Intelligent cabin performance evaluation method and system, and electronic device

Abstract

This application provides a method, system, and electronic device for evaluating the performance of an intelligent cockpit. The method includes: acquiring multi-source time-series data of the intelligent cockpit, including hardware status data, software status data, user interaction data, and operating environment data; extracting and fusing features from the multi-source time-series data to obtain a first fused feature vector; determining the target operating scenario of the intelligent cockpit based on the multi-source time-series data, and retrieving a second fused feature vector from historical performance evaluations under the same operating scenario; then assigning target weights to each feature component in the first fused feature vector based on the target operating scenario, the first fused feature vector, and the second fused feature vector; and weighting and aggregating each feature component according to its target weight to obtain a first comprehensive performance score for the intelligent cockpit. By evaluating the performance of the intelligent cockpit using multi-dimensional data and combining it with a dynamic weight adaptive mechanism, the comprehensiveness and reliability of the performance evaluation are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent vehicle technology, and in particular to an intelligent cockpit performance evaluation method, system and electronic device. Background Technology

[0002] As the core carrier of human-machine interaction in modern automobiles, the intelligent cockpit integrates multiple functional modules such as infotainment, human-machine interface, driver assistance, environmental perception, and personalized user services. Its operational performance directly affects the user's driving experience, operational safety, and system reliability. With the continuous improvement of vehicle intelligence and connectivity, cockpit systems are becoming increasingly complex, with high hardware-software coupling and variable operating environments, significantly increasing the requirements for their real-time performance, stability, and long-term reliability. Therefore, reliable, comprehensive, and continuous evaluation of intelligent cockpit performance has become a key link in ensuring the overall vehicle intelligence level and user experience.

[0003] In related technologies, the performance evaluation of smart cockpits often relies on single-dimensional data, which makes the evaluation results relatively one-sided. Although some evaluation systems have introduced multiple evaluation indicators, these indicators are still limited to a certain functional level. The overall evaluation dimension is still single and it is difficult to adapt to the dynamic changes in the performance influencing factors of vehicles under different usage stages, different operating scenarios and different user habits, which weakens the objectivity and reliability of the evaluation results. Summary of the Invention

[0004] This application discloses a method, system, and electronic device for evaluating the performance of intelligent cockpits, which addresses the technical problems of one-sided performance evaluation and poor reliability of evaluation results.

[0005] In a first aspect, this application provides a method for evaluating the performance of an intelligent cockpit. The method includes: acquiring multi-source time-series data of the intelligent cockpit, the multi-source time-series data including hardware status data, software status data, user interaction data, and operating environment data; extracting features from the multi-source time-series data and fusing the extracted features to obtain a first fused feature vector; determining a target operating scenario for the intelligent cockpit based on the multi-source time-series data, and retrieving a second fused feature vector under the same operating scenario from historical performance evaluations based on the target operating scenario; assigning target weights to each feature component in the first fused feature vector based on the target operating scenario, the first fused feature vector, and the second fused feature vector; and weighting and aggregating the feature components based on the target weights to obtain a first comprehensive performance score for the intelligent cockpit.

[0006] In a first possible implementation of the first aspect, determining the target operating scenario of the intelligent cockpit based on the multi-source time-series data includes: extracting scenario discrimination features from the multi-source time-series data; if the scenario discrimination features satisfy the rule conditions corresponding to any operating scenario in a preset scenario rule base, then determining the arbitrary operating scenario as the target operating scenario; if the arbitrary operating scenario does not exist in the scenario rule base, then inputting the scenario discrimination features into a pre-constructed scenario recognition model to obtain the probability distribution of various operating scenarios, and determining the target operating scenario based on the probability distribution of various operating scenarios; wherein, the scenario rule base includes the rule conditions that the scenario discrimination features must satisfy under various operating scenarios, and the scenario recognition model is trained based on first sample data labeled with operating scenario tags.

[0007] In the first possible implementation of the first aspect, rich scene discrimination features are extracted from multi-source time-series data, providing a comprehensive information foundation for scene recognition. Then, a hierarchical scene recognition mechanism combining a scene rule base and a scene recognition model is adopted. This enables rapid and deterministic recognition of known and well-defined operating scenes through the rule base, ensuring recognition efficiency and high confidence. For complex, ambiguous, or unknown scenes not covered by the rule base, probabilistic judgment can be made through the scene recognition model. This significantly improves the accuracy and robustness of target operating scene recognition in the intelligent cockpit, providing a basis for subsequent retrieval of historical performance evaluation data and allocation of feature component target weights, thereby ensuring the accuracy and relevance of the intelligent cockpit performance evaluation results.

[0008] In a second possible implementation of the first aspect, assigning target weights to each feature component in the first fused feature vector includes: determining the base weights corresponding to each feature component from a preset weight strategy library based on the target operating scenario, wherein the weight strategy library includes the base weights of each feature component under various operating scenarios; determining the weight adjustment ratio corresponding to each feature component based on the degree of offset of each feature component relative to the corresponding feature component in the second fused feature vector; and adjusting the base weights of the corresponding feature components in each feature component according to each weight adjustment ratio to obtain the target weights of each feature component.

[0009] In the second possible implementation of the first aspect, based on the target operating scenario of the intelligent cockpit, the basic weights of each feature component are obtained from a pre-set weight strategy library, thereby ensuring the scenario adaptability of the weight allocation. At the same time, by calculating the deviation of each feature component relative to the performance of the same historical scenario, and determining the weight adjustment ratio accordingly, dynamic correction of the basic weights is achieved. This weight adjustment mechanism, which combines a pre-set weight strategy library with real-time performance deviation information, enables the final allocated target weights to more accurately reflect the impact of each feature component on the overall performance of the intelligent cockpit under the current operating state, thereby making the comprehensive evaluation results more consistent with the sensitive points and risk points of the actual scenario.

[0010] In a third possible implementation of the first aspect, the second fused feature vector is multiple, and determining the weight adjustment ratio corresponding to each feature component includes: calculating the feature mean and feature standard deviation corresponding to each feature component based on the multiple second fused feature vectors; calculating the standard deviation multiple of each feature component relative to the corresponding feature mean based on each feature component and the corresponding feature mean and feature standard deviation; determining the weight adjustment ratio corresponding to each feature component based on the standard deviation multiple corresponding to each feature component and a preset first mapping relationship, wherein the first mapping relationship characterizes the correspondence between the standard deviation multiple corresponding to each feature component and the weight adjustment ratio.

[0011] In the third possible implementation of the first aspect, multiple historical second fusion feature vectors under the same operating scenario are fully utilized. By calculating the feature mean and feature standard deviation in the historical evaluation, the normal fluctuation range of each feature component in history is accurately captured, reducing the randomness and bias that may be brought about by a single historical data. By calculating the standard deviation multiple, the deviation between the current state and the historical state is quantified in a standardized way, making the deviation between different feature components comparable. Finally, combined with the preset first mapping relationship, the weight of each feature component is dynamically and accurately adjusted according to this quantified deviation, ensuring that the final first comprehensive performance score can more objectively and accurately reflect the actual scenario state of the intelligent cockpit.

[0012] In a fourth possible implementation of the first aspect, the second fused feature vector is multiple, and assigning target weights to each feature component in the first fused feature vector further includes: calculating the degree of offset of each feature component relative to the corresponding feature component in the multiple second fused feature vectors; inputting the degree of offset between the target running scene and each feature component into a pre-constructed weight allocation model to obtain the target weight of each feature component; wherein the degree of offset includes a standard deviation multiple, and the weight allocation model is trained based on second sample data labeled with running scene labels and weight labels.

[0013] In the fourth possible implementation of the first aspect, by calculating the degree of deviation of each feature component relative to multiple historical second fusion feature vectors and using the standard deviation multiple as a quantitative indicator, it is possible to more comprehensively and accurately reflect the degree of deviation of the current intelligent cockpit's various performance characteristics from the historical norm. On this basis, the target operating scenario and these refined degree of deviation are input into a pre-built weight allocation model, so that the allocation of target weights is no longer a simple preset, but is intelligently adjusted based on a lightweight neural network according to the specific operating scenario and the statistical deviation of each feature component, ensuring that the final first comprehensive performance score can more objectively and accurately reflect the actual scenario state of the intelligent cockpit.

