maximum vehicle speed was 90.5 km / h, the minimum was 89.8 km / h, and the maximum difference was 0.7 km / h (less than the second preset difference of 2 km / h). This stable state lasted for 10 seconds (greater than the preset duration of 5 seconds). Therefore, this 10-second data set was determined as the steady-state window data.

[0112] Another embodiment involves handling critical situations. If the standard deviation of lateral acceleration within the candidate window is 0.12 m / s² (greater than 0.1 m / s²), even if the duration exceeds 5 seconds and the vehicle speed is stable, the window is not determined to be a steady-state window due to excessive lateral force fluctuations. The engine will continue to slide the window backward to find a data segment that meets the conditions until a window that meets all three conditions is found or it is determined that there is no valid steady-state data for the operating condition.

[0113] Step S52: Perform multi-condition joint determination for each data frame in the steady-state window data.

[0114] In this embodiment of the application, the multi-condition joint determination refers to the logical judgment process that simultaneously checks four dimensions: the distance of the outermost point of the vehicle from the lane line, the activation status of the ACSF system, the 0.5-second moving average lateral acceleration, and the 1-second standard deviation of the filtered lateral acceleration.

[0115] The test control box performs multi-condition joint judgment on each data frame within the steady-state window, rather than relying on a single threshold or manual observation.

[0116] Specifically, the judgment logic includes, but is not limited to: calculating the distance d from the outermost point of the vehicle to the lane line, requiring d ≥ 0.3m to leave a safety margin; checking whether the ACSF system status signal is Active and without Override or Fail-safe flags to confirm that the system is working properly; calculating the 0.5-second moving average lateral acceleration Jerk_0.5s, requiring the absolute value ≤ 5m / s³ to limit drastic changes in control input; and calculating the standard deviation σ_ay_1s of the filtered lateral acceleration within a 1-second time window, requiring σ_ay_1s < 0.1m / s² to verify the smoothness of the acceleration output.

[0117] A data frame is marked as held if and only if all four conditions C1, C2, C3, and C4 are met simultaneously; otherwise, it is marked as deviated.

[0118] This multi-condition joint judgment mechanism transforms complex vehicle dynamics behavior into clear state labels through logical AND operations, eliminating misjudgments caused by subjective human factors and the limitations of single indicators.

[0119] like Figure 7 As shown, the multi-condition joint decision logic of the hold / deviation dual-state decision engine provided in this application introduces a joint decision engine composed of four specific physical quantities and system states.

[0120] Among them, the lane distance condition refers to the lateral distance between the outermost point of the vehicle and the edge of the lane line needing to meet a specific safety margin; the system state condition refers to the operating mode of the advanced driver assistance system needing to be in normal working condition; the lateral acceleration condition refers to the rate of change of lateral acceleration over time needing to be limited within a safe threshold; and the lateral acceleration stability condition refers to the degree of fluctuation of the lateral acceleration signal within a short time window needing to be within a controllable range.

[0121] Specifically, single-dimensional judgments (such as relying solely on whether a lane line has been crossed) are easily affected by sensor noise, uneven road surfaces, or transient interference, leading to misjudgments. Therefore, by using a multi-condition joint judgment, the vehicle status can be comprehensively evaluated from four dimensions: spatial location, system health, dynamic smoothness, and signal stability.

[0122] Specifically, the lane distance condition is calculated by setting d=W_lane / 2-|Y_vehicle|-W_vehicle / 2 and setting a threshold d≥0.3m to ensure that the vehicle not only does not cross the lane line, but also retains sufficient safety margin to prevent critical misjudgment caused by positioning error or vehicle body sway.

[0123] The system status condition requires the ACSF system to be in "Active" and without Override or Fail-safe status, eliminating data interference during driver intervention or system failure, and ensuring that the judgment result reflects the true level of the system's automatic control capability.

[0124] The lateral jerk condition was evaluated by calculating the 0.5-second moving average jerk and limiting |Jerk_0.5s|≤5m / s³. This assessed the smoothness of the steering input, preventing vehicle instability caused by sharp or abrupt steering maneuvers, and also filtering out abnormal data caused by oscillations in the control algorithm.

[0125] The stability condition of lateral acceleration was further verified by calculating the standard deviation of the filtered lateral acceleration σ_ay_1s < 0.1 m / s² within a 1-second window, thus eliminating the instantaneous pulse interference.

[0126] These four conditions are combined using a logical AND operation. The condition is marked as "hold" only when all conditions are met simultaneously; otherwise, it is marked as "off". This design achieves both rigor and robustness in the decision logic.

[0127] In one specific implementation, during the test, the vehicle travels with a lateral acceleration of 2.5 m / s². The data processing engine calculates various conditions in real time: the outermost point of the vehicle is 0.35 m from the lane line (meeting ≥0.3 m), the ACSF system displays Active and no driver intervention (meeting system status conditions), the lateral acceleration changes from 2.48 m / s² to 2.52 m / s² within 0.5 seconds, and the calculated jerk is 0.08 m / s³ (meeting ≤5 m / s³), and the standard deviation of the lateral acceleration within 1 second is 0.05 m / s² (meeting <0.1 m / s²). Since all four conditions are met, the data frame is marked as "held".

[0128] Another embodiment involves a critical situation: if, at a certain moment, the lateral acceleration fluctuates instantaneously due to road bumps, causing the standard deviation per second to be slightly higher than 0.1 m / s² (e.g., 0.11 m / s²), even though the lane distance and system state are normal, the data frame is still marked as "off-target" because the lateral acceleration stability condition is not met. This avoids misjudging the disturbed unstable state as a stable system state and improves the accuracy of the judgment.

[0129] Step S53: Mark data frames that meet the determination conditions as held state, and mark data frames that do not meet the determination conditions as off-state.

[0130] In this embodiment of the application, the "maintained state" refers to the state in which the data frame simultaneously meets the above four determination conditions; the "deviation state" refers to the state in which any one of the determination conditions in the data frame is not met.

