CN114384483B - Assessment method for radar sensor model fidelity based on deep learning - Google Patents
Assessment method for radar sensor model fidelity based on deep learning Download PDFInfo
- Publication number
- CN114384483B CN114384483B CN202210020097.7A CN202210020097A CN114384483B CN 114384483 B CN114384483 B CN 114384483B CN 202210020097 A CN202210020097 A CN 202210020097A CN 114384483 B CN114384483 B CN 114384483B
- Authority
- CN
- China
- Prior art keywords
- data
- evaluation
- radar
- point cloud
- real
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/40—Means for monitoring or calibrating
- G01S7/4052—Means for monitoring or calibrating by simulation of echoes
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
- G01S13/931—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of land vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Networks & Wireless Communication (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Electromagnetism (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a radar sensor model fidelity assessment method based on deep learning, which comprises the following steps: obtaining Lei Dadian cloud data, wherein the radar point cloud data comprises real data and simulation data; performing similarity evaluation based on the real data and the simulation data to obtain an evaluation result; based on the evaluation result, the fidelity evaluation of the radar sensor model is realized. According to the method, the radar model is subjected to fidelity evaluation through the traditional index manually specified and the implicit measurement index based on deep neural network learning, the deep neural network is utilized to learn the characteristics of the radar point cloud, and the authenticity of the radar model can be comprehensively evaluated by combining the traditional index. And the method can be applied to radar sensor models of various types, so that the effectiveness of the automatic driving virtual test method can be evaluated, and the method has high economic and social benefits.
Description
Technical Field
The invention belongs to the field of radar fidelity assessment, and particularly relates to a radar sensor model fidelity assessment method based on deep learning.
Background
A Radar (Radar) irradiates a target by emitting electromagnetic waves and receives echoes thereof, thereby obtaining information of a distance, a distance change rate (radial velocity), an azimuth, an altitude, and the like of the target to an electromagnetic wave emission point. The current automatic driving virtual test method based on simulation is gradually widely applied, and the difference between simulation and reality is necessary to be quantified, so that whether the fidelity of the adopted sensor model meets the expected purpose or not is verified. The generation of the simulation data mainly comprises two steps, namely, performing simulation based on recorded ground real data, and generating a virtual scene of the environment from the angle of a sensor to obtain a simulated radar point cloud. There is currently no reliable method to measure radar sensor model fidelity, nor is there an adequate measure. The traditional measurement method mainly comprises an evaluation method of an original data level and an evaluation method of a detection level after forming a laser point cloud. The evaluation method of the original data level can only detect simple scenes and basic functions, the evaluation method of the detection level is mainly qualitative evaluation, quantitative evaluation is lacking, the evaluation index is mainly manually formulated explicit type, and measurement of implicit index is lacking.
Disclosure of Invention
In order to solve the problems, the invention provides the following scheme: a radar sensor model fidelity assessment method based on deep learning comprises the following steps:
obtaining Lei Dadian cloud data, wherein the radar point cloud data comprises real data and simulation data;
Performing similarity evaluation based on the real data and the simulation data to obtain an evaluation result;
based on the evaluation result, the fidelity evaluation of the radar sensor model is realized.
Preferably, the acquiring of the real data comprises,
Performing real driving based on a test scene, generating real radar point cloud information, and obtaining the real data;
The acquisition of the analog data may include,
Simulating based on the real data to obtain the simulation data; or generating a virtual scene based on the radar sensor to obtain simulated radar point cloud information; and obtaining the simulation data based on the radar point cloud information.
Preferably, the similarity evaluation based on the real data and the simulation data includes a conventional index evaluation and a depth index evaluation.
Preferably, the conventional index evaluation calculates the similarity of the real data and the analog data based on the difference in two-dimensional distance and doppler velocity.
Preferably, the evaluation index of the similarity at least comprises a distance between point clouds and a gas distance;
the distance between the point clouds is the normalized sum of the minimum Euclidean distance from the real point cloud to the simulated point cloud.
