Intelligent cockpit evaluation method and system

Abstract

The invention relates to the technical field of smart cabins, and particularly discloses a smart cabin evaluation method and system, and the method comprises the steps: obtaining user feedback data after scene labeling; analyzing to obtain an evaluation tendency, and converting the evaluation tendency into scenarized feedback data; obtaining a multi-source data association analysis result; screening based on a multi-source data association analysis result to obtain an index system; distributing basic weights for indexes in the index system; scene weight factors are distributed to indexes in the index system; according to the index system, the basic weight and the scene weight factor, establishing an evaluation system for evaluation; according to the intelligent cockpit evaluation method provided by the invention, the balance of the evaluation indexes between the static importance and the dynamic scene demand is realized by dynamically fusing the scene analysis fed back by the user, the multi-source data association integration and the double-weight distribution mechanism; the problems that in a traditional method, the utilization rate of unstructured user feedback data is low, scene relevance is missing, data dimensions are isolated, and weights are solidified are solved.

Description

Technical Field

[0001] The present application relates to the field of smart cockpit technology, and more specifically, to a smart cockpit evaluation method and system. Background Art

[0002] With the rapid development of intelligent automotive technology, the smart cockpit, as the core vehicle for human-machine interaction, has become a crucial component of automotive intelligent systems. Modern smart cockpits integrate diverse functional modules, including voice interaction systems, digital key management modules, and connected car ecosystems. These systems utilize multimodal interaction technologies to create an intelligent cockpit environment. Their implementation involves multiple technical areas, including in-vehicle operating systems, artificial intelligence algorithms, and IoT communication protocols.

[0003] Smart cockpits have become the core vehicle for human-machine interaction, and their performance directly impacts user experience and driving safety. Traditional smart cockpit evaluation technology suffers from the following core flaws in complex and dynamic driving scenarios: First, the utilization rate of unstructured user feedback data is low, making it difficult for existing methods to parse the implicit scenario-related information contained in the text. This results in the submersion of scenario-specific pain points such as entertainment system freezes on congested roads and glare on the navigation interface at night. Second, evaluation dimensions are isolated, and there is a lack of a unified mapping framework between expert experience, historical data, and user feedback, which can easily lead to mismatched functional priorities. Third, static weighting mechanisms cannot adapt to changing scenario requirements. For example, the importance of parking assistance functions in narrow parking spaces is far higher than in open roads, but traditional weighting systems cannot be dynamically adjusted. These issues collectively lead to a deviation between evaluation results and user experience, making it difficult to guide the scenario-adaptive optimization of smart cockpit systems.

[0004] There is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this application is to provide a smart cockpit evaluation method and system to solve the problems of low utilization of unstructured user feedback data, lack of scene relevance, isolated data dimensions and solidified weights in traditional methods.

[0006] In a first aspect, the present application provides a smart cockpit evaluation method for evaluating the performance of a smart cockpit system in a complex dynamic driving environment. The method comprises the following steps: S1. Obtain user feedback data stream; S2. Label the user feedback data stream by driving scene category to obtain the labeled user feedback data; S3. Analyze and collect the user feedback data after scene annotation to determine the evaluation trends of users for different smart cockpit functional points in different scenarios, and convert the evaluation trends into scenario-based feedback data; S4. Align and integrate the scenario-based feedback data, expert opinion data, and historical evaluation data to obtain multi-source data correlation analysis results; S5. Screening the functional requirements list based on the multi-source data association analysis results to obtain an indicator system; S6. Assigning basic weights to the indicators in the indicator system based on expert opinion data and historical evaluation data; S7. Analyze the occurrence frequencies of function points corresponding to different indicators in different scenarios based on the scenario-based feedback data to assign scenario weight factors to the indicators in the indicator system; S8. Establish an evaluation system based on the indicator system, the basic weights and the scenario weight factors for evaluation.

[0007] The smart cockpit evaluation method of this application achieves a balance between the static importance and dynamic scenario requirements of evaluation indicators through scenario analysis of dynamic fusion of user feedback, multi-source data correlation integration, and dual weight allocation mechanism, solving the problems of low utilization of unstructured user feedback data, lack of scenario relevance, isolated data dimensions, and solidified weights in traditional methods.

[0008] The smart cockpit evaluation method further includes the following steps between step S5 and step S6: SA, refer to expert opinion data and historical evaluation data to refine and grade the indicator system.

[0009] Step SA effectively solves the problem that the indicator system still has a single dimension and insufficient evaluation granularity after screening, and avoids the situation where directly assigning basic weights may not accurately reflect the differences in functional performance under different vehicle models and usage scenarios, thereby preventing the evaluation results from deviating from the actual user experience.

[0010] In the smart cockpit evaluation method, step S6 includes: S61. Assign basic weights to the indicators in the indicator system based on the importance scores of the corresponding indicators from the expert opinion data and the scores of the corresponding indicators from the historical evaluation data.

[0011] This basic weight allocation method integrates subjective experience and objective statistics, so that the basic weight reflects both expert cognition and actual data patterns, avoids the bias of a single data source, effectively integrates subjective expert experience and objective data statistical results, and solves the problems of excessive human factors or insufficient data representativeness in the traditional weight allocation process.

[0012] In the smart cockpit evaluation method, the user feedback data stream includes unstructured text and scores, and step S2 includes: S21. Acquire vehicle sensor data based on vehicle sensors, where the vehicle sensor data includes vehicle speed, steering angle, ambient light intensity, and traffic flow data; S22. Determine a driving scene type by matching the vehicle sensor data based on a preset rule base, and annotate the user feedback data stream with the driving scene type to form scene-annotated user feedback data; S23. When the matching in step S22 fails, extract driving scene terms from the unstructured text based on a preset keyword library, and use the driving scene terms to annotate the user feedback data stream to form scene-annotated user feedback data.