[0131] Step S54: Determine the set of hold samples based on all data frames marked as hold states.

[0132] Step S55: Determine the set of biased samples based on all data frames marked as biased states.

[0133] In this embodiment, the test control box aggregates the lateral acceleration values ​​corresponding to all data frames marked as holding states to form a holding sample set, and aggregates the lateral acceleration values ​​corresponding to all data frames marked as deviating states to form a deviating sample set. This fine-grained data frame-based marking and aggregation method can more accurately capture the probability distribution characteristics of the system near the critical boundary, providing a high-quality data foundation for subsequent calculation of confidence intervals and declared values.

[0134] In one specific implementation, under test conditions of 60-100 km / h vehicle speed, 250m curve radius, and target lateral acceleration of 2.50 m / s², the driving robot controls the vehicle to enter the curve at 90 km / h. The data processing engine begins searching for a steady-state window 2 seconds after entering the curve, finding a 10-second data segment where the filtered lateral acceleration standard deviation is 0.08 m / s², and the vehicle speed fluctuation is ±0.5 km / h, meeting the steady-state extraction conditions. Each of the 1000 data frames within this 10-second period undergoes multi-condition evaluation. Assuming that 950 of these data frames have a distance greater than 0.3m from the lane line, are in Active ACSF state, have a 0.5-second moving average absolute value of lateral acceleration less than 5 m / s³, and have an acceleration standard deviation less than 0.1 m / s² within the 1-second window, these 950 data frames are marked as maintaining a steady-state state. Of the other 50 data frames, 30 were marked as off track because the distance to the lane line was less than 0.3m, and 20 were marked as off track because the lateral acceleration exceeded the limit. Ultimately, the mean lateral acceleration of the 950 hold-up data frames was assigned to the hold-up sample set, and the mean lateral acceleration of the 50 off track data frames was assigned to the off track sample set.

[0135] Another embodiment involves adjusting the judgment parameters at different vehicle speeds. For high-speed conditions above 150km / h, the distance threshold from the lane line can be appropriately relaxed to 0.4m to adapt to the changes in the lateral stability of vehicles at high speeds. However, the logical framework of multi-condition joint judgment remains unchanged. The state of maintaining or deviating is still marked by the intersection of the four conditions to ensure the consistency and scientific nature of the judgment criteria.

[0136] The vehicle testing method of this application extracts steady-state window data from several data frames after the vehicle enters a preset state, enabling the data involved in state determination to eliminate dynamic noise interference during the curve transition phase, thus improving the purity and representativeness of the data acquisition. By performing multi-condition joint determination on each data frame in the steady-state window data, a multi-dimensional quantitative assessment of the vehicle's lateral holding capability is achieved, eliminating the subjectivity and inconsistency caused by relying on manual interpretation or a single threshold, and improving the consistency and repeatability of the state determination results. A holding sample set is determined based on data frames marked as holding states, and a deviation sample set is determined based on data frames marked as deviation states. This allows the sample sets to more accurately reflect the probability distribution characteristics of the system near the critical boundary, providing a high-quality data foundation for subsequent statistical confidence interval calculations and declaration value determination, which is beneficial to improving the scientificity and reliability of the confirmation of the lateral acceleration declaration value of autonomous vehicles.

[0137] This application's embodiments are aimed at... Figure 1The vehicle testing method step S15 shown provides specific statistical calculation steps for determining the upper bound of the confidence interval. It introduces the calculation of the optimal threshold estimate, the acquisition of the sample standard deviation, the calculation of the standard error, and the determination of the final upper bound of the confidence interval, forming a complete link from raw data to statistical inference. Please refer to [link / reference] for details. Figure 8 , Figure 8 This is a flowchart illustrating the sixth embodiment of the vehicle testing method provided in this application.

[0138] like Figure 8 As shown, the specific steps are as follows: Step S61: Calculate the best estimate of the threshold based on the preserved sample set and the deviated sample set.

[0139] In the embodiments of this application, the optimal estimate of the threshold is the arithmetic mean of the 95th percentile of the retaining sample and the 5th percentile of the deviating sample, which is used to characterize the balance between retaining ability and deviating tendency.

[0140] The hold samples reflect the system's performance within the safety margin, while the outlier samples reflect the system's failure boundary. The test control box takes the average of the 95th percentile of the hold samples (representing the limit in the vast majority of hold cases) and the 5th percentile of the outlier samples (representing the limit in the very few outlier cases). This allows for a more accurate location of the center of the overlapping region of their probability distributions, i.e., the optimal estimate of the threshold. This design utilizes two-sided information, avoiding the estimation bias caused by using only one-sided data.

[0141] Step S62: Calculate the standard deviation of lateral acceleration for the retained sample set.

[0142] In the embodiments of this application, maintaining the lateral acceleration standard deviation of the sample set refers to maintaining the degree of dispersion of each data point in the sample from its mean.

[0143] The standard deviation quantifies the degree of dispersion within the held-up sample. The smaller the standard deviation, the more stable the system's performance in the held-up state and the more concentrated the data distribution.

[0144] Step S63: Calculate the standard error of the best estimate of the threshold based on the standard deviation of the lateral acceleration and the number of samples in the holding sample set.

[0145] In the embodiments of this application, the standard error refers to the standard deviation of the best estimate of the threshold under repeated sampling, which reflects the volatility of the estimate.

[0146] The standard error formula is the standard deviation divided by the square root of the sample size. It reflects the improvement in the accuracy of the optimal threshold estimate as the sample size increases. The larger the sample size, the smaller the standard error, and the more reliable the estimate.

[0147] Step S64: Determine the upper bound of the confidence interval for each test condition based on the standard error and the best estimate of the threshold.

[0148] In the embodiments of this application, the upper bound of the confidence interval refers to the upper limit value obtained by adding or subtracting a certain multiple of the standard error based on the best estimate of the threshold, which usually corresponds to a 95% confidence level.

