A method and system for directional acquisition of cockpit data

By collecting user interaction parameter information in the smart cockpit and quantifying the signal, and combining it with cloud-based causal reasoning to generate directional data collection commands, the problem of low data collection efficiency and high redundancy in existing technologies is solved, achieving efficient, low-redundancy data collection and optimized user experience.

CN123123384APending Publication Date: 2026-11-17HUIZHOU DESAY SV AUTOMOTIVE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611268265.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing smart cockpit data collection solutions suffer from problems such as a large proportion of low-value data, heavy cloud computing load, and a deviation between data and the direction of experience improvement, making it difficult to support iterative needs centered on user experience.

Method used

By collecting user interaction parameter information on the vehicle, determining the signal quantification score, triggering the experience degradation signal to the cloud, the cloud performs causal reasoning to determine the root cause contribution, generating targeted data collection instructions, and the vehicle performs targeted data collection, achieving efficient collaborative and low-redundancy data collection.

Benefits of technology

It enables precise quantification of user subjective experience, dynamically identifies high-value interaction samples, and improves the efficiency and quality of data collection in smart cockpits.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123384A_ABST
    Figure CN123123384A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for directional cockpit data acquisition. The method includes: the vehicle-side acquiring parameter information generated by user interactions within the cockpit and determining the corresponding signal quantization score; the vehicle-side determining a comprehensive experience score based on the signal quantization score, and triggering an experience degradation signal when the comprehensive experience score meets preset requirements, sending the experience degradation signal to the cloud; the cloud responding to the experience degradation signal, acquiring user interaction process information within the cockpit, and determining the root cause contribution causal effect value based on the cockpit interaction function dependency topology and interaction process information; the cloud generating a directional data acquisition instruction based on the root cause contribution causal effect value; and the vehicle-side determining the data acquisition range and performing data acquisition based on the directional data acquisition instruction. These technical features enable precise quantification of user subjective experience, achieving efficient collaboration and low-redundancy transmission between the vehicle-side and cloud, and realizing an intelligent cockpit data acquisition mechanism driven by experience optimization.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent cockpit technology, and in particular to a method and system for directional acquisition of cockpit data. Background Technology

[0002] With the large-scale application of large-scale model technology in smart cockpits, the multimodal interaction data generated by cockpit systems is growing exponentially, including voice commands, touch behaviors, in-cabin vision, and vehicle status signals. High-quality, highly relevant data acquisition has become a key prerequisite for supporting the continuous optimization and experience iteration of large-scale cockpit models. Currently, the mainstream data acquisition solutions in the smart cockpit field primarily use data scale and model uncertainty as triggering conditions for data acquisition, collecting data in a comprehensive and undifferentiated manner. Secondly, existing data acquisition strategies are decoupled from experience optimization goals. These methods suffer from problems such as a large proportion of low-value data and heavy cloud computing load. The collected data deviates from the actual direction of experience improvement, making it difficult to support the iterative needs of smart cockpits that are "user experience-centric." Summary of the Invention

[0003] This invention provides a method and system for directional cockpit data collection, which can accurately quantify the user's subjective experience. Based on achieving efficient collaboration and low-redundancy transmission between the vehicle and the cloud, it realizes an intelligent cockpit data collection mechanism driven by experience optimization.

[0004] In one embodiment, the present invention provides a method for directional cockpit data acquisition, applied to a cockpit data directional acquisition system, the system including a vehicle-side terminal and a cloud-based terminal; the method includes:

[0005] The vehicle-mounted device collects parameter information generated by user interactions within the cabin and determines the signal quantization score corresponding to the parameter information; wherein, the parameter information includes at least user physiological signals, user behavioral signals, and user feedback signals;

[0006] The vehicle determines a comprehensive experience score based on the signal quantization score, and if the comprehensive experience score meets the preset requirements, it triggers an experience degradation signal and sends the experience degradation signal to the cloud.

[0007] In response to the experience degradation signal, the cloud acquires information about the user's interaction process in the cabin and determines the root cause contribution causal effect value based on the pre-configured cabin interaction function dependency topology and the interaction process information.