[0013] In the smart cockpit evaluation method, step S3 includes: S31. Extracting text fragments related to smart cockpit functional points from the scene-annotated user feedback data, and determining the user's evaluation polarity of each functional point based on a pre-trained sentiment analysis model. The evaluation polarity is categorized as positive, negative, or neutral. S32. Counting the number of user feedbacks, the proportion of positive feedback, and the proportion of negative feedback for different functional points in each scenario according to the preset driving scenario categories; S33: Normalize the feedback quantity, positive feedback ratio, and negative feedback ratio to form the scenario-based feedback data.

[0014] The smart cockpit evaluation method, wherein step S3 further includes: S34. Calculate the ratio of the number of feedbacks for each functional point in each scenario to the total number of feedbacks, and use this ratio as a scenario relevance indicator. S35. Perform weighted summation of the scenario relevance index, positive feedback ratio, and negative feedback ratio based on a preset proportional coefficient to obtain a comprehensive evaluation score for each function point of the user in a specific scenario.

[0015] In the smart cockpit evaluation method, step S4 includes: S41. Align and integrate the scenario-based feedback data, expert opinion data, and historical evaluation data based on a preset indicator mapping relationship table to obtain multi-source data association analysis results. The indicator mapping relationship table includes mapping relationships between different functional points and data indicators in different data sources.

[0016] In the smart cockpit evaluation method, step S41 includes: S411. Align the scenario-based feedback data, the importance scores in the expert opinion data, and the scores in the historical evaluation data according to the indicator mapping relationship table to ensure that the data from different data sources correspond to the same functional point. S412. Integrate the aligned data to obtain multi-source data association analysis results.

[0017] The smart cockpit evaluation method, wherein step S5 includes: S51. Calculate the comprehensive score for each function point in different driving scenarios based on the results of the multi-source data correlation analysis. The comprehensive score is obtained by weighted average of the scenario-based feedback data, expert opinion data, and historical evaluation data. S52. Screen the function points whose comprehensive scores are higher than the preset multi-source data fusion threshold to obtain the initial indicator system; S53. Calculate the variance of the feedback quantity in the scenario-based feedback data for each functional point under different driving scenarios for the initial indicator system. If the variance is greater than a preset scenario difference threshold, retain the corresponding indicator in the initial indicator system; otherwise, remove the corresponding indicator, thereby determining the indicator system.

[0018] Secondly, the present application also provides a smart cockpit evaluation system for evaluating the performance of a smart cockpit system in a complex dynamic driving environment, the system comprising: Data acquisition module, used to obtain user feedback data stream; The scene labeling module is used to label the user feedback data stream with driving scene categories and obtain the user feedback data after scene labeling; A data quantification module is used to analyze the user feedback data after statistical scene annotation for the evaluation trends of different smart cockpit functional points in different scenarios, and convert the evaluation trends into scenario-based feedback data; A data integration module is used to align and integrate the scenario-based feedback data, expert opinion data, and historical evaluation data to obtain multi-source data correlation analysis results; An indicator screening module, configured to screen the functional requirements list based on the multi-source data association analysis results to obtain an indicator system; A weight allocation module, for allocating basic weights to the indicators in the indicator system based on expert opinion data and historical evaluation data; A factor configuration module is used to analyze the frequency of occurrence of function points corresponding to different indicators in different scenarios based on scenario-based feedback data to assign scenario weight factors to the indicators in the indicator system; The evaluation system module is used to establish an evaluation system based on the indicator system, the basic weight and the scenario weight factor for evaluation.

[0019] The smart cockpit evaluation system of this application achieves a balance between the static importance and dynamic scenario requirements of evaluation indicators through scenario analysis of dynamic integration of user feedback, multi-source data correlation integration, and dual weight distribution mechanism, solving the problems of low utilization of unstructured user feedback data, lack of scenario relevance, isolated data dimensions, and solidified weights in traditional systems.

[0020] From the above, it can be seen that the present application provides a smart cockpit evaluation method and system, wherein the smart cockpit evaluation method of the present application constructs a dynamic scenario-driven smart cockpit evaluation system, which achieves a balance between the static importance and dynamic scenario requirements of evaluation indicators through scenario analysis of dynamic fusion of user feedback, multi-source data correlation integration and dual weight distribution mechanism, and solves the problems of low utilization rate of unstructured user feedback data, lack of scenario relevance, isolated data dimensions and weight solidification in traditional methods. It has the advantages of accurately identifying performance defects of smart cockpit systems in complex dynamic driving environments, dynamically adjusting evaluation weights to adapt to different scenario requirements, and improving the consistency between evaluation results and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of a smart cockpit evaluation method provided for some embodiments of the present application.

[0022] Figure 2 A flowchart of the smart cockpit evaluation method provided for some further embodiments of the present application.

[0023] Figure 3 A schematic diagram of the structure of the smart cockpit evaluation system provided for some embodiments of the present application.

[0024] Figure 4 A schematic structural diagram of the smart cockpit evaluation system provided for some further embodiments of the present application.

[0025] Figure numerals: 201, data acquisition module; 202, scene labeling module; 203, data quantification module; 204, data integration module; 205, indicator screening module; 206, weight allocation module; 207, factor configuration module; 208, evaluation system module; 209, indicator optimization module. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0027] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0028] First, please refer to Figure 1 and Figure 2 Some embodiments of the present application provide a smart cockpit evaluation method for evaluating the performance of a smart cockpit system in a complex dynamic driving environment. The method includes the following steps: S1. Obtain user feedback data stream; S2. Label the user feedback data stream by driving scene category to obtain the labeled user feedback data; S3. Analyze and collect user feedback data after scene annotation for user evaluation trends for different smart cockpit functional points in different scenarios, and convert the evaluation trends into scenario-based feedback data. S4. Align and integrate scenario-based feedback data, expert opinion data, and historical evaluation data to obtain multi-source data correlation analysis results; S5. Screen the functional requirements list based on the results of multi-source data correlation analysis to obtain the indicator system; S6. Assign basic weights to the indicators in the indicator system based on expert opinion data and historical evaluation data; S7. Analyze the frequency of occurrence of functional points corresponding to different indicators in different scenarios based on scenario-based feedback data to assign scenario weight factors to the indicators in the indicator system; S8. Establish an evaluation system based on the indicator system, basic weights and scenario weight factors for evaluation.