[0014] In the fifth possible implementation of the first aspect, feature extraction is performed on the multi-source time-series data, and the extracted features are fused to obtain a first fused feature vector. This includes: determining the performance index sequences of multiple predefined performance indicators based on the multi-source time-series data; for each performance indicator, extracting indicator evolution features from the corresponding performance index sequence, wherein the indicator evolution features include statistical distribution features, trend features, and abnormal deviation features; standardizing the multiple indicator evolution features corresponding to each performance indicator, and concatenating the multiple indicator evolution features of all standardized performance indicators to obtain a concatenated feature vector; and performing dimensionality reduction on the concatenated feature vector to obtain the first fused feature vector.

[0015] In the fifth possible implementation of the first aspect, the process progresses from the original multi-source time series data to the multi-performance index sequence, and then to the index evolution characteristics and the standardization, splicing and dimensionality reduction of the features. This step-by-step approach can systematically extract and fuse multi-dimensional dynamic features reflecting the comprehensive performance of the intelligent cockpit from multi-source heterogeneous time series data, obtain a more refined and representative first fusion feature vector, reduce the information bias that may be caused by a single feature, provide high-quality input for subsequent weight allocation and comprehensive performance scoring, and thus make the entire performance evaluation method more robust and reliable.

[0016] In a sixth possible implementation of the first aspect, after obtaining the first comprehensive performance score of the intelligent cockpit, the method further includes: determining the contribution of each feature component to the first comprehensive performance score based on each feature component and its corresponding target weight; and, if the contribution of a first target feature component among the feature components is less than a preset contribution threshold, determining a first performance bottleneck currently affecting the performance of the intelligent cockpit based on the first target feature component and a preset second mapping relationship; determining the current health level of the intelligent cockpit based on the first comprehensive performance score and a preset third mapping relationship; and, if the health level reaches a preset warning level, generating first warning information based on the health level, the first performance bottleneck, and the maintenance measures for the first performance bottleneck; wherein the second mapping relationship characterizes the correspondence between each feature component and the performance bottleneck, and the third mapping relationship characterizes the correspondence between each comprehensive performance score and the health level.

[0017] In the sixth possible implementation of the first aspect, based on the comprehensive performance score of the intelligent cockpit, a quantitative correlation model between the comprehensive performance score and the health status of the intelligent cockpit is constructed to realize an interpretable mapping from score to health status. Based on contribution analysis and feature reverse mapping mechanism, it can accurately locate the key performance bottlenecks that lead to performance degradation and provide maintenance measures. This solves the problem of having only a single score but lacking specific root cause analysis, intelligent early warning and operation guidance when the score is low, and enhances the interpretability and operability of the evaluation results.

[0018] In the seventh possible implementation of the first aspect, after obtaining the first comprehensive performance score of the intelligent cockpit, the method further includes: acquiring multiple second comprehensive performance scores from historical performance evaluations, and determining a comprehensive performance score sequence for the intelligent cockpit in a future period based on the multiple second comprehensive performance scores and the first comprehensive performance score; determining the performance change trend of the intelligent cockpit and the remaining service life of the intelligent cockpit in the future period based on the comprehensive performance score sequence; if there is a period in the performance change trend where the rate of decline is greater than a preset rate threshold, then the period is determined as a performance degradation period, and a fourth fusion feature vector is determined under the performance degradation period based on multiple third fusion feature vectors from historical performance evaluations and the first fusion feature vector; calculating the difference between each feature component and the corresponding feature component in the fourth fusion feature vector, and determining the feature component whose difference is greater than a preset difference as a second target feature component, and then determining a second performance bottleneck that will affect the performance of the intelligent cockpit in the future based on the second target feature vector and the second mapping relationship; and generating second early warning information based on the performance degradation period, the second performance bottleneck, and the maintenance measures for the second performance bottleneck.

[0019] In the seventh possible implementation of the first aspect, by analyzing historical and current comprehensive performance scores, the performance change trend and remaining service life of the intelligent cockpit can be predicted. Before significant performance degradation occurs, specific characteristic components that may lead to performance decline can be identified in advance and mapped as maintainable performance bottlenecks. This provides a basis for proactive optimization and predictive maintenance of the intelligent cockpit, thereby improving the user experience.

[0020] Secondly, this application also provides an intelligent cockpit performance evaluation system, the system comprising: a data acquisition module for acquiring multi-source time-series data of the intelligent cockpit, the multi-source time-series data including hardware status data, software status data, user interaction data, and operating environment data; a data processing module for extracting features from the multi-source time-series data and fusing the extracted features to obtain a first fused feature vector; a weight allocation module for determining the target operating scenario of the intelligent cockpit based on the multi-source time-series data, retrieving a second fused feature vector under the same operating scenario in historical performance evaluations based on the target operating scenario, and then assigning target weights to each feature component in the first fused feature vector based on the target operating scenario, the first fused feature vector, and the second fused feature vector; and a performance scoring module for weighted aggregation of each feature component based on the target weights of each feature component to obtain a first comprehensive performance score for the intelligent cockpit.

[0021] Thirdly, this application also provides an electronic device, including: a processor; and a storage device for storing a program, which, when executed by the processor, causes the electronic device to implement the intelligent cockpit performance evaluation method as described above.

[0022] The beneficial effects of this application are as follows: This application provides a method, system, and electronic device for evaluating the performance of an intelligent cockpit. First, it acquires multi-source time-series data of the intelligent cockpit, including hardware status data, software status data, user interaction data, and operating environment data. Next, it extracts features from the multi-source time-series data and fuses the extracted features to obtain a first fused feature vector. Then, it determines the target operating scenario of the intelligent cockpit based on the multi-source time-series data and retrieves a second fused feature vector from historical performance evaluations of the same operating scenario based on the target operating scenario. Finally, based on the target operating scenario, the first fused feature vector, and the second fused feature vector, it generates a first fused feature vector. Each feature component in the feature vector is assigned a target weight. Finally, based on the target weight of each feature component, the feature components are weighted and aggregated to obtain the first comprehensive performance score of the intelligent cockpit. By acquiring multi-source time-series data of the intelligent cockpit and integrating multi-dimensional data such as hardware, software, users, and environment, the performance evaluation of the intelligent cockpit can be carried out. This can comprehensively capture the state of the intelligent cockpit. In addition, during the evaluation process, weights can be dynamically assigned to each state feature according to the operating scenario of the intelligent cockpit, thereby conducting performance scoring. This makes the evaluation results more scenario-adaptive, ensures the objectivity and reliability of the evaluation, and provides a reliable data foundation for the optimization and maintenance of the intelligent cockpit. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0024] In the attached diagram:

[0025] Figure 1 This is a schematic diagram illustrating the implementation environment of an optional intelligent cockpit performance evaluation system, as shown in an exemplary embodiment of this application. Figure 2 This is a schematic flowchart illustrating an optional smart cockpit performance evaluation method as an exemplary embodiment of this application; Figure 3 This is a schematic diagram illustrating an optional process for determining a target operating scenario, as shown in an exemplary embodiment of this application. Figure 4 This is a schematic diagram illustrating an optional process for assigning target weights, as shown in an exemplary embodiment of this application. Figure 5 This is a schematic diagram illustrating another optional process for assigning target weights, as shown in an exemplary embodiment of this application; Figure 6This is a schematic diagram illustrating an optional feature extraction and fusion process in an exemplary embodiment of this application; Figure 7 This is a schematic diagram illustrating an optional data preprocessing flow in an exemplary embodiment of this application; Figure 8 This is a block diagram illustrating an optional intelligent cockpit performance evaluation system, as shown in an exemplary embodiment of this application. Figure 9 This is a schematic diagram illustrating the structure of an optional electronic device, as shown in an exemplary embodiment of this application. Detailed Implementation

[0026] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0027] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the shape, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0028] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the present application. However, it will be apparent to those skilled in the art that embodiments of the present application may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the present application.