[0149] The upper bound of the confidence interval equals the best estimate of the threshold plus 1.96 times the standard error (corresponding to a 95% confidence level). This calculation process extends the point estimate to an interval estimate, taking into account the uncertainty brought about by the randomness of the sample, making the determined upper bound of the confidence interval statistically reliable.

[0150] In one specific implementation, within the vehicle speed range of 60-100 km / h, the maintain sample set K contains 50 data points, and the deviated sample set D contains 20 data points. Keep_P95 is calculated to be 2.85 m / s², and Deviate_P5 is calculated to be 2.75 m / s². Therefore, the optimal threshold estimate is (2.85 + 2.75) / 2 = 2.80 m / s². The standard deviation σ_keep of the maintain sample set K is calculated to be 0.08 m / s². The sample size n_k is 50. The standard error SE is calculated to be approximately 0.08 / sqrt(50) ≈ 0.0113 m / s². The upper bound of the 95% confidence interval is determined to be 2.80 + 1.96. 0.0113 ≈ 2.822 m / s². Another embodiment involves a smaller sample size. If n_k is only 10, the standard error increases accordingly, resulting in a more conservative upper bound on the confidence interval. This provides a higher safety margin when data is insufficient, ensuring the robustness of the reported values.

[0151] The vehicle testing method in this application calculates the optimal threshold estimate, integrating boundary information from both hold-up and outlier samples, resulting in more accurate threshold positioning and reflecting the true balance point of system performance. By calculating the standard deviation of the lateral acceleration of the hold-up sample set, the stability of the system's hold-up state is quantified, providing a basis for dispersion in subsequent statistical inference. The standard error is calculated based on the standard deviation and sample size, ensuring that the determination of the upper bound of the confidence interval reflects the impact of sample size; a larger sample size leads to more accurate estimation, improving the scientific rigor of statistical inference. Determining the upper bound of the confidence interval based on the standard error and the optimal threshold estimate ensures that the determination process for the declared value fully considers data uncertainty, guaranteeing reliability at a 95% confidence level, and enhancing the credibility and compliance of the declared value in certification testing.

[0152] This application's embodiments are aimed at... Figure 1The vehicle testing method shown in step S14 further evaluates the reliability of the judgment result by quantifying the degree of overlap between the two, introduces the statistical indicator of gray zone width, and triggers an early warning mechanism based on the indicator.

[0153] Please refer to details. Figure 9 , Figure 9 This is a flowchart illustrating the seventh embodiment of the vehicle testing method provided in this application.

[0154] like Figure 9 As shown, the specific steps are as follows: Step S71: Determine the width of the gray zone based on the maximum lateral acceleration in the retained sample set and the minimum lateral acceleration in the deviated sample set.

[0155] In this embodiment, the gray zone width refers to the difference between the maximum value in the hold sample set and the minimum value in the deviation sample set, used to characterize the overlapping range of the two states in terms of lateral acceleration values. The maximum lateral acceleration value in the hold sample set refers to the largest lateral acceleration value among all test data determined to be in the hold state; the minimum lateral acceleration value in the deviation sample set refers to the smallest lateral acceleration value among all test data determined to be in the deviation state.

[0156] After determining the hold-up and deviate-up sample sets, the test control box extracts the maximum value `Keep_max` from the hold-up samples and the minimum value `Deviate_min` from the deviate-up samples. These two extreme values ​​represent the lower limits of the system's ability to maintain stability and the lower limits of the system's ability to begin deviating, respectively. The difference between the two is calculated as `GreyZone = max(0, Keep_max - Deviate_min)`. If the result is positive, it indicates the existence of a lateral acceleration interval within which some tests exhibit hold-up behavior and some exhibit deviate behavior, i.e., there is an overlap in probability distributions. This overlap reflects the instability of the system at the critical state or the discreteness of the test data.

[0157] Step S72: When the width of the gray area is greater than the preset value, it is determined that there is a judgment ambiguity area in the test condition, and a warning prompt is output.

[0158] In this embodiment of the application, the ambiguous region refers to the uncertain state in which the vehicle may exhibit either a holding or a deviating state within the lateral acceleration range due to the overlap between the holding sample and the deviating sample.

[0159] When the width of the gray area exceeds a preset value, it indicates significant uncertainty. The current test sample size may be insufficient to accurately characterize the true boundaries of the system, or the system's control performance may fluctuate considerably in this region. In this case, the test control box determines that there is a region of ambiguity in the test condition and outputs a warning message to remind testers or evaluation experts that the confidence level of the data in this region is low, requiring an increase in the test sample size or a re-evaluation.

[0160] In one specific implementation, within a certain vehicle speed range, after repeated testing, the maximum lateral acceleration in the sample is kept at 2.65 m / s², while the minimum lateral acceleration deviating from the sample is 2.60 m / s², resulting in a gray area width of 0.05 m / s². If the preset value is 0.1 m / s², since 0.05 is less than the preset value, the system determines there is no significant blurry area. If the maximum sample value is kept at 2.70 m / s², and the minimum deviating sample value is 2.60 m / s², the gray area width is 0.10 m / s². If the preset value is set to 0.08 m / s², then 0.10 is greater than the preset value, and the system outputs a warning, suggesting further testing. Another embodiment involves differentiated warnings for different vehicle speed ranges. For high-speed ranges, due to the more complex dynamic characteristics, a smaller preset value (e.g., 0.05 m / s²) may be set, while a larger preset value (e.g., 0.15 m / s²) may be set for low-speed ranges to adapt to the system stability characteristics at different speeds.

[0161] The vehicle testing method of this application quantitatively assesses the degree of overlap between the maintained sample and the deviation sample by calculating the width of the gray zone, enabling intuitive identification of uncertain areas in the judgment results. When the gray zone width exceeds a preset value, a warning is output, allowing testers to promptly identify conditions with severe data overlap and low judgment confidence. This facilitates prompting an increase in the test sample size to improve the reliability of statistical results, thereby reducing the risk of deviations in declared values ​​due to insufficient samples or system fluctuations, and further enhancing the scientific rigor and precision of the declared value confirmation process.

