Multi-mode composite physical simulation method and system based on feature analysis

Through the feature analysis method, the environmental data of intelligent vehicle sensors is integrated, and the problem of insufficient environmental information in multi-mode composite real-world simulation test is solved, the accuracy and stability of the test are improved, and it is suitable for complex environmental simulation of intelligent driving vehicles.

CN118689129BActive Publication Date: 2024-11-05XIAN GAOSHANG INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411186245.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-11-05
Estimated Expiration
2044-08-28

AI Technical Summary

Technical Problem

In multi-mode composite real object simulation test, the environmental information in actual on-site or virtual arrangement cannot meet the testing requirements of intelligent driving vehicles, resulting in low accuracy of simulation tests.

Method used

Using a feature analysis-based method, by obtaining the environmental data of the sensors over different time periods, building feature vectors, and fusion of data based on the differences and similarity of the feature vectors is generated to improve the accuracy of the test.

Benefits of technology

It improves the accuracy and stability of simulation tests of smart vehicles in complex environments, and enhances the consistency and effectiveness of sensor data in different scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118689129B_ABST
    Figure CN118689129B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of vehicle testing technology, and in particular to a multi-mode composite physical simulation method and system based on feature analysis. The method obtains each feature performance index, and combines the similarity of the difference between the feature vectors of the sensor at each scene moment and the other scene moments to obtain a comprehensive scene association index of the feature vector of each sensor at each scene moment; combines the similarity of the feature vectors of each sensor at any two scene moments with the comprehensive scene association index, performs data fusion on the feature vector to obtain fused data, and uses it to perform simulation testing on the physical object to be tested. The present invention uses fused data for simulation testing to reduce redundant data in environmental data, makes simulation testing closer to actual conditions, and improves the authenticity of multi-mode composite physical simulation testing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle testing, and in particular to a multi-mode composite physical simulation method and system based on feature analysis. Background Art

[0002] Multi-mode composite physical simulation is an important test method for intelligent driving technology testing. Through simulation testing, the test cost can be reduced and the test safety can be guaranteed. Intelligent driving vehicles will face various complex environments, and the car needs to quickly handle complex environments during testing. In the process of simulation testing of composite physical objects, various simulation signals are simulated and input to the sensor equipment in the physical object, and the intelligent vehicle is debugged based on the vehicle's response.

[0003] When conducting simulation tests on multi-mode composite objects, simulated data is transferred to physical devices for simulation testing; because smart driving vehicles need to face various emergencies during actual driving, various complex environments need to be set up during simulation tests to test the performance of the vehicle. However, the environmental information that can be arranged on site is relatively simple, and arranging it through simulating virtual data may cause distortion of the simulated data. Both situations cannot meet the test requirements of smart vehicles, resulting in low accuracy of simulation tests on multi-mode composite objects. Summary of the invention

[0004] In order to solve the technical problem that the environmental information of the actual site or the virtual arrangement does not meet the test requirements of the physical object to be tested, resulting in low accuracy of simulation test of the physical object to be tested, the purpose of the present invention is to provide a multi-mode composite physical simulation method and system based on feature analysis, and the technical scheme adopted is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a multi-mode composite physical simulation method based on feature analysis, the method comprising:

[0006] Obtain the environmental data of each sensor installed on the object to be tested at each moment in a set time period; divide the set time period into different local time periods, record the last moment in each local time period as the scene moment, and use the environmental data of each sensor in each local time period to form a feature vector of each sensor at the scene moment corresponding to each local time period;

[0007] According to the difference between the feature vectors of each sensor and all sensors at each scene moment, the feature performance index of the feature vector of each sensor at each scene moment is obtained;

[0008] According to the similarity of the difference between the feature vectors of each sensor and the other sensors at each scene moment and the other scene moments, and the feature performance index of each sensor at each scene moment, a comprehensive scene association index of the feature vector of each sensor at each scene moment is obtained;

[0009] According to the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, the feature vectors are fused to obtain fused data of each sensor at any two scene moments; and the fused data is used to perform simulation test on the object to be tested.

[0010] Furthermore, the method for obtaining the feature performance index of the feature vector of each sensor at each scene moment includes:

[0011] The ratio of the distance between the feature vector of each sensor at each scene moment and the mean of the feature vectors of all sensors at each scene moment to the variance of the elements in the feature vectors of all sensors at each scene moment is used as the variance contribution index of the feature vector of each sensor at each scene moment;

[0012] Obtain similarity indexes of any two feature vectors; select the maximum value of the similarity indexes of the feature vectors of each sensor and the other sensors at each scene moment, and record it as the redundancy index of the feature vector of each sensor at each scene moment;

[0013] A feature performance index of a feature vector of each sensor at each scene moment is obtained according to the variance contribution index and the redundancy index.

[0014] Furthermore, the method for obtaining the comprehensive scene association index of the feature vector of each sensor at each scene moment includes:

[0015] Any two scene moments are recorded as analysis moments, any one sensor is recorded as a target sensor, and the analysis sensor of the target sensor at the two analysis moments is selected from the remaining sensors except the target sensor; the similarity index of the feature vectors of the target sensor and the analysis sensor at each analysis moment is greater than a preset similarity threshold;

[0016] For each of the analysis sensors, a negative correlation mapping is performed on the difference in the distance between the feature vectors of the target sensor and the analysis sensor at the two analysis moments to obtain a local difference index between the feature vectors of the target sensor and the analysis sensor at the two analysis moments; the cumulative sum of the local difference indexes of the target sensor and all the analysis sensors at the two analysis moments is recorded as a comprehensive difference index of the feature vectors of the target sensor at the two analysis moments;

[0017] One of the two analysis moments is recorded as the target moment, and the other analysis moment is recorded as the control moment, and a local scene association index of a feature vector of the target sensor at the target moment at the control moment is obtained according to the feature performance index of the target sensor at the target moment and the comprehensive difference index;

[0018] The mean values ​​of the local scene association indexes of the target sensor at the target moment at the remaining scene moments except the target moment are normalized to obtain a comprehensive scene association index of the feature vector of the target sensor at the target moment.

[0019] Furthermore, the method of performing data fusion on the feature vectors according to the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index to obtain fused data of each sensor at any two scene moments includes:

[0020] Obtaining a fusion judgment index of the feature vector of each sensor at the two analysis moments according to a difference between the similarity index of the feature vector of each sensor at the two analysis moments and the comprehensive scene association index;

[0021] If the fusion judgment index is greater than the preset fusion threshold, the comprehensive scene association index is used to perform weighted summation on the feature vectors of each sensor at the two analysis moments to obtain the fusion vector of each sensor at the two analysis moments;

[0022] If the fusion judgment index is less than or equal to the preset fusion threshold, then according to the feature performance index of each sensor at each analysis moment and the comprehensive scene association index, the fusion coefficient of the feature vector of each sensor at each analysis moment is obtained; the feature vectors of each sensor at two analysis moments are weightedly summed using the fusion coefficient to obtain the fusion vector of each sensor at the two analysis moments;

[0023] The fusion vector is recorded as the fusion data of each sensor at two analysis moments.

[0024] Furthermore, the method of obtaining a feature performance index of a feature vector of each sensor at each scene moment according to the variance contribution index and the redundancy index includes:

[0025] For the feature vector of each sensor at each scene moment, the product of the result of negative correlation mapping of the redundant index of the feature vector and the variance contribution index is normalized to obtain the feature performance index of the feature vector.

[0026] Furthermore, the method for obtaining the fusion judgment index of the feature vector of each sensor at two analysis moments includes:

[0027] The absolute value of the difference between the comprehensive scene association index of the feature vector of each sensor at two analysis moments and the product of the similarity index are normalized to obtain the fusion judgment index of the feature vector of each sensor at two analysis moments.

[0028] Furthermore, the similarity index between any two feature vectors is the cosine similarity between the corresponding two feature vectors.

[0029] Furthermore, the preset similarity threshold is 0.7.

[0030] Furthermore, the preset fusion threshold is 0.7.

[0031] In a second aspect, another embodiment of the present invention provides a multi-mode composite physical simulation system based on feature analysis, the system comprising:

[0032] The data acquisition module is used to obtain the environmental data of each sensor installed on the object to be tested at each moment in a set time period; the set time period is divided into different local time periods, the last moment in each local time period is recorded as the scene moment, and the environmental data of each sensor in each local time period constitutes the feature vector of each sensor at the scene moment corresponding to each local time period;

[0033] A feature performance analysis module is used to obtain a feature performance index of a feature vector of each sensor at each scene moment according to the difference between the feature vectors of each sensor and all sensors at each scene moment;

[0034] A scene association analysis module, used to obtain a comprehensive scene association index of the feature vector of each sensor at each scene moment according to the similarity of the difference between the feature vectors of each sensor and the other sensors at each scene moment and the feature performance index of each sensor at each scene moment;

[0035] The data fusion module is used to fuse the feature vectors according to the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, so as to obtain the fusion data of each sensor at any two scene moments; and use the fusion data to perform simulation test on the physical object to be tested.

[0036] The present invention has the following beneficial effects:

[0037] In an embodiment of the present invention, the feature performance index reflects the accuracy of the object to be tested in perceiving and understanding the surrounding environment. The difference between the feature vectors of each sensor and all sensors at each scene moment presents the significance of the feature vector of each sensor compared with the feature vectors of all sensors, thereby obtaining the feature performance index of the feature vector, and combining the similarity between the difference between the feature vectors of each sensor and the remaining sensors at each scene moment and the remaining scene moments, presents the consistency of changes in different feature vectors at each scene moment and the remaining scene moments, further analyzes the feature vector and the scene moment, and obtains a comprehensive scene association index, which is used to accurately measure the degree of association between the feature vector and the scene moment, which helps the object to be tested to understand and adapt to complex and changing environments; the similarity of the feature vectors of each sensor at two scene moments reflects the consistency of sensor data in different scenes, and combines the scene association index that measures the degree of matching between the feature vector and the scene to perform data fusion, fuses the features of different data into one scene, ensures the consistency and stability of sensor data in different scenes, effectively enhances the effectiveness and stability of sensor data in specific scenes, and uses the fused data to perform simulation tests on the object to be tested, which can improve the accuracy of simulation tests on multi-mode composite objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flowchart of a multi-mode composite physical simulation method based on feature analysis provided by an embodiment of the present invention;