[0008] The cloud platform generates a targeted data collection instruction based on the root cause contribution causal effect value and sends it to the vehicle terminal.

[0009] The vehicle-mounted device determines the data collection range according to the directional data collection instruction, and performs directional data collection within the data collection range.

[0010] In one embodiment, the present invention provides a cockpit data directional acquisition system, applied to the cockpit data directional acquisition method described in any embodiment of the present invention; the system includes a vehicle-side component and a cloud-side component; wherein, the vehicle-side component includes: a signal quantization and determination module, a degraded signal transmission module, and a directional data acquisition module; the cloud-side component includes: a root cause contribution determination module and an acquisition command issuance module;

[0011] The signal quantization score determination module is used to collect parameter information generated by user interaction in the cockpit and determine the signal quantization score corresponding to the parameter information; wherein the parameter information includes at least user physiological signals, user behavioral signals and user feedback signals;

[0012] The degradation signal sending module is used to determine the overall experience score based on the signal quantization score, and when the overall experience score meets the preset requirements, to trigger an experience degradation signal and send the experience degradation signal to the cloud.

[0013] The root cause contribution determination module responds to the experience degradation signal by acquiring user interaction process information in the cabin and determining the root cause contribution causal effect value based on the pre-configured cabin interaction function dependency topology and the interaction process information.

[0014] The data acquisition instruction issuing module generates a targeted data acquisition instruction based on the root cause contribution causal effect value and issues it to the vehicle terminal.

[0015] The directional data acquisition module determines the data acquisition range according to the directional data acquisition instruction, and performs directional data acquisition within the data acquisition range.

[0016] The technical solution of this invention collects parameter information generated by user interaction in the cockpit on the vehicle side, determines the signal quantization score corresponding to the parameter information, determines the comprehensive experience score based on the signal quantization score, and triggers an experience degradation signal when the comprehensive experience score meets preset requirements. The experience degradation signal is sent to the cloud, and the cloud responds to the experience degradation signal to obtain the user's interaction process information in the cockpit. Based on the pre-configured cockpit interaction function dependency topology map and interaction process information, the root cause contribution causal effect value is determined. Based on the root cause contribution causal effect value, a directional data collection instruction is generated and sent to the vehicle side. Finally, the vehicle side determines the data collection range according to the directional data collection instruction and performs directional data collection within the data collection range. The user's subjective experience is accurately quantified. On the basis of achieving efficient collaboration and low-redundancy transmission between the vehicle side and the cloud, an intelligent cockpit data collection mechanism driven by experience optimization is realized.

[0017] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a cockpit data orientation acquisition method provided in an embodiment of the present invention;

[0020] Figure 2 This is a flowchart of another method for directional acquisition of cockpit data provided in an embodiment of the present invention;

[0021] Figure 3 This is a flowchart illustrating another method for directional acquisition of cockpit data provided in an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of a cockpit data orientation acquisition system provided in an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] In one embodiment, Figure 1 This is a flowchart of a cockpit data directional acquisition method provided by an embodiment of the present invention. This embodiment is applicable to situations where cockpit data is directionally acquired. This method can be executed by the vehicle end and the cloud end in the cockpit data directional acquisition process.

[0026] like Figure 1 As shown, this cockpit data targeted acquisition method is applied to a cockpit data targeted acquisition system, which includes a vehicle-side terminal and a cloud-based system. It should be noted that the vehicle-side terminal in this embodiment can include multiple vehicle-side terminals, each capable of executing operations performed by the vehicle-side terminal and interacting with the cloud for data exchange. The vehicle and cloud collaborate to perform causal reasoning, deeply pinpoint the root causes of user experience degradation, and then implement targeted acquisition of high-value samples based on the contribution of these root causes. Specifically, the cockpit data targeted acquisition method includes:

[0027] S110: The vehicle-mounted system collects parameter information generated by user interactions within the cockpit and determines the corresponding signal quantization score.