Preferably, the depth index evaluation includes that the real data and radar model data are mixed randomly to obtain a first data set; enhancing the first data set, and disturbing the first data set through random Gaussian noise based on the enhanced data set to obtain a second data set; and carrying out depth evaluation measurement on the point cloud data of the second data set based on PointNet ++ network model to obtain a measurement result.
Preferably, the point cloud data input to the PointNet ++ network model includes at least two spatial coordinates and doppler velocity.
Preferably, the depth index evaluation further comprises sampling by adopting a random repetition method under the condition of oversampling and adopting a drawing method under the condition of undersampling, so as to realize the fixed input point number of the point cloud.
Preferably, the depth evaluation metric evaluates the depth index based on a predicted confidence score of the real radar point cloud class.
Preferably, the evaluation method further comprises standardization of evaluation results, including z-score standardization after scaling of the measurement results, and spatial numerical mapping of fidelity evaluation by adjusting importance coefficients of the conventional index and the depth index.
The invention discloses the following technical effects:
According to the assessment method of the radar sensor model fidelity based on deep learning, the radar model is subjected to fidelity assessment through the traditional index manually specified and the implicit measurement index based on deep neural network learning from the detection level. According to the method, the characteristics of the radar point cloud are learned by using the deep neural network, and the authenticity of the radar model can be comprehensively evaluated by combining the traditional indexes. And the method can be applied to radar sensor models of various types, so that the effectiveness of the automatic driving virtual test method can be evaluated, and the method has high economic and social benefits.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a diagram of a training process according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating a test procedure according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
With the development of automatic driving test technology, virtual testing is becoming more and more important. The radar sensor model is an important component of the virtual test, and the authenticity of the radar sensor model is decisive for the reliability of the virtual test. In order to determine the reliability of the sensor model, it is necessary to detect the gap between the simulated data and the real data of the radar sensor model. There are many radar sensor simulation methods at present, but the problem of verifying and quantitatively evaluating the overall fidelity of radar models remains unsolved. Based on the above, the patent provides a multi-level evaluation method combining the traditional index and the hidden metric evaluation based on deep learning, which is used for carrying out overall quantitative evaluation on the fidelity of the radar sensor model.
As shown in fig. 1, the invention provides a method for evaluating the fidelity of a radar sensor model based on deep learning, which comprises the following steps:
obtaining Lei Dadian cloud data, wherein the radar point cloud data comprises real data and simulation data;
Performing similarity evaluation based on the real data and the simulation data to obtain an evaluation result;
based on the evaluation result, the fidelity evaluation of the radar sensor model is realized.
The acquisition of the real data comprises the steps of,
Performing real driving based on a test scene, generating real radar point cloud information, and obtaining the real data;
The acquisition of the analog data may include,
Simulating based on the real data to obtain the simulation data; or generating a virtual scene based on the radar sensor to obtain simulated radar point cloud information; and obtaining the simulation data based on the radar point cloud information.
And performing similarity evaluation based on the real data and the simulation data comprises traditional index evaluation and depth index evaluation.
The conventional index evaluation is to calculate the similarity of the real data and the analog data based on the difference in the two-dimensional distance and the Doppler velocity.
The evaluation index of the similarity at least comprises the distance between point clouds and the distance between the points;
the distance between the point clouds is the normalized sum of the minimum Euclidean distance from the real point cloud to the simulated point cloud.
The depth index evaluation comprises the steps of randomly mixing the real data with radar model data to obtain a first data set; enhancing the first data set, and disturbing the first data set through random Gaussian noise based on the enhanced data set to obtain a second data set; and carrying out depth evaluation measurement on the point cloud data of the second data set based on PointNet ++ network model to obtain a measurement result.
The point cloud data input into the PointNet ++ network model at least includes two spatial coordinates and doppler velocity.
The depth index evaluation further comprises the steps of adopting a random repetition method under the condition of oversampling, and adopting a drawing method to sample under the condition of undersampling, so that the input point number of the point cloud is fixed.
And the depth evaluation measurement is used for evaluating the depth index based on the prediction confidence score of the real radar point cloud.
The evaluation method further comprises the step of normalizing the evaluation result, wherein the step of performing z-score normalization after scaling the measurement result, and the step of performing space numerical mapping of fidelity evaluation by adjusting importance coefficients of the traditional index and the depth index.
