Vehicle positioning track evaluation system and method and machine readable storage medium
Through the vehicle positioning trajectory evaluation system, the positioning trajectory of autonomous vehicles is calculated and evaluated using high-precision sensors and true-value equipment data, and the problem of fusion positioning and map matching evaluation is solved, and the effective evaluation of the performance of the positioning algorithm module and system performance adjustment is achieved. excellent.
Patent Information
- Application Number
- CN202310592328.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-24
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to accurately evaluate the fusion positioning of autonomous driving vehicles and the relative positioning of map matching results, and it is difficult to effectively evaluate the performance of each positioning algorithm module, affecting the performance tuning and road measurement efficiency of the system.
Provide a vehicle positioning trajectory evaluation system, including information acquisition vehicle and cloud data platform, use high-precision sensor data and true-value equipment data to calculate the comparison and evaluation of target positioning trajectory and high-precision reference trajectory, and generate it through segmentation and translation rotation matrix conversion. Evaluate the results and compare the algorithm performance using CEP68 and CEP95 indicators.
Accurate evaluation of relative positioning of fusion positioning and performance evaluation of various functional modules is achieved, evaluation efficiency and accuracy are improved, and performance tuning and algorithm optimization can be better guided.
Smart Images

Figure CN119027787A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of autonomous driving, and in particular to a system, method and machine-readable storage medium for evaluating vehicle positioning trajectory. Background Art
[0002] In the autonomous driving system, the vehicle positioning system is particularly critical, which affects the performance of almost all downstream links. The development of high-precision positioning technology has extended to various types of solutions. Fusion positioning technology combines the advantages of various solutions and can still output stable and reliable positioning information under complex working conditions through reasonable strategy adjustment.
[0003] Usually, the positioning evaluation scheme will use high-precision positioning reference equipment as the basis for the true value, but high-precision absolute positioning as the basis cannot accurately determine whether the relative positioning after map matching can meet the application requirements. Another optional method is the evaluation method based on the known vehicle driving trajectory, but this method requires the vehicle to drive along the center line of the lane to know the relative position relationship between the vehicle and the map in advance. This method is difficult to operate and is difficult to implement due to road conditions, and it is impossible to conduct large-scale road tests.
[0004] On the other hand, fusion positioning is based on the fusion of multiple sensor positioning technologies. Therefore, in addition to the final performance verification, it is also necessary to accurately evaluate the performance of each positioning algorithm submodule in order to correctly guide the performance tuning of each module itself. Therefore, the evaluation of fusion positioning is not only the evaluation of the final result, but also should include the evaluation of multiple process data, and present it in the form of a complete set of positioning evaluation products. Summary of the invention
[0005] An object of the present invention is to meet the relative positioning evaluation requirements of fused positioning and map matching results.
[0006] A further object of the present invention is to perform performance evaluation on each functional module of fusion positioning.
[0007] A further object of the present invention is to improve the efficiency of evaluation and acceptance.
[0008] In particular, the present invention provides a vehicle positioning trajectory evaluation system, which includes:
[0009] The information collection vehicle includes at least one or more sensors selected from the group consisting of a camera, a navigation and positioning module, and a radar sensor, and also includes at least one or more true value devices selected from the group consisting of a high-precision inertial navigation system and a high-precision radar, and is capable of uploading data of the at least one or more sensors and the true value device data, wherein the true value device is a high-precision positioning reference device with higher measurement accuracy than ordinary sensors;
[0010] Cloud data platform, which includes:
[0011] A data interface configured to receive sensor data and true value device data uploaded by the information collection vehicle, wherein the sensor data includes one or more of the following data: vehicle speed, IMU data, RTK data, camera data, and high-precision map data, and the true value device data includes high-precision inertial navigation data;
[0012] An algorithm playback module, configured to calculate the sensor data to obtain the target positioning trajectory;
[0013] The truth value preprocessing module is configured to solve the truth value device data to obtain a high-precision reference trajectory; the high-precision reference trajectory is converted by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory truth value corresponding to the target positioning trajectory;
[0014] The evaluation calculation module is configured to compare and evaluate the target positioning trajectory with the true value of the high-precision reference trajectory to obtain an evaluation result.
[0015] Optionally, the target positioning trajectory includes a dead reckoning trajectory;
[0016] The truth value preprocessing module is further configured to: segment the high-precision reference trajectory according to a first preset length to obtain a high-precision reference trajectory segment; obtain a dead reckoning trajectory segment corresponding to the high-precision reference trajectory segment; align the high-precision reference trajectory segment with the dead reckoning trajectory segment until feature points overlap and calculate a translation rotation matrix, wherein the feature points include a starting point or an end point in the high-precision reference trajectory segment and the dead reckoning trajectory segment; calculate the high-precision reference trajectory segment and the translation rotation matrix to obtain a high-precision reference trajectory truth value corresponding to the dead reckoning trajectory.
[0017] Optionally, the target localization track includes a semantic localization track;
[0018] The truth value preprocessing module is further configured to: segment the high-precision reference trajectory according to a second preset length to obtain a high-precision reference trajectory segment; obtain a semantic positioning trajectory segment corresponding to the high-precision reference trajectory segment; align the high-precision reference trajectory segment with the semantic positioning trajectory segment through a preset algorithm, the preset algorithm includes a least squares method; calculate a translation rotation matrix; convert the high-precision reference trajectory segment through the translation rotation matrix to obtain a high-precision reference trajectory truth value corresponding to the semantic positioning trajectory segment; and / or
[0019] The target positioning trajectory includes a fused positioning trajectory;
[0020] The truth value preprocessing module is also configured to: segment the high-precision reference trajectory according to a third preset length to obtain high-precision reference trajectory segments; obtain the fused positioning trajectory segments corresponding to the high-precision reference trajectory segments; align the high-precision reference trajectory segments with the fused positioning trajectory segments through a preset algorithm, the preset algorithm including a least squares method; calculate a translation and rotation matrix; and convert the high-precision reference trajectory segments through the translation and rotation matrix to obtain the high-precision reference trajectory truth value corresponding to the fused positioning trajectory segments.
[0021] Optionally, the truth value preprocessing module is further configured to:
[0022] Obtain the standard deviation value of the high-precision reference trajectory in horizontal positioning;
[0023] Determine whether the standard deviation value is less than a preset threshold;
[0024] If it is greater than the preset threshold, the high-precision reference trajectory of the section where the standard deviation value is located will be eliminated.
[0025] Optionally, the cloud data platform also includes a result display module;
[0026] The result display module is configured to generate different types of renderings according to the evaluation results, and the renderings are used to display the error change process between the target positioning trajectory and the true value of the high-precision reference trajectory; the positioning deviation CEP68 and / or CEP95 indicators are calculated according to the evaluation results, so as to use the positioning deviation CEP68 and / or CEP95 indicators to compare the performance of the algorithms.
[0027] According to another aspect of the present invention, a vehicle positioning trajectory evaluation method is also provided, which includes:
[0028] Acquire sensor data and true value device data. The accuracy of true value device data is higher than that of sensor data. Sensor data includes one or more of the following data: vehicle speed, IMU data, RTK data, camera data, and high-precision map data. True value device data includes high-precision inertial navigation data.
[0029] Calculate the sensor data to obtain the target positioning trajectory;
[0030] Solve the true value equipment data to obtain a high-precision reference trajectory;
[0031] The high-precision reference trajectory is converted by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a true value of the high-precision reference trajectory corresponding to the target positioning trajectory;
[0032] The target positioning trajectory is compared and evaluated with the true value of the high-precision reference trajectory to obtain the evaluation result.