[0029] Specifically, the user feedback data stream refers to a collection of real-time or offline data collected from users about their experience using smart cockpit functions. This can be achieved by combining unstructured text, rating data, and vehicle sensor data. The ability to parse unstructured user feedback can be improved by integrating multi-dimensional data sources.

[0030] More specifically, driving scene category labeling refers to the operation of classifying user feedback into scenarios based on vehicle operating status and environmental parameters. This can be achieved by matching sensor data with a preset rule library or identifying text information with a keyword library, and establishing an association between user feedback and the driving environment through scene labels.

[0031] More specifically, evaluation tendency conversion refers to the process of quantifying users' subjective feedback into analyzable structured data. This can be achieved by using a sentiment analysis model to determine text polarity and combining it with statistical feedback ratios. Quantification can improve the computability of user feedback data.

[0032] More specifically, the results of multi-source data association analysis can be obtained by using an indicator mapping relationship table to align data, which serves to enhance the systematic nature of the evaluation dimensions through multi-dimensional data cross-validation.

[0033] More specifically, basic weight allocation refers to the process of determining the initial importance of indicators based on expert experience and historical data. It can be achieved by weighted calculation of the correlation between importance scores and historical scores, and by balancing the rationality of indicator weight allocation through the integration of subjective judgment and objective data.

[0034] More specifically, scenario weight factor allocation refers to a mechanism for adjusting indicator weights based on the frequency of occurrence of functional points in different scenarios. The method of this application improves the adaptability of the evaluation system to complex driving environments through dynamic adjustment of weights.

[0035] More specifically, the smart cockpit evaluation method of the present embodiment includes a full process of data collection, scenario annotation, evaluation analysis, multi-source data integration, and dynamic weight assignment. Its operating principle is as follows: First, raw data containing unstructured text and ratings is acquired from the user feedback data stream. Next, driving scenario categories are annotated, and user feedback is associated with specific driving scenarios. A sentiment analysis model is used to extract the polarity of function point evaluations from the annotated data. The number of feedback points and the positive / negative ratio for different scenarios are counted to generate normalized scenario-based feedback data. This scenario-based feedback data is aligned and integrated with expert opinion data and historical evaluation data using an indicator mapping table to generate multi-source data correlation analysis results. Based on this multi-source data correlation analysis, an indicator system is generated to screen functional requirements. Basic weights are assigned based on the expert importance ratings and historical data relevance. Scenario weight factors are then calculated based on the frequency of function points in the corresponding scenarios. Finally, a dynamic evaluation system is established by combining the indicator system, basic weights, and scenario weight factors to implement scenario adaptability assessment. This approach effectively addresses the low utilization rate of unstructured user feedback data, enabling accurate identification of functional requirements through scenario annotation and multi-source data alignment. A dynamic weighting mechanism enables the evaluation system to adapt to changes in function importance in different scenarios, for example, automatically enhancing the sensitivity of visual interaction indicators in low-light conditions. Multi-dimensional data integration eliminates information silos, ensuring that performance fluctuations of newly deployed functions are captured promptly, avoiding the misjudgment of function priorities that can occur with traditional static weighting systems.

[0036] The smart cockpit evaluation method of the embodiment of the present application constructs a dynamic scenario-driven smart cockpit evaluation system. It achieves a balance between the static importance and dynamic scenario requirements of evaluation indicators through scenario analysis of dynamic integration of user feedback, multi-source data correlation integration, and a dual-weight distribution mechanism. It solves the problems of low utilization of unstructured user feedback data, lack of scenario relevance, isolated data dimensions, and weight solidification in traditional methods. It has the advantages of accurately identifying performance defects of smart cockpit systems in complex dynamic driving environments, dynamically adjusting evaluation weights to adapt to different scenario requirements, and improving the consistency between evaluation results and user experience.

[0037] In some preferred embodiments, the following steps are further included between step S5 and step S6: SA, refer to expert opinion data and historical evaluation data to refine and grade the indicator system.

[0038] Specifically, step SA effectively solves the problem that the indicator system still has a single dimension and insufficient evaluation granularity after screening, and avoids the situation where directly assigning basic weights may not accurately reflect the differences in functional performance under different vehicle models and usage scenarios, thereby preventing the evaluation results from deviating from the actual user experience.

[0039] More specifically, the main purpose of this step is to expand the dimensions and divide the levels of the indicator system with reference to expert opinion data and historical evaluation data.

[0040] In some preferred embodiments, step SA comprises: SA1. Refer to expert opinion data to determine the detailed dimensions of each indicator in the indicator system, including performance, security, and user experience; SA2. Based on historical evaluation data, analyze the score distribution of each indicator on different vehicle models, identify the indicators that need to be graded in the indicator system, set grading thresholds based on the score distribution, and divide the indicators into multiple evaluation levels to form a detailed and graded indicator system.

[0041] Specifically, step SA1 selects three core dimensions of performance, security and user experience when determining the refined dimensions. These three dimensions correspond to technical implementation reliability, system security assurance capability and human-computer interaction comfort, respectively. The performance dimension can cover objective technical parameters such as voice interaction response time, the security dimension can evaluate functional reliability such as emergency alarm trigger rate, and the user experience dimension can target subjective perception indicators such as interface operation smoothness to form a multi-dimensional evaluation system.