[0162] This application's embodiments are aimed at... Figure 1 The vehicle testing method shown in step S16 further conducts system-level margin assessment and availability verification, introduces the global maximum declared value as a benchmark, calculates the functional test margin to evaluate the stability of the low declared value condition relative to the high declared value condition within the functional test range, and marks the unusable condition accordingly.

[0163] Please refer to details. Figure 10 , Figure 10 This is a flowchart illustrating the eighth embodiment of the vehicle testing method provided in this application.

[0164] like Figure 10 As shown, the specific steps are as follows: Step S81: Obtain the maximum declared value of lateral acceleration for all test conditions.

[0165] In this embodiment of the application, the maximum declared value of lateral acceleration refers to the maximum value among all the declared values ​​of lateral acceleration determined in the final calculation of all test conditions.

[0166] The test control box acquires the maximum declared value of lateral acceleration across all test conditions to determine the upper limit of the system's capabilities globally. Since declared values ​​may differ across different speed ranges or corner radii, the maximum declared value represents the most conservative or stringent performance boundary benchmark for the system.

[0167] Step S82: Calculate the functional test margin for each test condition based on the maximum declared value of lateral acceleration and the declared value of lateral acceleration for each test condition.

[0168] In this embodiment of the application, the functional test margin refers to the difference between the declared value of lateral acceleration for a specific test condition and the maximum declared value multiplied by a preset coefficient (such as 0.9), which is used to quantify the safety margin of the test condition in the functional test.

[0169] The test control box calculates the functional test margin based on the maximum declared value and the declared value of the current test condition. Functional testing requires the vehicle to remain stable within 80% to 90% of the declared value range. If the declared value for a certain condition is significantly lower than the maximum declared value, then the 90% declared value for that condition may still be lower than 80% or even lower than other conditions, or its absolute value may be too small to maintain stability during dynamic testing. This relative difference can be quantified by calculating the difference.

[0170] Step S83: When the functional test margin is less than or equal to the preset value, output a warning signal that the functional test margin is insufficient, and mark the declared value of the test condition as unavailable.

[0171] In the embodiments of this application, the preset value is usually zero or a small positive number, which is used to define whether the margin is sufficient; unusable means that the declared value of the working condition is determined to be invalid by the system due to insufficient margin and is no longer a candidate for the final certification declaration value.

[0172] When the functional test margin is less than or equal to the preset value, it indicates that the declared value for that operating condition is too low. This may lead to a failure to maintain consistency with the global performance benchmark within the functional test range, or there is a risk of functional test failure due to insufficient margin. In this case, a warning signal is output and marked as unavailable. This automatically eliminates declared values ​​that are statistically valid but have defects in overall system consistency, ensuring that the final submitted declared values ​​are comparable and reliable under all operating conditions.

[0173] In one specific implementation, it is assumed that the declared value is 2.5 m / s² in the 60-100 km / h range, and 2.0 m / s² in the 100-130 km / h range, with a maximum declared value of 2.5 m / s². If the preset coefficient is 0.9, then the maximum declared value for 90% of the 100-130 km / h range is 2.25 m / s². The functional test margin for this condition is 2.0 - 2.25 = -0.25 m / s². If the preset value is 0, then -0.25 is less than 0, the system determines that the margin is insufficient, outputs a warning, and marks the 2.0 m / s² declared value as unusable. Another embodiment involves adjusting different preset values. For scenarios with high safety requirements, the preset value can be set to 0.1 m / s². In this case, even if the margin is 0.05 m / s², it will still be determined as unusable, thereby further tightening the screening criteria and ensuring the absolute robustness of the declared value.

[0174] The vehicle testing method of this application obtains the maximum declared value of lateral acceleration in all test conditions and uses this as a benchmark to calculate the functional test margin. This allows for the assessment of the relative stability of the declared values ​​for each test condition within the global performance framework, which is beneficial for identifying and eliminating abnormal declared values ​​that are statistically valid locally but have poor global consistency. By outputting a warning signal and marking the declared value as unusable when the functional test margin is less than or equal to a preset value, automatic screening and quality control of declared values ​​are achieved. This makes the final set of declared values ​​more rigorous and reliable, avoiding the risk of functional test failure due to insufficient margin in individual test conditions, and improving the consistency and credibility of certification test results for the steering control system of autonomous vehicles.

[0175] This application's embodiments are aimed at... Figure 1 Following step S16 of the vehicle testing method shown, an over-limit trigger margin assessment mechanism based on the optimal threshold estimate is further introduced, forming a secondary verification closed loop for the reliability of the declared value. Please refer to [link / reference] for details. Figure 11 , Figure 11 This is a flowchart illustrating the ninth embodiment of the vehicle testing method provided in this application.

[0176] like Figure 11 As shown, the specific steps are as follows: Step S91: Calculate the over-limit trigger margin based on the declared lateral acceleration value, the preset over-limit amplitude value, and the best estimated threshold value for each test condition.

[0177] In this embodiment, the preset over-limit amplitude value refers to the acceleration increment allowed for short-term over-limit as specified in regulations or testing standards, such as 0.3 m / s²; the optimal threshold estimate refers to the statistical value representing the system's equilibrium position between maintaining and deviating, calculated based on the 95th percentile of the maintaining sample and the 5th percentile of the deviating sample; the over-limit trigger margin refers to the difference between the target acceleration under the normal over-limit amplitude and the optimal threshold estimate, used to quantify the degree of certainty that the system will trigger a deviating state in the over-limit test.

[0178] The test control box calculates the threshold using the formula Margin_TR1 = (a_y_smax + preset over-limit amplitude value) - best estimated threshold value. The preset over-limit amplitude value can be set to 0.3 m / s². This calculation logic aims to assess how far the vehicle's lateral acceleration reaches a_y_smax + 0.3 from the statistical center (best estimated threshold value) where the system actually deviates. If the margin is too small, it means that in regular over-limit tests, the system may fail to reliably trigger the deviation due to statistical fluctuations, leading to certification test failure. By quantifying this distance, the risk of declared values ​​being too close to the system's instability boundary can be detected in advance.