[0033] Optionally, the target positioning trajectory includes a dead reckoning trajectory;
[0034] The step of converting the high-precision reference trajectory by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory includes: cutting the high-precision reference trajectory according to a first preset length to obtain a high-precision reference trajectory segment; obtaining a dead reckoning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the dead reckoning trajectory segment until feature points overlap and calculating a translation rotation matrix, wherein the feature points include a starting point or an end point in the high-precision reference trajectory segment and the dead reckoning trajectory segment; and calculating the high-precision reference trajectory segment and the translation rotation matrix to obtain a high-precision reference trajectory true value corresponding to the dead reckoning trajectory.
[0035] Optionally, the target localization track includes a semantic localization track;
[0036] The step of converting the high-precision reference trajectory by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory includes: cutting the high-precision reference trajectory according to a second preset length to obtain a high-precision reference trajectory segment; obtaining a semantic positioning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the semantic positioning trajectory segment through a preset algorithm, the preset algorithm including a least squares method; calculating a translation rotation matrix; converting the high-precision reference trajectory segment through the translation rotation matrix to obtain a high-precision reference trajectory true value corresponding to the semantic positioning trajectory segment; and / or
[0037] The target positioning track also includes a fusion positioning track;
[0038] The step of converting the high-precision reference trajectory by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory includes: cutting the high-precision reference trajectory according to a third preset length to obtain high-precision reference trajectory segments; obtaining a fused positioning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment and the fused positioning trajectory segment through a preset algorithm, the preset algorithm including a least squares method; calculating a translation rotation matrix; and converting the high-precision reference trajectory segment through the translation rotation matrix to obtain a high-precision reference trajectory true value corresponding to the fused positioning trajectory segment.
[0039] Optionally, after the step of solving the true value device data to obtain a high-precision reference trajectory, the following steps are included:
[0040] Obtain the standard deviation value of the high-precision reference trajectory in horizontal positioning;
[0041] Determine whether the standard deviation value is less than a preset threshold;
[0042] If it is greater than the preset threshold, the high-precision reference trajectory of the section where the standard deviation value is located will be eliminated.
[0043] Optionally, the step of obtaining the evaluation result includes:
[0044] Generate different types of effect diagrams based on the evaluation results. The effect diagrams are used to show the error change process between the target positioning trajectory and the true value of the high-precision reference trajectory.
[0045] The positioning deviation CEP68 and / or CEP95 indicators are calculated according to the evaluation results, so as to compare the performance of the target algorithm using the positioning deviation CEP68 and / or CEP95 indicators.
[0046] According to another aspect of the present invention, a machine-readable storage medium is provided, on which a machine-executable program is stored. When the machine-executable program is executed by a processor, any of the above-mentioned vehicle positioning trajectory evaluation methods is implemented.
[0047] In the scheme of the present invention, after receiving the sensor data and true value device data uploaded by the information collection vehicle, the sensor data is calculated to obtain the target positioning trajectory; the true value device data is solved to obtain the high-precision parameter trajectory; the high-precision reference trajectory is converted by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain the high-precision reference trajectory true value corresponding to the target positioning trajectory, thereby meeting the relative positioning evaluation requirements of matching the target positioning trajectory with the high-precision reference trajectory on the same map, and the performance of different functional modules is evaluated according to the type of the target positioning trajectory.
[0048] Furthermore, in the solution of the present invention, after obtaining the evaluation results, different types of effect diagrams are generated according to the evaluation results, and the effect diagrams are used to show the error change process between the target positioning trajectory and the true value of the high-precision reference trajectory; the positioning deviation CEP68 and / or CEP95 indicators are calculated according to the evaluation results, so as to compare the performance of the algorithm using the positioning deviation CEP68 and / or CEP95 indicators. Through this method, the efficiency of evaluation and acceptance can be improved.
[0049] Based on the following detailed description of specific embodiments of the present invention in conjunction with the accompanying drawings, those skilled in the art will become more aware of the above and other objects, advantages and features of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Hereinafter, some specific embodiments of the present invention will be described in detail in an exemplary and non-limiting manner with reference to the accompanying drawings. The same reference numerals in the accompanying drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. In the accompanying drawings:
[0051] Figure 1 It is a schematic diagram of a vehicle positioning trajectory evaluation system;
[0052] Figure 2 yes Figure 1Schematic diagram of the architecture of the vehicle positioning trajectory evaluation system;
[0053] Figure 3 is a schematic diagram of the architecture of a vehicle positioning trajectory evaluation system according to an embodiment of the present invention;
[0054] Figure 4 is a schematic diagram of a cloud data platform architecture of a vehicle positioning trajectory evaluation system according to an embodiment of the present invention;
[0055] Figure 5 is a schematic diagram of the positions of a fused positioning trajectory and a true value of the fused positioning trajectory of the same segment according to a vehicle positioning trajectory evaluation method according to an embodiment of the present invention;
[0056] Figure 6 The vehicle positioning trajectory evaluation method according to an embodiment of the present invention is a method for evaluating the vehicle positioning trajectory. Figure 5 Schematic diagram of the position after the fused positioning trajectory and the true value of the fused positioning trajectory;
[0057] Figure 7 is a schematic diagram of the positions of a combined navigation trajectory and a true value of the combined navigation trajectory of the same segment according to a vehicle positioning trajectory evaluation method according to an embodiment of the present invention;
[0058] Figure 8 The vehicle positioning trajectory evaluation method according to an embodiment of the present invention is a method for evaluating the vehicle positioning trajectory. Figure 7 Schematic diagram of the position of the combined navigation trajectory and the true value of the combined navigation trajectory of the same segment;
[0059] Fig. 9 is a flow chart of a vehicle positioning trajectory evaluation method according to an embodiment of the present invention;
[0060] Fig.10 is a schematic diagram of a machine-readable storage medium in a method for evaluating vehicle positioning trajectory according to an embodiment of the present invention; and
[0061] Fig.11 It is a schematic diagram of a computer device in a method for evaluating vehicle positioning trajectory according to an embodiment of the present invention. DETAILED DESCRIPTION
[0062] It should be understood by those skilled in the art that the embodiments described below are only some embodiments of the present invention, rather than all embodiments of the present invention, and these embodiments are intended to explain the technical principles of the present invention, rather than to limit the protection scope of the present invention. Based on the embodiments provided by the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should still fall within the protection scope of the present invention.
[0063] Figure 1The schematic diagram of a vehicle positioning trajectory evaluation system is shown in FIG. The vehicle positioning trajectory evaluation system may include an information collection vehicle 110 and a cloud data platform 120 .
[0064] The information collection vehicle 110 may be a vehicle equipped with a variety of sensors, capable of uploading one or more vehicle data such as vehicle speed, IMU (Inertial Measurement Unit), RTK (Real-Time Kinematic), camera, and high-precision map to the cloud data platform 120. In some embodiments, the information collection vehicle 110 may be an autonomous driving vehicle, which sends the above vehicle data to the cloud data platform 120 for processing through sensors; in other embodiments, the information collection vehicle 110 may be a professional collection vehicle or a crowdsourcing collection vehicle, which is collected and produced by professional surveying and mapping personnel through surveying and mapping methods, and has geographic information data such as lane-level road topology and element information; and crowdsourcing collection vehicles include private cars, buses, etc., which can use on-board sensors to monitor environmental changes in real time and upload vehicle data to the cloud data platform 120, thereby realizing the upload of vehicle data comprehensively and quickly.