[0042] More specifically, the grading process of step SA2 can use the percentile method to set the grading threshold, for example, the interval with the indicator score in the top 25% is set as grade A, the middle 50% is set as grade B, and the bottom 25% is set as grade C.

[0043] More specifically, the aforementioned approach, through the refinement of dimensions and a data-driven grading mechanism, enables the index system to comprehensively cover evaluation factors while dynamically adapting to the technical characteristics of different vehicle models, effectively resolving the issue of insufficient granularity in indicators caused by isolated evaluation dimensions. The method of this application ensures systematic coverage of key factors such as performance, safety, and user experience through a multi-dimensional refinement framework determined through expert collaboration (with reference to expert opinion data). Furthermore, the method of this application, based on the dynamic division of indicator levels based on historical data, objectively reflects the differences in technical capabilities of different vehicle models, enabling the evaluation results to accurately identify weaknesses in the voice interaction system in terms of response speed and false trigger control, providing manufacturers with precise improvement directions for optimizing voice interaction performance in specific scenarios.

[0044] In some preferred embodiments, step S6 includes: S61. Assign basic weights to the indicators in the indicator system based on the importance scores of the corresponding indicators from expert opinion data and the scores of the corresponding indicators from historical evaluation data.

[0045] Specifically, in step S61, the basic weight allocation process involves two data dimensions: the subjective rating of the expert on the importance of the indicator and the objective statistical results of the historical evaluation data. The importance rating is obtained through an expert questionnaire or the Delphi method, and each indicator is independently rated by multiple experts; the correlation between the indicator score and the overall score in the historical evaluation data is quantified by the Pearson correlation coefficient, and the absolute value of the correlation coefficient reflects the degree of influence of the indicator on the overall performance. This basic weight allocation method integrates subjective experience and objective statistics, so that the basic weight reflects both expert cognition and actual data patterns, avoids the bias of a single data source, effectively integrates subjective expert experience and objective data statistical results, and solves the problem of excessive human factors or insufficient data representativeness in the traditional weight allocation process.

[0046] In some preferred embodiments, step S61 includes: S611. Obtain experts' scores on the importance of each indicator in the indicator system, as well as the Pearson correlation coefficient between the scores of each indicator and the overall score in the historical evaluation data; S612. Perform weighted calculation on the normalized result of the importance score and the absolute value of the Pearson correlation coefficient based on the preset weights to generate the basic weight of each indicator in the indicator system.

[0047] Specifically, the normalization of importance scores can use the linear normalization method to map the original scores to the [0,1] interval. The Pearson correlation coefficient is calculated by statistically analyzing the covariance and standard deviation of the single indicator scores and the overall scores in the historical evaluation data. Its absolute value reflects the linear correlation strength of the indicator to the overall performance.

[0048] More specifically, the preset weights are set to a fixed ratio based on expert experience, for example, the expert weight accounts for 70% and the correlation coefficient weight accounts for 30%, to ensure a balance between subjective judgment and objective data.

[0049] More specifically, the normalized importance score and the absolute value of the correlation coefficient are used to generate the final basic weight through a weighted summation formula, which can be expressed as: basic weight = α×normalized importance score + β×|Pearson correlation coefficient|, where α and β are preset weights, satisfying α+β=1.

[0050] More specifically, this approach combines expert experience with statistical data, preserving domain knowledge while quantifying actual impact, ensuring both professionalism and objectivity in weight allocation. Adjusting the α and β ratios can adapt to the needs of different evaluation stages, and the weight of the correlation coefficient can be increased when it is necessary to enhance the actual user experience.

[0051] More specifically, the method in this application effectively integrates expert subjective judgment with objective data correlation, retaining the professional understanding of function importance held by domain experts while verifying the objective correlation between indicators and overall system performance through historical data. This weight allocation method avoids the subjective bias that may arise from relying solely on expert experience, while overcoming the shortcomings of purely data-driven methods in insufficiently evaluating emerging functions. This method ensures that the basic weights are both professionally guided and data-supported, establishing a scientific and reliable benchmark for the dynamic adjustment of subsequent scenario weights.

[0052] In some preferred embodiments, the user feedback data stream includes unstructured text and ratings, and step S2 includes: S21. Acquire vehicle sensor data based on vehicle sensors, where the vehicle sensor data includes vehicle speed, steering angle, ambient light intensity, and traffic flow data; S22. Determine the driving scene type by matching the vehicle sensor data based on a preset rule base, and annotate the user feedback data stream with the driving scene type to form scene-annotated user feedback data; S23. When the matching in step S22 fails, unstructured text is identified based on a preset keyword library to extract driving scene terms, so as to annotate the user feedback data stream with the driving scene terms to form scene-annotated user feedback data.

[0053] Specifically, vehicle sensor data includes four types of parameters: vehicle speed, steering angle, ambient light intensity, and traffic flow data. The vehicle speed parameter is in kilometers per hour, the steering angle is recorded in degrees, the steering wheel angle is quantified using lux values, and the traffic flow data is obtained by counting the surrounding vehicle density through on-board cameras.

[0054] More specifically, the preset rule base stores the correspondence between the vehicle speed threshold range and the driving scenario. For example, a vehicle speed below 15 km / h and a continuously changing steering angle correspond to a congested road scenario, and a vehicle speed of 0 and an ambient light intensity below 10 lux correspond to a nighttime parking scenario.

[0055] More specifically, the keyword library includes phrases containing scene elements such as "tunnel navigation reflection" and "touch failure in rainy days", and the named entity recognition algorithm in natural language processing technology is used to extract scene terms from unstructured text.