[0040] Figure 2 A structural diagram of a multi-mode composite physical simulation system based on feature analysis provided by one embodiment of the present invention;

[0041] Figure 3 A schematic diagram of a computer device of a multi-mode composite physical simulation device based on feature analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a multi-mode composite physical simulation method and system based on feature analysis proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0044] The following is a detailed description of a method and system for multi-mode composite physical simulation based on feature analysis provided by the present invention in conjunction with the accompanying drawings.

[0045] Embodiment 1:

[0046] The present invention proposes a multi-mode composite physical simulation method based on feature analysis, please refer to Figure 1 , which shows a flowchart of a multi-mode composite physical simulation method based on feature analysis provided by an embodiment of the present invention, the method comprising:

[0047] Step S1: Obtain the environmental data of each sensor installed on the physical object to be tested at each moment within a set time period; divide the set time period into different local time periods, record the last moment in each local time period as the scene moment, and use the environmental data of each sensor in each local time period to form a feature vector of each sensor at the scene moment corresponding to each local time period.

[0048] This solution uses smart cars as the physical object to be tested for analysis; drones are used to collect video data of roads, and 3D modeling software such as Blender modeling software, SketchUp, etc. are used in combination with video data as a reference to create models of environmental elements such as roads, buildings, and traffic signs, and obtain physical models in real traffic scenarios.

[0049] Each sensor installed on the smart car collects environmental data at each moment in a set time period. A single piece of environmental data is not sufficient to reflect the environmental characteristics of the smart car during driving. The embodiment of the present invention evenly divides the set time period into multiple local time periods, and takes the last moment in each local time period as a scene, and records it as the scene moment. Since the environmental data collected by the sensor has redundant data, it is impossible to directly generate decision data. Generally, it is necessary to extract features from the environmental data collected by the sensor. This embodiment arranges the environmental data of each sensor at all moments in each local time period in time series to form a feature vector of each sensor at the scene moment corresponding to each local time period. The feature vector provides a data basis for the control system of the smart car to make decisions.

[0050] It should be noted that the duration of the set time period is 1 hour, the duration of the local time period is 1 minute, the data collection frequency of all sensors is once per second and the sampling time is the same, which can be set by the implementer according to the specific situation; the number of elements in all feature vectors is the same. The sensors installed on smart cars include: cameras, radar sensors and global positioning system (GPS), etc.

[0051] The simulation test process of smart cars is as follows: drones collect video data of the road and environmental data collected by sensors on the car, and computers are used to generate virtual traffic environment models to simulate real-world roads, traffic signs, other vehicles and pedestrians; the control system of the smart car (such as the adaptive headlight system controller, etc.) receives simulated sensor data and virtual environment information, makes decisions and generates control instructions; the dynamic response of the car is adjusted through control instructions, including acceleration, braking, steering and other behaviors, and the impact of these behaviors on the vehicle status is analyzed; control instructions are sent to the simulated vehicle model, affecting the vehicle's behavior in the virtual environment, forming a closed feedback loop.

[0052] This solution performs data fusion preprocessing on the environmental data collected by sensors before inputting them into the control system of the smart car, and uses the fused data for car decision-making.

[0053] Step S2: According to the difference between the feature vectors of each sensor and all sensors at each scene moment, the feature performance index of the feature vector of each sensor at each scene moment is obtained.

[0054] When intelligently controlling smart cars, it is necessary to obtain external environmental information through on-board sensors, and then extract feature information from the environmental information for decision-making. Since the environmental data collected by the sensors may contain redundant data, data fusion is performed before decision-making, and the features of different environmental data are integrated into one scene. The fused data is used for simulation testing to simulate and evaluate the performance and response capabilities of the automotive system in different scenes and situations, making the simulation test closer to the actual situation and improving the authenticity and fidelity of the simulation test.

[0055] The performance degree of feature information reflects the importance and influence of feature information on the decision-making process of smart cars, which can help smart cars perceive and understand the surrounding environment more accurately, such as identifying and classifying different types of road signs, identifying pedestrians and vehicles, etc. Therefore, it is necessary to analyze the performance degree of feature information to obtain feature performance indicators.

[0056] The difference between the feature vectors of each sensor and all sensors at each scene moment presents the significance of the feature vector of each sensor compared with the feature vectors of all sensors. If the above difference is larger, the feature vector of each sensor is easier to distinguish from the feature vectors of all sensors at that scene moment, and the feature information represented by the feature vector of each sensor is more obvious, thereby obtaining the feature performance index of the feature vector.