[0179] Step S92: When the over-limit trigger margin is less than or equal to the preset value, output a warning signal indicating insufficient over-limit trigger margin, and mark the declared value of the test condition as unavailable.

[0180] In this embodiment, the test control box compares the calculated over-limit trigger margin with a preset safety threshold (e.g., 0). The preset value can be 0 or a small positive number. When the margin is less than or equal to the preset value, it indicates that the declared value has a risk of failure during the over-limit test. At this time, a warning signal is output to alert the tester, and the declared value is marked as unusable, preventing it from entering the final report or certification submission stage. This mechanism automatically intercepts unreliable declared values, avoiding repeated certifications and increased costs caused by improperly set declared values.

[0181] In one specific implementation, within a vehicle speed range of 60-100 km / h, the calculated lateral acceleration declaration value for a certain test condition is 2.50 m / s², the preset over-limit value is 0.3 m / s², and the optimal estimated threshold value for this condition is 2.75 m / s². The calculated over-limit trigger margin is (2.50 + 0.3) - 2.75 = 0.05 m / s². If the preset value is 0.1 m / s², then 0.05 is less than 0.1, and the system outputs a warning signal indicating insufficient over-limit trigger margin, marks the declaration value as unusable, and suggests that the tester lower the declaration value or increase the test sample to narrow the confidence interval.

[0182] The vehicle testing method of this application calculates the over-limit trigger margin based on the declared lateral acceleration value, the preset over-limit amplitude value, and the optimal estimate of the threshold, thereby achieving a quantitative assessment of the trigger reliability of the declared value in conventional over-limit testing, making the verification dimensions of the declared value more comprehensive. By outputting a warning signal and marking the declared value as unusable when the over-limit trigger margin is less than or equal to the preset value, it is beneficial to intercept unreliable declared values ​​at the end of the testing process, avoiding the risk of certification failure due to overly aggressive declaration value settings, and improving the rigor of the test results and the certification pass rate.

[0183] This application's embodiments are aimed at... Figure 1 The vehicle testing method shown in step S16 further determines the final declared value of the lateral acceleration control capability of the entire autonomous vehicle. A global minimum value selection logic is introduced to solve the vehicle certification compliance problem caused by inconsistent declared values ​​in multiple speed ranges.

[0184] First, the advanced steering control system of autonomous vehicles exhibits varying performance at different speeds. For instance, it may possess high lateral acceleration holding capability at low speeds, while at high speeds, its holding capability may decrease due to limitations in tire grip or system response delays. Consequently, the declared lateral acceleration values ​​calculated for different speed ranges often differ. Directly adopting the declared values ​​for each operating condition could lead to excessively high values ​​in some conditions, exceeding the system's actual capabilities, or excessively low values ​​in others, limiting vehicle performance.

[0185] By aggregating the reported lateral acceleration values ​​for all test conditions and selecting the minimum value as the final reported value, it can be ensured that the selected final reported value does not exceed the upper limit of the lateral acceleration that the system can actually maintain stably under all test conditions. This design is based on the most conservative principle, that is, the overall capability of the system is determined by its weakest link.

[0186] In a practical implementation, the test control box may include storing the calculated lateral acceleration declaration values ​​for five vehicle speed ranges (such as 0-60km / h, 60-100km / h, etc.) into a data storage module. The main control unit uses a comparison algorithm to identify the declaration value with the smallest value and marks it as the final lateral acceleration control capability declaration value. For example, if the declaration value for the 60-100km / h range is 2.5m / s², and the declaration value for the 100-130km / h range is 2.2m / s², then the final declaration value is determined to be 2.2m / s².

[0187] Another embodiment involves corrections for different curve radii. Although the test sequence includes multiple radii, the declared value is usually based on a specific baseline radius or the average effect. If there are differences in the declared values ​​obtained under different radii, the minimum value selection logic is also applied to cover the worst-case scenario.

[0188] Furthermore, the test control box of this application can also verify whether the above declared values ​​meet all of the following conditions: the normal over-limit amplitude a_y_smax+0.3 does not exceed 3.0 m / s²; the short-term over-limit amplitude min(a_y_smax×1.4, 3.3) does not exceed 3.3 m / s² and the duration does not exceed 2.0 seconds; the functional test range from 80%a_y_smax to 90%a_y_smax is completely within the maintenance area of ​​all test conditions in the vehicle speed range; the outermost point of the vehicle must not cross the outer edge of the lane line within 4 seconds after the short-term over-limit is triggered.

[0189] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0190] To implement the above vehicle testing method, this application also proposes a test control box, for details please refer to [link / reference needed]. Figure 12 , Figure 12 This is a schematic diagram of an embodiment of the test control box provided in this application.

[0191] The test control box 500 in this embodiment includes: The test case management module 51 is used to generate test sequences for autonomous vehicles.

[0192] The main control unit 52 is used to drive the unmanned vehicle to execute each test condition in sequence according to the test sequence.

[0193] The signal receiving module 53 is used to acquire vehicle status data during the execution of various test conditions by the unmanned vehicle.

[0194] The data processing engine 54 is used to determine the hold-up sample set and the deviation sample set for each test condition in the vehicle status data; to determine the upper bound of the confidence interval for each test condition based on the hold-up sample set and the deviation sample set; and to determine the declared lateral acceleration value for each test condition based on the upper bound of the confidence interval for each test condition.

[0195] The test control box 500 provided in this application is a combination of dedicated hardware and functional modules for field certification testing of unmanned vehicles. Its protection boundary covers the complete data processing chain from test sequence generation, vehicle drive control, multi-source data acquisition, statistical sample classification to final declaration value calculation.