[0056] More specifically, when the vehicle sensor data meets the preset rules, the driving scene type is directly determined by the physical quantity threshold. For example, if the vehicle speed is detected to be in the range of 0-5 km / h and the steering angle change frequency exceeds 20 times per minute, it is marked as a narrow parking space parking scene. When the sensor data fails to trigger the rule match, the user feedback text content is analyzed. For example, if "voice recognition delay during heavy rain" appears in the user comment, "heavy rain" is identified as an extreme weather scene term through the keyword library. Furthermore, the light intensity data is cross-validated with the description of "screen glare" in the text to improve the accuracy of night scene annotation. This technical solution combines objective sensor data with subjective user descriptions through multimodal data fusion to ensure that scene annotation can still be completed in complex driving environments. For example, when the sensor fails or the rule library does not cover emerging scenarios, supplementary annotation is achieved through text keywords, thereby solving the problem of missing scene associations caused by low utilization of unstructured data.

[0057] More specifically, through the above-mentioned processing, the method of this application effectively solves the problem of low utilization of unstructured user feedback data and realizes the accurate extraction of driving scene information. Through the dual verification mechanism of sensor data and text analysis, it ensures the accurate identification of various scenario problems in the complex and changing actual driving environment, and provides a reliable data foundation for the subsequent allocation of scenario-based indicator weights. It further avoids the omission of user pain point analysis caused by the lack of scene annotation in traditional methods, and significantly improves the adaptability of the evaluation system to dynamic driving environments.

[0058] In some preferred embodiments, step S3 includes: S31. Extract text snippets related to smart cockpit features from the scenario-annotated user feedback data. Use a pre-trained sentiment analysis model to determine the polarity of user evaluations of each feature, categorizing evaluation polarity as positive, negative, or neutral. S32. Counting the number of user feedbacks, the proportion of positive feedback, and the proportion of negative feedback for different functional points in each scenario according to the preset driving scenario categories; S33. Normalize the number of feedbacks, the positive feedback ratio, and the negative feedback ratio to form scenario-based feedback data.

[0059] Specifically, the extraction of text fragments is achieved through natural language processing technology, and the named entity recognition algorithm is used to locate functional point keywords.

[0060] More specifically, the sentiment analysis model uses a pre-trained model based on the Transformer architecture. After inputting a text fragment, it outputs a probability distribution and divides it into three categories of evaluation polarity according to a preset threshold.

[0061] More specifically, the statistics of the number of feedbacks are implemented based on a structured database, and a two-dimensional statistical table is established according to scenario categories and functional points.

[0062] More specifically, the normalization process may use a maximum and minimum value scaling method to map values ​​of different dimensions to the range of 0-1.

[0063] More specifically, in the stage of extracting text fragments, the method of the present application first identifies the specific functional module names involved in user feedback through an entity recognition model, such as the navigation module in "night navigation interface glare". The text fragment containing the functional point is then input into the sentiment analysis model, and the model outputs the sentiment polarity probability value of the fragment. When the positive probability exceeds 0.6, it is judged as a positive evaluation, and when the negative probability exceeds 0.55, it is judged as a negative evaluation. The rest are classified as neutral. For each driving scenario category, the method of the present application aggregates all feedback records for the same functional point, and calculates the total number of feedback, the proportion of positive evaluations, and the proportion of negative evaluations in the scenario respectively. Finally, the original statistical value is converted into a dimensionless value through a normalization formula to eliminate the impact of data magnitude differences between different scenarios on subsequent analysis. This process converts subjective user feedback into standardized indicators that can be quantified and analyzed through structured processing of non-text data, providing data support for dynamically adjusting evaluation weights.

[0064] More specifically, through the above-mentioned processing, the method of the present application effectively solves the problem of insufficient extraction of implicit scene-related information in unstructured user feedback data, and realizes the quantitative conversion of user evaluation tendencies. Specifically, by combining sentiment polarity analysis with multidimensional statistics, it accurately captures the differentiated evaluation characteristics of users on functional points in specific driving scenarios, providing data support for the subsequent establishment of a dynamic weight allocation mechanism. This technical solution avoids the loss of important information caused by the mixed processing of feedback from different scenarios in traditional methods, thereby improving the adaptability of the evaluation system to real driving environments.

[0065] In some preferred embodiments, step S3 further includes: S34. Calculate the ratio of the number of feedbacks for each functional point in each scenario to the total number of feedbacks, and use this ratio as a scenario relevance indicator. S35. Based on a preset proportional coefficient, the scenario relevance index, the positive feedback ratio, and the negative feedback ratio are weighted and summed to obtain a comprehensive evaluation score for each functional point of the user in a specific scenario. The comprehensive evaluation score is used for evaluation in combination with the evaluation system.

[0066] Specifically, the scenario relevance index is calculated by the ratio of the number of feedbacks, which can be obtained by dividing the number of times a function point appears in a specific scenario by the total number of feedbacks in the scenario.

[0067] More specifically, the preset proportional coefficient includes three components, corresponding to the weight distribution of scene relevance, positive feedback and negative feedback, for example, it can be set to [0.4, 0.3, 0.3] to achieve balance in different dimensions.

[0068] More specifically, the weighted summation process combines the three normalized indicators into a single score value through linear combination, which retains the distribution characteristics of the original data while emphasizing the importance of high-frequency functional points.

[0069] More specifically, the comprehensive score obtained in step S35 reflects both the frequency of problem occurrence and the level of user satisfaction, providing a quantitative basis for the subsequent dynamic adjustment of the parking assistance function's weighting in narrow scenarios. By incorporating the comprehensive evaluation score into the evaluation system, the weighting can be allocated to balance the contextual relevance of the function point and user satisfaction, improving the adaptability of the evaluation results to complex scenarios.

[0070] More specifically, through the aforementioned processing, the method of this application achieves a deep coupling of user feedback data with the dynamic characteristics of scenarios, resolving the technical issue of traditional static weighting mechanisms being unable to quantify the scenario-dependence of function points. By introducing a weighted calculation mechanism combining scenario relevance indicators with multi-dimensional feedback ratios, the function point evaluation can automatically adapt to the weight distribution patterns of different driving scenarios, effectively improving evaluation accuracy in highly dynamic scenarios such as complex lighting environments and congested roads.