[0065] The cloud data platform 120 can read the vehicle data uploaded by the information collection vehicle 110 and perform calculations according to a preset algorithm, thereby calculating the vehicle trajectory result and realizing the positioning of the vehicle.
[0066] Figure 2 yes Figure 1 Schematic diagram of the vehicle positioning trajectory evaluation system. The information collection vehicle 110 is configured to upload at least part of the collected vehicle data to the cloud data platform 120 through the sensor 130, and the collected vehicle data includes one or more sensor data such as vehicle speed, IMU, RTK, camera, and high-precision map.
[0067] The cloud data platform 120 includes a calculation module 140, and the calculation module 140 is configured to calculate the vehicle data uploaded by the information collection vehicle 110 through a preset algorithm, thereby obtaining a fused positioning result and outputting positioning information.
[0068] In the process of collecting and producing high-precision maps, or in the process of converting coordinate frames, it is inevitable that certain fixed deviations will occur. Therefore, the industry's demand for fusion positioning has changed from simply pursuing absolute positioning accuracy to requiring accurate and reliable relative positioning after map matching. Therefore, a vehicle positioning trajectory evaluation system and method are proposed.
[0069] Figure 31 is a schematic diagram of the architecture of a vehicle positioning trajectory evaluation system according to an embodiment of the present invention. This schematic diagram includes: an information collection vehicle 110 and a cloud data platform 120, wherein the information collection vehicle 110 can be a vehicle equipped with multiple sensors, capable of uploading one or more vehicle data such as vehicle speed, wheel speed, IMU (Inertial Measurement Unit), RTK (Real-Time Kinematic, real-time dynamic differential positioning), camera and high-precision map to the cloud data platform 120. In some embodiments, the information collection vehicle 110 can be an autonomous driving vehicle, which sends the above vehicle data to the cloud data platform 120 for processing through the sensor 130; in other embodiments, the information collection vehicle 110 can be a professional collection vehicle or a crowdsourcing collection vehicle, which is collected and produced by professional surveying and mapping personnel through surveying and mapping methods. Geographic information data such as lane-level road topology and element information; and crowdsourcing collection vehicles include private cars, buses, etc., which can use on-board sensors to monitor environmental changes in real time and upload vehicle data to the cloud data platform 120, thereby fully and quickly realizing the upload of vehicle data.
[0070] The information collection vehicle 110 also includes a true value device 131, which is a high-precision positioning reference device with higher measurement accuracy than ordinary sensors, such as millimeter-wave radar, laser radar, high-precision combined inertial navigation, and high-precision wheel speed meter. Through the true value device 131, high-precision true value device data can be obtained as reference data and uploaded to the cloud data platform 120.
[0071] In some embodiments, an optional device for the true value device is a high-precision inertial navigation device, which has higher accuracy than ordinary sensors.
[0072] The cloud data platform 120 includes an algorithm playback module 141, a truth preprocessing module 142, an evaluation calculation module 143, and a result display module 144. The algorithm playback module 141 is configured to cluster the sensor data uploaded by the information collection vehicle 110 according to the scene when the information collection vehicle 110 collects data and create a corresponding data set, then select the required data set and the algorithm version code required by the instruction according to the instruction, and perform environment deployment and code compilation on the cloud data platform, and finally calculate and process the data set according to the algorithm required by the instruction to obtain the processing result. In some embodiments, the vehicle speed, wheel speed and IMU data in the sensor data are calculated according to the algorithm required by the instruction to obtain a dead reckoning trajectory, and the dead reckoning trajectory is corrected according to the GNSS data parsed from the true value device data to obtain a combined navigation trajectory; the map geometry layer data in the high-precision map data in the vehicle data is loaded as the initialization position according to the GNSS data parsed from the true value device data, and the map geometry layer data is compared with the semantic elements obtained by visual perception to obtain a semantic positioning trajectory; the above-mentioned combined navigation trajectory is fused with the semantic positioning trajectory using Kalman filtering, and matched again with the roads and lanes in the high-precision map to finally obtain a fused positioning trajectory.
[0073] The truth value preprocessing module 142 is configured to process and solve the truth value device data uploaded by the information collection vehicle 110, so as to obtain a high-precision reference trajectory; in some embodiments, the truth value preprocessing module 142 parses the GNSS (Global Navigation Satellite System) and IMU (Inertial Measurement Unit) raw data based on the truth value device data, and combines the base station data and ephemeris data stored offline to obtain the coordinates and vehicle attitude of the vehicle in WGS84 (World Geodetic System-1984, an internationally adopted geocentric coordinate system); the truth value device data is processed by a bidirectional smoothing algorithm to improve the accuracy and smoothness of the truth value data, and finally a high-precision reference trajectory is obtained. Then, it is determined whether the accuracy of the high-precision reference trajectory meets the preset conditions, and the high-precision reference trajectory that does not meet the preset conditions is processed. In other embodiments, the Std (Standard Deviation) value in the high-precision reference trajectory is compared with a preset threshold. When it is less than the preset threshold, it is determined that the accuracy of the high-precision reference trajectory meets the conditions and can be used as a measurement standard. If it is greater than the preset threshold, it means that the accuracy of the high-precision reference trajectory here is not sufficient to be used as a measurement standard. Therefore, the high-precision reference trajectory here is eliminated, thereby screening out the high-precision reference trajectory whose accuracy meets the preset conditions as the measurement standard for subsequent other vehicle data.
[0074] The truth value preprocessing module 142 is also configured to perform multiple transformations on the high-precision reference trajectory obtained in the algorithm playback module 141, thereby serving as the truth value for evaluating the fused positioning trajectory, combined navigation trajectory, dead reckoning trajectory and semantic positioning trajectory calculated by the algorithm playback module 141.
[0075] In some embodiments, when calculating the true value of the fused positioning trajectory, the high-precision reference trajectory is segmented according to a set distance, such as 200 meters, and then the fused positioning trajectory corresponding to this section of the high-precision reference trajectory in the same time period is found, and the characteristic lines in the two trajectory segments are extracted. The two trajectory segments are moved to the position where the characteristic lines overlap, and the difference between the two segments is compared after the two trajectory segments are aligned according to the least squares method, and then the translation and rotation matrix is calculated by the SVD (Singular Value Decomposition) decomposition method, and the true value of the fused positioning trajectory is obtained after the high-precision reference trajectory of this section is transformed by the translation and rotation matrix. The characteristic line can be one of the characteristic lines such as the roadside line and the center line of the high-precision reference trajectory and the fused positioning trajectory.
[0076] Another optional method for obtaining the true value of the fused positioning trajectory is to directly use the high-precision reference trajectory as the true value of the fused positioning trajectory, but this method will lose objectivity when there is a fixed deviation in the local map. Alternatively, the method of this embodiment can also be based on the pre-acquired precise trajectory of the vehicle, but this method often requires the vehicle to travel along the center line of the lane to determine whether the final positioning result is on the center line of the lane on the map, and this method has extremely high requirements on the driver and the test site environment, and cannot be tested in batches.
[0077] When calculating the true value of the integrated navigation trajectory, the high-precision reference trajectory will be directly used as the true value of the integrated navigation trajectory.