[0029] The performance of a smart cockpit directly impacts the user's driving experience, operational safety, and system reliability. Reliable, comprehensive, and continuous evaluation of smart cockpit performance has become crucial for ensuring the overall intelligence level of a vehicle and the user experience. However, the inventors of this application have found that performance evaluations of smart cockpits often rely on single-dimensional data, leading to biased assessment results. While some evaluation systems have introduced multiple evaluation indicators, these indicators remain limited to a specific functional level, such as functional verification during the factory phase or instantaneous performance evaluations of system startup time, interface rendering frame rate, or touch response latency in specific scenarios. The overall evaluation dimension remains singular. Furthermore, the weighting of each performance indicator in the evaluation is often fixed, making it difficult to adapt to the dynamic changes in performance influencing factors under different usage stages, operating scenarios, and user habits, thus weakening the objectivity and reliability of the evaluation results.

[0030] Therefore, please see Figure 1 , Figure 1 This is a schematic diagram illustrating an implementation environment for an optional intelligent cockpit performance evaluation system, as shown in an exemplary embodiment of this application. Figure 1 As shown, the implementation environment may include an intelligent cockpit performance evaluation system 110 and a computer device 120. The intelligent cockpit performance evaluation system 110 can be installed within the computer device 120 to perform performance evaluation of the intelligent cockpit. The computer device 120 can be at least one of a desktop GPU (Graphics Processing Unit) computer, a GPU computing cluster, or a neural network computer. This intelligent cockpit performance evaluation system 110 acquires multi-source time-series data from the intelligent cockpit and integrates multi-dimensional data from hardware, software, users, and the environment to evaluate the intelligent cockpit's performance. It can comprehensively capture the state of the intelligent cockpit and, during the evaluation process, dynamically assign weights to each state feature according to the intelligent cockpit's operating scenario, thereby generating a performance score. This makes the evaluation results more scenario-adaptive, ensuring the objectivity and reliability of the evaluation and providing data support for the optimization and maintenance of the intelligent cockpit.

[0031] Please see Figure 2 , Figure 2 This is a schematic flowchart illustrating an optional smart cockpit performance evaluation method, as shown in an exemplary embodiment of this application. This method can be applied to... Figure 1 The implementation environment is shown, and the method is specifically executed by the intelligent cockpit performance evaluation system 110 within that implementation environment. It should be understood that this method can also be applied to other exemplary implementation environments and executed by devices in other implementation environments; this embodiment does not limit the implementation environment to which the method is applicable.

[0032] like Figure 2As shown, in an exemplary embodiment, the intelligent cockpit performance evaluation method includes at least steps S210 to S240, which are described in detail below: Step S210: Obtain multi-source time-series data of the intelligent cockpit. The multi-source time-series data includes hardware status data, software status data, user interaction data, and operating environment data.

[0033] Hardware status data refers to time-series data reflecting the operational status of the underlying hardware of the intelligent cockpit; software status data refers to time-series data reflecting the operational status of the intelligent cockpit's operating system and application software; user interaction data refers to time-series data reflecting the interaction between the user and the intelligent cockpit; and operating environment data refers to time-series data reflecting the status of the physical and communication environments in which the intelligent cockpit operates. Hardware status data, software status data, user interaction data, and operating environment data are all time-series data within the same time period. These data are indexed by timestamps and can reflect the operational status of the intelligent cockpit at a specific point in time.

[0034] Step S220: Extract features from multi-source time-series data and fuse the extracted features to obtain a first fused feature vector.

[0035] The first fusion feature vector refers to a numerical vector that can comprehensively characterize the current performance status of the intelligent cockpit after feature extraction from multi-source time-series data and fusion processing of the extracted features. Each feature component in the first fusion feature vector represents the performance characteristics of the intelligent cockpit in different dimensions or at different levels.

[0036] Step S230: Determine the target operating scenario of the intelligent cockpit based on multi-source time series data, retrieve the second fused feature vector under the same operating scenario in the historical performance evaluation based on the target operating scenario, and then assign target weights to each feature component in the first fused feature vector based on the target operating scenario, the first fused feature vector and the second fused feature vector.

[0037] The target operating scenario refers to the specific working environment or usage situation in which the intelligent cockpit is evaluated during performance assessment. The second fused feature vector is a fused feature vector obtained from historical performance evaluation data under the same historical operating scenario as the current target operating scenario. This fused feature vector is used to provide a benchmark or reference for comparing and evaluating the performance of the current intelligent cockpit. The target weight refers to the weight value assigned to each feature component in the first fused feature vector when calculating the comprehensive performance score of the intelligent cockpit. This weight value reflects the degree of influence of different feature components on the overall performance evaluation of the intelligent cockpit under a specific target operating scenario.

[0038] Step S240: Based on the target weight of each feature component, perform weighted aggregation on each feature component to obtain the first comprehensive performance score of the intelligent cockpit.

[0039] The first comprehensive performance score refers to a numerical value that comprehensively reflects the current overall performance level of the intelligent cockpit, calculated by weighted aggregation based on each feature component and its corresponding target weight in the first fusion feature vector. It is used to intuitively measure the performance status of the intelligent cockpit.

[0040] In this embodiment, by acquiring multi-source time-series data of the intelligent cockpit and integrating multi-dimensional data such as hardware, software, users, and environment, the performance evaluation of the intelligent cockpit can be carried out. This can comprehensively capture the state of the intelligent cockpit. Furthermore, during the evaluation process, weights can be dynamically assigned to each state feature according to the operating scenario of the intelligent cockpit, thereby performing performance scoring. This makes the evaluation results more scenario-adaptive, ensuring the objectivity and reliability of the evaluation, and providing data support for the optimization and maintenance of the intelligent cockpit.

[0041] For example, hardware status data includes the CPU (Central Processing Unit) / GPU / NPU (Neural Processing Unit) core utilization, frequency, and temperature of the intelligent cockpit domain controller and associated ECUs (Electronic Control Units); memory usage and bandwidth; storage I / O performance, read / write speed, and remaining space; bus load rate; power supply voltage, current, and power consumption of key chips / sensors; and cooling fan speed. Software status data includes the operating system / middleware / application process status and resource usage status (such as the number of threads and handles) of the intelligent cockpit domain controller and associated ECUs; system and key application logs (such as errors, warnings, and timeout events); and the response latency and success rate of services (such as voice recognition, navigation, and entertainment). User interaction data includes the frequency, type, and response time of touch / button / voice commands, user feedback data (such as satisfaction ratings and complaint content), and screen operation heatmaps. Operating environment data includes in-vehicle / outdoor temperature, vehicle status (such as driving / stationary and vehicle speed), and network connection status (such as signal strength, bandwidth, latency, and packet loss rate).

[0042] In this exemplary embodiment, multi-dimensional data on hardware, software, user and environment are provided for the performance evaluation of the smart cockpit, constructing a multi-dimensional and dynamic performance profile that goes beyond single-point static evaluation, enabling the evaluation results to be more comprehensive.

[0043] For example, let each feature component in the first fused feature vector be... The corresponding target weight is In the process of weighted aggregation of each feature component according to its target weight, element-wise weighting is first performed, that is, each feature component in the first fused feature vector is... Multiply by its corresponding target weight ,get Then, the weighted feature vectors are aggregated into scores. The obtained scores are then mapped to a range of 0-100 to obtain the first comprehensive performance score. The score mapping function is as follows: ,in, This indicates the first overall performance score. This represents the Sigmoid function. , These represent the scaling and translation parameters, respectively.

[0044] In one embodiment, please refer to Figure 3 , Figure 3 This is a schematic diagram illustrating an optional process for determining a target operating scenario, as shown in an exemplary embodiment of this application. Figure 3 As shown, the steps for determining the target operating scenario of the intelligent cockpit based on multi-source time-series data include at least the following: Step S310, extracting scenario discrimination features from multi-source time-series data; Step S320, if the scenario discrimination features satisfy the rule conditions corresponding to any operating scenario in the preset scenario rule base, then the arbitrary operating scenario is determined as the target operating scenario; Step S330, if there is no arbitrary operating scenario in the scenario rule base, then the scenario discrimination features are input into a pre-built scenario recognition model to obtain the probability distribution of various operating scenarios, and the target operating scenario is determined based on the probability distribution of various operating scenarios; wherein, the scenario rule base includes the rule conditions that the scenario discrimination features must satisfy under various operating scenarios, and the scenario recognition model is trained based on the first sample data labeled with the operating scenario label.