Example 1
Further, the evaluation method of the radar sensor model fidelity based on the deep learning provided by the invention comprises the following steps:
Step one, selecting a test scene, such as a front vehicle lane change cut scene.
Step two: and carrying out real driving and generating real radar point cloud information.
Step three: and simulating driving and generating radar point cloud data by a simulation method to be evaluated.
Step four: the conventional index of the detection level is evaluated.
The similarity of the two-dimensional distance and the Doppler speed of the real and simulated radar detection data is compared, and two evaluation indexes are the distance Dpp between point clouds and the Neisseria distance EMD respectively.
The method evaluates the fidelity of the radar model through two indexes, namely traditional indexes and depth indexes, from the detection level.
Evaluation of traditional indexes:
The traditional index evaluation calculates the similarity between a real radar point cloud and a simulated radar point cloud through the difference of two components of a two-dimensional distance and a Doppler speed, and mainly comprises two calculation metrics.
The first metric is the normalized sum of the minimum Euclidean distances of the real point cloud to the simulated point cloud, represented by Dpp (Point cloud to point cloud distance). The real point cloud is represented by x= (X 1,x2,...,xM), the simulated point cloud by y= (Y 1,y2,...,yN), and X m,yn e R3 are three-dimensional points. Since Dpp is asymmetric, division by the respective points is standardized and a worst case scenario is assumed.
The second metric is the Neisserian distance (WASSERSTEIN DISTANCE), also known as the Earth Mover distance (EMD, earth Mover' S DISTANCE), which is used to compare the point distributions of a real radar point cloud to a simulated radar point cloud, determined by the optimal cost of rearranging one distribution to another. In addition to the three-dimensional point clouds X and Y, m and n describe the number of points in the point set, and the solution of the transmission problem between the two point cloud distributions is represented by the optimal flow f m,n, and the euclidean distance is selected as the ground distance d m,n. EMD naturally extends the concept of distance between single points to the distance between point distributions.
Step five: a PointNet ++ network model was trained and evaluated from depth indices. The training data is processed in the following way: and (3) carrying out random mixing on the real data and the data of multiple radar models, and enhancing a data set to avoid model overfitting. And testing the point cloud data generated by the specific simulation method by using the trained classifier model, and taking the prediction confidence score of the 'real radar point cloud' class as a Depth Evaluation Measure (DEM).
And (3) evaluating depth indexes:
Since conventional evaluation methods rely on manually customized metrics to evaluate a particular feature, some implicit features are not considered and the comprehensiveness of the metrics by manually defining the metrics cannot be guaranteed. To solve the problem of correct metrology, depth index evaluation classifies real and simulated radar data by training a neural network. Compared with the traditional index evaluation, the method aims at learning and distinguishing potential characteristics of real and simulated radar point clouds. The metric is a predictive confidence score for the "real radar point cloud" class of classifier. The network architecture uses a hierarchical neural network PointNet ++ architecture that enables direct use of point clouds and learning of local features.
In order to simplify the data acquisition, the invention only considers radar monitoring around the target vehicle, namely, the sensor data are recorded in an empty test scene, and the real driving conditions are reproduced in the simulation to generate simulated radar data. When the radar model contains random components to approximate true, averaging can be performed multiple times to reduce the impact. The resulting data set is relatively balanced between the real point cloud and the simulated point cloud. The radar point cloud of the input network contains two spatial coordinates and Doppler velocity (Doppler velocity).
A PointNet ++ network model was trained and evaluated from depth indices. The training data is processed in the following way: and (3) carrying out random mixing on the real data and the data of multiple radar models, and enhancing a data set to avoid model overfitting. And testing the point cloud data generated by the specific simulation method by using the trained classifier model, and taking the prediction confidence score of the 'real radar point cloud' class as a Depth Evaluation Measure (DEM).