[0071] It should be noted that the scenario relevance index can also be regarded as the frequency of occurrence of the function point in step S7 in different scenarios.

[0072] In some preferred embodiments, step S4 includes: S41. Align and integrate scenario-based feedback data, expert opinion data, and historical evaluation data based on a preset indicator mapping relationship table to obtain multi-source data association analysis results. The indicator mapping relationship table includes mapping relationships between different functional points and data indicators in different data sources.

[0073] Specifically, a pre-defined indicator mapping table utilizes a pre-defined data structure that records the relationships between various functional points in the smart cockpit and specific data indicators from different data sources (such as scenario-based feedback data, expert opinion data, and historical evaluation data). This can be implemented using a database table, XML file, or JSON file.

[0074] More specifically, aligning and integrating scenario-based feedback data, expert opinion data, and historical evaluation data refers to matching and aggregating data describing the same functional point from different data sources based on the relationships defined in the indicator mapping relationship table.

[0075] More specifically, this mapping table-based structured alignment and integration avoids errors and omissions that may be caused by manual judgment or simple keyword matching, ensures the accuracy of multi-source data association analysis, and thus improves the reliability of subsequent evaluation system construction and evaluation results.

[0076] More specifically, the method of this application solves the problems of inconsistent indicator systems and different description methods that may exist in the alignment and integration process of data from different sources (scenario-based feedback, expert opinions, historical evaluations), ensuring the accurate association and effective integration of multi-source data, thereby improving the accuracy and reliability of the multi-source data association analysis results, and providing a data basis for the construction of the subsequent evaluation system and the reliability of the evaluation results.

[0077] In some preferred implementations, the indicator mapping relationship table includes function point names, user feedback data indicators, expert opinion data indicators, and historical evaluation data indicators. Step S41 includes: S411. Based on the indicator mapping table, align the number of feedbacks, the proportion of positive feedback, and the proportion of negative feedback in the scenario-based feedback data, the importance scores in the expert opinion data, and the scores in the historical evaluation data to ensure that data from different data sources correspond to the same functional point. S412. Integrate the aligned data to obtain multi-source data association analysis results.

[0078] Specifically, the data alignment process uses a mapping table to match the number of feedback, positive feedback ratio, and negative feedback ratio in the scenario-based feedback data with the importance scores in the expert opinion data and the scores in the historical evaluation data. The data integration process aggregates the matched data according to the functional point dimension to generate multi-source data correlation analysis results containing multi-dimensional indicators.

[0079] More specifically, when establishing the indicator mapping table, the function point name serves as the core association key. User feedback data indicators include the number of feedback points extracted from unstructured text and their positive and negative ratios. Expert opinion data indicators include expert ratings of the function point's importance. Historical evaluation data indicators include the function point's scores in historical evaluations. During the data alignment phase, indicator fields from different data sources are mapped using the function point name. For example, the number of user feedback points for "voice interaction response speed" is mapped to the corresponding "voice interaction module importance score" in expert opinion and "voice interaction score" in historical evaluations. During the data integration phase, multi-source data indicators for the same function point are merged to form comprehensive analysis data containing user feedback statistics, expert ratings, and historical scores. Step S412 integrates the proportion of negative feedback for that scenario in user feedback, expert safety ratings, and the average score for that function in historical evaluations to generate multidimensional association analysis results, which are used for subsequent weight allocation of the indicator system and optimization of the evaluation system. This process standardizes data mapping relationships, eliminates semantic differences between data sources, and improves the accuracy of multi-source data fusion.

[0080] More specifically, through the above processing, the method of the present application effectively solves the mapping misalignment problem caused by differences in indicator definitions of multi-source data. By establishing a standardized mapping table, precise alignment across data sources is achieved, ensuring the integrity and consistency of the functional point evaluation dimensions, avoiding misjudgment of functional priorities caused by data silos, and providing a reliable data foundation for subsequent weight allocation and scenario adaptability optimization.

[0081] In some preferred embodiments, step S5 includes: S51. Calculate the comprehensive score for each function point in different driving scenarios based on the results of the multi-source data correlation analysis. The comprehensive score is obtained by weighted average of the scenario-based feedback data, expert opinion data, and historical evaluation data. S52. Screen the function points whose comprehensive scores are higher than the preset multi-source data fusion threshold to obtain the initial indicator system; S53. For the initial index system, calculate the variance of the feedback quantity in the scenario-based feedback data of each functional point under different driving scenarios. If the variance is greater than a preset scenario difference threshold, retain the corresponding index in the initial index system; otherwise, eliminate the corresponding index, thereby determining the index system.

[0082] Specifically, the comprehensive score is obtained by weighted averaging scenario-based feedback data, expert opinion data, and historical evaluation data, and the weight distribution is set based on the confidence level of the data source.

[0083] More specifically, the multi-source data fusion threshold is a preset value.

[0084] More specifically, the scenario difference threshold is set according to the variance distribution law of typical functional points in historical evaluation data, and the benchmark value is determined by counting the percentiles of the variance values ​​of each functional point. For example, the percentile with a variance value in the top 30% is taken as the threshold benchmark.

[0085] More specifically, the dual screening mechanism of step S52 and step S53 ensures that the final indicator system includes both high-value functional points and retains indicators with strong scenario differentiation capabilities, so that the evaluation system can accurately reflect the performance differences of the smart cockpit under different driving environments. The final indicator system focuses on core functions with scenario sensitivity, ensuring that the evaluation system can dynamically reflect the performance differences of the smart cockpit under different driving environments.