[0078] When calculating the true value of the dead reckoning trajectory, the high-precision reference trajectory will be segmented according to another set distance, such as 100 meters, and then the dead reckoning trajectory corresponding to this high-precision reference trajectory in the same time period will be found, and the high-precision reference trajectory segment and the dead reckoning trajectory segment will be aligned. One optional way to align is to overlap the feature points of the high-precision reference trajectory segment and the dead reckoning trajectory segment and move them to the starting point of the coordinate system while keeping the direction and angle of the high-precision reference trajectory segment and the dead reckoning trajectory segment unchanged, where the feature point can be the starting point or end point of the high-precision reference trajectory segment and the dead reckoning trajectory segment, and obtain the translation and rotation matrix of the high-precision reference trajectory segment and the dead reckoning trajectory segment through a preset algorithm, and transform the high-precision reference trajectory segment through the translation and rotation matrix to obtain the true value of the dead reckoning trajectory. In this way, the dead reckoning trajectory is post-processed by the high-precision reference device based on the segmentation, and the processing algorithm itself applies a bidirectional smoothing algorithm, which has high smoothness and trajectory accuracy; and the segmentation alignment method eliminates the negative impact caused by the cumulative error, making the calculation result more fair and reasonable.
[0079] Another optional method for obtaining the true value of the dead reckoning trajectory is to stitch the dead reckoning trajectory through a preset stitching algorithm to obtain a final map. However, when jitter glitches appear in the final map, it is impossible to determine whether the problem lies in the abnormal dead reckoning trajectory or the stitching algorithm.
[0080] When calculating the true value of the semantic localization trajectory, the semantic localization trajectory is segmented according to the set distance, and then the semantic localization trajectory corresponding to this segment of the high-precision reference trajectory in the same time period is found. After the two trajectory segments are aligned, the difference between the two segments is compared to obtain the translation and rotation matrix, and this segment of the high-precision reference trajectory is transformed through the translation and rotation matrix to obtain the true value of the semantic localization trajectory.
[0081] The evaluation and calculation module 143 is configured to evaluate and calculate the difference between the target positioning trajectory and the true value of the positioning trajectory. The target positioning trajectory may include: dead reckoning trajectory, integrated navigation trajectory, semantic positioning trajectory and fused positioning trajectory; the true value of the positioning trajectory may include: the true value of the dead reckoning trajectory, the true value of the integrated navigation trajectory, the true value of the semantic positioning trajectory and the true value of the fused positioning trajectory; the calculated difference content may include: the deviation of the latitude and longitude, heading angle, pitch angle, roll angle, etc. between the target positioning trajectory and the true value of the positioning trajectory.
[0082] In some embodiments, the method for evaluating the dead reckoning trajectory and the dead reckoning trajectory truth value may be: while keeping the directions and angles of the dead reckoning trajectory segment and the dead reckoning trajectory truth value segment unchanged, overlap the feature points of the two and move them to the starting point of the coordinate system, wherein the feature point may be the starting point or the end point of the dead reckoning trajectory segment and the dead reckoning trajectory segment, and then calculate the deviations of the latitude and longitude, heading angle, pitch angle, and roll angle of the dead reckoning trajectory segment and the end point of the dead reckoning trajectory truth value segment in the coordinate system.
[0083] The method for evaluating the combined navigation trajectory and the true value of the combined navigation trajectory can be: while keeping the direction and angle of the combined navigation trajectory segment and the true value segment of the combined navigation trajectory unchanged, the feature points of the two are overlapped and moved to the starting point of the coordinate system, where the feature point can be the starting point or end point of the combined navigation trajectory segment and the combined navigation trajectory segment, and then the deviations of the longitude and latitude, heading angle, pitch angle and roll angle of the end point of the combined navigation trajectory segment and the true value segment of the combined navigation trajectory in the coordinate system are calculated.
[0084] The method for evaluating the fused positioning trajectory and the true value of the fused positioning trajectory can be: while keeping the direction and angle of the fused positioning trajectory segment and the true value segment unchanged, overlap the feature lines of the two, where the feature line can be one of the feature lines such as the roadside line, center line, etc. of the fused positioning trajectory and the true value of the fused positioning trajectory, and then calculate the deviation of the longitude and latitude, heading angle, pitch angle, and roll angle between the two trajectories.
[0085] The method for evaluating the semantic positioning trajectory and the true value of the semantic positioning trajectory can be: while keeping the direction and angle of the semantic positioning trajectory segment and the semantic positioning trajectory true value segment unchanged, overlap the feature lines of the two, where the feature line can be one of the feature lines such as the roadside line, center line, etc. of the semantic positioning trajectory and the semantic positioning trajectory true value, and then calculate the deviation of the longitude and latitude, heading angle, pitch angle, and roll angle between the two trajectories.
[0086] The result display module 144 is configured to plot the deviations of the latitude and longitude, heading angle, pitch angle, and roll angle between each positioning trajectory calculated by the evaluation calculation module 143 and its true value into error time series diagrams and error distribution diagrams of various indicators and calculate CEP (Circular Error Probable) 68 and CEP95 accuracy indicators. In some embodiments, the above deviations are decomposed into lateral (perpendicular to the heading direction) and longitudinal (along the heading direction) according to the heading angle of the true value trajectory, and then the error time series diagrams and error distribution diagrams of various indicators are plotted and the CEP68 and CEP95 accuracy indicators are calculated.
[0087] Among them, drawing the error timing diagram is to sort the above deviations in chronological order, and then draw it as a line graph, in which the time axis supports scaling, so that the deviation of the entire time period can be observed and the flying point can be quickly located, and the deviation change of the entire time period can be viewed, which is convenient for technical personnel in this field to analyze the cause of the error.
[0088] To draw the error distribution graph, all the deviation values are treated as a list, the frequency of deviations is counted at intervals of increasing preset distances, and a histogram of the error distribution is drawn. In this way, the average level of deviation and the probability of exceeding the required range can be quantitatively reflected, which is used to examine the robustness of the algorithm for calculating the above positioning trajectory.
[0089] The calculation of CEP68 and CEP95 accuracy indicators is to statistically calculate the above deviations according to the upper limit of the range of 68% and 95% confidence intervals, which is used to quantitatively describe the performance of the algorithm. For the same set of data, running multiple algorithm versions, the cross-version climbing of CEP68 and CEP95 indicators can be plotted. Because each algorithm update will lead to an increase in some indicators and a decrease in some indicators, through the comprehensive analysis of the climbing graphs of multiple algorithm performance indicators, the optimization and degradation of the algorithm can be obtained, which can be used as a criterion for judging whether the algorithm is allowed to be released.
[0090] Through the result display module 144, technicians in this field can view the error time series distribution diagram of the detailed fusion positioning trajectory, combined navigation trajectory, dead reckoning trajectory and semantic analysis trajectory, so as to find the module with poor performance and find the corresponding scene according to the timestamp for targeted optimization. For multiple sets of data and multiple versions of algorithms, technicians in this field can draw an algorithm performance climbing diagram based on the CEP68 or CEP95 performance indicators, and view the advantages and disadvantages of the algorithm version from different evaluation indicator dimensions.
[0091] Figure 4 1 is a schematic diagram of the cloud data platform architecture of a vehicle positioning trajectory evaluation system according to an embodiment of the present invention. In this schematic diagram, the cloud data platform 120 includes sensor data 410, true value device data 411, algorithm playback module 141, true value preprocessing module 142, evaluation calculation module 143 and result display module 144.
[0092] Sensor data 410 is vehicle data collected on the road by the information collection vehicle 110 through sensors on the vehicle body. The vehicle data may include one or more vehicle data such as vehicle speed, wheel speed, IMU, RTK, camera, and high-precision map. True value device data 411 is high-precision data collected on the road by the information collection vehicle 110 through a true value device, where the true value device is a high-precision positioning reference device with higher measurement accuracy than ordinary sensors, such as millimeter wave radar, laser radar, high-precision combined inertial navigation and other devices.
[0093] The algorithm playback module 141 is configured to cluster the sensor data 410 uploaded by the information collection vehicle 110 and create a corresponding data set according to the scenario when the information collection vehicle 110 collects data, and then select the required data set and the algorithm version code required by the instruction according to the instruction, and perform environment deployment and code compilation on the cloud data platform, and finally calculate and process the data set according to the algorithm required by the instruction to obtain the processing result.
[0094] The algorithm playback module 141 includes a dead reckoning algorithm module 420 , an integrated navigation algorithm module 421 , a semantic positioning algorithm module 422 and a fusion positioning algorithm module 423 .
[0095] In some embodiments, the dead reckoning algorithm module 420 is configured to calculate the vehicle speed, wheel speed and IMU data in the sensor data 410 according to the algorithm required by the instruction to obtain a dead reckoning trajectory; the integrated navigation algorithm module 421 is configured to correct the dead reckoning trajectory calculated by the dead reckoning algorithm module 420 according to the GNSS data parsed from the true value device data 411 to obtain a combined navigation trajectory; the semantic positioning algorithm module 422 is configured to load the map geometry layer data in the high-precision map data in the sensor data 410 according to the GNSS data parsed from the true value device data 411 as the initialization position and compare it with the semantic elements obtained by visual perception to obtain a semantic positioning trajectory; the fusion positioning algorithm module 423 is configured to use Kalman filtering to fuse the dead reckoning trajectory calculated by the dead reckoning algorithm module 420, the combined navigation trajectory calculated by the integrated navigation algorithm module 421 and the semantic positioning trajectory calculated by the semantic positioning algorithm module 422 according to the selected algorithm, and match them with the roads and lanes in the high-precision map data again to finally obtain a fused positioning trajectory.
[0096] The truth value preprocessing module 142 is configured to parse the GNSS and IMU raw data according to the truth value device data 411, and calculate the high-precision reference trajectory according to the preset algorithm, and select the high-precision reference trajectory with accuracy that meets the preset conditions as the measurement standard for subsequent other vehicle data.
[0097] The truth value preprocessing module 142 also includes a dead reckoning trajectory processing module 430, a combined navigation trajectory processing module 431, a semantic positioning trajectory processing module 432, and a fused positioning trajectory processing module 433. The dead reckoning trajectory processing module 430 is configured to obtain a high-precision reference trajectory, segment the high-precision reference trajectory according to a set distance, such as 100 meters, and then find the dead reckoning trajectory corresponding to this high-precision reference trajectory in the same time period, and align the high-precision reference trajectory segment with the dead reckoning trajectory segment. An optional way to align is to overlap the feature points of the high-precision reference trajectory segment and the dead reckoning trajectory segment and move them to the starting point of the coordinate system while keeping the direction and angle of the high-precision reference trajectory segment and the dead reckoning trajectory segment unchanged. The feature point can be the starting point or end point of the high-precision reference trajectory segment and the dead reckoning trajectory segment, obtain the translation and rotation matrix of each of the high-precision reference trajectory segment and the dead reckoning trajectory segment through a preset algorithm, and transform the high-precision reference trajectory segment through the translation and rotation matrix to obtain the true value of the dead reckoning trajectory. In this way, the dead reckoning trajectory is post-processed by a high-precision reference device based on the segmentation. The processing algorithm itself applies a two-way smoothing algorithm, which has high smoothness and trajectory accuracy. The segmentation alignment method eliminates the negative impact of the cumulative error, making the calculation result more fair and reasonable.
[0098] The integrated navigation trajectory processing module 431 is configured to obtain a high-precision reference trajectory when calculating the true value of the integrated navigation trajectory, and directly use the high-precision reference trajectory as the true value of the integrated navigation trajectory.
[0099] The semantic positioning trajectory processing module 432 is configured to obtain a high-precision reference trajectory, compare the semantic positioning trajectory with the high-precision reference trajectory, and transform the semantic positioning trajectory to obtain a true value of the semantic positioning trajectory.
[0100] The fused positioning trajectory processing module 433 is configured to obtain a high-precision reference trajectory, segment the high-precision reference trajectory according to a set distance, such as 200 meters, and then find the fused positioning trajectory corresponding to this section of the high-precision reference trajectory in the same time period, extract the characteristic lines in the two trajectory segments, and move the two trajectory segments to the position where the characteristic lines overlap while keeping the direction and angle of the high-precision reference trajectory segment and the fused positioning trajectory segment unchanged, and compare the difference between the two segments after aligning the two trajectory segments according to the least squares method, and then calculate the translation and rotation matrix by the SVD decomposition method, and transform this section of the high-precision reference trajectory through the translation and rotation matrix to obtain the true value of the fused positioning trajectory. The characteristic line can be one of the characteristic lines such as the roadside line and the center line of the high-precision reference trajectory and the fused positioning trajectory.
[0101] The positioning trajectory true value corresponding to each positioning trajectory is obtained through the above modules, and those skilled in the art can add or delete processing modules and algorithms of different positioning trajectory types according to actual conditions.
[0102] The evaluation calculation module 143 is configured to compare and evaluate the vehicle trajectory calculated by the algorithm playback module 141 with the trajectory truth value calculated by the truth value preprocessing module 142. The evaluation calculation module 143 includes a dead reckoning trajectory evaluation module 440, an integrated navigation trajectory evaluation module 441, a semantic positioning trajectory evaluation module 442, and a fusion positioning trajectory evaluation module 443.
[0103] In some embodiments, the dead reckoning trajectory evaluation module 440 is configured to overlap the feature points of the dead reckoning trajectory segment and the dead reckoning trajectory truth segment and move them to the starting point of the coordinate system while keeping the directions and angles of the dead reckoning trajectory segment and the dead reckoning trajectory truth segment unchanged, wherein the feature point may be the starting point or the end point of the dead reckoning trajectory segment and the dead reckoning trajectory segment, and then calculate the deviations of the latitude and longitude, heading angle, pitch angle, and roll angle of the dead reckoning trajectory segment and the end point of the dead reckoning trajectory truth segment in the coordinate system.
[0104] The combined navigation trajectory evaluation module 441 is configured to overlap the feature points of the combined navigation trajectory segment and the combined navigation trajectory true value segment and move them to the starting point of the coordinate system while keeping the directions and angles of the combined navigation trajectory segment and the combined navigation trajectory true value segment unchanged, where the feature points can be the starting point or end point of the combined navigation trajectory segment and the combined navigation trajectory segment, and then calculate the deviations of the longitude and latitude, heading angle, pitch angle and roll angle of the end point of the combined navigation trajectory segment and the combined navigation trajectory true value segment in the coordinate system.
[0105] The semantic positioning trajectory evaluation module 442 is configured to overlap the feature lines of the semantic positioning trajectory segment and the semantic positioning trajectory truth segment while keeping the direction and angle of the two segments unchanged, where the feature line can be one of the feature lines such as the roadside line, center line, etc. of the semantic positioning trajectory and the semantic positioning trajectory truth, and then calculate the deviation of the longitude and latitude, heading angle, pitch angle, and roll angle between the two trajectories.
[0106] The fused positioning trajectory evaluation module 443 is configured to overlap the characteristic lines of the fused positioning trajectory segment and the fused positioning trajectory true value segment while keeping the directions and angles of the two segments unchanged, where the characteristic line can be one of the characteristic lines such as the roadside line, center line, etc. of the fused positioning trajectory and the fused positioning trajectory true value, and then calculate the deviations of the longitude and latitude, heading angle, pitch angle, and roll angle between the two trajectories.