Training and testing. In training the model, it is first necessary to randomly mix the real data with the data of various radar models. The data set then needs to be enhanced to avoid model overfitting. The data set is perturbed with random gaussian noise with a mean value of 0 and a standard deviation of 0.1, and the spatial coordinates and doppler velocity are changed. Meanwhile, in order to ensure that the input point number of each point cloud is fixed, a random repetition method is adopted under the condition of oversampling, and a drawing method is adopted under the condition of undersampling for sampling. The entire dataset was randomly split into 7: training and test set for 3 scale. The simulated radar data is classified by a classifier obtained through training, and the prediction confidence score of the 'real radar point cloud' class is a depth evaluation metric (DEM, deep evaluation metric). The process block diagrams of training and testing are shown in fig. 2-3.
Step six: each metric is normalized. The units of the traditional index and the depth index which are directly calculated are different, scaling is needed, and after the z-score standardization is carried out on the result, the numerical value is mapped in a 0-1 space, and the importance coefficients of the traditional index and the depth index can be adjusted according to the needs.
According to the assessment method of the radar sensor model fidelity based on deep learning, the radar model is subjected to fidelity assessment through the traditional index manually specified and the implicit measurement index based on deep neural network learning from the detection level. According to the method, the characteristics of the radar point cloud are learned by using the deep neural network, and the authenticity of the radar model can be comprehensively evaluated by combining the traditional indexes. And the method can be applied to radar sensor models of various types, so that the effectiveness of the automatic driving virtual test method can be evaluated, and the method has high economic and social benefits.
The above embodiments are only illustrative of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope defined by the claims of the present invention without departing from the design spirit of the present invention.
Claims (7)
1. A method for evaluating the fidelity of a radar sensor model based on deep learning, which is characterized by comprising the following steps:
obtaining Lei Dadian cloud data, wherein the radar point cloud data comprises real data and simulation data;
Performing similarity evaluation based on the real data and the simulation data to obtain an evaluation result;
Based on the evaluation result, realizing the fidelity evaluation of the radar sensor model;
Performing similarity evaluation based on the real data and the simulation data, wherein the similarity evaluation comprises traditional index evaluation and depth index evaluation;
The conventional index evaluation is used for calculating the similarity of the real data and the simulation data based on the difference of the two-dimensional distance and the Doppler speed;
The depth index evaluation comprises the steps of randomly mixing the real data with radar model data to obtain a first data set; enhancing the first data set, and disturbing the first data set through random Gaussian noise based on the enhanced data set to obtain a second data set; and carrying out depth evaluation measurement on the point cloud data of the second data set based on PointNet ++ network model to obtain a measurement result.
2. The method for evaluating the fidelity of a deep learning-based radar sensor model according to claim 1, wherein,
The acquisition of the real data comprises the steps of,
Performing real driving based on a test scene, generating real radar point cloud information, and obtaining the real data;
The acquisition of the analog data may include,
Simulating based on the real data to obtain the simulation data; or generating a virtual scene based on the radar sensor to obtain simulated radar point cloud information; and obtaining the simulation data based on the radar point cloud information.
3. The method for evaluating the fidelity of the radar sensor model based on the deep learning according to claim 1, wherein the evaluation index of the similarity at least comprises the distance between point clouds and the distance between the points of the radar sensor model;
the distance between the point clouds is the normalized sum of the minimum Euclidean distance from the real point cloud to the simulated point cloud.
4. The method for evaluating the fidelity of a deep learning based radar sensor model of claim 1, wherein the point cloud data input to the PointNet ++ network model comprises at least two spatial coordinates and doppler velocity.
5. The method for evaluating the fidelity of the radar sensor model based on the deep learning according to claim 1, wherein the depth index evaluation further comprises the steps of adopting a random repetition method in the case of oversampling and adopting a drawing method to sample in the case of undersampling, so as to realize the fixation of the input points of the point cloud.
6. The depth learning-based radar sensor model fidelity assessment method of claim 1, wherein the depth evaluation metric performs depth index evaluation based on a predictive confidence score of a real radar point cloud.