[0196] Specifically, the test case management module 51 is used to generate test sequences. This module automatically generates test cases containing different combinations of vehicle speed and radius by pre-loading parameters such as the curve radius list, vehicle speed range configuration, and target lateral acceleration step value, using a combination matrix. It can automatically calculate the target lateral acceleration for each condition based on the formula a_y=V² / (3.6²×R), where V is the target vehicle speed and R is the curve radius. For example, in the 60-100km / h speed range, for a 250m radius curve, test cases with a target vehicle speed of 90km / h and a target lateral acceleration of 2.50m / s² can be generated, and the number of repetitions can be set.

[0197] The main control unit 52 is used to drive the vehicle to perform tests. As the core controller, the main control unit coordinates the work of each module and outputs control commands to the driving robot or vehicle control system. During test execution, the main control unit controls the vehicle to accelerate to the target speed on the straight before entering the curve, and applies pre-steering control parameters when entering the curve, such as a pre-aiming distance of 15m and a lateral gain increase of 20%, to ensure that the vehicle enters the set test conditions.

[0198] The signal receiving module 53 is used to acquire vehicle status data. This module includes at least a GPS / GNSS signal receiving module and a CAN bus communication module. The GPS / GNSS module receives position, speed, and timestamp information via an external antenna at a frequency of at least 100Hz, supporting RTK differential positioning to improve accuracy. The CAN bus communication module reads inertial measurement unit data, ACSF system status signals, and vehicle dynamics signals, such as lateral acceleration, yaw rate, and steering wheel angle, via the vehicle's OBD port at a frequency of at least 100Hz. Both data streams are accompanied by high-precision timestamps for subsequent time alignment.

[0199] The data processing engine 54 is used to determine the hold-up and out-of-line sample sets. The engine first performs time alignment on the dual-source data, using the GNSS timestamp as a reference, requiring a frame drop rate of less than 1%. Then, a fourth-order Butterworth low-pass filter with a cutoff frequency of 0.5Hz is applied to the lateral acceleration signal. Steady-state window data is extracted starting 2 seconds after cornering, requiring the standard deviation of the filtered lateral acceleration within the steady-state window to be less than 0.1 m / s² and the vehicle speed fluctuation to not exceed ±1 km / h, with a duration of at least 5 seconds. Within the steady-state window, each data frame undergoes a multi-condition joint judgment: the distance d from the outermost point of the vehicle to the lane line is calculated; if d ≥ 0.3 m, the distance condition is met; the ACSF system status is checked to see if it is Active; the 0.5-second moving average lateral acceleration is calculated; if its absolute value ≤ 5 m / s³, the acceleration condition is met; and the standard deviation of the filtered lateral acceleration within a 1-second time window is calculated; if it is less than 0.1 m / s², the stability condition is met. When all four conditions are met, the state is marked as "maintained"; when any one condition is not met, the state is marked as "deviation". The results of all repeated tests under the same test condition are summarized, and the mean lateral acceleration values ​​of the steady-state windows marked as "maintained" are compiled into a "maintained" sample set, and the values ​​of those marked as "deviation" are compiled into a "deviation" sample set.

[0200] The data processing engine 54 determines the upper bound of the confidence interval for each test condition based on the hold-up sample set and the out-of-range sample set. Specifically, it calculates the 95th percentile of the hold-up sample and the 5th percentile of the out-of-range sample, and takes the average of the two as the best estimate of the threshold. It calculates the standard deviation of the hold-up sample, and then calculates the 95% confidence interval for the best estimate of the threshold; the upper limit of this interval is the upper bound of the confidence interval. The width of the gray zone, i.e., the difference between the maximum value of the hold-up sample and the minimum value of the out-of-range sample, can be calculated to assess the uncertainty of the decision.

[0201] The data processing engine 54 determines the declared lateral acceleration value for each test condition based on the upper bound of the confidence interval for each test condition. The declared value is calculated by subtracting 0.3 m / s² from the upper bound of the confidence interval. This design ensures that the deviation state is triggered under normal over-limit conditions, while verifying the stability of the functional test interval through margin evaluation.

[0202] Please continue reading. Figure 13 , Figure 13 This is a schematic diagram of another embodiment of the test control box provided in this application.

[0203] like Figure 13 As shown, the test control box provided in this application embodiment includes a housing, a main control unit, a GPS / GNSS signal receiving module, a CAN bus communication module, a test case management module, a data processing engine, a report generation module, and a data storage module.

[0204] The housing is a portable, sealed housing, equipped with a GPS / GNSS antenna interface, a CAN bus interface, an Ethernet interface, and a USB (Universal Serial Bus) data export interface for connecting external devices and transmitting data.

[0205] The main control unit, housed within the enclosure, runs the test control program and coordinates the operation of each module. It employs an embedded processor with sufficient computing power to run the data processing engine and report generation module.

[0206] The GPS / GNSS signal receiving module is located inside the housing and communicates with the main control unit. It connects to an external GNSS antenna through the GPS / GNSS antenna interface and supports RTK (Real-Time Kinematic) differential positioning. It is used to receive vehicle location information, speed information and timestamp information at a sampling frequency of not less than 100Hz.

[0207] The CAN bus communication module is housed within the housing and communicates with the main control unit. It connects to the vehicle's OBD (On-Board Diagnostics) port or CAN bus diagnostic interface via a CAN bus interface to read inertial measurement unit (IMU) data, ACSF system status signals, and vehicle dynamics signals at a sampling frequency of at least 100Hz. The signals read by the CAN bus communication module include at least: longitudinal acceleration, lateral acceleration (Ay), yaw rate, ACSF system activation status, ACSF system override status, ACSF system fail-safe status, steering wheel angle, steering wheel torque, and vehicle speed.