[0045] Among them, scene discrimination features refer to key indicators extracted from multi-source time series data that can effectively characterize the operation scenarios of the intelligent cockpit and are used to distinguish different operation scenarios; the scene rule base can be a pre-built database, configuration file or lookup table, which stores the rule conditions that scene discrimination features must meet under various operation scenarios; the first sample data refers to historical multi-source time series data with clear operation scenario labels, which can extract scene discrimination features from historical multi-source time series data for training the scene recognition model.

[0046] In this embodiment, considering that simply relying on a single scene judgment logic may not cover all complex situations, resulting in insufficient accuracy of scene recognition and thus affecting the accuracy of subsequent performance evaluation, a hierarchical scene recognition mechanism combining hard matching (i.e., using a scene rule engine to match scenes) and soft matching (i.e. using a scene recognition model to recognize scenes) is proposed to ensure the accuracy of scene recognition.

[0047] In this embodiment, scene discrimination features are extracted from multi-source time-series data. Specifically, short-term scene discrimination features are extracted from hardware status data, software status data, user interaction data, and operating environment data from multi-source time-series data. For example, the highest CPU temperature, foreground applications, voice command frequency, and average vehicle speed are extracted. By comprehensively considering multiple data dimensions, the physical environment, driving tasks, and user needs and behavior patterns of the intelligent cockpit are fully reflected, such as high temperature weather + urban congestion + frequent voice navigation, low temperature cold start + after system update, high-speed driving + background download + multi-screen display, etc. When determining whether the scene discrimination features meet the rule conditions in the pre-set scene rule base, the scene rule base has a series of clearly defined rules pre-stored. That is, corresponding judgment conditions are set for various predefined operating scenarios to quickly and accurately identify the operating scenarios of the intelligent cockpit. These rule conditions are usually formulated based on expert experience and a large amount of actual data analysis, and are implemented through threshold judgment, logical combination or state machine, such as IF (avg_speed_2min<20km / h AND max_cpu_temp_5min>85°C AND active_app_id='Navigation') THEN Cs='Scenario_High Temperature Congestion Navigation', IF(voice_cmd_freq_1min>20times) THEN Cs='Scenario_Dense Voice Interaction'. If a suitable operating scenario cannot be matched in the scenario rule base, that is, if the current scenario discrimination features do not meet any preset rule conditions, the scenario discrimination features are input into a pre-built scenario recognition model. This scenario recognition model can be built based on machine learning or deep learning techniques and is a lightweight classification model, such as GBDT (Gradient Boosting Decision Tree). Through the complex mapping relationships learned internally, it outputs the probability distribution of the smart cockpit in various predefined operating scenarios, such as Cs={('Scenario_High Temperature Congestion Navigation',0.75), ('Scenario_Dense Voice Interaction',0.20)...}. Then, based on this probability distribution, the target operating scenario of the smart cockpit is determined, usually selecting the operating scenario with the highest probability as the target operating scenario.

[0048] In this embodiment, rich scene discrimination features are extracted from multi-source time-series data, providing a comprehensive information foundation for scene recognition. Then, a hierarchical scene recognition mechanism combining a scene rule base and a scene recognition model is adopted. This enables rapid and deterministic recognition of known and well-defined operating scenes through the rule base, ensuring recognition efficiency and high confidence. For complex, ambiguous, or unknown scenes not covered by the rule base, probabilistic judgment can be made through the scene recognition model. This significantly improves the accuracy and robustness of target operating scene recognition in the intelligent cockpit, providing a basis for subsequent retrieval of historical performance evaluation data and allocation of feature component target weights, thereby ensuring the accuracy and relevance of the intelligent cockpit performance evaluation results.

[0049] In one possible embodiment, updating the scene rule base includes: monitoring historical multi-source time-series data of target recognition scenarios that fail to match from the scene rule base; if the occurrence of target historical multi-source time-series data is more frequent than a preset frequency threshold, then clustering all target historical multi-source time-series data for smart cockpit operation scenarios and rule conditions using a clustering algorithm to obtain new operation scenarios and rule conditions associated with the new operation scenarios; adding the new operation scenarios and associated rule conditions to the scene rule base to complete the update of the scene rule base.

[0050] In one embodiment, please refer to Figure 4 , Figure 4 This is a schematic diagram illustrating an optional process for assigning target weights, as shown in an exemplary embodiment of this application. Figure 4 As shown, the steps for assigning target weights to each feature component in the first fused feature vector include at least the following: Step S410, determining the basic weights corresponding to each feature component from a preset weight strategy library based on the target operating scenario, wherein the weight strategy library includes the basic weights of each feature component under various operating scenarios; Step S420, determining the weight adjustment ratio corresponding to each feature component based on the degree of offset of each feature component relative to the corresponding feature component in the second fused feature vector; Step S430, adjusting the basic weights of the corresponding feature components in each feature component according to the weight adjustment ratios to obtain the target weights of each feature component.

[0051] The weight strategy library can be a pre-built database, configuration file, or lookup table, which stores the basic weights of each feature component under various operating scenarios. These basic weights can be obtained based on expert experience, historical data statistical analysis, or by training a large amount of labeled data through machine learning models to characterize the degree of influence of each feature component on the performance score of the smart cockpit in a specific scenario. The offset degree reflects the difference between the current smart cockpit performance (characterized by the first fused feature vector) and the historical performance under the same scenario (characterized by the second fused feature vector). This offset degree can be determined in a variety of ways, such as the difference between the current feature component value and the historical average component value, percentage change, or standardized difference.

[0052] In this embodiment, considering that the performance of the intelligent cockpit varies under different operating scenarios and the user's focus is different, if the weight of each performance indicator remains unchanged, the evaluation results will lack objectivity and reliability. Therefore, a dynamic weight allocation mechanism that integrates the operating scenario and the changes in the intelligent cockpit status is proposed to achieve adaptive adjustment of indicator weights.

[0053] In this embodiment, once the target operating scenario of the intelligent cockpit is determined, the weight strategy library is queried to match the predefined base weights of each feature component under that scenario. For example, in the operating scenario 'Scenario_High Temperature Congestion Navigation', the base weight of the feature component representing the slope of the CPU temperature change trend is 0.3. When determining the weight adjustment ratio for each feature component, if a feature component deviates significantly from its historical performance, it indicates that its impact on the current performance evaluation needs to be amplified, thus requiring an increase in its base weight. The weight adjustment ratio can be determined based on the magnitude of the deviation. For example, if the current value of a feature component is much lower than the historical average, a positive adjustment ratio is needed to highlight its negative impact. After obtaining the weight adjustment ratios corresponding to each feature component, the basic weights of the corresponding feature components in each feature component are adjusted. The adjustment method can be multiplicative adjustment (basic weight multiplied by (1 + weight adjustment ratio)), additive adjustment (basic weight plus weight adjustment ratio), or a more complex nonlinear combination. The adjusted weights are the final target weights used for weighted aggregation. In addition, the adjusted weights need to be normalized as a whole so that the sum of the target weights of each feature component is 1, to ensure that the adjusted target weights still meet the rationality of weight allocation.

[0054] In this embodiment, based on the target operating scenario of the intelligent cockpit, the basic weights of each feature component are obtained from a pre-set weight strategy library, thereby ensuring the scenario adaptability of the weight allocation. At the same time, by calculating the deviation of each feature component relative to the performance of the same historical scenario, and determining the weight adjustment ratio accordingly, dynamic correction of the basic weights is achieved. This weight adjustment mechanism, which combines a pre-set weight strategy library with real-time performance deviation information, enables the final allocated target weights to more accurately reflect the impact of each feature component on the overall performance of the intelligent cockpit under the current operating state, thereby making the comprehensive evaluation results more consistent with the sensitive points and risk points of the actual scenario.