[0086] In some preferred embodiments, step S7 includes: S71. Extract the number of user feedbacks for each function point corresponding to each indicator in the scenario-based feedback data in different scenarios, and calculate the proportion of the number of feedbacks for each function point in each scenario to the total number of feedbacks, to obtain the frequency of occurrence of the function point in different scenarios. S72. Calculate the scenario weight factor of each indicator in different scenarios based on the preset scenario importance score and occurrence frequency, where scenario weight factor = scenario mention frequency * scenario importance score; S73. Normalize the scenario weight factors of each indicator in all scenarios.

[0087] Specifically, step S71 can be implemented in the following manner: First, from the scenario-annotated user feedback data, specific function points associated with each indicator are identified according to a preset indicator system. Then, the total number of user feedbacks generated for these function points in different driving scenario categories is counted. Next, for each function point, the proportion of feedback received in a specific scenario to the total number of feedback received for that function point in all scenarios, or the proportion of feedback received for all function points in that specific scenario, is calculated, and this is used as the frequency of occurrence of the function point in that scenario.

[0088] More specifically, step S72 can be implemented as follows: First, pre-set importance scores for different driving scenarios are obtained. These scores can be based on expert experience, risk assessments, or user surveys, and reflect the criticality or potential risk of each scenario in the overall driving process. Then, the frequency of occurrence of each function point in each scenario, calculated in step S71, is multiplied by the pre-set importance score for that scenario to determine a preliminary weight factor for that function point in that scenario.

[0089] More specifically, step S73 can be implemented as follows: For each metric in the indicator system, the scenario weight factors calculated for all identified driving scenarios are summed. The scenario weight factor for each specific scenario is then divided by this sum to obtain a normalized scenario weight factor. After normalization, the sum of the weight factors for the same metric across different scenarios equals 1, making these weight factors comparable and consistent in subsequent weighted calculations.

[0090] More specifically, through the above processing, the method of the present application provides a method for calculating the scenario weight factor of the smart cockpit evaluation index based on scenario feedback data and scenario importance, so that the indicator weight can dynamically reflect its importance in different driving scenarios, solving the problem that traditional static weights cannot adapt to changes in scenario requirements, and improving the scenario adaptability and effectiveness of the evaluation system.

[0091] In some preferred embodiments, the steps of establishing an evaluation system based on the indicator system, basic weights, and scenario weight factors include: S81. Based on the indicator system, select indicators that match the indicators in the indicator system from a preset evaluation indicator library to construct a scenario-based evaluation indicator set; S82. According to the basic weight and the scenario weight factor, adjust the weight coefficient of each indicator in the scenario-based evaluation indicator set to obtain the final evaluation system, wherein the adjusted weight coefficient is equal to the product of the basic weight and the scenario weight factor.

[0092] Specifically, the preset evaluation index library stores historically accumulated evaluation indicators and their metadata. The metadata includes information such as the indicator name, applicable scenario type, and functional module affiliation. The screening process uses a combination of indicator name semantic matching and metadata filtering. For example, the semantic similarity between the indicator name in the indicator system and the candidate indicator in the indicator library is calculated through a natural language processing model. When the similarity exceeds the matching threshold and there is an intersection in the applicable scenario type, it is determined to be a valid match. In the weight adjustment process, the basic weight is used as the benchmark value and is coupled with the weight factor that reflects the dynamic needs of the scenario for calculation.

[0093] More specifically, through the above-mentioned processing, the method of the present application realizes dynamic scenario adaptation of the evaluation weights, effectively solving the problem of insufficient applicability of the traditional static weight system in complex driving environments. It ensures that the evaluation results accurately reflect the differences in functional importance in different scenarios by multiplying the basic weights with the scenario weight factors. At the same time, the indicator screening mechanism based on semantic matching ensures the precise correspondence between the evaluation system and functional requirements, avoids evaluation bias caused by indicator mismatch, and provides accurate data support for the scenario-based optimization of the smart cockpit.

[0094] Second, please refer to Figure 3 and Figure 4 Some embodiments of the present application further provide a smart cockpit evaluation system for evaluating the performance of a smart cockpit system in a complex dynamic driving environment. The system includes: Data acquisition module 201, used to obtain user feedback data stream; A scene labeling module 202 is used to label the user feedback data stream with driving scene categories and obtain user feedback data after scene labeling; The data quantification module 203 is used to analyze the evaluation trends of users for different smart cockpit functional points in different scenarios in the user feedback data after statistical scene annotation, and convert the evaluation trends into scenario-based feedback data; Data integration module 204, used to align and integrate scenario-based feedback data, expert opinion data, and historical evaluation data to obtain multi-source data correlation analysis results; The indicator screening module 205 is used to screen the functional requirements list based on the multi-source data association analysis results to obtain the indicator system; The weight allocation module 206 is used to allocate basic weights to the indicators in the indicator system based on the expert opinion data and the historical evaluation data; A factor configuration module 207 is used to analyze the occurrence frequency of function points corresponding to different indicators in different scenarios based on the scenario-based feedback data to assign scenario weight factors to the indicators in the indicator system; The evaluation system module 208 is used to establish an evaluation system based on the indicator system, basic weights and scenario weight factors for evaluation.

[0095] The smart cockpit evaluation system of the embodiment of the present application constructs a dynamic scenario-driven smart cockpit evaluation system. It achieves a balance between the static importance and dynamic scenario requirements of evaluation indicators through scenario analysis of dynamic integration of user feedback, multi-source data correlation integration, and dual weight distribution mechanism. It solves the problems of low utilization of unstructured user feedback data, lack of scenario relevance, isolated data dimensions and solidified weights in traditional systems. It has the advantages of accurately identifying performance defects of smart cockpit systems in complex dynamic driving environments, dynamically adjusting evaluation weights to adapt to different scenario requirements, and improving the consistency between evaluation results and user experience.

[0096] In some preferred embodiments, the system further comprises: The indicator optimization module 209 is used to refer to expert opinion data and historical evaluation data to refine and classify the indicator system.