[0208] The test case management module is deployed in the main control unit and is used to preload test parameters and automatically generate test sequences. The test parameters preloaded by the test case management module include at least: a list of curve radii, containing six standard radius values: 100m, 125m, 200m, 250m, 500m, and 600m; five speed range configurations: 0-60km / h, 60-100km / h, 100-130km / h, 130-150km / h, and above 150km / h; a target lateral acceleration step value of 0.1m / s²; and the regulatory limits for a_y_smax in each speed range according to ECE R79. The preset values ​​in Table 1 of Rev.5 are as follows: 0.30 m / s² for speeds 10-60 km / h, 0.5-3.0 m / s² for speeds greater than 60-100 km / h, 0.8-3.0 m / s² for speeds greater than 100-130 km / h, 0.8-3.0 m / s² for speeds greater than 130-150 km / h, and 0.3-3.0 m / s² for speeds greater than 150 km / h. The test case management module automatically generates test sequences based on the combination matrix of speed ranges and curve radii, and automatically calculates the target lateral acceleration for each test condition according to the formula a_y=V² / (3.6²×R), where V is the target speed (in km / h), R is the curve radius (in m), and a_y is the target lateral acceleration (in m / s²).

[0209] The data processing engine is deployed in the main control unit and communicates with the GPS / GNSS signal receiving module and the CAN bus communication module. It is used to receive dual-source data and perform time alignment, filtering, hold / off dual-state determination, threshold probability distribution modeling, and optimized calculation of the declared value. The data processing engine includes the following functional units: a data alignment unit, which performs time alignment of GPS / GNSS data and CAN bus data based on the GNSS timestamp and checks the frame loss rate, requiring the frame loss rate to be less than 1%; a filtering unit, which applies a fourth-order Butterworth low-pass filter with a cutoff frequency of 0.5Hz to the lateral acceleration signal and yaw rate signal; a steady-state segment extraction unit, which extracts steady-state window data starting from the 2nd second after entering the curve, requiring the standard deviation of the filtered lateral acceleration within the steady-state window to be less than 0.1m / s² and the vehicle speed fluctuation to not exceed ±1km / h, with the steady-state window duration not less than 5 seconds; a hold / deviation dual-state determination unit, which performs multi-condition joint determination on each data frame within the steady-state window; a threshold probability distribution modeling unit, which establishes a threshold probability distribution model based on hold samples and deviation samples; and a declaration value optimization calculation unit, which automatically calculates the optimal declaration value based on the threshold statistical results.

[0210] The report generation module is deployed in the main control unit and communicates with the data processing engine to generate standardized test reports. The report generated by this module includes the following parts: basic test information, a threshold statistics table summarized by vehicle speed range, the best estimated value and 95% confidence interval for each speed range, gray zone width assessment, calculation process and final recommended value for the declared value, regulatory compliance verification conclusion table, margin assessment results, test condition coverage list, and abnormal condition records. The report supports one-click export to PDF (Portable Document Format) format or export via USB data export interface to a CSV (Comma-Separated Values) raw data file.

[0211] The data storage module is located inside the housing and communicates with the main control unit to store test configuration parameters, raw acquisition data, and processing result data.

[0212] To implement the above vehicle testing method, this application also proposes a vehicle testing device, for details please refer to [link / reference needed]. Figure 14 , Figure 14 This is a schematic diagram of an embodiment of the vehicle testing device provided in this application.

[0213] The vehicle testing device 400 in this embodiment includes a processor 41, a memory 42, an input / output device 43, and a bus 44.

[0214] The processor 41, memory 42, and input / output device 43 are respectively connected to the bus 44. The memory 42 stores program data, and the processor 41 is used to execute the program data to implement the vehicle testing method described in the above embodiments.

[0215] In this embodiment, processor 41 can also be referred to as a CPU (Central Processing Unit). Processor 41 may be an integrated circuit chip with signal processing capabilities. Processor 41 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor, or processor 41 can be any conventional processor.

[0216] This application also provides a computer storage medium; please refer to the following: Figure 15 , Figure 15 This is a schematic diagram of a computer storage medium according to an embodiment of the present application. The computer storage medium 600 stores a computer program 61, which, when executed by a processor, is used to implement the vehicle testing method of the above embodiment.

[0217] When the embodiments of this application are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0218] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A vehicle testing method, characterized in that, The vehicle testing method is applied to a test control box, which is used for the automated control, data acquisition, and test evaluation of unmanned vehicles. The vehicle testing method includes: Generate a test sequence for the autonomous vehicle; The driverless vehicle is driven to perform each test condition sequentially according to the test sequence. Acquire vehicle status data during the execution of each test condition by the unmanned vehicle; Determine the hold-up sample set and the off sample set for each test condition in the vehicle state data; The upper bound of the confidence interval for each test condition is determined based on the maintained sample set and the deviated sample set. The lateral acceleration declaration value for each test condition is determined based on the upper bound of the confidence interval for each test condition.

2. The vehicle testing method according to claim 1, characterized in that, The generation of test sequences for the autonomous vehicle includes: Obtain the test configuration file, which includes at least one of the following fields: a list of corner radii, a vehicle speed range configuration, a target lateral acceleration step value, regulatory limits for each vehicle speed range, the number of test repetitions, a test direction configuration, and corner entry control parameters; The test configuration file is analyzed to determine the target lateral acceleration calculated for each curve radius and each target vehicle speed; Test instances are generated by combining test fields whose target lateral acceleration is less than or equal to the maximum limit value of the speed range in which the target vehicle speed is located. All test cases are combined into a test sequence for the autonomous vehicle.

3. The vehicle testing method according to claim 1, characterized in that, The acquisition of vehicle status data during the execution of various test conditions by the autonomous vehicle includes: The location data and vehicle operation data of the unmanned vehicle during the execution of each test condition are acquired, wherein the vehicle operation data includes at least a lateral acceleration signal; Using the timestamp of the location data as a reference time axis, the vehicle operation data is aligned with the location data according to the timestamp to form the vehicle status data.

4. The vehicle testing method according to claim 3, characterized in that, After aligning the vehicle operation data with the location data according to the timestamp, the vehicle testing method further includes: The theoretical number of data frames for the location data is determined based on the reference time axis. Determine the number of valid data frames in the aligned vehicle operation data that correspond to the data frame timestamps of the location data; The frame drop rate of the test condition is determined based on the theoretical number of data frames and the effective number of data frames. When the frame loss rate exceeds a preset threshold, the autonomous vehicle is re-driven to perform the test condition and data is collected again.

