An unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data

Abstract

The present invention provides an unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data, comprising: building a simulation test environment and a measured environment under the same scenario according to the test vehicle, and performing at least one project test respectively, recording the simulation data and the measured data; processing the obtained data to obtain target simulation data and target measured data for building a true value training model; back-injecting the target simulation data and target measured data into the true value training model, and obtaining a true value simulation model through fitting, comparison and deduction; obtaining sample simulation data and sample measured data under multiple scenarios, and optimizing the true value simulation model; determining the evaluation dimension of the test vehicle, constructing a test evaluation system model, combining the optimized true value simulation model to obtain a true value evaluation system, and generating an evaluation report for the test vehicle. The present invention can efficiently and accurately evaluate vehicle simulation test results, providing reliable data support for the research and development of unmanned vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned vehicle simulation evaluation, and in particular to an unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data. Background Art

[0002] Functional unmanned vehicles, a subcategory of intelligent connected vehicles, are primarily used for off-road transportation operations, such as campus delivery and shuttle services. Compared to the development of intelligent connected cloud-based control platforms for highways, these vehicles lag behind. Consequently, with the development of functional unmanned vehicles, traditional simulation systems suffer from issues such as a lack of reference or inaccuracy in their simulation results, raising questions about the reliability of simulation data. To ensure the safe and efficient production and use of functional unmanned vehicles, a method for efficiently and accurately evaluating simulation test results for functional unmanned vehicles is urgently needed. Summary of the Invention

[0003] Based on this, it is necessary to provide an unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data to address the above technical issues.

[0004] A method for evaluating an unmanned vehicle based on simultaneous reinjection of simulation and measured data, comprising the following steps: constructing a simulation test environment and a measured environment under the same scenario according to the test standards and regulations of the test vehicle; conducting at least one project test in each of the constructed simulation test environment and the measured environment, and recording the corresponding simulation data and measured data; processing the simulation data and the measured data to obtain target simulation data and target measured data, respectively injecting them into a simulation database and a measured database, and constructing a true value training model based on the target simulation data and the target measured data; simultaneously reinjecting the target simulation data and the target measured data into the true value training model, and obtaining a true value simulation model through fitting, comparison and deduction; obtaining sample simulation data and sample measured data under multiple scenarios, injecting them into the true value simulation model, and optimizing the true value simulation model; determining the evaluation dimensions of the test vehicle, and constructing a test evaluation system model for the test vehicle based on the evaluation dimensions; combining the test evaluation system model and the optimized true value simulation model to obtain a true value evaluation system, and generating an evaluation report for the test vehicle based on the true value evaluation system.

[0005] In one embodiment, the simulation test environment and the actual test environment under the same scenario are built according to the test standards and procedural requirements of the test vehicle, including: building a corresponding actual test environment according to the test standards and procedural requirements of the test vehicle; constructing a corresponding simulation test environment according to the actual test environment, the scenarios of the simulation test environment and the actual test environment are the same, and both contain test elements and states of corresponding test items; and setting a dynamic model corresponding to the simulation test environment based on the parameters of the test vehicle.

[0006] In one embodiment, the simulation data and the measured data correspond one-to-one, and both include vehicle position, speed, acceleration, heading angle, expected collision time, obstacle recognition trigger time, distance to the obstacle, deceleration, lateral angle, angular velocity and lateral distance to the obstacle after circumventing the obstacle.

[0007] In one embodiment, the simulation data and the measured data are processed to obtain target simulation data and target measured data, which are respectively injected into a simulation database and a measured database, and a true value training model is constructed based on the target simulation data and the target measured data, including: under the same test conditions, multiple project tests are carried out through the constructed simulation environment and the measured environment to obtain the simulation data and the measured data; the simulation data and the measured data are matched one by one to obtain several data groups, and the corresponding data difference is calculated based on the data in the data groups; a preset maximum deviation value is obtained, and when the data difference exceeds the maximum deviation value, the corresponding data group is eliminated to obtain a cleaned data group; the cleaned data group is subjected to median average filtering to obtain target simulation data and target measured data; a true value training model is constructed based on the target simulation data and the target measured data, and the target simulation data and the target measured data are respectively injected into the simulation database and the measured database.