[0057] Preferably, in some possible implementation modes of the embodiments of the present invention, the method for obtaining the feature performance index of the feature vector includes: taking the ratio of the distance between the feature vector of each sensor and the mean of the feature vectors of all sensors at each scene moment, and the variance of the elements in the feature vectors of all sensors at each scene moment, as the variance contribution index of the feature vector of each sensor at each scene moment; obtaining the similarity index of any two feature vectors; selecting the maximum value of the similarity index between the feature vectors of each sensor and the remaining sensors at each scene moment, and recording it as the redundancy index of the feature vector of each sensor at each scene moment; and obtaining the feature performance index of the feature vector of each sensor at each scene moment according to the variance contribution index and the redundancy index.

[0058] It should be noted that the mean of the feature vectors of all sensors at each scene moment is still a vector, and the element of each dimension in the mean of the above vector is equal to the mean of the elements of the corresponding dimension in the feature vectors of all sensors at each scene moment.

[0059] This embodiment analyzes the degree of difference of feature vectors from two aspects, specifically: First, the concept of variance contribution index refers to variance contribution rate. Variance contribution index measures the contribution degree or proportion of the feature vector of each sensor in the overall variance of the feature vectors of all sensors at each scene moment. If the variance contribution index is larger, the difference between the feature vector of each sensor and the feature vectors of all sensors is larger, indicating that the feature vector of each sensor at each scene moment is more obvious in the feature vectors of all sensors. Second, in the embodiment of the present invention, the cosine similarity index between two vectors measures the degree of difference between the two vectors, and the cosine similarity between any two feature vectors is used as the similarity index of the corresponding two feature vectors; if the similarity index between the feature vectors of each sensor and all sensors at each scene moment is smaller, the difference between the feature vector of each sensor and all feature vectors is more obvious.

[0060] Therefore, the variance contribution index is positively correlated with the feature performance index, and the redundancy index is negatively correlated with the feature performance index. In an embodiment of the present invention, for the feature vector of each sensor at each scene moment, the result of negative correlation mapping of the redundant index of the feature vector is normalized by the product of the variance contribution index to obtain the feature performance index of the feature vector.

[0061] In the embodiment of the present invention, the correlation between the variance contribution index and the redundancy index and the feature performance index can also be constructed through other basic mathematical operations, which are not limited or elaborated here. It should be noted that the optional normalization methods in the embodiment of the present invention include: normalization methods such as Norm function and function transformation, which are not limited here.

[0062] In a specific implementation of the embodiment of the present invention, taking the bth feature vector at the ath scene moment in a set time period as an example, the feature performance index is expressed by the formula:

[0063]

[0064] In the formula, is the characteristic performance index of the feature vector of the bth sensor at the ath scene moment; is the feature vector of the bth sensor at the ath scene moment; is the mean of the feature vectors of all sensors at the a-th scene moment; is the variance of the elements in the feature vector of all sensors at the a-th scene moment; is the redundancy index of the feature vector of the bth sensor at the ath scene moment, that is, the maximum value of the similarity indexes of the feature vectors of the bth sensor and the other sensors at the ath scene moment; is the variance contribution index of the feature vector of the bth sensor at the ath scene moment; Norm is the normalization function; is the symbol of the two-norm; It is a preset positive number, and the empirical value is 0.01, which is used to prevent the denominator from being zero, resulting in meaningless fractions. In the embodiment of the present invention, negative correlation mapping is achieved by taking the reciprocal of the redundant index. In other embodiments, negative correlation mapping can also be performed using an exponential function with a natural constant as the base.

[0065] It should be noted that Measures the overall variance of the feature vector of the bth sensor in the feature vector of all sensors at the ath scene moment The contribution of The bigger, The more obvious the characteristic information represented is in the characteristic vectors of all sensors, the higher the characteristic performance index The larger the The bigger, The more similar it is to the feature information represented by all feature vectors, The less obvious the characteristic information represented, the higher the characteristic performance index The smaller.

[0066] Step S3: According to the similarity between the difference between the feature vectors of each sensor and the other sensors at each scene moment and the feature performance index of each sensor at each scene moment, a comprehensive scene association index of the feature vector of each sensor at each scene moment is obtained.

[0067] Smart cars face different requirements in different scenarios, and specific feature information (such as pedestrians, obstacles, road conditions, etc.) may have different importance and influence in different scenarios. Analyzing the relationship between feature information and scenarios is conducive to dynamically adjusting the decision-making strategy of smart cars during operation to cope with various real-time changes and situations. Therefore, it is necessary to analyze the correspondence between feature information and different scenarios to help smart cars understand and adapt to complex and changing environments.

[0068] The feature performance index reflects the particularity of the feature information represented by the feature vector in the scene. The larger the feature performance index, the more particular the feature information represented by the feature vector is at the scene moment, and the closer the association between the feature vector and the scene moment is. The difference between the feature vectors of each sensor and the other sensors at each scene moment reflects the degree of change of the feature information represented by each feature vector and the other feature vectors at each scene moment. The more similar or consistent the above degree of change between each scene moment and the other scene moments is, the closer the association between the feature vector of each sensor and the corresponding scene moment is. Therefore, the above two factors are combined to analyze the degree of association between the feature vector and the scene moment to obtain a comprehensive scene association index.