[0055] In one embodiment, there are multiple second fused feature vectors. Determining the weight adjustment ratio corresponding to each feature component includes: calculating the feature mean and feature standard deviation corresponding to each feature component based on the multiple second fused feature vectors; calculating the standard deviation multiple of each feature component relative to the corresponding feature mean based on each feature component and its corresponding feature mean and feature standard deviation; and determining the weight adjustment ratio corresponding to each feature component based on the standard deviation multiple corresponding to each feature component and a preset first mapping relationship, wherein the first mapping relationship characterizes the correspondence between the standard deviation multiple corresponding to each feature component and the weight adjustment ratio.

[0056] The standard deviation multiple is a standardized metric that can intuitively represent the degree to which the current feature component deviates from its historical mean, expressed in units of standard deviation. The first mapping relationship is used to convert the standard deviation multiple of the feature component into the corresponding weight adjustment ratio. This first mapping relationship can be a lookup table that defines the weight adjustment ratio corresponding to different standard deviation multiples, or it can be a function model, such as a linear function or a nonlinear function, which calculates the weight adjustment ratio by inputting the standard deviation multiple.

[0057] In this embodiment, to more accurately measure the performance of each feature component in the first fused feature vector, for each feature component in the first fused feature vector, the corresponding feature value is extracted from all the multiple second fused feature vectors, and these historical feature values ​​are statistically analyzed to calculate the historical average performance level (i.e., feature mean) and fluctuation range (i.e., feature standard deviation) of each feature component. Based on this, to quantify the degree to which the current feature component deviates from the historical average level, the standard deviation multiple of each feature component relative to the corresponding feature mean is calculated. This value can be positive (above average) or negative (below average). The larger the absolute value, the greater the deviation. That is, when the absolute value of the standard deviation multiple is large, it means that the feature component is performing abnormally, and its weight adjustment ratio will be increased accordingly to highlight its impact on the overall performance evaluation. Conversely, if the absolute value of the standard deviation multiple is small, the adjustment ratio may be close to 1, indicating that its impact on the overall performance is consistent with the historical average level. Subsequently, based on the standard deviation multiple corresponding to each feature component and the preset first mapping relationship, the weight adjustment ratio corresponding to each feature component is determined.

[0058] Continuing with the example of the slope of the CPU temperature change trend as a feature component, with a base weight of 0.3, if the corresponding standard deviation is 3, then its base weight is increased by 50%, and the final target weight can be 0.3. 1.5 = 0.45.

[0059] In this embodiment, multiple historical second fusion feature vectors under the same operating scenario are fully utilized. By calculating the feature mean and feature standard deviation in historical evaluations, the normal fluctuation range of each feature component in history is accurately captured, reducing the randomness and bias that may be brought about by single historical data. By calculating the standard deviation multiple, the deviation between the current state and the historical state is quantified in a standardized way, making the deviation between different feature components comparable. Finally, combined with the preset first mapping relationship, the weight of each feature component is dynamically and accurately adjusted according to this quantified deviation, ensuring that the final first comprehensive performance score can more objectively and accurately reflect the actual scenario state of the intelligent cockpit.

[0060] In another embodiment, please refer to Figure 5 , Figure 5 This is a schematic diagram illustrating another optional process for assigning target weights, as shown in an exemplary embodiment of this application. Figure 5As shown, there are multiple second fusion feature vectors. The step of assigning target weights to each feature component in the first fusion feature vector includes at least the following steps: Step S510, calculating the degree of offset of each feature component relative to the corresponding feature component in the multiple second fusion feature vectors; Step S520, inputting the degree of offset between the target running scene and each feature component into a pre-built weight allocation model to obtain the target weight of each feature component; wherein, the degree of offset includes the standard deviation multiple, and the weight allocation model is trained based on second sample data labeled with running scene labels and weight labels.

[0061] The weight allocation model is a pre-trained lightweight neural network that intelligently outputs the target weights for each feature component based on the target operating scenario of the intelligent cockpit and the corresponding offset of each feature component. The second sample data refers to historical multi-source time-series data with clear operating scenario and weight labels. This data is used to extract feature components for training the weight allocation model, enabling it to learn the mapping relationship from the operating scenario and feature offset to the target weights. Furthermore, the weight calculation model is a multilayer perceptron, whose output layer uses the Softmax activation function to ensure that the sum of the target weights for each feature component is 1, thus guaranteeing that the adjusted target weights still satisfy the rationality of the weight allocation.

[0062] In this embodiment, by calculating the degree of deviation of each feature component relative to multiple historical second fusion feature vectors and using the standard deviation multiple as a quantitative indicator, the degree of deviation of the current intelligent cockpit's various performance characteristics from historical norms can be reflected more comprehensively and accurately. On this basis, the target operating scenario and these refined degree of deviation are input into a pre-built weight allocation model, so that the allocation of target weights is no longer a simple preset, but is intelligently adjusted based on a lightweight neural network according to the specific operating scenario and the statistical deviation of each feature component, ensuring that the final first comprehensive performance score can more objectively and accurately reflect the actual scenario state of the intelligent cockpit.

[0063] In one embodiment, please refer to Figure 6 , Figure 6 This is a schematic diagram illustrating an optional feature extraction and fusion process, as shown in an exemplary embodiment of this application. Figure 6As shown, the steps of extracting features from multi-source time-series data and fusing the extracted features to obtain a first fused feature vector include at least the following: Step S610, determining the performance index sequences of multiple predefined performance indicators based on the multi-source time-series data; Step S620, extracting indicator evolution features from the corresponding performance index sequences for each performance indicator, wherein the indicator evolution features include statistical distribution features, trend features, and abnormal deviation features; Step S630, standardizing the multiple indicator evolution features corresponding to each performance indicator, and concatenating the multiple indicator evolution features of all standardized performance indicators to obtain a concatenated feature vector; Step S640, reducing the dimensionality of the concatenated feature vector to obtain the first fused feature vector.

[0064] Among them, predefined performance indicators can cover various aspects of the intelligent cockpit, such as the overall fluency index, voice interaction success rate, map rendering frame rate, application startup time, and system stability index; statistical distribution features are used to describe the overall distribution characteristics of the performance indicator sequence, including mean, median, variance, standard deviation, maximum value, minimum value, and coefficient of variation, reflecting the central tendency, dispersion, and distribution pattern of the performance indicators; change trend features are used to capture the pattern of performance indicator sequence changes over time, including linear / nonlinear trend slope, trend change direction (rising / falling / stable), and periodicity, reflecting whether the performance indicators are in a rising, falling, stable, or periodic state, and reflecting the rate of change; anomaly deviation features are used to identify whether there are significantly deviated data points (i.e., outliers) in the performance indicator sequence and the number of significantly deviated data points, reflecting the frequency of anomalies in the performance indicators.

[0065] In this embodiment, when determining the performance index sequences of multiple predefined performance indicators based on multi-source time-series data, the aim is to transform the raw and complex multi-source time-series data of the intelligent cockpit into a more structured and analyzable performance index sequence. This involves parsing, calculating, or aggregating the multi-source time-series data to generate sequence data showing how the indicators change over time. Based on this, for each performance indicator, statistical distribution characteristics, trend characteristics, and abnormal deviation characteristics are extracted from the corresponding performance index sequence to uncover deeper and more representative features from a single performance index sequence, thus comprehensively characterizing the dynamic behavior and performance of each indicator. Subsequently, due to the potentially significant differences in the dimensions, numerical ranges, and distributions of different performance indicators, direct concatenation could lead to some features having an excessively large or insufficient impact on subsequent analysis. Therefore, the evolution characteristics of multiple indicators corresponding to each performance indicator are first standardized. After standardization, the evolution characteristics of multiple indicators extracted from each performance indicator are horizontally concatenated, thus combining all these standardized features into a single, high-dimensional concatenated feature vector. This concatenated feature vector integrates all the key information regarding the statistical distribution, trends, and abnormal deviations of the various performance indicators of the intelligent cockpit. Finally, considering that the concatenated feature vector may still have a very high dimensionality—for example, if there are 10 performance indicators, each corresponding to 15 features, resulting in a 150-dimensional feature vector—the dimensionality reduction process is further performed on the concatenated feature vector, reducing its dimensionality while retaining the main information and variability.