[0097] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0098] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0099] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0100] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A smart cockpit evaluation method is used to evaluate the performance of a smart cockpit system in a complex dynamic driving environment, characterized by: The method comprises the following steps: S1. Obtain user feedback data stream; S2. Label the user feedback data stream by driving scene category to obtain the labeled user feedback data; S3. Analyze and collect the user feedback data after scene annotation to determine the evaluation trends of users for different smart cockpit functional points in different scenarios, and convert the evaluation trends into scenario-based feedback data; S4. Align and integrate the scenario-based feedback data, expert opinion data, and historical evaluation data to obtain multi-source data correlation analysis results; S5. Screening the functional requirements list based on the multi-source data association analysis results to obtain an indicator system; S6. Assigning basic weights to the indicators in the indicator system based on expert opinion data and historical evaluation data; S7. Analyze the occurrence frequencies of function points corresponding to different indicators in different scenarios based on the scenario-based feedback data to assign scenario weight factors to the indicators in the indicator system; S8. Establish an evaluation system based on the indicator system, the basic weights and the scenario weight factors for evaluation.

2. The smart cockpit evaluation method according to claim 1, characterized in that: The following steps are included between step S5 and step S6: SA, refer to expert opinion data and historical evaluation data to refine and grade the indicator system.

3. The smart cockpit evaluation method according to claim 1, characterized in that: Step S6 includes: S61. Assign basic weights to the indicators in the indicator system based on the importance scores of the corresponding indicators from the expert opinion data and the scores of the corresponding indicators from the historical evaluation data.

4. The smart cockpit evaluation method according to claim 1, characterized in that: The user feedback data stream includes unstructured text and ratings. Step S2 includes: S21. Acquire vehicle sensor data based on vehicle sensors, where the vehicle sensor data includes vehicle speed, steering angle, ambient light intensity, and traffic flow data; S22. Determine a driving scene type by matching the vehicle sensor data based on a preset rule base, and annotate the user feedback data stream with the driving scene type to form scene-annotated user feedback data; S23. When the matching in step S22 fails, extract driving scene terms from the unstructured text based on a preset keyword library, and use the driving scene terms to annotate the user feedback data stream to form scene-annotated user feedback data.

5. The smart cockpit evaluation method according to claim 1, characterized in that: Step S3 includes: S31. Extracting text fragments related to smart cockpit functional points from the scene-annotated user feedback data, and determining the user's evaluation polarity of each functional point based on a pre-trained sentiment analysis model. The evaluation polarity is categorized as positive, negative, or neutral. S32. Counting the number of user feedbacks, the proportion of positive feedback, and the proportion of negative feedback for different functional points in each scenario according to the preset driving scenario categories; S33: Normalize the feedback quantity, positive feedback ratio, and negative feedback ratio to form the scenario-based feedback data.

6. The smart cockpit evaluation method according to claim 5, characterized in that: Step S3 further includes: S34. Calculate the ratio of the number of feedbacks for each functional point in each scenario to the total number of feedbacks, and use this ratio as a scenario relevance indicator. S35. Perform weighted summation of the scenario relevance index, positive feedback ratio, and negative feedback ratio based on a preset proportional coefficient to obtain a comprehensive evaluation score for each function point of the user in a specific scenario.

7. The smart cockpit evaluation method according to claim 1, characterized in that: Step S4 includes: S41. Align and integrate the scenario-based feedback data, expert opinion data, and historical evaluation data based on a preset indicator mapping relationship table to obtain multi-source data association analysis results. The indicator mapping relationship table includes mapping relationships between different functional points and data indicators in different data sources.

8. The smart cockpit evaluation method according to claim 7, characterized in that: Step S41 includes: S411. Align the scenario-based feedback data, the importance scores in the expert opinion data, and the scores in the historical evaluation data according to the indicator mapping relationship table to ensure that the data from different data sources correspond to the same functional point. S412. Integrate the aligned data to obtain multi-source data association analysis results.

9. The smart cockpit evaluation method according to claim 1, characterized in that: Step S5 includes: S51. Calculate the comprehensive score for each function point in different driving scenarios based on the results of the multi-source data correlation analysis. The comprehensive score is obtained by weighted average of the scenario-based feedback data, expert opinion data, and historical evaluation data. S52. Screen the function points whose comprehensive scores are higher than the preset multi-source data fusion threshold to obtain the initial indicator system; S53. Calculate the variance of the feedback quantity in the scenario-based feedback data for each functional point under different driving scenarios for the initial indicator system. If the variance is greater than a preset scenario difference threshold, retain the corresponding indicator in the initial indicator system; otherwise, remove the corresponding indicator, thereby determining the indicator system.

10. A smart cockpit evaluation system for evaluating the performance of a smart cockpit system in a complex dynamic driving environment, characterized by: The system comprises: Data acquisition module, used to obtain user feedback data stream; The scene labeling module is used to label the user feedback data stream with driving scene categories and obtain the user feedback data after scene labeling; A data quantification module is used to analyze the user feedback data after statistical scene annotation for the evaluation trends of different smart cockpit functional points in different scenarios, and convert the evaluation trends into scenario-based feedback data; A data integration module is used to align and integrate the scenario-based feedback data, expert opinion data, and historical evaluation data to obtain multi-source data correlation analysis results; An indicator screening module, configured to screen the functional requirements list based on the multi-source data association analysis results to obtain an indicator system; A weight allocation module, for allocating basic weights to the indicators in the indicator system based on expert opinion data and historical evaluation data; A factor configuration module is used to analyze the frequency of occurrence of function points corresponding to different indicators in different scenarios based on scenario-based feedback data to assign scenario weight factors to the indicators in the indicator system; The evaluation system module is used to establish an evaluation system based on the indicator system, the basic weight and the scenario weight factor for evaluation.

Images