[0008] In one embodiment, the target simulation data and the target measured data are simultaneously injected into the true value training model, and the true value simulation model is obtained through fitting, comparison, and deduction, including: injecting the target simulation data and the target measured data into the true value training model at the same time; and establishing a high-order correspondence model between the target simulation data and the target measured data through fitting and comparison, which is:

[0009] X act =αXsim 3 +βXsim 2 +γXsim+X offset

[0010] Where, Xsim is the simulation data, X act are measured data, α, β, γ and X offset Parameter matrix values; deriving the parameter matrix values ​​by injecting multiple groups of the target simulation data and the target measured data into the high-order corresponding model; obtaining a parameter matrix according to the parameter matrix values, and obtaining a true value simulation model by superimposing the parameter matrix on the basic simulation model.

[0011] In one embodiment, the obtaining of sample simulation data and sample measured data in a variety of scenarios and injecting them into the true value simulation model to optimize the true value simulation model includes: obtaining corresponding sample simulation data and sample measured data by conducting project tests in a variety of scenarios; inputting the sample simulation data and sample measured data into the true value simulation model, training and optimizing the true value simulation model, and obtaining an optimized true value simulation model.

[0012] In one embodiment, the test evaluation system model includes: obtaining evaluation dimensions of the test vehicle, the evaluation dimensions including safety, compliance, serviceability and timeliness, and configuring the weight corresponding to each evaluation dimension; determining the corresponding test items according to each dimension, which may be one or more test items; extracting key technical indicators from the test items, and scoring them according to the test results of the key technical indicators; obtaining the score of each evaluation dimension to obtain the total score of the test vehicle; and determining the overall situation of the test vehicle according to the range of the total score.

[0013] In one embodiment, the test evaluation system model and the optimized true value simulation model are combined to obtain a true value evaluation system, and an evaluation report of the test vehicle is generated based on the true value evaluation system, including: combining the test evaluation system model and the optimized true value simulation model to obtain a true value evaluation system; obtaining simulation test data of the test vehicle, and inputting the simulation test data into the true value evaluation system; and obtaining an evaluation report of the simulation test of the test vehicle based on the true value evaluation system.

[0014] Compared with the existing technology, the advantages and beneficial effects of the present invention are as follows: based on the test standards and regulations of the test vehicle, a simulation test environment and a test environment under the same scenario are built, and at least one project test is carried out respectively, and the corresponding simulation data and test data are recorded; the simulation data and the test data are processed to obtain the target simulation data and the target test data, and a true value training model is constructed, and the target simulation data and the test data are injected into the simulation database and the test database respectively, thereby improving the accuracy of the vehicle test data and ensuring the authenticity and reliability of the simulation data; the target simulation data and the target test data are simultaneously injected back into the true value training model, and the target test data are compared and inferred through fitting, comparison and inference. Guide, get the true value simulation model, obtain sample simulation data and sample measured data in various scenarios, and inject the true value simulation model to train and optimize it to ensure that the simulation data can better fit the measured data; determine the evaluation dimensions of the test vehicle, build a test evaluation system model based on the evaluation dimensions, combine with the optimized true value simulation model to get the true value evaluation system, generate the test vehicle evaluation report, thereby realizing the automatic generation of the evaluation report, improving the evaluation efficiency and accuracy of the vehicle simulation test results, and the data is real and reliable, which can be applied to the simulation test evaluation of unmanned vehicles in various scenarios, providing data support for the research and development of unmanned vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 1 is a flow chart of an unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data in one embodiment;

[0016] Figure 2 This is the overall process architecture of an unmanned vehicle evaluation method based on simultaneous reinjection of simulation and measured data in one embodiment. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] In one embodiment, Figure 1 and Figure 2 As shown, a self-driving car evaluation method based on simultaneous back-injection of simulation and measured data is provided, including the following steps:

[0019] Step S101: Building a simulation test environment and a real test environment under the same scenario according to the test standards and regulations of the test vehicle.

[0020] Specifically, a realistic test environment for the test items must be established according to the testing standards and procedures for functional unmanned vehicles. This test environment must include the relevant test elements and states for the corresponding test items. For example, the cone recognition and response test requires a long, straight, flat road with a cone placed along the edge of the road. The test vehicle is set to travel along the center of the road toward the cone at a constant speed. The vehicle-in-the-loop simulation test environment must be constructed to replicate the actual test environment. The dynamics model of the test vehicle in the simulation environment must also be configured based on the test vehicle's actual parameters.