[0066] In this embodiment, the process progresses step by step, from the original multi-source time-series data to the multi-performance index sequence, and then to the index evolution characteristics and the standardization, splicing and dimensionality reduction of the features. This allows for the systematic extraction and fusion of multi-dimensional dynamic features reflecting the comprehensive performance of the intelligent cockpit from the multi-source heterogeneous time-series data, resulting in a more refined and representative first fusion feature vector. This reduces the information bias that may result from a single feature, providing high-quality input for subsequent weight allocation and comprehensive performance scoring, thereby making the entire performance evaluation method more robust and reliable.

[0067] For example, let the time window length of each performance index sequence be... The data points within the window are The formula for calculating the trend slope is: Equation (1) in, Indicates the slope of the trend; Indicates the length of the time window; Indicates a time index; This represents the mean of the time index; Indicates the first Performance metrics values ​​at each time point; This represents the average of the performance metrics across all points in time within the time window.

[0068] For example, the formula for calculating the coefficient of variation is: Equation (2) in, Indicates the coefficient of variation; This represents the standard deviation of the performance metric across all points in time within the time window. This represents the average of the performance metrics across all time points.

[0069] For example, the outlier identification method uses the four-part interval method, and the condition for being identified as an outlier is: Equation (3) in, Indicates the first Performance metrics values ​​at each time point; Indicates the first quartile; Indicates the interquartile range; Indicates the third quartile; Indicates the boundary of low-side outliers; This indicates the boundary of high-side outliers.

[0070] For example, dimensionality reduction methods include PCA (Principal Component Analysis), LDA (Linear Discriminant Analysis), and t-SNE (t-Distributed Stochastic Neighbor Embedding).

[0071] In one possible embodiment, see Figure 7 , Figure 7 This is a schematic diagram illustrating an optional data preprocessing flow, as shown in an exemplary embodiment of this application. Figure 7 As shown, before feature extraction from multi-source time-series data, the process includes: data cleaning of the multi-source time-series data, including noise reduction and missing value processing; timestamp alignment of the cleaned multi-source time-series data; and normalization of the timestamp-aligned multi-source time-series data to obtain structured multi-source time-series data for feature extraction.

[0072] As a possible implementation, considering that multi-source time-series data of intelligent cockpits typically have characteristics such as high dimensionality, strong heterogeneity, and complex time dependencies, direct feature extraction and fusion can easily lead to the loss of key performance information, increased noise interference, or difficulty in effectively capturing the dynamic evolution of the intelligent cockpit's operating status in different time periods, thus significantly reducing the accuracy and reliability of subsequent performance evaluation. Therefore, the multi-source time-series data is first cleaned, and then the cleaned multi-source time-series data is timestamped to strictly synchronize data from different sources such as hardware status, software status, user interaction, and operating environment to a unified time base, constructing a highly consistent spatiotemporally aligned dataset. On this basis, the timestamped multi-source time-series data is normalized to eliminate the influence of differences in units and inconsistencies in numerical scales, thereby obtaining high-quality raw data.

[0073] As one possible implementation, in handling missing values, a spatiotemporal correlation compensation method (such as neighboring sensor data + historical pattern completion) is adopted. That is, on the one hand, the effective data of neighboring sensors or related subsystems at the same time are used to complete the missing values, and on the other hand, the historical operation mode is combined for collaborative inference to achieve high-precision reconstruction of missing values ​​and complete the missing value.

[0074] In one embodiment, after obtaining the first comprehensive performance score of the intelligent cockpit, the method further includes: determining the contribution of each feature component to the first comprehensive performance score based on each feature component and its corresponding target weight; and, under the condition that the contribution of a first target feature component is less than a preset contribution threshold, determining the first performance bottleneck currently affecting the performance of the intelligent cockpit based on the first target feature component and a preset second mapping relationship; determining the current health level of the intelligent cockpit based on the first comprehensive performance score and a preset third mapping relationship; and, under the condition that the health level reaches a preset warning level, generating a first warning message based on the health level, the first performance bottleneck, and the maintenance measures for the first performance bottleneck; wherein, the second mapping relationship represents the correspondence between each feature component and the performance bottleneck, and the third mapping relationship represents the correspondence between each comprehensive performance score and the health level.

[0075] The second mapping relationship can be a predefined rule base, lookup table, or knowledge graph built based on expert experience. It maps specific feature components (such as high CPU utilization, network latency, low screen refresh rate, etc.) to specific performance bottlenecks (such as insufficient computing resources, communication module failure, display driver problems, etc.). The third mapping relationship divides the continuous comprehensive performance score range into discrete health levels, such as healthy, sub-healthy, etc.

[0076] In this embodiment, after obtaining the first comprehensive performance score of the intelligent cockpit, the contribution of each feature component to the first comprehensive performance score is determined based on each feature component and its corresponding target weight. The contribution can be the product of the feature component and the target weight. Features with negative contributions and large absolute values ​​are the main factors causing a decrease in performance score. After calculating the contribution of each feature component to the first comprehensive performance score, all first target feature components with contributions less than a preset contribution threshold are identified. The first target feature components are the key factors causing poor overall performance score. Once the first target feature components are identified, the second mapping relationship is used to associate them with specific performance bottlenecks, thereby determining the first performance bottleneck currently affecting the performance of the intelligent cockpit. This system transforms abstract performance data into concrete problem descriptions. Simultaneously, based on the first comprehensive performance score and the third mapping relationship, it determines the current health level of the intelligent cockpit. This allows users or maintenance personnel to quickly grasp the overall operational status of the intelligent cockpit without needing to deeply understand the complex scoring mechanism, thus facilitating decision-making. Once the intelligent cockpit's health level reaches a preset warning level, a first warning message is automatically generated. This first warning message not only includes the current health level but also clearly identifies the first performance bottleneck and provides maintenance measures for that bottleneck. These maintenance measures are predefined and associated with specific performance bottlenecks. For example, for a bottleneck due to insufficient computing resources, maintenance measures could include cleaning up background applications or upgrading the processor.

[0077] In this embodiment, based on the comprehensive performance score of the intelligent cockpit, a quantitative correlation model between the comprehensive performance score and the health status of the intelligent cockpit is constructed to achieve an interpretable mapping from score to health status. Furthermore, based on contribution analysis and feature reverse mapping mechanism, the key performance bottlenecks that lead to performance degradation can be accurately located and maintenance measures can be provided. This solves the problem of having only a single score but lacking specific root cause analysis, intelligent early warning, and operation guidance when the score is low, thus enhancing the interpretability and operability of the evaluation results.

[0078] In one embodiment, after obtaining the first comprehensive performance score of the intelligent cockpit, the method further includes: acquiring multiple second comprehensive performance scores from historical performance evaluations, and determining a comprehensive performance score sequence for the intelligent cockpit in future periods based on the multiple second comprehensive performance scores and the first comprehensive performance score; determining the performance change trend and remaining service life of the intelligent cockpit in future periods based on the comprehensive performance score sequence; if there is a period in the performance change trend where the rate of decline is greater than a preset rate threshold, then the period is determined as a performance degradation period, and a fourth fusion feature vector is determined under the performance degradation period based on multiple third fusion feature vectors and the first fusion feature vector from historical performance evaluations; calculating the difference between each feature component and the corresponding feature component in the fourth fusion feature vector, and determining the feature component with a difference greater than a preset difference as a second target feature component, and then determining the second performance bottleneck that will affect the performance of the intelligent cockpit in the future based on the second target feature vector and the second mapping relationship; and generating second early warning information based on the performance degradation period, the second performance bottleneck, and the maintenance measures for the second performance bottleneck.