[0069] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the comprehensive scene association index of the feature vector includes: randomly selecting two scene moments as analysis moments, randomly selecting one sensor as the target sensor, and selecting the analysis sensor of the target sensor at the two analysis moments from the remaining sensors except the target sensor; the similarity index of the feature vectors of the target sensor and the analysis sensor at the two analysis moments is greater than a preset similarity threshold; for each analysis sensor, negatively correlating the difference in the distance between the feature vectors of the target sensor and the analysis sensor at the two analysis moments to obtain a local difference index of the feature vectors of the target sensor and the analysis sensor at the two analysis moments ; The cumulative sum of the local difference indicators of the target sensor at the two analysis moments and all the analysis sensors is recorded as the comprehensive difference indicator of the feature vector of the target sensor at the two analysis moments; One of the two analysis moments is recorded as the target moment, and the other analysis moment is recorded as the control moment. According to the characteristic performance indicator and the comprehensive difference indicator of the target sensor at the target moment, the local scene association indicator of the feature vector of the target sensor at the control moment is obtained; The mean of the local scene association indicators of the target sensor at the target moment at the remaining scene moments except the target moment is normalized as the comprehensive scene association indicator of the feature vector of the target sensor at the target moment.

[0070] In order to reduce the amount of calculation of the degree of association between feature information and different scenes, the feature vector with a greater degree of redundancy with the feature vector of the target sensor is selected, namely the feature vector of the analysis sensor. The analysis sensor is specifically: for each analysis moment, the similarity index of the feature vectors of the target sensor and the analysis sensor; the above similarity index at the two analysis moments is greater than the preset similarity threshold. It should be noted that the preset similarity index takes the empirical value of 0.7, and the implementer can set it according to the specific situation.

[0071] In a specific implementation of the embodiment of the present invention, taking the bth feature vector at the ath scene moment in a set time period as an example, the comprehensive scene association index is expressed by a formula:

[0072]

[0073]

[0074] In the formula, is the comprehensive scene association index of the feature vector of the bth sensor at the ath scene moment; C is the total number of scene moments in the set time period; is the characteristic performance index of the feature vector of the bth sensor at the ath scene moment; is the comprehensive difference index of the feature vector of the bth sensor at the ath scene moment and the cth scene moment; is the total number of analytical sensors to be tested, and the analytical sensors to be tested are the analytical sensors of the bth sensor at the ath scene moment at the ath scene moment and at the cth scene moment; is the feature vector of the bth sensor at the ath scene moment; is the feature vector of the nth sensor to be tested and analyzed at the ath scene moment; is the feature vector of the bth sensor at the cth scene moment; is the feature vector of the nth sensor to be tested and analyzed at the cth scene moment; is the local difference index of the feature vectors of the bth sensor and the nth sensor to be tested and analyzed at the ath scene moment and the cth scene moment; is the local scene association index of the bth feature vector at the ath target moment at the cth scene moment; Norm is the normalization function; is the symbol of the two-norm; exp is the exponential function with the natural constant e as the base.

[0075] It should be noted that when When the value is larger, the feature information represented by the feature vector of the bth sensor at the ath scene moment is more special in the scene moment, and the correlation between the feature vector and the ath scene moment is closer, then the comprehensive scene correlation index The larger the When the value is larger, the difference between the feature vectors of the bth sensor and all the sensors to be tested and analyzed at the ath scene moment and the cth scene moment is smaller, and the change of the feature vectors of the bth sensor and all the sensors to be tested and analyzed is more consistent. Compared with the cth scene moment, the feature vector of the bth sensor at the ath scene moment is more closely associated with the ath scene moment, and the comprehensive scene association index is The bigger.

[0076] The method for obtaining the comprehensive scene correlation index of the feature vector of each sensor at any two scene moments is the same as the method for obtaining the comprehensive scene correlation index of the feature vector of each sensor at two analysis moments.

[0077] Step S4: Based on the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, the feature vectors are fused to obtain the fused data of each sensor at any two scene moments; and the fused data is used to perform simulation tests on the object to be tested.

[0078] The similarity of the feature vectors of each sensor at any two scenes reflects the consistency of the sensor data in different scenes. If the feature vectors are more similar in two scenes, the measurement results of the sensor will not change much in different scenes, and the data stability is higher. The comprehensive scene association index presents the matching degree between the feature vector and the scene. If the comprehensive scene association index is larger, the feature vector can effectively capture the important features of the scene. The above factors are analyzed to obtain the fusion judgment index, which is used to weigh the data stability and scene matching degree in the data fusion process.

[0079] According to the difference between the similarity index of the feature vector of each sensor at the two analysis moments and the comprehensive scene association index, the fusion judgment index of the feature vector of each sensor at the two analysis moments is obtained. Preferably, in the embodiment of the present invention, the product of the difference absolute value of the comprehensive scene association index of the feature vector of each sensor at the two analysis moments and the similarity index is normalized to obtain the fusion judgment index of the feature vector of each sensor at the two analysis moments.