[0107] The deviation between the positioning trajectory and the corresponding positioning trajectory true value can be obtained through the above evaluation module. Those skilled in the art can add or delete evaluation modules and algorithms of different positioning trajectory types according to actual conditions.
[0108] The result display module 144 is configured to display the deviation value between the positioning trajectory calculated in the evaluation calculation module 143 and the high-precision reference trajectory in multiple aspects. The result display module 144 includes a performance indicator display module 450, a performance climbing version comparison module 451 and a playback module 452.
[0109] In some embodiments, the performance indicator display module 450 is configured to sort the deviations between the positioning trajectory and the corresponding positioning trajectory true value in chronological order, and then draw them as a line graph, where the time axis supports scaling, so that the deviations of the entire time period can be observed and the positioning flying point can be quickly found, and the deviation changes of the entire time period can be viewed, which is convenient for technical personnel in this field to analyze the causes of errors.
[0110] In other embodiments, the performance indicator display module 450 is further configured to treat all deviation values as a list, count the frequency of deviations according to the intervals of increasing preset distances, and draw a histogram of the error distribution. In this way, the average level of the deviation and the probability of exceeding the required range can be quantitatively reflected, which is used to examine the robustness of the algorithm for calculating the above positioning trajectory.
[0111] The performance climbing version comparison module 451 is configured to statistically calculate the deviation between the positioning trajectory and the corresponding positioning trajectory true value according to the upper limit of the range of the 68% and 95% confidence intervals, which is used to quantitatively describe the performance of the algorithm. For the same set of data, multiple algorithm versions can be run, and the cross-version climbing of the CEP68 and CEP95 indicators can be plotted. Because each algorithm update will lead to an increase in several indicators and a decrease in some indicators, through the comprehensive analysis of the climbing graphs of multiple algorithm performance indicators, the optimization and degradation of the algorithm can be obtained, which can be used as a criterion for judging whether the algorithm is allowed to be released.
[0112] The playback module 612 is configured to playback the trajectory content of the corresponding positioning trajectory type in a specified time period and its positioning trajectory true value and deviation analysis according to the instruction.
[0113] Through the result display module 144, technicians in this field can view the detailed error time series distribution diagram of the fusion positioning trajectory, combined navigation trajectory, dead reckoning trajectory and semantic analysis trajectory, thereby finding the module with poor performance, and find the corresponding scene according to the timestamp to perform targeted optimization. For multiple sets of data and multiple versions of algorithms, technicians in this field can draw an algorithm performance climbing diagram based on the CEP68 or CEP95 performance indicators, and show the advantages and disadvantages of the algorithm version from different evaluation indicator dimensions.
[0114] Through this cloud data platform, true device data can be added to the original sensor data, and the processed sensor data can be compared and evaluated with the true device data, so as to find the modules with poor performance and determine the feasibility of different versions of algorithms based on the evaluation results, thereby improving the efficiency of evaluation and acceptance. In addition, the high-precision reference trajectory can be transformed in a variety of other forms to meet the needs of performance tuning and comparison of various functional modules of fusion positioning.
[0115] Figure 5 The schematic diagram is a position diagram of the fused positioning trajectory and the true value of the fused positioning trajectory of the same segment according to the vehicle positioning trajectory evaluation method of an embodiment of the present invention. The schematic diagram includes a fused positioning trajectory segment 510 of a preset distance and a fused positioning trajectory true value segment 511 corresponding to the fused positioning trajectory of the same time period. Since the errors of the two are not in the same position, the fused positioning trajectory segment 510 and the fused positioning trajectory true value segment 511 are aligned by extracting characteristic lines such as roadside lines 520 in the two road sections, as shown in FIG. Figure 6 shown. Figure 6 The vehicle positioning trajectory evaluation method according to an embodiment of the present invention is a method for evaluating the vehicle positioning trajectory. Figure 5 The schematic diagram of the position after the fusion positioning trajectory and the fusion positioning trajectory truth value are shown in FIG. This method provides data support for the subsequent calculation of the fusion positioning trajectory segment 510 and the fusion positioning trajectory truth value segment 511.
[0116] Figure 7 The present invention is a schematic diagram of the positions of the combined navigation trajectory and the true value of the combined navigation trajectory of the same segment of the vehicle positioning trajectory evaluation method according to an embodiment of the present invention. The schematic diagram includes a combined navigation trajectory segment 710 of a preset distance and a combined navigation trajectory true value segment 711 corresponding to the combined navigation trajectory segment of the same time period. Since the two are not in the same position due to errors, the feature points in the two road sections, such as the starting point of the segment, are extracted, and the combined navigation trajectory segment 710 and the combined navigation trajectory true value segment 711 are aligned to coincide with each other while ensuring that the direction and angle remain unchanged, and the starting point is moved to the origin of the coordinate system, as shown in FIG. Figure 8 shown. Figure 8 The vehicle positioning trajectory evaluation method according to an embodiment of the present invention is a method for evaluating the vehicle positioning trajectory. Figure 7 The schematic diagram of the positions of the combined navigation track and the true value of the combined navigation track in the same segment. The dotted line 810 in the schematic diagram is the deviation between the combined navigation track segment 710 and the true value segment 711 of the combined navigation track. The deviation value between the combined navigation track segment 710 and the true value segment 711 of the combined navigation track is obtained by this method.
[0117] Fig. 9 1 is a flow chart of a vehicle positioning trajectory evaluation method according to an embodiment of the present invention. The flow chart includes:
[0118] Step S901, obtain sensor data and true value device data. The sensor data may include vehicle speed, IMU data, RTK data, camera data, and high-precision map data, etc.; the true value device data may include high-precision inertial navigation data and high-precision GNSS data, etc.; the accuracy of the true value device data is higher than that of the sensor data.
[0119] Step S902, the sensor data is calculated to obtain the target positioning trajectory. In some embodiments of the present application, the sensor data is clustered and a corresponding data set is created according to the scene when the information collection vehicle collects data, and then the required data set and the algorithm version code required by the instruction are selected according to the instruction, and the environment is deployed and the code is compiled on the cloud data platform, and then the vehicle speed, wheel speed and IMU data in the sensor data are calculated according to the algorithm required by the instruction to obtain the dead reckoning trajectory, and the dead reckoning trajectory is corrected according to the GNSS data parsed from the true value device data to obtain the combined navigation trajectory; the map geometry layer data in the high-precision map data in the vehicle data is loaded as the initialization position according to the GNSS data parsed from the true value device data, and the semantic elements obtained by visual perception are compared to obtain the semantic positioning trajectory; the above-mentioned combined navigation trajectory is fused with the semantic positioning trajectory using Kalman filtering, and matched again with the roads and lanes in the high-precision map, and finally a fused positioning trajectory is obtained.
[0120] Step S903, the true value device data is solved to obtain a high-precision reference trajectory. In some embodiments of the present application, the GNSS and IMU raw data are parsed according to the true value device data, and the coordinates and vehicle attitude of the vehicle in WGS84 are obtained by combining the base station data and ephemeris data saved offline; the true value device data is processed by a two-way smoothing algorithm to improve the accuracy and smoothness of the true value data, and finally a high-precision reference trajectory is obtained. Then, it is determined whether the accuracy of the high-precision reference trajectory meets the preset conditions, and the high-precision reference trajectory that does not meet the preset conditions is processed. In other embodiments, the Std value in the high-precision reference trajectory is compared with the preset threshold. When it is less than the preset threshold, it is determined that the accuracy of the high-precision reference trajectory meets the conditions and can be used as a measurement standard. If it is greater than the preset threshold, it means that the accuracy of the high-precision reference trajectory here is not enough to be used as a measurement standard, so the high-precision reference trajectory here is eliminated, so as to screen out the high-precision reference trajectory whose accuracy meets the preset conditions as the measurement standard for subsequent other vehicle data.