[0079] In this embodiment, considering that evaluating only the current performance of the smart cockpit may not be able to predict potential performance degradation trends and future performance bottlenecks in a timely manner, thus leading to delays in maintenance work, after obtaining the first comprehensive performance score of the smart cockpit, the system continues to predict the performance change trend and remaining service life of the smart cockpit in the future, and performs corresponding potential performance bottleneck analysis and early warning when the performance deteriorates significantly in the future, thereby achieving predictive maintenance of the smart cockpit.

[0080] In this embodiment, after obtaining the first comprehensive performance score of the smart cockpit, multiple second comprehensive performance scores from historical performance evaluations are obtained to form a known comprehensive performance score sequence. Based on this known comprehensive performance score sequence, the comprehensive performance score sequence of the smart cockpit in future periods is predicted. The prediction method can be implemented using exponential smoothing or a lightweight long short-term memory network model. Based on the comprehensive performance score sequence for future periods, the performance trend of the intelligent cockpit in future periods can be determined (this performance trend is a long-term trend prediction result). For example, by calculating the slope or rate of change of the score sequence (i.e., the trend slope), the overall future performance trend (such as stabilization, improvement, or decline) can be judged. When the score drops to a certain level, it indicates the end of the intelligent cockpit's lifespan, thus obtaining the remaining lifespan of the intelligent cockpit. Subsequently, if there are periods in the performance trend where the rate of decline is greater than a preset threshold, these periods are identified as performance degradation periods. Multiple third-level fusion feature vectors from historical performance evaluations are obtained. Based on these third-level fusion feature vectors and the first-level fusion feature vector, a fourth-level fusion feature vector is predicted for the performance degradation period. This prediction method can be implemented using exponential smoothing or a lightweight long short-term memory network model. This fourth-level fusion feature vector is used for subsequent bottleneck localization. After determining the fourth fused feature vector for the performance degradation period, to identify the specific causes of future performance degradation, each current feature component is compared with its corresponding feature component in the fourth fused feature vector. This involves calculating the difference between them and quantifying the deviation of each feature component. When the difference for a feature component exceeds a preset value, it indicates that the feature component exhibits a significant anomaly during the performance degradation period and is therefore identified as the second target feature component. Next, using a second mapping relationship, at least one second target feature component is mapped to a specific second performance bottleneck. Once the performance degradation period and the specific second performance bottleneck are predicted, a forward-looking second early warning message is automatically generated. This message includes the potential performance degradation time window, clearly identifies potential performance bottlenecks, and provides corresponding maintenance measures, thus achieving a shift from passive response to proactive prediction and maintenance.

[0081] In this embodiment, by analyzing historical and current comprehensive performance scores, the performance change trend and remaining service life of the smart cockpit can be predicted. Before significant performance degradation occurs, specific characteristic components that may lead to performance decline can be identified in advance and mapped as maintainable performance bottlenecks. This provides a basis for proactive optimization and predictive maintenance of the smart cockpit, thereby improving the user experience.

[0082] Furthermore, effectively mapping the performance trends of the intelligent cockpit to its health status and remaining service life provides key data support for vehicle lifecycle management and value assessment, enabling intelligent decision-making. This includes dynamically allocating resources for the intelligent cockpit (e.g., allocating computing power based on predicted load), optimizing OTA (Over-The-Air) strategies (e.g., prioritizing the push of patches with significant performance impact), and optimizing personalized user experiences (e.g., adjusting function recommendations based on performance predictions). By combining real-time performance monitoring, performance trend prediction, health status assessment, remaining service life prediction, performance early warning, and performance bottleneck maintenance, the system extends from diagnosis to predictive maintenance, forming a complete closed loop.

[0083] In one possible embodiment, it also includes short-term situational prediction of the intelligent cockpit's performance score, i.e., based on a known sequence of comprehensive performance scores. Perform linear regression and fit a straight line. Obtain the trend slope , This indicates improved performance. It indicates performance degradation and calculates the slope difference between two consecutive windows to determine the rate of degradation or improvement.

[0084] In one possible embodiment, updating the scene recognition model includes: if the misjudgment rate of the target running scene is greater than a preset probability threshold, constructing a first training sample based on the multi-source time-series data at the time of misjudgment and the corresponding real running scene label; performing online training on the scene recognition model based on the first training sample to iteratively optimize the running scene recognition strategy of the scene recognition model, thereby obtaining a new scene recognition model.

[0085] As one possible implementation, by automatically constructing labeled training samples and triggering online training when the misjudgment rate of the target operating scenario exceeds the standard, the scene recognition model can continuously adapt to changes in the state of the intelligent cockpit, significantly improving the accuracy and generalization ability of scene recognition and avoiding performance evaluation deviations caused by scene misjudgment.

[0086] In one possible embodiment, updating the weight allocation model includes: if the prediction error between the first comprehensive performance score obtained based on the target weight and the actual comprehensive performance score exceeds a preset error threshold a certain number of times, then constructing a second training sample based on the target running scenario, the offset degree of each feature component and the corresponding true weight label; performing online training on the weight allocation model based on the second training sample to iteratively optimize the weight allocation strategy of the weight allocation model, thereby obtaining a new weight allocation model.

[0087] As one possible implementation, when the prediction error of the comprehensive performance score frequently exceeds the limit, the weight allocation strategy is dynamically optimized based on real feedback, so that the weight of each feature component can adapt to the precise changes in different operating scenarios, significantly improving the accuracy and generalization ability of performance evaluation.

[0088] The aforementioned intelligent cockpit performance evaluation method first acquires multi-source time-series data of the intelligent cockpit, including hardware status data, software status data, user interaction data, and operating environment data. Next, it extracts features from the multi-source time-series data and fuses these features to obtain a first fused feature vector. Then, it determines the target operating scenario of the intelligent cockpit based on the multi-source time-series data and retrieves a second fused feature vector from historical performance evaluations of the same operating scenario. Furthermore, based on the target operating scenario, the first fused feature vector, and the second fused feature vector, it assigns target weights to each feature component in the first fused feature vector. Finally, it weights and aggregates each feature component according to its target weight to obtain a first comprehensive performance score for the intelligent cockpit. By acquiring multi-source time-series data of the intelligent cockpit and fusing multi-dimensional data from hardware, software, users, and the environment for performance evaluation, it can comprehensively capture the state of the intelligent cockpit. Moreover, during the evaluation process, it can dynamically assign weights to each state feature based on the intelligent cockpit's operating scenario, thereby generating a performance score. This makes the evaluation results more scenario-adaptive, ensuring the objectivity and reliability of the evaluation and providing a reliable data foundation for the optimization and maintenance of the intelligent cockpit.

[0089] Please see Figure 8 , Figure 8 This is a block diagram illustrating an optional intelligent cockpit performance evaluation system, as shown in an exemplary embodiment of this application. This system can be applied to... Figure 1 The implementation environment shown is intended to illustrate the system, but it should be understood that the system can also be applied to other exemplary implementation environments. This embodiment does not limit the implementation environment to which the system is applicable.

[0090] like Figure 8 As shown, in an exemplary embodiment, the intelligent cockpit performance evaluation system 800 includes at least a data acquisition module 810, a data processing module 820, a weight allocation module 830, and a performance scoring module 840, which are described in detail below: The data acquisition module 810 is used to acquire multi-source time-series data of the intelligent cockpit, including hardware status data, software status data, user interaction data and operating environment data. The data processing module 820 is used to extract features from multi-source time-series data and fuse the extracted features to obtain a first fused feature vector. The weight allocation module 830 is used to determine the target operating scenario of the intelligent cockpit based on multi-source time series data, and retrieve the second fusion feature vector under the same operating scenario in the historical performance evaluation based on the target operating scenario. Then, based on the target operating scenario, the first fusion feature vector and the second fusion feature vector, the target weight is assigned to each feature component in the first fusion feature vector. The performance scoring module 840 is used to perform weighted aggregation of each feature component according to the target weight of each feature component to obtain the first comprehensive performance score of the smart cockpit.

[0091] It should be noted that the intelligent cockpit performance evaluation system provided in the above embodiments and the intelligent cockpit performance evaluation method provided in the above embodiments belong to the same concept. The content of the operation performed by each module has been described in detail in the method embodiments, and will not be repeated here.