[0080] In a specific implementation of the embodiment of the present invention, taking the feature vector of the bth sensor at the ath scene moment and the eth scene moment as an example, the fusion judgment index is expressed by the formula:

[0081]

[0082] In the formula, is the fusion judgment index of the feature vector of the bth sensor at the ath scene moment and the eth scene moment; is the similarity index of the feature vector of the bth sensor at the ath scene moment and the eth scene moment; is the absolute value of the difference between the comprehensive scene association index of the feature vector of the bth sensor at the ath scene moment and the eth scene moment; Norm is the normalization function.

[0083] If the fusion judgment index is larger, although the feature vectors are more similar in different scenarios, the correlation degree between the scene and the sensor feature vector is more different. In order to ensure the consistency and stability of sensor data in different scenarios, the comprehensive scene correlation index is used as the weight for data fusion, which can help reduce the excessive impact on scenes with large data changes, thereby ensuring the stability and reliability of the test results; if the fusion judgment index is smaller, the performance of the feature vectors in different scenarios is close, and the correlation degree between the scene and the sensor is high. The fusion weight obtained by combining the feature performance index with the comprehensive scene correlation index analysis can more accurately reflect the effectiveness and adaptability of sensor data in specific scenarios, thereby improving the comprehensive effect after data fusion.

[0084] If the fusion judgment index is greater than the preset fusion threshold, the comprehensive scene association index is used to perform weighted summation on the feature vectors of each sensor at the two analysis moments to obtain the fusion vector of each sensor at the two analysis moments; if the fusion judgment index is less than or equal to the preset fusion threshold, in the embodiment of the present invention, the product of the feature performance index of each sensor at each analysis moment and the comprehensive scene association index is normalized to obtain the fusion coefficient of each sensor at each analysis moment, and the fusion coefficient is used to perform weighted summation on the feature vectors of each sensor at the two analysis moments to obtain the fusion vector of each sensor at the two analysis moments; the fusion vector is recorded as the fusion data of each sensor at the two analysis moments. It should be noted that in the embodiment of the present invention, the preset fusion threshold is an empirical value of 0.7, and the implementer can set it according to the specific situation.

[0085] The method of weighting the two vectors is as follows: for the feature vector of each sensor at the two analysis moments, all elements in the feature vector of the sensor are weighted according to the fusion coefficient of each sensor at each analysis moment to obtain the weighted feature vector, and then the sum of the elements of the same dimension in the two weighted feature vectors constitutes the fusion vector of each sensor at the two analysis moments. The features of different data are fused into one scene to ensure the validity and stability of sensor data in different scenes.

[0086] According to the method of data fusion of the feature vectors of each sensor at two analysis moments, data fusion is performed on the feature vectors of each sensor at each scene moment, and the data fusion methods of the two are the same.

[0087] The dynamic data of the virtual vehicle, including vehicle weight, tire friction coefficient, suspension system characteristics, etc., as well as the vehicle information in the virtual model, are obtained, and the virtual environment data obtained above, i.e., the fusion data of each sensor at the time of pairwise analysis, is input into the smart terminal in the physical object to be tested, which effectively enhances the effectiveness and adaptability of sensor data in specific scenarios, and improves the accuracy of simulation testing of multi-mode composite physical objects; the smart terminal generates control instructions after analysis, and sends the control instructions to the simulation model, and observes vehicle behavior and algorithm performance by constructing vehicle dynamics, which is closer to the actual situation, improves the understanding of driving scenarios and decision-making capabilities, and thus increases the accuracy of simulation testing of multi-mode composite physical objects. The specific steps for simulation testing of physical objects are well known to those skilled in the art and will not be introduced in detail here.

[0088] So far, the present invention is completed.

[0089] In summary, in an embodiment of the present invention, each feature performance index is obtained, and combined with the similarity of the difference between the feature vectors of the sensor at each scene moment and the other scene moments, a comprehensive scene association index of the feature vector of each sensor at each scene moment is obtained; combined with the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, the feature vectors are data fused to obtain fused data, and the fused data is used to perform simulation tests on the physical object to be tested. The present invention uses fused data for simulation tests to reduce redundant data in environmental data, make simulation tests closer to actual conditions, and improve the authenticity of multi-mode composite physical simulation tests.

[0090] Embodiment 2:

[0091] The present invention proposes a multi-mode composite physical simulation system based on feature analysis, please refer to Figure 2 , which shows a structure diagram of a multi-mode composite physical simulation system based on feature analysis provided by an embodiment of the present invention, the system comprising:

[0092] The data acquisition module 510 is used to obtain the environmental data of each sensor installed on the physical object to be tested at each moment in a set time period; the set time period is divided into different local time periods, and the last moment in each local time period is recorded as a scene moment, and the environmental data of each sensor in each local time period constitutes a feature vector of each sensor at the scene moment corresponding to each local time period;

[0093] A feature performance analysis module 520 is used to obtain a feature performance index of the feature vector of each sensor at each scene moment according to the difference between the feature vectors of each sensor and all sensors at each scene moment;

[0094] A scene association analysis module 530, configured to obtain a comprehensive scene association index of the feature vector of each sensor at each scene moment according to the similarity of the difference between the feature vectors of each sensor and the other sensors at each scene moment and the feature performance index of each sensor at each scene moment;

[0095] The data fusion module 540 is used to fuse the feature vectors according to the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, so as to obtain the fusion data of each sensor at any two scene moments; and use the fusion data to perform simulation test on the physical object to be tested.

[0096] It should be noted that: the device provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the multi-mode composite physical simulation system based on feature analysis and the multi-mode composite physical simulation method embodiment based on feature analysis provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0097] Embodiment 3:

[0098] Figure 3 A computer device schematic diagram of a multi-mode composite physical simulation device based on feature analysis provided by an embodiment of the present invention. Figure 3 As shown, the computer device includes: a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and running on the processor 602, wherein when the processor 602 executes the computer program 603, the computer device can execute any one of the multi-mode composite physical simulation methods based on feature analysis introduced above.

[0099] In addition, an embodiment of the present application also protects a device, which may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to execute a multi-mode composite physical simulation method based on feature analysis provided in an embodiment of the present application.

[0100] In this embodiment, the functional modules of the device can be divided according to the above method example. For example, each functional module can be corresponded, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0101] It should be understood that the device provided in this embodiment is used to execute the above-mentioned multi-mode composite physical simulation method based on feature analysis, and thus can achieve the same effect as the above-mentioned implementation method.

[0102] In the case of an integrated unit, the device may include a processing module and a storage module. When the device is applied to a device, the processing module may be used to control and manage the actions of the device. The storage module may be used to support the device to execute mutual program codes, etc. Mutual program codes refer to program codes that can communicate and execute between different devices or systems.

[0103] The processing module may be a processor or a controller, which may implement or execute various exemplary logic blocks, modules and circuits included in the disclosure of this application. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc. The storage module may be a memory.

[0104] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-mode composite physical simulation method based on feature analysis, characterized in that: The method includes: Obtain the environmental data of each sensor installed on the object to be tested at each moment in a set time period; divide the set time period into different local time periods, record the last moment in each local time period as the scene moment, and use the environmental data of each sensor in each local time period to form a feature vector of each sensor at the scene moment corresponding to each local time period; According to the difference between the feature vectors of each sensor and all sensors at each scene moment, the feature performance index of the feature vector of each sensor at each scene moment is obtained; According to the similarity of the difference between the feature vectors of each sensor and the other sensors at each scene moment and the other scene moments, and the feature performance index of each sensor at each scene moment, a comprehensive scene association index of the feature vector of each sensor at each scene moment is obtained; According to the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, the feature vectors are data-fused to obtain fused data of each sensor at any two scene moments; and the fused data is used to perform simulation tests on the object to be tested; The method for obtaining the feature performance index of the feature vector of each sensor at each scene moment includes: The ratio of the distance between the feature vector of each sensor at each scene moment and the mean of the feature vectors of all sensors at each scene moment to the variance of the elements in the feature vectors of all sensors at each scene moment is used as the variance contribution index of the feature vector of each sensor at each scene moment; Obtain similarity indexes of any two feature vectors; select the maximum value of the similarity indexes of the feature vectors of each sensor and the other sensors at each scene moment, and record it as the redundancy index of the feature vector of each sensor at each scene moment; Acquire a feature performance index of a feature vector of each sensor at each scene moment according to the variance contribution index and the redundancy index; The method for obtaining the comprehensive scene association index of each sensor at each scene moment includes: Any two scene moments are recorded as analysis moments, any one sensor is recorded as a target sensor, and the analysis sensor of the target sensor at the two analysis moments is selected from the remaining sensors except the target sensor; the similarity index of the feature vectors of the target sensor and the analysis sensor at each analysis moment is greater than a preset similarity threshold; For each of the analysis sensors, a negative correlation mapping is performed on the difference in the distance between the feature vectors of the target sensor and the analysis sensor at the two analysis moments to obtain a local difference index between the feature vectors of the target sensor and the analysis sensor at the two analysis moments; the cumulative sum of the local difference indexes of the target sensor and all the analysis sensors at the two analysis moments is recorded as a comprehensive difference index of the feature vectors of the target sensor at the two analysis moments; One of the two analysis moments is recorded as the target moment, and the other analysis moment is recorded as the control moment, and a local scene association index of a feature vector of the target sensor at the target moment at the control moment is obtained according to the feature performance index of the target sensor at the target moment and the comprehensive difference index; The mean values ​​of the local scene association indexes of the target sensor at the target moment at the remaining scene moments except the target moment are normalized to serve as the comprehensive scene association index of the feature vector of the target sensor at the target moment.