5. The vehicle testing method according to claim 3, characterized in that, After aligning the vehicle operation data with the location data according to the timestamp, the vehicle testing method further includes: The aligned lateral acceleration and yaw rate signals are processed using digital low-pass filtering. The filtering method used is zero-phase filtering.

6. The vehicle testing method according to claim 1, characterized in that, After acquiring the vehicle status data during the execution of each test condition by the autonomous vehicle, the vehicle testing method further includes: Steady-state window data is extracted from several data frames after the vehicle enters a preset state in the vehicle state data. Perform multi-condition joint determination on each data frame in the steady-state window data; Data frames that meet the judgment criteria are marked as held, and data frames that do not meet the judgment criteria are marked as deviated. The set of held samples is determined based on all data frames marked as held states; The set of biased samples is determined based on all data frames marked as biased.

7. The vehicle testing method according to claim 6, characterized in that, Extracting steady-state window data from several data frames after the vehicle enters a preset state in the vehicle state data includes: Obtain candidate window data after the vehicle enters a preset state from the vehicle status data; If the candidate window data simultaneously meets the following conditions, then the candidate window data is determined as steady-state window data: The standard deviation of lateral acceleration in all data frames of the candidate window data is less than a first preset difference. The maximum difference in vehicle speed among all data frames of the candidate window data is less than or equal to the second preset difference. The duration of the candidate window data is greater than or equal to a preset duration.

8. The vehicle testing method according to claim 7, characterized in that, The determination criteria used in the multi-condition joint determination include at least one of the following: lane distance condition, system state condition, lateral acceleration condition, and lateral acceleration stability condition.

9. The vehicle testing method according to claim 6 or 8, characterized in that, The step of performing a multi-condition joint determination on each data frame in the steady-state window data includes: Based on the position information in the data frame, determine the lateral offset of the vehicle center relative to the lane center; The distance from the outermost point of the vehicle to the lane line is determined based on the lateral offset, the lane width of the test site, and the vehicle width. When the distance between the outermost point of the vehicle and the lane line is greater than a preset distance, the data frame is determined to meet condition one; When the system status signal of the data frame is an active and control signal, it is determined that the data frame satisfies condition two. When the average lateral acceleration within the first preset window where the data frame is located is less than or equal to a preset average threshold, the data frame is determined to satisfy condition three. When the standard deviation of the lateral acceleration within the second preset window where the data frame is located is less than the preset standard deviation threshold, the data frame is determined to meet condition four. Data frames that meet all the above conditions are marked as held, and other data frames are marked as off-target.

10. The vehicle testing method according to claim 1, characterized in that, The step of determining the upper bound of the confidence interval for each test condition based on the maintained sample set and the deviated sample set includes: The optimal estimate of the threshold is calculated based on the maintained sample set and the deviated sample set; Calculate the standard deviation of the lateral acceleration of the maintained sample set; Calculate the standard error of the best estimate of the threshold based on the standard deviation of the lateral acceleration and the number of samples in the hold-up sample set; The upper bound of the confidence interval for each test condition is determined based on the standard error and the optimal estimate of the threshold.

11. The vehicle testing method according to claim 1, characterized in that, After determining the hold-up sample set and the deviation sample set for each test condition in the vehicle state data, the vehicle testing method further includes: The width of the gray zone is determined based on the maximum lateral acceleration in the maintained sample set and the minimum lateral acceleration in the deviated sample set; When the width of the gray area is greater than a preset value, it is determined that there is a ambiguity area in the test condition, and a warning message is output.

12. The vehicle testing method according to claim 1, characterized in that, After determining the declared lateral acceleration values ​​for each test condition, the vehicle testing method further includes: Obtain the maximum declared value of lateral acceleration for all test conditions; Based on the maximum declared value of lateral acceleration and the declared value of lateral acceleration for each test condition, calculate the functional test margin for each test condition. When the functional test margin is less than or equal to the preset value, a warning signal indicating insufficient functional test margin is output, and the declared value of the test condition is marked as unavailable.

13. The vehicle testing method according to claim 1 or 12, characterized in that, After determining the declared lateral acceleration values ​​for each test condition, the vehicle testing method further includes: Calculate the over-limit trigger margin based on the declared lateral acceleration value, the preset over-limit amplitude value, and the best estimated threshold value for each test condition. When the over-limit trigger margin is less than or equal to a preset value, an over-limit trigger margin insufficient warning signal is output, and the declared value of the test condition is marked as unavailable.

14. The vehicle testing method according to claim 1, characterized in that, After determining the declared lateral acceleration values ​​for each test condition, the vehicle testing method further includes: The minimum value among all the declared lateral acceleration values ​​for all test conditions is determined as the final declared value of the autonomous vehicle's lateral acceleration control capability.

15. A test control box, characterized in that, The test control box includes: The test case management module is used to generate test sequences for autonomous vehicles; The main control unit is used to drive the unmanned vehicle to execute each test condition sequentially according to the test sequence; The signal receiving module is used to acquire vehicle status data during the execution of various test conditions by the unmanned vehicle; A data processing engine is used to determine the hold-up sample set and the deviation sample set for each test condition in the vehicle state data; determine the upper bound of the confidence interval for each test condition based on the hold-up sample set and the deviation sample set; and determine the declared lateral acceleration value for each test condition based on the upper bound of the confidence interval for each test condition.

16. A vehicle testing device, characterized in that, The vehicle testing device includes a memory and a processor coupled to the memory; The memory is used to store program data, and the processor is used to execute the program data to implement the vehicle testing method as described in any one of claims 1 to 14.

17. A computer storage medium, characterized in that, The computer storage medium is used to store program data, which, when executed by the computer, is used to implement the vehicle testing method as described in any one of claims 1 to 14.

Images