[0092] Please see Figure 9 , Figure 9 This is a schematic diagram illustrating the structure of an optional electronic device, as shown in an exemplary embodiment of this application. Figure 9 A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown. It should be noted that... Figure 9 The computer system 900 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0093] like Figure 9 As shown, the computer system 900 includes a Central Processing Unit (CPU) 901, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on programs stored in Read-Only Memory (ROM) 902 or programs loaded from storage portion 908 into Random Access Memory (RAM) 903. The RAM 903 also stores various programs and data required for system operation. The CPU 901, ROM 902, and RAM 903 are interconnected via a bus 904. An Input / Output (I / O) interface 905 is also connected to the bus 904.

[0094] The following components are connected to I / O interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to I / O interface 905 as needed. Removable media 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 910 as needed so that computer programs read from them can be installed into storage section 908 as needed.

[0095] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 909, and / or installed from removable medium 911. When the computer program is executed by central processing unit (CPU) 901, it performs various functions defined in the system of this application.

[0096] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for evaluating the performance of an intelligent cockpit, characterized in that, The method includes: Acquire multi-source time-series data of the intelligent cockpit, including hardware status data, software status data, user interaction data, and operating environment data; Feature extraction is performed on the multi-source time-series data, and the extracted features are fused to obtain a first fused feature vector; The target operating scenario of the intelligent cockpit is determined based on the multi-source time series data, and the second fusion feature vector under the same operating scenario in the historical performance evaluation is retrieved based on the target operating scenario. Then, target weights are assigned to each feature component in the first fusion feature vector based on the target operating scenario, the first fusion feature vector and the second fusion feature vector. Based on the target weights of each feature component, the feature components are weighted and aggregated to obtain the first comprehensive performance score of the smart cockpit.

2. The intelligent cockpit performance evaluation method according to claim 1, characterized in that, The target operating scenario of the intelligent cockpit is determined based on the multi-source time-series data, including: Extract scene discrimination features from the multi-source time-series data; If the scene discrimination feature satisfies the rule conditions corresponding to any running scene in the preset scene rule base, then the arbitrary running scene is determined as the target running scene; If the arbitrary running scenario does not exist in the scenario rule base, the scenario discrimination feature is input into the pre-built scenario recognition model to obtain the probability distribution of various running scenarios, and the target running scenario is determined based on the probability distribution of various running scenarios. The scenario rule base includes rule conditions that scenario discrimination features must meet under various operating scenarios, and the scenario recognition model is trained based on the first sample data labeled with the operating scenario label.

3. The intelligent cockpit performance evaluation method according to claim 1, characterized in that, Assigning target weights to each feature component in the first fused feature vector, including: Based on the target operating scenario, the basic weights corresponding to each feature component are determined from a preset weight strategy library, wherein the weight strategy library includes the basic weights of each feature component under various operating scenarios. The weight adjustment ratio corresponding to each feature component is determined based on the degree of offset of each feature component relative to the corresponding feature component in the second fused feature vector; According to the weight adjustment ratio, the basic weights of the corresponding feature components in each feature component are adjusted to obtain the target weights of each feature component.

4. The intelligent cockpit performance evaluation method according to claim 3, characterized in that, The second fused feature vector consists of multiple components. Determining the weight adjustment ratio corresponding to each feature component includes: Based on multiple second fusion feature vectors, calculate the feature mean and feature standard deviation corresponding to each feature component; Based on each feature component and its corresponding feature mean and feature standard deviation, calculate the standard deviation multiple of each feature component relative to its corresponding feature mean; Based on the standard deviation multiple corresponding to each feature component and the preset first mapping relationship, the weight adjustment ratio corresponding to each feature component is determined, wherein the first mapping relationship characterizes the correspondence between the standard deviation multiple corresponding to each feature component and the weight adjustment ratio.

5. The intelligent cockpit performance evaluation method according to claim 1, characterized in that, The second fused feature vector comprises multiple vectors, and assigning target weights to each feature component in the first fused feature vector further includes: Calculate the degree of offset of each feature component relative to the corresponding feature component in the plurality of second fused feature vectors; The target running scenario and the degree of offset corresponding to each feature component are input into a pre-built weight allocation model to obtain the target weight of each feature component; The degree of offset includes the standard deviation multiple, and the weight allocation model is trained based on second sample data labeled with running scenario labels and weight labels.

6. The intelligent cockpit performance evaluation method according to claim 1, characterized in that, Feature extraction is performed on the multi-source time-series data, and the extracted features are fused to obtain a first fused feature vector, including: Based on the multi-source time-series data, determine the performance index sequence of each of the predefined multiple performance indices; For each performance indicator, the indicator evolution features are extracted from the corresponding performance indicator sequence, wherein the indicator evolution features include statistical distribution features, trend features and abnormal deviation features. The evolution characteristics of various performance indicators are standardized, and the evolution characteristics of various performance indicators after standardization are concatenated to obtain a concatenated feature vector. The spliced ​​feature vector is subjected to dimensionality reduction processing to obtain the first fused feature vector.

7. The intelligent cockpit performance evaluation method according to any one of claims 1 to 6, characterized in that, After receiving the first overall performance score for the smart cockpit, it also includes: Based on each feature component and its corresponding target weight, the contribution of each feature component to the first comprehensive performance score is determined. If the contribution of the first target feature component among the feature components is less than a preset contribution threshold, the first performance bottleneck affecting the performance of the intelligent cockpit is determined based on the first target feature component and a preset second mapping relationship. Based on the first comprehensive performance score and the preset third mapping relationship, the current health level of the intelligent cockpit is determined, and when the health level reaches the preset warning level, a first warning message is generated based on the health level, the first performance bottleneck and the maintenance measures for the first performance bottleneck. The second mapping relationship represents the correspondence between each feature component and the performance bottleneck, and the third mapping relationship represents the correspondence between each comprehensive performance score and the health level.

8. The intelligent cockpit performance evaluation method according to claim 7, characterized in that, After receiving the first overall performance score for the smart cockpit, it also includes: Obtain multiple second comprehensive performance scores from historical performance evaluations, and determine the comprehensive performance score sequence of the intelligent cockpit in future time periods based on the multiple second comprehensive performance scores and the first comprehensive performance score; Based on the comprehensive performance scoring sequence, determine the performance change trend of the intelligent cockpit and the remaining service life of the intelligent cockpit in the future period; If there is a period in the performance change trend where the rate of decline is greater than a preset rate threshold, then the period is determined as a performance degradation period, and a fourth fusion feature vector is determined based on multiple third fusion feature vectors in historical performance evaluations and the first fusion feature vector. Calculate the difference between each feature component and the corresponding feature component in the fourth fused feature vector, and determine the feature component whose difference is greater than a preset difference as the second target feature component. Then, based on the second target feature vector and the second mapping relationship, determine the second performance bottleneck that will affect the performance of the smart cockpit in the future. A second early warning message is generated based on the performance degradation period, the second performance bottleneck, and the maintenance measures for the second performance bottleneck.

9. A smart cockpit performance evaluation system, characterized in that, The system includes: The data acquisition module is used to acquire multi-source time-series data of the intelligent cockpit, including hardware status data, software status data, user interaction data, and operating environment data. The data processing module is used to extract features from the multi-source time-series data and fuse the extracted features to obtain a first fused feature vector. The weight allocation module is used to determine the target operating scenario of the intelligent cockpit based on the multi-source time series data, retrieve the second fusion feature vector under the same operating scenario in the historical performance evaluation based on the target operating scenario, and then allocate target weights to each feature component in the first fusion feature vector based on the target operating scenario, the first fusion feature vector and the second fusion feature vector. The performance scoring module is used to perform weighted aggregation of the feature components according to the target weight of each feature component to obtain the first comprehensive performance score of the smart cockpit.

10. An electronic device, characterized in that, include: processor; A storage device for storing a program that, when executed by the processor, causes the electronic device to implement the intelligent cockpit performance evaluation method as described in any one of claims 1 to 8.

Images