[0121] Step S904: The high-precision reference trajectory is converted by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory. The target positioning trajectory includes a dead reckoning trajectory, a combined navigation trajectory, a semantic positioning trajectory, and a fusion positioning trajectory.
[0122] The step of calculating the true value of the dead reckoning trajectory includes: cutting the high-precision reference trajectory into segments according to a first preset length, where the first preset length may be 100 meters; obtaining a dead reckoning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the dead reckoning trajectory segment until feature points overlap and calculating a translation rotation matrix, where the feature points include a starting point or an end point in the high-precision reference trajectory segment and the dead reckoning trajectory segment; and calculating the high-precision reference trajectory segment with the translation rotation matrix to obtain the true value of the high-precision reference trajectory corresponding to the dead reckoning trajectory.
[0123] The step of calculating the true value of the integrated navigation trajectory includes: directly using the high-precision reference trajectory as the true value of the integrated navigation trajectory.
[0124] The steps of calculating the true value of the semantic positioning trajectory include: cutting the high-precision reference trajectory into segments according to a second preset length, where the second preset length can be 200 meters; obtaining the semantic positioning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the semantic positioning trajectory segment through a preset algorithm, where the preset algorithm includes a least squares method; calculating a translation rotation matrix; and converting the high-precision reference trajectory segment through the translation rotation matrix to obtain the true value of the high-precision reference trajectory corresponding to the semantic positioning trajectory segment.
[0125] The step of calculating the true value of the fused positioning trajectory includes: cutting the high-precision reference trajectory into segments according to a third preset length, where the third preset length can be 200 meters; obtaining a fused positioning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the fused positioning trajectory segment through a preset algorithm, where the preset algorithm includes a least squares method; calculating a translation rotation matrix; and converting the high-precision reference trajectory segment through the translation rotation matrix to obtain the true value of the high-precision reference trajectory corresponding to the fused positioning trajectory segment.
[0126] Step S905, compare and evaluate the target positioning trajectory with the true value of the high-precision reference trajectory to obtain an evaluation result. This step also includes: generating different types of effect diagrams according to the evaluation results, the effect diagrams are used to show the error change process between the target positioning trajectory and the true value of the high-precision reference trajectory;
[0127] The positioning deviation CEP68 and / or CEP95 indicators are calculated according to the evaluation results, so as to compare the performance of the target algorithm using the positioning deviation CEP68 and / or CEP95 indicators. The technology in this field can expand the application of the evaluation results according to actual conditions, so as to better discover problems in the positioning trajectory process.
[0128] Through this method, the modules with poor performance can be found according to the evaluation results and the feasibility of different versions of algorithms can be determined, thereby improving the efficiency of evaluation and acceptance. In addition, the high-precision reference trajectory can be transformed in a variety of other forms to meet the needs of performance tuning and comparison of various functional modules of fusion positioning.
[0129] This embodiment also provides a machine-readable storage medium and a computer device. Fig.10 is a schematic diagram of a machine-readable storage medium 1001 according to an embodiment of the present invention, Fig.11 is a schematic diagram of a computer device 1103 according to one embodiment of the present invention.
[0130] The machine-readable storage medium 1001 stores a machine-executable program 1002 thereon. When the machine-executable program 1002 is executed by a processor, the method for evaluating the vehicle positioning trajectory in any of the above-mentioned embodiments can be implemented.
[0131] The computer device 1103 may include a memory 1101, a processor 1102, and a machine executable program 1002 stored in the memory 1101 and running on the processor 1102, and the method for vehicle positioning trajectory evaluation of any of the above embodiments is implemented when the processor 1102 executes the machine executable program 1002.
[0132] It should be noted that the logic and / or steps represented in the flowchart or described in other ways herein, for example, calculating the fusion positioning trajectory, can be specifically implemented in any machine-readable storage medium for use by an instruction execution system, device or equipment (such as a computer-based system, a processor-based system or other system that can fetch instructions from an instruction execution system, device or equipment and execute instructions), or used in combination with these instruction execution systems, devices or equipment.
[0133] For the description of this embodiment, the machine-readable storage medium 1001 can be any device that can contain, store, communicate, propagate or transmit a program for use in an instruction execution system, device or equipment or in combination with these instruction execution systems, devices or equipment. More specific examples (non-exhaustive list) of the machine-readable storage medium 1001 include the following: an electrical connection portion (electronic device) with one or more wirings, a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the machine-readable storage medium 1001 can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting or processing in other suitable ways as necessary, and then stored in a computer memory.
[0134] It should be understood that each part of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system.
[0135] Computer device 1103 can be, for example, a server, a desktop computer, a notebook computer, a tablet computer, or a smart phone. In some examples, computer device 1103 can be a cloud computing node. Computer device 1103 can be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. Typically, a program module can include routines, programs, target programs, components, logic, data structures, etc. that perform specific tasks or implement specific abstract data types. Computer device 1103 can be implemented in a distributed cloud computing environment where remote processing devices linked through a communication network perform tasks. In a distributed cloud computing environment, program modules can be located on a local or remote computing system storage medium including a storage device.
[0136] The computer device 1103 may include a processor 1102 adapted to execute stored instructions, and a memory 1101 providing temporary storage space for the operation of the instructions during operation. The processor 1102 may be a single-core processor, a multi-core processor, a computing cluster, or any number of other configurations. The memory 1101 may include a random access memory (RAM), a read-only memory, a flash memory, or any other suitable storage system.
[0137] The processor 1102 may be connected to an I / O interface (input / output interface) suitable for connecting the computer device 1003 to one or more I / O devices (input / output devices) through a system interconnect (e.g., PCI, PCI-Express, etc.). The I / O devices may include, for example, a keyboard and a pointing device, wherein the pointing device may include a touch pad or a touch screen, etc. The I / O devices may be built-in components of the computer device 1003, or may be devices externally connected to the computing device.
[0138] Processor 1102 can also be linked to a display interface suitable for connecting computer device 1103 to a display device through a system interconnection. Display device can include a display screen as a built-in component of computer device 1103. Display device can also include a computer monitor, a television or a projector, etc., which are externally connected to computer device 1103. In addition, a network interface controller (network interface controller, NIC) can be suitable for connecting computer device 1103 to a network through a system interconnection. In some embodiments, NIC can use any suitable interface or protocol (such as Internet Small Computer System Interface, etc.) to transmit data. The network can be a cellular network, a radio network, a wide area network (WAN), a local area network (LAN) or the Internet, etc. Remote devices can be connected to computing devices through a network.
[0139] The flow chart provided by the present embodiment is not intended to indicate that the operation of the method will be performed in any particular order, or that all operations of the method are included in all every case. In addition, the method may include additional operations. Within the scope of the technical thinking provided by the present embodiment method, additional changes may be made to the above method.
[0140] At this point, those skilled in the art should recognize that, although multiple exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications that conform to the principles of the present invention can still be directly determined or derived based on the content disclosed in the present invention without departing from the spirit and scope of the present invention. Therefore, the scope of the present invention should be understood and recognized as covering all these other variations or modifications.