7. The method for evaluating the fidelity of a deep learning-based radar sensor model according to claim 1, wherein the evaluating method further comprises the step of normalizing the evaluation result, wherein the step of scaling the measurement result and then performing z-score normalization is performed, and wherein the spatial numerical mapping of the fidelity evaluation is performed by adjusting the importance coefficients of the conventional index and the depth index.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210020097.7A CN114384483B (en) | 2022-01-10 | 2022-01-10 | Assessment method for radar sensor model fidelity based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210020097.7A CN114384483B (en) | 2022-01-10 | 2022-01-10 | Assessment method for radar sensor model fidelity based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114384483A CN114384483A (en) | 2022-04-22 |
| CN114384483B true CN114384483B (en) | 2024-07-02 |
Family
ID=81199676
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210020097.7A Active CN114384483B (en) | 2022-01-10 | 2022-01-10 | Assessment method for radar sensor model fidelity based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114384483B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119339187A (en) * | 2024-12-23 | 2025-01-21 | 光轮智能(北京)科技有限公司 | Authenticity evaluation method, equipment and medium for 3D digital assets |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108107413A (en) * | 2018-01-09 | 2018-06-01 | 中国空空导弹研究院 | A kind of radar simulator calibration system |
| CN110246112A (en) * | 2019-01-21 | 2019-09-17 | 厦门大学 | Three-dimensional point cloud quality evaluating method in the room laser scanning SLAM based on deep learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| SE1551370A1 (en) * | 2015-10-22 | 2017-02-07 | Uniquesec Ab | Testing method with virtual radar signatures for an automotive safety radar system |
-
2022
- 2022-01-10 CN CN202210020097.7A patent/CN114384483B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108107413A (en) * | 2018-01-09 | 2018-06-01 | 中国空空导弹研究院 | A kind of radar simulator calibration system |
| CN110246112A (en) * | 2019-01-21 | 2019-09-17 | 厦门大学 | Three-dimensional point cloud quality evaluating method in the room laser scanning SLAM based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114384483A (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hanke et al. | Generation and validation of virtual point cloud data for automated driving systems | |
| CN111369003B (en) | A method and device for determining the fidelity of a qubit read signal | |
| CN112257605B (en) | 3D target detection method, system and device based on self-labeled training samples | |
| CN113486961A (en) | Radar RD image target detection method and system based on deep learning under low signal-to-noise ratio and computer equipment | |
| US20200065664A1 (en) | System and method of measuring the robustness of a deep neural network | |
| Ngo et al. | A multi-layered approach for measuring the simulation-to-reality gap of radar perception for autonomous driving | |
| CN113204903A (en) | Method for predicting thunder and lightning | |
| Jiang et al. | Efficient and unbiased safety test for autonomous driving systems | |
| Rajasekaran et al. | Ptrm: Perceived terrain realism metric | |
| Ghavidel et al. | Sonar data classification by using few-shot learning and concept extraction | |
| CN114384483B (en) | Assessment method for radar sensor model fidelity based on deep learning | |
| CN117371200A (en) | Parameter sensitivity evaluation method for millimeter wave radar simulation model | |
| CN112348030A (en) | Solar photovoltaic panel identification method and device, electronic equipment and storage medium | |
| CN116994084A (en) | Regional intrusion detection model training method and regional intrusion detection method | |
| CN116958468A (en) | Mountain snow environment simulation method and system based on SCycleGAN | |
| Zeng et al. | ATGAN: A SAR target image generation method for automatic target recognition | |
| Lambertenghi et al. | Assessing quality metrics for neural reality gap input mitigation in autonomous driving testing | |
| Müller et al. | Robustness evaluation and improvement for vision-based advanced driver assistance systems | |
| Ngo et al. | Deep evaluation metric: Learning to evaluate simulated radar point clouds for virtual testing of autonomous driving | |
| CN116263735A (en) | Robustness assessment method, device, equipment and storage medium for neural network | |
| Wang et al. | Property existence inference against generative models | |
| CN114298299A (en) | Model training method, device, equipment and storage medium based on course learning | |
| CN116449807B (en) | Simulation test method and system for automobile control system of Internet of things | |
| KR102558609B1 (en) | Method for evaluating wind speed patterns to ensure structural integrity of buildings, and computing apparatus for performing the method | |
| CN117274649A (en) | Radar target cluster recognition method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |