An EMC test-based intelligent driving system reliability evaluation method

Abstract

The application discloses a kind of based on EMC test intelligent driving system reliability evaluation method, it is related to EMC test technical field, including: obtaining the intelligent driving system software and hardware interaction request data under the closed electromagnetic interference of target vehicle, the prior distribution screening of the function fault feedback data of known intelligent driving system is carried out to the intelligent driving system software and hardware interaction state of target vehicle, and the known fault chain development network of the intelligent driving system software and hardware component of target vehicle is formed;According to the fault chain development network of the intelligent driving system software and hardware component of target vehicle, the posterior distribution of the intelligent driving system software and hardware component interaction request list of target vehicle open road real-time is analyzed, and the intelligent driving system fault chain development trend evaluation model is established, and the intelligent driving system software and hardware component function failure risk score of target vehicle is predicted.The application has the advantages that the safety and robustness of EMC test intelligent driving system are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of EMC test, in particular to a reliability evaluation method of intelligent driving system based on EMC test. BACKGROUND

[0002] The existing intelligent driving system reliability scheme based on EMC test cannot accurately reflect the dynamic electromagnetic interference characteristics in the real road environment through the static interaction data obtained in the closed environment, resulting in a large deviation between the test results and the actual situation. At the same time, the existing method lacks the accurate modeling capability of multi-component coupling faults in complex electromagnetic environment, can only identify single fault mode and is difficult to predict the fault chain propagation path, so that the accuracy and reliability of the evaluation results are seriously insufficient. In addition, the disconnection between the test data and the functional safety evaluation further reduces the effectiveness of the risk warning, and cannot provide accurate reliability feedback for the intelligent driving system. SUMMARY

[0003] To solve the above technical problems, an intelligent driving system reliability evaluation method based on EMC test is provided, which solves the problem that the existing intelligent driving system reliability scheme based on EMC test cannot accurately reflect the dynamic electromagnetic interference characteristics in the real road environment through the static interaction data obtained in the closed environment, resulting in a large deviation between the test results and the actual situation. At the same time, the existing method lacks the accurate modeling capability of multi-component coupling faults in complex electromagnetic environment, can only identify single fault mode and is difficult to predict the fault chain propagation path, so that the accuracy and reliability of the evaluation results are seriously insufficient. In addition, the disconnection between the test data and the functional safety evaluation further reduces the effectiveness of the risk warning, and cannot provide accurate reliability feedback for the intelligent driving system.

[0004] To achieve the above purpose, the technical scheme adopted by the present application is:

[0005] An intelligent driving system reliability evaluation method based on EMC test, comprising:

[0006] S1, based on the closed environment EMC test of the target vehicle, obtaining the intelligent driving system software and hardware interaction request data of the target vehicle under the closed electromagnetic interference, and evaluating the intelligent driving system software and hardware interaction state of the target vehicle;

[0007] S2, based on the Internet and the vehicle enterprise improvement plan, obtaining the functional fault feedback data of the known intelligent driving system and the prior distribution screening of the intelligent driving system software and hardware component interaction state of the target vehicle, and assembling the known fault chain development network of the intelligent driving system software and hardware components of the target vehicle;

[0008] S3, based on the target vehicle's open environment EMC test, according to the intelligent driving system of the target vehicle Soft and hard component fault chain development network, analysis of the target vehicle open road real-time intelligent driving system Soft and hard component interaction request list of posterior distribution, determine the target vehicle intelligent driving system Soft and hard component fault chain development propagation path;

[0009] S4, based on the target vehicle's intelligent driving system Soft and hard component fault chain development propagation path, establish intelligent driving system fault chain development trend evaluation model, predict the target vehicle intelligent driving system Soft and hard component function failure risk score.

[0010] Preferably, based on the target vehicle's closed environment anechoic chamber several rounds of EMC test, obtain the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction request data;

[0011] According to the hardware interaction synchronization and hardware interaction asynchronous, for the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction request data division, get the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction synchronous request data and intelligent driving system Soft and hardware interaction asynchronous request data;

[0012] Using linear interpolation, for the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction asynchronous request data completion;

[0013] Using intelligent driving system Soft and hardware interaction transmission protocol, for the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction synchronous request data and intelligent driving system Soft and hardware interaction asynchronous request data space-time data alignment, get the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction synchronous- asynchronous request space-time alignment data;

[0014] With intelligent driving system Soft and hardware interaction synchronous- asynchronous request time step as observation window, with intelligent driving system Soft and hardware interaction synchronous- asynchronous request interaction data as observation object, obtain the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction synchronous- asynchronous request timing data.

[0015] Preferably, based on ARIMA autoregressive integral moving average, establish intelligent driving system Soft and hardware interaction synchronous- asynchronous request deviation analysis model;

[0016] Using partial autocorrelation function and autocorrelation function, verify the target vehicle closed electromagnetic interference under intelligent driving system Soft and hardware interaction synchronous- asynchronous request timing data, determine the autoregressive order and moving average order;

[0017] Based on the intelligent driving system software and hardware interaction synchronous-asynchronous request deviation analysis model, the autoregressive order and the moving average order are used as the training limit conditions, the target vehicle closed electromagnetic interference under the intelligent driving system software and hardware interaction synchronous-asynchronous request time series data observation value is used as the input, and the target vehicle closed electromagnetic interference under the intelligent driving system software and hardware interaction synchronous-asynchronous request time series data prediction value is used as the output;

[0018] The difference between the target vehicle closed electromagnetic interference under the intelligent driving system software and hardware interaction synchronous-asynchronous request time series data observation value and the prediction value is calculated, and the target vehicle closed electromagnetic interference under the intelligent driving system software and hardware interaction synchronous-asynchronous request time series data residual sequence is established.

[0019] The target vehicle intelligent driving system software and hardware interaction synchronous-asynchronous request standardized time series data is determined, which is substituted into the intelligent driving system software and hardware interaction synchronous-asynchronous request deviation analysis model to obtain the target vehicle intelligent driving system software and hardware interaction synchronous-asynchronous request standardized time series data residual sequence.

[0020] Preferably, according to the three standard deviation threshold method, it is verified whether the standard deviation of the target vehicle closed electromagnetic interference under the intelligent driving system software and hardware interaction synchronous-asynchronous request time series data residual sequence is within the three standard deviation range interval based on the target vehicle intelligent driving system software and hardware interaction synchronous-asynchronous request standardized time series data residual sequence, if yes, it is marked as normal driving software and hardware interaction synchronous-asynchronous request, if not, it is marked as abnormal driving software and hardware interaction synchronous-asynchronous request;

[0021] Determine the normal driving software and hardware interaction synchronous-asynchronous request and the abnormal driving software and hardware interaction synchronous-asynchronous request error coding type;

[0022] Using One-Hot unique heat coding, the normal driving software and hardware interaction synchronous-asynchronous request and the abnormal driving software and hardware interaction synchronous-asynchronous request error coding type are mapped and converted to obtain the target vehicle closed electromagnetic interference under the intelligent driving system normal driving software and hardware interaction synchronous-asynchronous request error coding type vector and the abnormal driving software and hardware interaction synchronous-asynchronous request error coding type vector;

[0023] The target vehicle closed electromagnetic interference under the intelligent driving system normal driving software and hardware interaction synchronous-asynchronous request error coding type vector and the abnormal driving software and hardware interaction synchronous-asynchronous request error coding type vector are normalized;

[0024] According to the DBSCAN density clustering, the Mahalanobis distance between the target vehicle closed electromagnetic interference under the intelligent driving system normal driving software and hardware interaction synchronous- asynchronous request error coding type vector and the abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector is taken as the domain radius, and the abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector is divided, to obtain each abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector set;

[0025] Using the Fourier transform function, the frequency components corresponding to each abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector set are calculated, and each abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector is given a weight;

[0026] Using each abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector set and each abnormal driving software and hardware interaction synchronous- asynchronous request error coding type vector weight, a weighted comprehensive evaluation formula is established, and the intelligent driving system software and hardware component interaction state index of the target vehicle is calculated.

[0027] Preferably, based on the Internet and the vehicle enterprise improvement plan, the functional failure feedback data of the known intelligent driving system is obtained;

[0028] Using hierarchical clustering, according to each functional failure type, the functional failure feedback data of the known intelligent driving system is merged according to the Levenshtein edit distance, and the functional failure type label of the intelligent driving system of the target vehicle is generated;

[0029] Using linear mapping, the functional failure type label of the intelligent driving system of the target vehicle and the intelligent driving system software and hardware component interaction state index of the target vehicle are spatially mapped to obtain the functional failure type label vector of the intelligent driving system of the target vehicle and the intelligent driving system software and hardware component interaction state vector of the target vehicle;

[0030] Using the correlation coefficient, the correlation degree between the functional failure type label vector of the intelligent driving system of the target vehicle and the intelligent driving system software and hardware component interaction state vector of the target vehicle is calculated, and the positive correlation is screened to obtain the intelligent driving system software and hardware component interaction functional failure type label set of the target vehicle.

[0031] Preferably, using a sliding window, the frequency of the functional failure type label in the intelligent driving system software and hardware component interaction functional failure type label set of the target vehicle in a unit time is counted, and the prior probability of the functional failure type label of the intelligent driving system of the target vehicle is calculated;

[0032] Based on the function fault type label of the intelligent driving system of the target vehicle, taking each function fault type label as a node, a function fault undirected graph of the initialized intelligent driving system of the target vehicle is established;

[0033] By using the chi-square test, the dependency relationship between each adjacent edge node in the initialized complete undirected graph is verified, and the non-positive dependency associated edge is deleted to obtain a function fault directed graph of the intelligent driving system of the target vehicle.

[0034] According to the PageRank algorithm, the transition probability between each node in the function fault directed graph of the intelligent driving system of the target vehicle is calculated, and the weight of the associated edge of each node is assigned to obtain a function fault directed weighted graph of the intelligent driving system of the target vehicle.

[0035] Based on the function fault directed weighted graph of the intelligent driving system of the target vehicle, taking each fault point as a starting node, a fault propagation path is generated by using CPT probability random walk to obtain a known fault chain development network of the soft and hard components of the intelligent driving system of the target vehicle.

[0036] Preferably, based on the EMC test of the open environment of the target vehicle, the interactive request list of the soft and hard components of the real-time intelligent driving system on the open road under each test scene of the target vehicle is obtained.

[0037] By using the wavelet change method, the corresponding interactive request list spectrum graph of the soft and hard components of the intelligent driving system on the open road under each test scene is extracted from the interactive request list of the soft and hard components of the real-time intelligent driving system on the open road under each test scene of the target vehicle.

[0038] Based on the interactive request list spectrum graph of the soft and hard components of the intelligent driving system on the open road under each test scene, a single soft and hard component interaction request task is divided according to the unit timestamp, and an interactive request spectrum feature matrix of the soft and hard components of the real-time intelligent driving system on the open road under each test scene is established.

[0039] Each element in the interactive request spectrum feature matrix of the soft and hard components of the real-time intelligent driving system on the open road under each test scene is calculated, and the Euclidean distance of each node in the known fault chain development network of the soft and hard components of the intelligent driving system of the target vehicle is substituted therein to verify the transfer probability of the corresponding network node after the verification of the spectrum feature of the interactive request of the soft and hard components of the real-time intelligent driving system on the open road under each test scene, and the fault chain development propagation path of the intelligent driving system of the target vehicle under each test scene is generated.

[0040] Preferably, based on the intelligent driving system soft and hard component fault chain development propagation path of each test scene of the target vehicle, the intelligent driving system soft and hard component fault chain development multi-dimensional feature parameters are marked, and the intelligent driving system soft and hard component fault chain development feature vector matrix is established; the intelligent driving system soft and hard component fault chain development multi-dimensional feature parameters include: component fault propagation transition probability, component fault propagation path length, component fault propagation soft and hard component type;

[0041] Based on the intelligent driving system soft and hard component fault chain development feature vector matrix, the soft and hard component fault chain development observation function is established, and the intelligent driving system soft and hard component fault chain development trend multi-dimensional vector under each test scene of the target vehicle is generated, in the following way:

[0042] ;

[0043] Among them, is the state binary vector of the intelligent driving system soft and hard component in the tth unit time, is the intelligent driving system soft and hard component fault chain development state transition matrix; is the EMC interference value of the intelligent driving system soft and hard component in the tth unit time, is the intelligent driving system soft and hard component fault chain development state transition input matrix, is the development noise of the intelligent driving system soft and hard component in the tth unit time, is the intelligent driving system soft and hard component fault chain development state observation matrix, is the observation noise of the intelligent driving system soft and hard component in the tth unit time, is the intelligent driving system soft and hard component fault chain development activity, is the total number of times of activating the intelligent driving system soft and hard component fault chain development path in the time window T, is the observation time length, is the overall risk score cumulative value of the intelligent driving system soft and hard component in the tth unit time, is the activation probability of the kth fault chain development path of the intelligent driving system soft and hard component, is the time interval between the tth unit time and the last activation of the fault chain development path, is the forgetting speed coefficient of control history risk, is the decay rate coefficient of the influence of time interval on risk;

[0044] Based on logistics logistic regression, an intelligent driving system fault chain development trend evaluation model is established;

[0045] The analytic hierarchy process is used to give each dimension vector weight of the intelligent driving system soft and hard component fault chain development trend of the target vehicle in each test scene;

[0046] Based on the intelligent driving system fault chain development trend evaluation model, the multi-dimensional vector of the intelligent driving system soft and hard component fault chain development trend of the target vehicle in each test scene and the weight of each dimension vector of the intelligent driving system soft and hard component fault chain development trend of the target vehicle in each test scene are used as influence factors to predict the intelligent driving system soft and hard component function failure risk score of the target vehicle.

[0047] Compared with the prior art, the beneficial effects of the present application are that:

[0048] The present application proposes an intelligent driving system reliability evaluation scheme based on EMC test, which can accurately identify the abnormal interaction mode of software and hardware under electromagnetic interference by fusing static test data and dynamic road data, combining prior fault knowledge base and real-time interaction features, and constructing a multi-dimensional fault development network model, and can dynamically predict function failure risk, providing a quantitative evaluation means and risk warning capability for intelligent driving system in complex electromagnetic environment, and significantly improving the safety and robustness of EMC test intelligent driving system monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 It is a flow chart of an intelligent driving system reliability evaluation method based on EMC test. DETAILED DESCRIPTION

[0050] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0051] Referring to Figure 1 The intelligent driving system reliability evaluation method based on EMC test comprises the following steps:

[0052] Step one, based on the target vehicle's closed environment EMC test, obtaining the intelligent driving system software and hardware interaction request data of the target vehicle under closed electromagnetic interference, evaluating the intelligent driving system software and hardware interaction state of the target vehicle;

[0053] The step one comprises the following contents:

[0054] Based on the target vehicle's closed environment EMC test, obtaining the intelligent driving system software and hardware interaction request data of the target vehicle under closed electromagnetic interference;

[0055] As a further development, since the target vehicle is stationary during EMC testing in a closed anechoic chamber, the backend API interface of the intelligent driving system can still be used to call function functions to deceive the function interface calls, thereby simulating the software and hardware interaction of the vehicle in motion.

[0056] Based on the synchronous and asynchronous hardware interaction, the intelligent driving system software and hardware interaction request data under closed electromagnetic interference of the target vehicle is divided into synchronous request data and asynchronous request data of intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle.

[0057] Using linear interpolation, asynchronous request data completion is achieved for the software and hardware interaction of the intelligent driving system of the target vehicle under closed electromagnetic interference.

[0058] Using the intelligent driving system software and hardware interaction transmission protocol, spatiotemporal data alignment is performed on the synchronous request data and asynchronous request data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle, to obtain spatiotemporal aligned data of synchronous-asynchronous request of intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle.

[0059] Using the time step of synchronous-asynchronous request interaction between intelligent driving system software and hardware as the observation window and the interaction data of synchronous-asynchronous request interaction between intelligent driving system software and hardware as the observation object, the timing data of synchronous-asynchronous request interaction between intelligent driving system software and hardware under closed electromagnetic interference of target vehicle is obtained.

[0060] Based on ARIMA autoregressive integral moving average, a deviation analysis model for synchronous-asynchronous request interaction between software and hardware of intelligent driving system is established.

[0061] Using partial autocorrelation function and autocorrelation function, we verify the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle, and determine the autoregression order and the moving average order.

[0062] Based on the analysis model of synchronous-asynchronous request deviation of software and hardware interaction of intelligent driving system, the autoregressive order and the moving average order are used as training constraints. The time series data of synchronous-asynchronous request of software and hardware interaction of intelligent driving system under closed electromagnetic interference of target vehicle are used as input and the predicted data of synchronous-asynchronous request of software and hardware interaction of target vehicle under closed electromagnetic interference of target vehicle are used as output.

[0063] Calculate the difference between the observed and predicted values ​​of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle, and establish the residual sequence of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle.

[0064] The standardized time-series data of synchronous-asynchronous requests between the hardware and software of the intelligent driving system of the target vehicle are substituted into the deviation analysis model of synchronous-asynchronous requests between the hardware and software of the intelligent driving system to obtain the residual sequence of the standardized time-series data of synchronous-asynchronous requests between the hardware and software of the intelligent driving system of the target vehicle.

[0065] Using the three-standard-deviation threshold method, verify whether the standard deviation of the residual sequence of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle is within the range of three standard deviations of the residual sequence of the standardized time series data of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction of the target vehicle. If yes, mark it as normal driving software and hardware interaction synchronous-asynchronous request; if no, mark it as abnormal driving software and hardware interaction synchronous-asynchronous request.

[0066] Determine the error coding types for normal driving software and hardware interaction synchronous-asynchronous requests and abnormal driving software and hardware interaction synchronous-asynchronous requests;

[0067] Using One-Hot coding, a mapping conversion is performed on the synchronous-asynchronous request error coding types of normal driving software and hardware interaction and abnormal driving software and hardware interaction synchronous-asynchronous request error coding types to obtain the vector of normal driving software and hardware interaction synchronous-asynchronous request error coding types and the vector of abnormal driving software and hardware interaction synchronous-asynchronous request error coding types of intelligent driving system under closed electromagnetic interference of target vehicle.

[0068] Normalization processing is performed on the synchronous-asynchronous request error code type vector of normal driving software and hardware interaction and the synchronous-asynchronous request error code type vector of abnormal driving software and hardware interaction under closed electromagnetic interference of the target vehicle.

[0069] Based on DBSCAN density clustering, the Mahalanobis distance between the normal driving software and hardware interaction synchronous-asynchronous request error code type vector and the abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector under closed electromagnetic interference of the target vehicle is used as the neighborhood radius. The abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector is divided to obtain the set of each abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector.

[0070] Using the Fourier transform function, calculate the frequency components corresponding to the sets of synchronous-asynchronous request error code type vectors for each abnormal driving software and hardware interaction, and assign weights to each abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector.

[0071] By utilizing the set of synchronous-asynchronous request error code type vectors for various abnormal driving software and hardware interactions and their respective weights, a weighted comprehensive evaluation formula is established to calculate the interaction status index of the intelligent driving system software and hardware components of the target vehicle.

[0072] When using it, refer to the steps outlined above.

[0073] As a further point, traditional EMC testing is usually conducted in a static environment, which cannot truly reflect the dynamic electromagnetic interference during vehicle operation. This leads to discrepancies between test results and actual scenarios. Furthermore, existing methods lack the ability to differentiate between synchronous and asynchronous hardware interaction requests, resulting in insufficient timing data alignment accuracy and difficulty in capturing microsecond-level signal delays or packet loss issues under electromagnetic interference. Secondly, statistical analysis cannot identify intermittent anomalies caused by electromagnetic interference (such as occasional error frames on the CAN bus) and does not consider the frequency domain characteristics of error types. Most solutions do not incorporate data parsing based on communication protocols specific to intelligent driving systems (such as the PDU routing mechanism under the AutoSAR architecture), leading to a disconnect between interaction state assessment and actual hardware behavior.

[0074] This solution combines synchronous / asynchronous classification, spatiotemporal alignment, and ARIMA residual analysis to significantly improve the anomaly detection accuracy and temporal resolution of intelligent driving systems under electromagnetic interference. By using DBSCAN clustering and frequency domain weighting to construct an error feature library, a full-stack evaluation from electromagnetic interference modes to hardware components is established, enhancing the diagnostic granularity of intelligent driving systems and solving the problems of large static and dynamic differences and low detection rate of latent errors in traditional EMC testing.

[0075] Step 2: Based on the Internet and the vehicle manufacturer's improvement plan, obtain the functional fault feedback data of the known intelligent driving system and the interaction status of the intelligent driving system software and hardware components of the target vehicle, perform prior distribution screening, and build the known fault chain development network of the intelligent driving system software and hardware components of the target vehicle.

[0076] Step two includes the following:

[0077] Based on the Internet and automakers' improvement plans, obtain functional fault feedback data of known intelligent driving systems;

[0078] Using hierarchical clustering, based on each functional fault type, and for known intelligent driving system functional fault feedback data, similar functional fault feedback parameters of intelligent driving systems are merged according to Levenshtein edit distance to generate functional fault type labels for the intelligent driving system of the target vehicle.

[0079] Using linear mapping, the functional fault type labels of the intelligent driving system of the target vehicle are spatially mapped to the interaction state indicators of the hardware and software components of the intelligent driving system of the target vehicle, so as to obtain the functional fault type label vector and the interaction state vector of the hardware and software components of the intelligent driving system of the target vehicle.

[0080] Using correlation coefficients, the degree of correlation between the functional fault type label vector of the intelligent driving system of the target vehicle and the interaction state vector of the hardware and software components of the intelligent driving system of the target vehicle is calculated. Positive correlation is then used to filter and obtain the set of functional fault type labels for the interaction of the hardware and software components of the intelligent driving system of the target vehicle.

[0081] Using a sliding window, the frequency of occurrence of functional fault type labels in the set of interaction function fault type labels of the intelligent driving system of the target vehicle per unit time is statistically analyzed, and the prior probability of functional fault type labels of the intelligent driving system of the target vehicle is calculated.

[0082] Based on the functional fault type labels of the intelligent driving system of the target vehicle, an undirected graph of functional faults of the intelligent driving system of the target vehicle is established with each functional fault type label as a node.

[0083] Using the chi-square test, the dependency relationships between each neighboring node in the initial completely undirected graph are verified. Non-positive dependency edges are deleted to obtain the functional fault directed graph of the intelligent driving system of the target vehicle.

[0084] According to the PageRank algorithm, the transition probability between each node in the directed graph of the intelligent driving system of the target vehicle is calculated, and the associated edge weights of each node are assigned to obtain the directed weighted graph of the intelligent driving system of the target vehicle.

[0085] Based on the directed weighted graph of functional faults of the intelligent driving system of the target vehicle, with each fault point as the starting node, the fault propagation path is generated by CPT probabilistic random walk, and the known fault chain development network of the intelligent driving system software and hardware components of the target vehicle is obtained.

[0086] When using it, refer to the steps outlined above.

[0087] As a further point, traditional intelligent driving system fault analysis relies on a single data source, resulting in incomplete coverage of fault samples; simple methods such as keyword matching are insufficient to accurately classify semantically similar faults (such as "AEB false trigger" and "automatic emergency braking abnormality"); and static analysis based on expert experience or fixed rules cannot effectively uncover the dynamic statistical dependencies between faults, making it difficult to identify key fault nodes. Overall, this restricts the comprehensiveness, accuracy, and timeliness of fault analysis.

[0088] A comprehensive fault sample library is constructed by fusing multi-source data (Internet + vehicle enterprise data). Intelligent fault merging is achieved by combining Levenshtein edit distance and hierarchical clustering, which significantly improves the classification accuracy. Chi-square test and PageRank algorithm are used to dynamically mine fault associations and quantify key nodes to accurately identify high-impact faults. Based on CPT random walk simulation of fault propagation path, the sliding window mechanism is used to realize the adaptive update of fault path, improve the fault analysis and identification rate of intelligent driving system, improve the efficiency of fault development path discovery, and support the real-time inclusion of new fault modes.

[0089] Step 3: Based on the open environment EMC test of the target vehicle, according to the fault chain development network of the intelligent driving system hardware and software components of the target vehicle, analyze the posterior distribution of the interaction request list of the real-time intelligent driving system hardware and software components of the target vehicle on open roads, and determine the fault chain development propagation path of the intelligent driving system hardware and software components of the target vehicle.

[0090] Step three includes the following:

[0091] Based on open environment EMC testing of the target vehicle, obtain a list of software and hardware interaction requests of the real-time intelligent driving system on open roads under various test scenarios of the target vehicle; the various test scenarios include: urban road scenario, highway scenario, and special scenario of high-voltage substation.

[0092] Using wavelet transform, the spectrum diagram of the interaction request list of the software and hardware components of the open road intelligent driving system in each test scenario is extracted from the interaction request list of the software and hardware components of the open road intelligent driving system in each test scenario of the target vehicle.

[0093] Based on the spectrum of the interaction request list of the software and hardware components of the open road intelligent driving system under various test scenarios, the single software and hardware component interaction request task is divided according to the unit timestamp, and a spectrum feature matrix of the interaction request of the software and hardware components of the open road real-time intelligent driving system under various test scenarios is established.

[0094] Calculate each element in the spectrum feature matrix of the interaction requests between the hardware and software components of the real-time intelligent driving system on open roads under each test scenario. Substitute the Euclidean distance of each node in the known fault chain development network of the intelligent driving system hardware and software components of the target vehicle into the matrix to verify the spectrum features of the interaction requests between the hardware and software components of the real-time intelligent driving system on open roads under each test scenario. Verify the probabilities after the transition of the corresponding network nodes and generate the fault chain development propagation path of the intelligent driving system hardware and software components of the target vehicle under each test scenario.

[0095] As a further step, each element in the spectrum feature matrix of the interaction requests between the hardware and software components of the open road real-time intelligent driving system under various test scenarios is calculated. The Euclidean distance of each node in the known fault chain development network of the intelligent driving system hardware and software components of the target vehicle is then substituted into this matrix. This means that the Euclidean distance is used to substitute the spectrum feature matrix of the interaction requests between the hardware and software components of the open road real-time intelligent driving system under various test scenarios into the known fault chain development network of the intelligent driving system hardware and software components of the target vehicle.

[0096] When using it, refer to the steps outlined above.

[0097] As a further point, the existing EMC testing scenarios for intelligent driving systems are limited to standard open road environments, lacking coverage of special electromagnetic scenarios such as high-voltage substations, making it difficult to capture instantaneous interference characteristics; moreover, the fault propagation modeling adopts static assumptions, which cannot reflect the dynamic evolution characteristics under different electromagnetic environments; the analysis methods rely on post-processing, lacking real-time monitoring capabilities; and secondly, the test data of each scenario are analyzed in isolation, making it impossible to perform cross-scenario interference coupling correlation, which restricts the comprehensiveness and accuracy of the test.

[0098] This solution constructs an EMC interference database covering real driving environments through multi-dimensional scenario testing, uses wavelet transform to achieve millisecond-level time-frequency analysis, and accurately captures transient interference characteristics; based on dynamic probability transition analysis, it improves the accuracy of fault chain location and enhances the precision of intelligent driving systems in EMC testing.

[0099] Step 4: Based on the development and propagation path of the fault chain of the intelligent driving system hardware and software components of the target vehicle, establish an assessment model for the development trend of the fault chain of the intelligent driving system and predict the risk score of functional failure of the hardware and software components of the intelligent driving system of the target vehicle.

[0100] Step four includes the following:

[0101] Based on the development and propagation paths of fault chains in the hardware and software components of the intelligent driving system under various test scenarios of the target vehicle, multi-dimensional feature parameters of fault chain development in the hardware and software components of the intelligent driving system are marked, and a feature vector matrix of fault chain development in the hardware and software components of the intelligent driving system is constructed. The multi-dimensional feature parameters of fault chain development in the hardware and software components of the intelligent driving system include: component fault propagation and transfer probability, component fault propagation path length, and component fault propagation hardware and software component type.

[0102] Based on the feature vector matrix of fault chain development of hardware and software components in intelligent driving system, a fault chain development observation function is established to generate a multi-dimensional vector of the fault chain development trend of hardware and software components in intelligent driving system under various test scenarios of the target vehicle, as follows:

[0103] ;

[0104] in, Let be the state binary vector of the hardware and software components of the intelligent driving system at the t-th unit of time. This is the state transition matrix for the development of fault chains in the hardware and software components of an intelligent driving system. Let be the EMC interference value of the hardware and software components of the intelligent driving system in the t-th unit of time. This is the input matrix for the state transition of the fault chain development of the hardware and software components in the intelligent driving system. The noise level of the hardware and software components of the intelligent driving system in the t-th unit of time is the development noise. This is an observation matrix for the development status of fault chains in the hardware and software components of an intelligent driving system. The noise level of the hardware and software components of the intelligent driving system in the t-th unit of time is... The activity level of the fault chain development of intelligent driving system hardware and software components. This represents the total number of times the development path of the fault chain of the intelligent driving system's hardware and software components is activated within the time window T. The length of the observation period. This represents the cumulative overall risk score of the hardware and software components of the intelligent driving system at the t-th unit of time. Let be the activation probability of the k-th fault chain development path of the intelligent driving system's hardware and software components. Let t be the time interval between the t-th unit of time and the last activation of the fault chain development path. To control the forgetting rate coefficient of historical risk, To determine the decay rate coefficient of the risk impact of the time interval;

[0105] A fault chain development trend assessment model for intelligent driving systems is established based on logistic regression.

[0106] Using the analytic hierarchy process, vector weights are assigned to each dimension of the fault chain development trend of the intelligent driving system's hardware and software components in various test scenarios of the target vehicle.

[0107] Based on the intelligent driving system fault chain development trend assessment model, the multi-dimensional vector of the fault chain development trend of the intelligent driving system hardware and software components under various test scenarios of the target vehicle and the weight of each dimension vector of the fault chain development trend of the intelligent driving system hardware and software components under various test scenarios of the target vehicle are used as influencing factors to predict the functional failure risk score of the intelligent driving system hardware and software components of the target vehicle.

[0108] When using it, refer to the steps outlined above.

[0109] As a further point, traditional intelligent driving system risk assessment relies solely on a single fault probability dimension (such as RPN score), ignoring key features such as propagation path length and component type; the use of fixed weight allocation (such as ASIL level) cannot adapt to the dynamic needs of different scenarios; simple analysis based on linear models is difficult to capture the nonlinear synergistic effects between faults; inconsistent assessment standards across scenarios make it impossible to compare risk priorities horizontally; and the lack of real-time data processing capabilities allows for only post-event analysis, failing to achieve risk warnings, resulting in biased assessment results, poor adaptability, and delayed response of existing methods.

[0110] This solution integrates key parameters such as fault propagation probability, path length, and component type, and combines dynamic weight optimization and nonlinear modeling techniques to achieve unified risk quantification assessment across scenarios. This enables the EMC testing method for intelligent driving systems to have real-time risk warning capabilities. Furthermore, by leveraging visualization analysis tools, it significantly improves the comprehensiveness and accuracy of risk assessment, providing support for the safety design and quantitative numerical decision-making of intelligent driving systems.

[0111] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A reliability assessment method for intelligent driving systems based on EMC testing, characterized in that, include: S1. Based on the closed environment EMC test of the target vehicle, obtain the software and hardware interaction request data of the intelligent driving system of the target vehicle under closed electromagnetic interference, and evaluate the interaction status of the software and hardware components of the intelligent driving system of the target vehicle. S2. Based on the Internet and the vehicle company's improvement plan, obtain the functional fault feedback data of known intelligent driving systems and the interaction status of the intelligent driving system software and hardware components of the target vehicle, perform prior distribution screening, and build a known fault chain development network of the intelligent driving system software and hardware components of the target vehicle. S3. Based on the open environment EMC test of the target vehicle, according to the fault chain development network of the intelligent driving system software and hardware components of the target vehicle, analyze the posterior distribution of the interaction request list of the real-time intelligent driving system software and hardware components of the target vehicle on open roads, and determine the fault chain development propagation path of the intelligent driving system software and hardware components of the target vehicle. S4. Based on the development and propagation path of the fault chain of the intelligent driving system software and hardware components of the target vehicle, establish an assessment model for the development trend of the fault chain of the intelligent driving system and predict the functional failure risk score of the intelligent driving system software and hardware components of the target vehicle. Wherein, S1 includes: Based on several rounds of EMC testing in a closed anechoic chamber for the target vehicle, data on the software and hardware interaction requests of the intelligent driving system under closed electromagnetic interference were obtained. Based on the synchronous and asynchronous hardware interaction, the intelligent driving system software and hardware interaction request data under closed electromagnetic interference of the target vehicle is divided into synchronous request data and asynchronous request data of intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle. Using linear interpolation, asynchronous request data completion is achieved for the software and hardware interaction of the intelligent driving system of the target vehicle under closed electromagnetic interference. Using the intelligent driving system software and hardware interaction transmission protocol, spatiotemporal data alignment is performed on the synchronous request data and asynchronous request data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle, to obtain spatiotemporal aligned data of synchronous-asynchronous request of intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle. Using the time step of synchronous-asynchronous request interaction between the intelligent driving system's software and hardware as the observation window and the interaction data of synchronous-asynchronous request interaction between the intelligent driving system's software and hardware as the observation object, the timing data of synchronous-asynchronous request interaction between the intelligent driving system's software and hardware under closed electromagnetic interference of the target vehicle is obtained.

2. The reliability assessment method for an intelligent driving system based on EMC testing according to claim 1, characterized in that, S1 further includes: Based on ARIMA autoregressive integral moving average, a deviation analysis model for synchronous-asynchronous request interaction between software and hardware of intelligent driving system is established. Using partial autocorrelation function and autocorrelation function, we verify the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle, and determine the autoregression order and the moving average order. Based on the analysis model of synchronous-asynchronous request deviation of software and hardware interaction of intelligent driving system, the autoregressive order and the moving average order are used as training constraints. The time series data of synchronous-asynchronous request of software and hardware interaction of intelligent driving system under closed electromagnetic interference of target vehicle are used as input and the predicted data of synchronous-asynchronous request of software and hardware interaction of target vehicle under closed electromagnetic interference of target vehicle are used as output. Calculate the difference between the observed and predicted values ​​of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle, and establish the residual sequence of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle. The standardized time-series data of synchronous-asynchronous requests between the software and hardware of the intelligent driving system of the target vehicle are determined and then substituted into the deviation analysis model of synchronous-asynchronous requests between the software and hardware of the intelligent driving system to obtain the residual sequence of the standardized time-series data of synchronous-asynchronous requests between the software and hardware of the target vehicle's intelligent driving system.

3. The reliability assessment method for an intelligent driving system based on EMC testing according to claim 2, characterized in that, S1 further includes: Using the three-standard-deviation threshold method, verify whether the standard deviation of the residual sequence of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction under closed electromagnetic interference of the target vehicle is within the range of three standard deviations of the residual sequence of the standardized time series data of the synchronous-asynchronous request time series data of the intelligent driving system software and hardware interaction of the target vehicle. If yes, mark it as normal driving software and hardware interaction synchronous-asynchronous request; if no, mark it as abnormal driving software and hardware interaction synchronous-asynchronous request. Determine the error coding types for normal driving software and hardware interaction synchronous-asynchronous requests and abnormal driving software and hardware interaction synchronous-asynchronous requests; Using One-Hot coding, a mapping conversion is performed on the synchronous-asynchronous request error coding types of normal driving software and hardware interaction and abnormal driving software and hardware interaction synchronous-asynchronous request error coding types to obtain the vector of normal driving software and hardware interaction synchronous-asynchronous request error coding types and the vector of abnormal driving software and hardware interaction synchronous-asynchronous request error coding types of intelligent driving system under closed electromagnetic interference of target vehicle. Normalization processing is performed on the synchronous-asynchronous request error code type vector of normal driving software and hardware interaction and the synchronous-asynchronous request error code type vector of abnormal driving software and hardware interaction under closed electromagnetic interference of the target vehicle. Based on DBSCAN density clustering, the Mahalanobis distance between the normal driving software and hardware interaction synchronous-asynchronous request error code type vector and the abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector under closed electromagnetic interference of the target vehicle is used as the neighborhood radius. The abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector is divided to obtain the set of each abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector. Using the Fourier transform function, calculate the frequency components corresponding to the sets of synchronous-asynchronous request error code type vectors for each abnormal driving software and hardware interaction, and assign weights to each abnormal driving software and hardware interaction synchronous-asynchronous request error code type vector. By utilizing the set of synchronous-asynchronous request error code type vectors for various abnormal driving software and hardware interactions and their respective weights, a weighted comprehensive evaluation formula is established to calculate the interaction status index of the intelligent driving system software and hardware components of the target vehicle.

4. The reliability assessment method for an intelligent driving system based on EMC testing according to claim 3, characterized in that, S2 includes: Based on the Internet and automakers' improvement plans, obtain functional fault feedback data of known intelligent driving systems; Using hierarchical clustering, based on each functional fault type, and for known intelligent driving system functional fault feedback data, similar functional fault feedback parameters of intelligent driving systems are merged according to Levenshtein edit distance to generate functional fault type labels for the intelligent driving system of the target vehicle. Using linear mapping, the functional fault type labels of the intelligent driving system of the target vehicle are spatially mapped to the interaction state indicators of the hardware and software components of the intelligent driving system of the target vehicle, so as to obtain the functional fault type label vector and the interaction state vector of the hardware and software components of the intelligent driving system of the target vehicle. Using correlation coefficients, the degree of correlation between the functional fault type label vector of the intelligent driving system of the target vehicle and the interaction state vector of the hardware and software components of the intelligent driving system of the target vehicle is calculated. Positive correlation is then used to filter and obtain the set of functional fault type labels for the interaction of hardware and software components of the intelligent driving system of the target vehicle.

5. The reliability assessment method for an intelligent driving system based on EMC testing according to claim 4, characterized in that, S2 further includes: Using a sliding window, the frequency of occurrence of functional fault type labels in the set of interaction function fault type labels of the intelligent driving system of the target vehicle per unit time is statistically analyzed, and the prior probability of functional fault type labels of the intelligent driving system of the target vehicle is calculated. Based on the functional fault type labels of the intelligent driving system of the target vehicle, an undirected graph of functional faults of the intelligent driving system of the target vehicle is established with each functional fault type label as a node. Using the chi-square test, the dependency relationships between each neighboring node in the initial completely undirected graph are verified. Non-positive dependency edges are deleted to obtain the functional fault directed graph of the intelligent driving system of the target vehicle. According to the PageRank algorithm, the transition probability between each node in the directed graph of the intelligent driving system of the target vehicle is calculated, and the associated edge weights of each node are assigned to obtain the directed weighted graph of the intelligent driving system of the target vehicle. Based on the directed weighted graph of functional faults of the intelligent driving system of the target vehicle, with each fault point as the starting node, the fault propagation path is generated by CPT probabilistic random walk, and the known fault chain development network of the intelligent driving system software and hardware components of the target vehicle is obtained.

6. The reliability assessment method for an intelligent driving system based on EMC testing according to claim 5, characterized in that, S3 includes: Based on open environment EMC testing of the target vehicle, obtain a list of software and hardware interaction requests of the real-time intelligent driving system on open roads under various test scenarios of the target vehicle. Using wavelet transform, the spectrum diagrams of the interaction request lists of the hardware and software components of the open road intelligent driving system under each test scenario are extracted from the interaction request list of the hardware and software components of the open road intelligent driving system under each test scenario of the target vehicle. Based on the spectrum of the interaction request list of the software and hardware components of the open road intelligent driving system under various test scenarios, the single software and hardware component interaction request task is divided according to the unit timestamp, and a spectrum feature matrix of the interaction request of the software and hardware components of the open road real-time intelligent driving system under various test scenarios is established. Calculate each element in the spectrum feature matrix of the interaction requests between the hardware and software components of the real-time intelligent driving system on open roads under each test scenario. Substitute the Euclidean distance of each node in the known fault chain development network of the intelligent driving system hardware and software components of the target vehicle into the matrix to verify the spectrum features of the interaction requests between the hardware and software components of the real-time intelligent driving system on open roads under each test scenario. Verify the probabilities after the transition of the corresponding network nodes and generate the fault chain development propagation path of the intelligent driving system hardware and software components of the target vehicle under each test scenario.

Images