Claims
1. A vehicle positioning trajectory evaluation system, include: The information collection vehicle includes at least one or more sensors selected from the group consisting of a camera, a navigation and positioning module, and a radar sensor, and also includes at least one or more true value devices selected from the group consisting of a high-precision inertial navigation system and a high-precision radar, and is capable of uploading the data of the at least one or more sensors and the true value device data, wherein the true value device is a high-precision positioning reference device having a higher measurement accuracy than an ordinary sensor; Cloud data platform, which includes: A data interface configured to receive the sensor data and the true value device data uploaded by the information collection vehicle, wherein the sensor data includes one or more of the following data: vehicle speed, IMU data, RTK data, camera data, and high-precision map data, and the true value device data includes high-precision inertial navigation data; An algorithm playback module, configured to calculate the sensor data to obtain a target positioning trajectory; A truth value preprocessing module is configured to solve the truth value device data to obtain a high-precision reference trajectory; the high-precision reference trajectory is converted by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory truth value corresponding to the target positioning trajectory; The evaluation calculation module is configured to compare and evaluate the target positioning trajectory with the true value of the high-precision reference trajectory to obtain an evaluation result.
2. The vehicle positioning trajectory evaluation system according to claim 1, in, The target positioning trajectory includes a dead reckoning trajectory; The truth value preprocessing module is further configured to: segment the high-precision reference trajectory according to a first preset length to obtain a high-precision reference trajectory segment; obtain a dead reckoning trajectory segment corresponding to the high-precision reference trajectory segment; align the high-precision reference trajectory segment with the dead reckoning trajectory segment until feature points overlap and calculate a translation and rotation matrix, wherein the feature points include a starting point or an end point in the high-precision reference trajectory segment and the dead reckoning trajectory segment; The high-precision reference trajectory segment and the translation rotation matrix are calculated to obtain a true value of the high-precision reference trajectory corresponding to the dead reckoning trajectory.
3. The vehicle positioning trajectory evaluation system according to claim 1, in, The target positioning track includes a semantic positioning track; The truth value preprocessing module is further configured to: segment the high-precision reference trajectory according to a second preset length to obtain a high-precision reference trajectory segment; obtain a semantic positioning trajectory segment corresponding to the high-precision reference trajectory segment; align the high-precision reference trajectory segment with the semantic positioning trajectory segment through a preset algorithm, wherein the preset algorithm includes a least squares method; Calculate the translation and rotation matrix; The high-precision reference trajectory segment is converted by the translation and rotation matrix to obtain the high-precision reference trajectory true value corresponding to the semantic positioning trajectory segment; and / or The target positioning trajectory includes a fusion positioning trajectory; The truth value preprocessing module is further configured to: segment the high-precision reference trajectory according to a third preset length to obtain a high-precision reference trajectory segment; obtain the fused positioning trajectory segment corresponding to the high-precision reference trajectory segment; align the high-precision reference trajectory segment with the fused positioning trajectory segment through a preset algorithm, wherein the preset algorithm includes a least squares method; Calculate the translation and rotation matrix; The high-precision reference trajectory segment is transformed by the translation and rotation matrix to obtain the high-precision reference trajectory true value corresponding to the fused positioning trajectory segment.
4. The vehicle positioning trajectory evaluation system according to claim 1, in, The truth value preprocessing module is also configured to: Obtaining a standard deviation value of the high-precision reference trajectory in horizontal positioning; Determining whether the standard deviation value is less than a preset threshold; If it is greater than the preset threshold, the high-precision reference trajectory of the section where the standard deviation value is located is eliminated.
5. The vehicle positioning trajectory evaluation system according to claim 1, in, The cloud data platform also includes a result display module; The result display module is configured to generate different types of renderings according to the evaluation results, and the renderings are used to display the error change process between the target positioning trajectory and the true value of the high-precision reference trajectory; calculate the positioning deviation CEP68 and / or CEP95 indicators according to the evaluation results, so as to use the positioning deviation CEP68 and / or CEP95 indicators to compare the performance of the algorithm.
6. A method for evaluating vehicle positioning trajectory, include: Acquire sensor data and true value device data, wherein the accuracy of the true value device data is higher than that of the sensor data, the sensor data includes one or more of the following data: vehicle speed, IMU data, RTK data, camera data, and high-precision map data, and the true value device data includes high-precision inertial navigation data; Calculating the sensor data to obtain a target positioning trajectory; Solving the true value device data to obtain a high-precision reference trajectory; The high-precision reference trajectory is converted by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a true value of the high-precision reference trajectory corresponding to the target positioning trajectory; The target positioning trajectory is compared and evaluated with the true value of the high-precision reference trajectory to obtain an evaluation result.
7. The method for evaluating vehicle positioning trajectory according to claim 6, in, The target positioning trajectory includes a dead reckoning trajectory; The step of converting the high-precision reference trajectory by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory includes: cutting the high-precision reference trajectory into segments according to a first preset length; obtaining a dead reckoning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the dead reckoning trajectory segment until feature points overlap and calculating a translation rotation matrix, wherein the feature points include the high-precision reference trajectory segment and a starting point or an end point in the dead reckoning trajectory segment; and calculating the high-precision reference trajectory segment with the translation rotation matrix to obtain a high-precision reference trajectory true value corresponding to the dead reckoning trajectory.
8. The method for evaluating vehicle positioning trajectory according to claim 6, in, The target positioning track includes a semantic positioning track; The step of converting the high-precision reference trajectory by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory includes: cutting the high-precision reference trajectory according to a second preset length to obtain a high-precision reference trajectory segment; obtaining a semantic positioning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the semantic positioning trajectory segment through a preset algorithm, the preset algorithm including a least squares method; calculating a translation rotation matrix; converting the high-precision reference trajectory segment through the translation rotation matrix to obtain the high-precision reference trajectory true value corresponding to the semantic positioning trajectory segment; and / or The target positioning trajectory also includes a fusion positioning trajectory; The step of converting the high-precision reference trajectory by selecting a corresponding conversion method according to the type of the target positioning trajectory to obtain a high-precision reference trajectory true value corresponding to the target positioning trajectory includes: cutting the high-precision reference trajectory according to a third preset length to obtain a high-precision reference trajectory segment; obtaining the fused positioning trajectory segment corresponding to the high-precision reference trajectory segment; aligning the high-precision reference trajectory segment with the fused positioning trajectory segment through a preset algorithm, and the preset algorithm includes a least squares method; calculating a translation rotation matrix; and converting the high-precision reference trajectory segment through the translation rotation matrix to obtain the high-precision reference trajectory true value corresponding to the fused positioning trajectory segment.
9. The method for evaluating vehicle positioning trajectory according to claim 6, in, After the step of solving the true value device data to obtain a high-precision reference trajectory, the following steps are included: Obtaining a standard deviation value of the high-precision reference trajectory in horizontal positioning; Determining whether the standard deviation value is less than a preset threshold; If it is greater than the preset threshold, the high-precision reference trajectory of the section where the standard deviation value is located is eliminated.
10. The method for evaluating vehicle positioning trajectory according to claim 6, in, The step of obtaining the evaluation result includes: Generate different types of effect diagrams according to the evaluation results, wherein the effect diagrams are used to show the error change process between the target positioning trajectory and the true value of the high-precision reference trajectory; The positioning deviation CEP68 and / or CEP95 indicators are calculated according to the evaluation results, so as to compare the performance of the algorithms using the positioning deviation CEP68 and / or CEP95 indicators.
11. A machine-readable storage medium having a machine-executable program stored thereon, wherein the machine-executable program, when executed by a processor, implements the method for evaluating vehicle positioning trajectory according to any one of claims 6 to 10.
Citation Information
Cited By
Parking evaluation method and system, electronic equipment and vehicle
CN121415621A