2. The multi-mode composite physical simulation method based on feature analysis according to claim 1, characterized in that: The method of performing data fusion on the feature vectors according to the difference of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index to obtain fusion data of each sensor at any two scene moments includes: Obtaining a fusion judgment index of the feature vector of each sensor at the two analysis moments according to a difference between a similarity index of the feature vector of each sensor at the two analysis moments and the comprehensive scene association index; If the fusion judgment index is greater than the preset fusion threshold, the comprehensive scene association index is used to perform weighted summation on the feature vectors of each sensor at the two analysis moments to obtain a fusion vector of each sensor at the two analysis moments; If the fusion judgment index is less than or equal to the preset fusion threshold, the fusion coefficient of the feature vector of each sensor at each analysis moment is obtained according to the feature performance index of each sensor at each analysis moment and the comprehensive scene association index; the feature vectors of each sensor at two analysis moments are weightedly summed using the fusion coefficient to obtain the fusion vector of each sensor at the two analysis moments; and the fusion vector is recorded as the fusion data of each sensor at the two analysis moments.

3. The multi-mode composite physical simulation method based on feature analysis according to claim 1, characterized in that: The method of obtaining the feature performance index of the feature vector of each sensor at each scene moment according to the variance contribution index and the redundancy index includes: For the feature vector of each sensor at each scene moment, the product of the result of negative correlation mapping of the redundant index of the feature vector and the variance contribution index is normalized to obtain the feature performance index of the feature vector.

4. The multi-mode composite physical simulation method based on feature analysis according to claim 2 is characterized in that: The method for obtaining the fusion judgment index of the feature vector of each sensor at two analysis moments includes: The absolute value of the difference between the comprehensive scene association index of the feature vector of each sensor at two analysis moments and the product of the similarity index are normalized to obtain the fusion judgment index of the feature vector of each sensor at two analysis moments.

5. The multi-mode composite physical simulation method based on feature analysis according to claim 1, characterized in that: The method for obtaining similarity indexes of any two feature vectors includes: The cosine similarity between any two feature vectors is used as the similarity index of the corresponding two feature vectors.

6. The multi-mode composite physical simulation method based on feature analysis according to claim 1, characterized in that: The preset similarity threshold is 0.

7.

7. The multi-mode composite physical simulation method based on feature analysis according to claim 2 is characterized in that: The preset fusion threshold is 0.

7.

8. A multi-mode composite physical simulation system based on feature analysis, characterized in that: The system includes: The data acquisition module is used to obtain the environmental data of each sensor installed on the object to be tested at each moment in a set time period; the set time period is divided into different local time periods, the last moment in each local time period is recorded as the scene moment, and the environmental data of each sensor in each local time period constitutes the feature vector of each sensor at the scene moment corresponding to each local time period; A feature performance analysis module is used to obtain a feature performance index of a feature vector of each sensor at each scene moment according to the difference between the feature vectors of each sensor and all sensors at each scene moment; A scene association analysis module, used to obtain a comprehensive scene association index of the feature vector of each sensor at each scene moment according to the similarity of the difference between the feature vectors of each sensor and the other sensors at each scene moment and the feature performance index of each sensor at each scene moment; A data fusion module is used to fuse the feature vectors according to the similarity of the feature vectors of each sensor at any two scene moments and the comprehensive scene association index, so as to obtain the fusion data of each sensor at any two scene moments; and use the fusion data to perform simulation test on the object to be tested; The method for obtaining the feature performance index of the feature vector of each sensor at each scene moment includes: The ratio of the distance between the feature vector of each sensor at each scene moment and the mean of the feature vectors of all sensors at each scene moment to the variance of the elements in the feature vectors of all sensors at each scene moment is used as the variance contribution index of the feature vector of each sensor at each scene moment; Obtain similarity indexes of any two feature vectors; select the maximum value of the similarity indexes of the feature vectors of each sensor and the other sensors at each scene moment, and record it as the redundancy index of the feature vector of each sensor at each scene moment; Acquire a feature performance index of a feature vector of each sensor at each scene moment according to the variance contribution index and the redundancy index; The method for obtaining the comprehensive scene association index of each sensor at each scene moment includes: Any two scene moments are recorded as analysis moments, any one sensor is recorded as a target sensor, and the analysis sensor of the target sensor at the two analysis moments is selected from the remaining sensors except the target sensor; the similarity index of the feature vectors of the target sensor and the analysis sensor at each analysis moment is greater than a preset similarity threshold; For each of the analysis sensors, a negative correlation mapping is performed on the difference in the distance between the feature vectors of the target sensor and the analysis sensor at the two analysis moments to obtain a local difference index between the feature vectors of the target sensor and the analysis sensor at the two analysis moments; the cumulative sum of the local difference indexes of the target sensor and all the analysis sensors at the two analysis moments is recorded as a comprehensive difference index of the feature vectors of the target sensor at the two analysis moments; One of the two analysis moments is recorded as the target moment, and the other analysis moment is recorded as the control moment, and a local scene association index of a feature vector of the target sensor at the target moment at the control moment is obtained according to the feature performance index of the target sensor at the target moment and the comprehensive difference index; The mean values ​​of the local scene association indexes of the target sensor at the target moment at the remaining scene moments except the target moment are normalized to serve as the comprehensive scene association index of the feature vector of the target sensor at the target moment.