[0021] Among them, step S101 includes: building a corresponding actual measurement environment according to the test standards and regulations of the test vehicle; building a corresponding simulation test environment according to the actual measurement environment, the scenes of the simulation test environment and the actual measurement environment are the same, and both contain test elements and states of the corresponding test items; based on the parameters of the test vehicle, setting the dynamic model of the corresponding simulation test environment.

[0022] Specifically, when it is necessary to test a functional unmanned vehicle, a corresponding actual test environment is set up according to the test standards and procedures of the test vehicle, and a corresponding simulation test environment is constructed based on the established actual test environment. The simulation test environment corresponds to the actual test environment one by one, and both include the test elements and status of the corresponding test items. Before the test, the dynamic model of the vehicle in the simulation test environment is set according to the relevant parameters of the test vehicle to ensure that the vehicle power in the simulation test is the same as that in the actual test environment to avoid interference with the test results.

[0023] Step S102: Perform at least one project test in the constructed simulation test environment and actual measurement environment, and record corresponding simulation data and actual measurement data.

[0024] Specifically, the simulation test environment and actual measurement environment are used to complete the relevant project tests according to the test requirements. The number of tests can be one or more as needed, and the simulation data and actual measurement data generated during the test are collected and recorded.

[0025] Among them, the simulation data and the measured data correspond one to one, including vehicle position, speed, acceleration, heading angle, expected collision time, obstacle recognition trigger time, distance to the obstacle, deceleration, lateral angle, angular velocity and lateral distance to the obstacle after circumventing the obstacle.

[0026] Specifically, during the test process, relevant data of the simulation test and the actual measurement are collected, and the simulation data and the actual measurement data correspond one to one, so as to determine the difference between the actual measurement data and the simulation data, and then judge the result of the simulation test.

[0027] Step S103 , performing data processing on the simulation data and the measured data to obtain target simulation data and target measured data, and injecting them into the simulation database and the measured database respectively, and constructing a true value training model based on the target simulation data and the target measured data.

[0028] Specifically, data processing is performed on the collected simulation data and measured data, such as noise processing and data filtering, to obtain more accurate target simulation data and target measured data, which are stored in the simulation database and the measured database respectively. At the same time, a true value training model is constructed based on the target simulation data and the target measured data, so that the data obtained in the simulation test process is closer to the measured data.

[0029] Among them, step S103 includes: under the same test conditions, multiple project tests are carried out through the constructed simulation environment and actual measurement environment to obtain simulation data and actual measurement data; the simulation data and the actual measurement data are matched one by one to obtain several data groups, and the corresponding data difference is calculated according to the data in the data group; a preset maximum deviation value is obtained, and when the data difference exceeds the maximum deviation value, the corresponding data group is eliminated to obtain a cleaned data group; the cleaned data group is subjected to median average filtering to obtain target simulation data and target actual measurement data; a true value training model is constructed based on the target simulation data and the target actual measurement data, and the target simulation data and the target actual measurement data are respectively injected into the simulation database and the actual measurement database.

[0030] Specifically, in order to obtain a more accurate true value training model, multiple project tests are carried out in a well-established test environment under the same test conditions, and corresponding simulation data and measured data are obtained. The collected simulation data and measured data are matched one by one to obtain a corresponding data group, and the measured data is subtracted from the simulation data in the data group to calculate the data difference between the simulation data and the measured data. The preset deviation is obtained. Since the simulation data includes several data, there may be corresponding differences in the preset deviations corresponding to different data. The data difference is judged based on the preset deviation to see whether it meets the requirements. When the data deviation exceeds the preset deviation, the data group is identified as noise and is eliminated to complete the cleaning of the noise data, thereby improving the accuracy of the data and reducing the impact of errors.

[0031] The cleaned data group is subjected to median average filtering. For example, when 10 data are sampled, the maximum and minimum values ​​are removed, and the average value of the remaining 8 data is calculated to complete the filtering processing of the data group, and the target simulation data and target measured data are obtained. A true value training model is constructed based on the target simulation data and the target measured data, and the models are stored in the simulation database and the measured database respectively for the convenience of subsequent data calls.

[0032] In step S104, the target simulation data and the target measured data are simultaneously injected into the true value training model, and the true value simulation model is obtained through fitting, comparison and deduction.

[0033] Specifically, the target simulation data and target measured data obtained by data processing are simultaneously injected back into the true value training model. Through fitting and comparison, the target simulation data and the target measured data are matched one by one, and the true value simulation model is obtained through deduction. The simulation data can be processed by the true value simulation model, which effectively solves the problems of no reference and inaccuracy in traditional simulation results, so that the obtained simulation data can be closer to the measured data, thereby improving the reliability of the simulation test.