[0028] The parameter information includes at least physiological indicators, behavioral indicators, and user feedback signals. Physiological indicators may include, but are not limited to, facial expressions, heart rate, respiratory rate, and blink rate. Behavioral indicators include at least the number of voice interruptions, manual operations, command repetitions, page dwell time, and operation cancellations. User feedback can be understood as signals provided by the user through interactive components.

[0029] In this embodiment, the user's physiological indicators and other core features can directly reflect the user's internal state such as fatigue, irritability, and discomfort. The user's behavioral indicators can directly reflect the smoothness and ease of use of the cockpit interaction. The user's feedback signals can reflect the user's explicit subjective evaluation, such as multiple evaluation dimensions such as very satisfied, satisfied, average, and dissatisfied.

[0030] In this embodiment, the user performs interactive behaviors within the vehicle's cockpit domain, such as voice, touch, and other interactive behaviors. The vehicle collects physiological index information, behavioral index information, and user feedback signal information generated by the user's interaction within the cockpit. In some embodiments, the user's physiological index information can be collected through sensing devices within the cockpit, the user's behavioral index information can be collected through the vehicle's signal system, such as the vehicle's log information and CAN bus, and the feedback signal information fed back by the vehicle's in-vehicle interactive components can be collected through the user feedback system. In this embodiment, the signal quantization score calculation is different for different interactive information, and the physiological index information, behavioral index information, and user feedback signal information each correspond to a corresponding signal quantization score calculation.

[0031] In this embodiment, physiological indicator information corresponds to a corresponding physiological signal quantification score, behavioral indicator information corresponds to a corresponding behavioral signal quantification score, and user feedback signal information corresponds to a corresponding feedback signal quantification score. In some embodiments, the user behavioral signal quantification score can be determined using behavioral indicator information, preset behavioral indicator weights, and preset behavioral indicator benchmark values; the user feedback signal quantification score can be determined using feedback signal information; and the user physiological signal quantification score can be determined using physiological indicator information, preset physiological indicator weights, and preset physiological indicator benchmark values. Of course, physiological signal quantification scores, behavioral signal quantification scores, and feedback signal quantification scores can also be calculated in other ways, and this embodiment does not impose specific limitations.

[0032] S120: The vehicle determines the overall experience score based on the signal quantization score, and if the overall experience score meets the preset requirements, it triggers the experience degradation signal and sends the experience degradation signal to the cloud.

[0033] Among them, the experience degradation signal can be understood as a signal of reduced user experience.

[0034] In this embodiment, the preset requirements include: the overall experience score being below a preset threshold for a consecutive preset time window, or the overall experience score decreasing by more than a preset rate of change threshold within a preset time period. The preset time window is a time window set by the user based on their needs or experience; similarly, the preset rate of change threshold is a rate of change threshold set by the user based on their needs or experience. This can be understood as follows: when the overall experience score is below a dynamic threshold for at least two consecutive time windows (e.g., a time window length of 3 seconds), or when the overall experience score decreases by more than the preset rate of change threshold within a preset time period (e.g., within 10 seconds), an experience degradation signal is triggered and sent to the cloud.

[0035] In this embodiment, the comprehensive experience score can be calculated using the user's physiological signal quantification score, user behavior signal quantification score, and user feedback signal quantification score. This can be understood as the final score obtained by combining the scores of these three modalities (physiological, behavioral, and feedback signals), assigning dynamic confidence based on signal acquisition stability, and then introducing outlier penalties to remove interference. In some embodiments, the vehicle-side performs a weighted fusion of the physiological signal quantification score, behavioral signal quantification score, and feedback signal quantification score with the dynamic confidence score, introducing outlier penalty correction to calculate the comprehensive experience score. Specifically, abnormal and normal parameters can be selected from historical parameter information generated by the user's historical interactions to obtain sample labeling results. Based on this, the comprehensive experience score is determined using the user's physiological signal quantification score, user behavior signal quantification score, user feedback signal quantification score, sample labeling results, and a preset dynamic confidence score. Of course, the specific calculation method for the comprehensive experience score is not limited to the above method; it can also be obtained through other combined calculations. This embodiment does not impose specific limitations here.