[0034] Among them, step S104 includes: simultaneously injecting the target simulation data and the target measured data into the true value training model; and establishing a high-order correspondence model between the target simulation data and the target measured data through fitting comparison, which is:

[0035] X act =αXsim 3 +βXsim 2 +γXsim+X offset

[0036] Where, Xsim is the simulation data, X act are measured data, α, β, γ and X offset Parameter matrix values; inject multiple sets of target simulation data and target measured data into the high-order corresponding model to derive parameter matrix values; obtain the parameter matrix according to the parameter matrix values, and obtain the true value simulation model by superimposing the parameter matrix on the basic simulation model.

[0037] Specifically, after obtaining the processed target simulation data and target measured data, they are simultaneously injected back into the true value training model, and fitting comparison is performed to establish a high-order correspondence model between the target simulation data and the target measured data, so as to facilitate training the target simulation data through the target measured data, and inject multiple groups of target simulation data and target measured data into the high-order correspondence model to obtain parameter matrix values, and construct parameter data according to the parameter matrix values. With the basic simulation model as the framework, the parameter matrix is ​​superimposed to obtain the true value simulation model. The true value simulation model can be used to obtain simulation data that is closer to the measured data, thereby improving the reliability of the simulation data.

[0038] Step S105 , obtaining sample simulation data and sample measured data under various scenarios, injecting the data into the true value simulation model, and optimizing the true value simulation model.

[0039] Specifically, through a large number of project tests in different scenarios, a variety of sample simulation data and sample measured data are obtained, and injected into the true value simulation model, and the true value simulation model of different scenarios is corrected and calibrated, so as to continuously iterate and optimize the true value simulation model, so that the true value simulation model can be applied to different scenarios, and the simulation test data can be processed to improve the accuracy of the simulation test data.

[0040] Among them, step S105 includes: obtaining corresponding sample simulation data and sample measured data by conducting project tests in various scenarios; inputting the sample simulation data and sample measured data into the true value simulation model, training and optimizing the true value simulation model, and obtaining an optimized true value simulation model.

[0041] Specifically, sample simulation data and sample measured data of project tests in various scenarios are obtained and used to train and optimize the true value simulation model, so that the true value simulation model can be applied to simulation data processing in various scenarios, ensuring that the data obtained from the simulation test can be closer to the measured data, thereby improving the accuracy of the simulation test.

[0042] Step S106 , determining the evaluation dimensions of the test vehicle, and constructing a test evaluation system model for the test vehicle based on the evaluation dimensions.

[0043] Specifically, based on the test vehicle, its evaluation dimensions are determined, and a vehicle test evaluation system model is constructed based on the evaluation dimensions of the test vehicle, so as to facilitate scientific and accurate evaluation of the vehicle test results.

[0044] Among them, the test evaluation system model includes: obtaining the evaluation dimensions of the test vehicle, which include safety, compliance, serviceability and timeliness, and configuring the corresponding weights for each evaluation dimension; determining the corresponding test items according to each dimension, which can be one or more test items; extracting key technical indicators from the test items, and scoring them according to the test results of the key technical indicators; obtaining the scores for each evaluation dimension to obtain the total score of the test vehicle; and determining the overall situation of the test vehicle based on the range of the total score.

[0045] Specifically, the evaluation dimensions of the test vehicle are obtained, such as safety, compliance, serviceability, and timeliness, and corresponding weights are assigned to each evaluation dimension, with weights of 30, 30, 20, and 20 respectively. The total evaluation score is the sum of the four evaluation dimensions, with a full score of 100 points.

[0046] For each evaluation dimension, corresponding test items are selected. For example, the test items for safety include: self-driving (including vehicle starting 3, parking 3 and following 4), recognition and response 10 (including vehicle recognition and response 4, pedestrian and non-motor vehicle recognition and response 4 and obstacle recognition and response 2) and failure takeover (including failure detection 4, failure response 2 and takeover operation 4), with a weight of 10. The test items for compliance include: traffic light recognition and response, lane line recognition and response and traffic sign recognition and response, with a weight of 10. The test items for serviceability include: convenience of human-computer interaction and functional task achievement, with a weight of 10. The test items for timeliness include: cloud monitoring timeliness and remote driving timeliness, with a weight of 10.

[0047] A simulation test is performed according to the determined test items to obtain simulation test data. Based on the simulation test data, all test item indicators of each evaluation dimension are scored to obtain the score of each dimension, and the total score of the test vehicle is obtained by combining all dimensions. The performance of the test vehicle is judged according to the range of the total score. For example, a total score of 95 points or above is set as excellent, a total score of 80-95 points is set as good, a total score of 60-80 points is set as medium, and a total score below 60 points is set as poor. In this way, the simulation situation of the vehicle can be judged according to the score of the test vehicle, thereby determining the vehicle performance.

[0048] Through the above steps, an evaluation system for all evaluation indicators of the test vehicle is established, and the evaluation system is imported into the evaluation system in the form of a model to form a test evaluation system model, which can be used to evaluate the simulation results of the test vehicle to judge the performance of the test vehicle.

[0049] Step S107 , combining the test evaluation system model and the optimized true value simulation model to obtain a true value evaluation system, and generating an evaluation report for the test vehicle based on the true value evaluation system.

[0050] Specifically, a true-value evaluation system is constructed based on the obtained test evaluation system model and the optimized true-value simulation model. When evaluating vehicle simulation test results, the test data is input into the true-value evaluation system, which outputs realistic test evaluation results and generates an evaluation report. This enables efficient and scientific evaluation of simulation test results and improves their accuracy. Furthermore, the true-value evaluation system can be used to evaluate unmanned vehicle simulation tests in batches, improving work efficiency.

[0051] Among them, step S107 includes: combining the test evaluation system model and the optimized true value simulation model to obtain a true value evaluation system; obtaining simulation test data of the test vehicle, and inputting the simulation test data into the true value evaluation system; and obtaining an evaluation report of the test vehicle simulation test based on the true value evaluation system.

[0052] Specifically, by combining the constructed test evaluation system model and the optimized true value simulation model, a true value evaluation system is constructed. When the performance of the test vehicle needs to be tested in the future, the test vehicle can be simulated and tested directly by setting up a simulation test environment to obtain simulation test data. The simulation test data is input into the true value evaluation system to obtain an evaluation report of the test vehicle simulation test, thereby achieving efficient and accurate evaluation of the vehicle simulation test results and providing effective data support for the research and development of unmanned vehicles.

[0053] In this embodiment, based on the test standards and regulations of the test vehicle, a simulation test environment and a measurement environment are set up under the same scenario, and at least one project test is performed on each of them, and the corresponding simulation data and measurement data are recorded; the simulation data and measurement data are processed to obtain target simulation data and target measurement data, and a true value training model is constructed and injected into the simulation database and the measurement database respectively, thereby improving the accuracy of the vehicle test data and ensuring the authenticity and reliability of the simulation data; the target simulation data and the target measurement data are simultaneously injected back into the true value training model, and a true value simulation model is obtained through fitting, comparison and deduction. Sample simulation data and sample measurement data under various scenarios are obtained and injected into the true value simulation model for training and optimization, ensuring that the simulation data can better fit the measurement data; the evaluation dimensions of the test vehicle are determined, and a test evaluation system model is constructed based on the evaluation dimensions. Combined with the optimized true value simulation model, a true value evaluation system is obtained, and an evaluation report for the test vehicle is generated, thereby realizing the automatic generation of the evaluation report, improving the evaluation efficiency and accuracy of the vehicle simulation test results, and the data is authentic and reliable, which can be applied to the simulation test and evaluation of unmanned vehicles in various scenarios, providing data support for unmanned vehicle research and development.

[0054] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0055] Obviously, those skilled in the art should understand that the modules or steps of the present invention described above can be implemented using a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices. Alternatively, they can be implemented using program codes executable by the computing device, so that they can be stored in a computer storage medium (ROM / RAM, magnetic disk, optical disk) and executed by the computing device. In some cases, the steps shown or described can be performed in a different order than herein, or they can be made into individual integrated circuit modules, or multiple modules or steps can be made into a single integrated circuit module for implementation. Therefore, the present invention is not limited to any specific combination of hardware and software.

[0056] The above content is a further detailed description of the present invention in conjunction with specific embodiments, and the specific implementation of the present invention cannot be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A self-driving car evaluation method based on simultaneous back-injection of simulation and measured data, characterized in that: The following steps are involved: According to the test standards and regulations of the test vehicle, build a simulation test environment and a real test environment under the same scenario; Conduct at least one project test in the built simulation test environment and actual measurement environment, and record the corresponding simulation data and actual measurement data; Processing the simulation data and the measured data to obtain target simulation data and target measured data, injecting the data into a simulation database and a measured database, respectively, and constructing a true value training model based on the target simulation data and the target measured data; The target simulation data and the target measured data are simultaneously injected into the true value training model, and the true value simulation model is obtained through fitting comparison and deduction, including: injecting the target simulation data and the target measured data into the true value training model at the same time; and establishing a high-order correspondence model between the target simulation data and the target measured data through fitting comparison, which is: ； Where, is the simulation data, is the measured data, 、 、 and Parameter matrix values; injecting multiple sets of target simulation data and target measured data into the high-order corresponding model to derive the parameter matrix values; obtaining a parameter matrix according to the parameter matrix values, and superimposing the parameter matrix on a basic simulation model to obtain a true value simulation model; Acquire sample simulation data and sample measured data under various scenarios, inject the data into the true value simulation model, and optimize the true value simulation model; Determine the evaluation dimensions of the test vehicle, and construct a test evaluation system model for the test vehicle based on the evaluation dimensions. The test evaluation system model is used to: obtain the evaluation dimensions of the test vehicle, which include safety, compliance, serviceability, and timeliness, and configure the weight corresponding to each evaluation dimension; determine the corresponding test items based on each dimension, which can be one or more test items; extract key technical indicators from the test items, and score them based on the test results of the key technical indicators; obtain the score of each evaluation dimension to obtain the total score of the test vehicle; and determine the overall condition of the test vehicle based on the range of the total score; The test evaluation system model and the optimized true value simulation model are combined to obtain a true value evaluation system, and an evaluation report of the test vehicle is generated based on the true value evaluation system.

2. The unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data according to claim 1 is characterized in that: According to the test vehicle test standards and regulations, the simulation test environment and the actual test environment under the same scenario are built, including: Build the corresponding testing environment according to the test standards and regulations of the test vehicle; Constructing a corresponding simulation test environment based on the actual test environment, wherein the simulation test environment and the actual test environment have the same scenario and both contain test elements and status of corresponding test items; Based on the parameters of the test vehicle, a dynamic model corresponding to the simulation test environment is set.

3. The unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data according to claim 1 is characterized in that: The simulation data corresponds to the measured data one by one, and both include vehicle position, speed, acceleration, heading angle, expected collision time, obstacle recognition trigger time, distance to the obstacle, deceleration, lateral angle, angular velocity, and lateral distance to the obstacle after circumventing the obstacle.

4. The unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data according to claim 2 is characterized in that: The data processing is performed on the simulation data and the measured data to obtain target simulation data and target measured data, and the target simulation data and target measured data are injected into the simulation database and the measured database respectively, and a true value training model is constructed according to the target simulation data and the target measured data, including: Under the same test conditions, multiple project tests are carried out in the constructed simulation environment and actual measurement environment to obtain the simulation data and actual measurement data; Matching the simulated data with the measured data one by one to obtain a plurality of data groups, and calculating corresponding data differences based on the data in the data groups; Obtaining a preset maximum deviation value, and when the data difference exceeds the maximum deviation value, eliminating the corresponding data group to obtain a cleaned data group; Performing median average filtering on the cleaned data group to obtain target simulation data and target measured data; A true value training model is constructed based on the target simulation data and the target measured data, and the target simulation data and the target measured data are respectively injected into a simulation database and a measured database.

5. The unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data according to claim 1 is characterized in that: The acquiring of sample simulation data and sample measured data in a variety of scenarios, injecting the data into the true value simulation model, and optimizing the true value simulation model includes: By conducting project tests in various scenarios, corresponding sample simulation data and sample measured data are obtained; The sample simulation data and the sample measured data are input into the true value simulation model, and the true value simulation model is trained and optimized to obtain an optimized true value simulation model.

6. The unmanned vehicle evaluation method based on simultaneous back-injection of simulation and measured data according to claim 1 is characterized in that: The test evaluation system model and the optimized true value simulation model are combined to obtain a true value evaluation system, and an evaluation report of the test vehicle is generated based on the true value evaluation system, including: Combining the test evaluation system model with the optimized true value simulation model to obtain a true value evaluation system; Acquiring simulation test data of a test vehicle and inputting the simulation test data into the true value evaluation system; An evaluation report of the test vehicle simulation test is obtained according to the true value evaluation system.

Images