[0036] S130 and cloud respond to experience degradation signals, acquire information about the user's interaction process in the cockpit, and determine the root cause contribution and causal effect value based on the pre-configured cockpit interaction function dependency topology and interaction process information.

[0037] The interaction process information refers to the complete interaction flow of the current cabin generated during the user's interaction with the vehicle within the cabin domain. For example, the complete interaction flow of a user's voice interaction may include, but is not limited to, voice input, ASR recognition, intent understanding, dialogue generation, and TTS output. The root cause contribution causal effect value can be understood as the numerical value that quantifies how much each functional node "contributes" to the final experience degradation result after the experience degradation is triggered.

[0038] In this embodiment, the cockpit interaction function depends on the topology graph constructed by the complete interaction process of the user's historical interactions within the cockpit domain. Specifically, the interaction process is decomposed layer by layer, breaking down the complete interaction process from top to bottom until it reaches indivisible atomic functional units. Taking the voice interaction link as an example, it is decomposed into: voice input acquisition, audio noise reduction and enhancement, ASR speech recognition, semantic understanding / NLU, dialogue state tracking, dialogue strategy decision-making, response generation / NLG, TTS speech synthesis, and audio playback output. A unique functional node identifier is established for each atomic functional unit in the topology graph, and metadata is recorded. Functional dependencies are extracted, and the dependencies between nodes are determined one by one. If the output data of functional node A is directly used as the input of functional node B, or the completion of functional node A is a prerequisite for the start of functional node B, then a directed edge is established from A to B. This directed edge can be represented as an abnormality of functional node A that can propagate along this edge and affect functional node B, which is irreversible. All functional nodes and the identified set of directed edges E are combined to form an initial directed graph. DAG loop verification and correction are performed, and then direct pre-dependencies are identified to finally obtain the cockpit interaction function dependency topology graph.

[0039] In this embodiment, after receiving the experience degradation trigger signal, the cloud performs a causal reasoning process. First, the current complete cockpit interaction process is broken down into independent functional nodes, and the operating status and standard experience benchmark score of each functional node are obtained simultaneously. The breakdown is based on a preset cockpit interaction function dependency topology graph, which is a directed acyclic graph. The complete interaction process is decomposed into the smallest functional unit node. The directed edge represents the upstream and downstream dependency relationship between data flow and control flow. Node anomalies only propagate downstream along the directed edge and cannot be reversed. Based on the real-time operating status, standard experience benchmark score, and causal propagation weight of each functional node, an integrated quantitative evaluation is performed on the anomalies of a single functional node and the coupling anomalies of two nodes. The evaluation calculates the anomaly of the functional node itself and the propagation impact of the anomaly along the interaction link. Combined with the status identifier and coupling correlation of each functional node, the root cause contribution causal effect value of each functional node is finally obtained.

[0040] S140: The cloud generates a targeted data collection instruction based on the root cause contribution and causal effect value and sends it to the vehicle.

[0041] In this embodiment, the directional data acquisition instruction may include, but is not limited to, the abnormal root cause node number, the associated node of the abnormal root cause node, the data acquisition type to be acquired during data acquisition, the data acquisition time window length, the data format and data transmission requirements required for data upload after data acquisition.

[0042] In this embodiment, after the root cause contribution effect value is calculated in the cloud, it is directly used as the basis for root cause contribution, and prioritized according to the value to filter out high-priority abnormal root causes. Based on the located high-priority root cause nodes, a targeted data collection instruction is generated and sent to the vehicle terminal that triggered the experience degradation. This can be understood as the vehicle terminal that triggered the experience degradation sending an experience degradation signal to the cloud to obtain information about the user's interaction process in the cabin. Based on the pre-configured cabin interaction function dependency topology and the interaction process information, the root cause contribution causal effect value is determined. Based on this root cause contribution causal effect value, a targeted data collection instruction is generated and then sent to the corresponding vehicle terminal. It should be noted that each vehicle terminal corresponds to a unique vehicle identifier, and the cloud can use this vehicle identifier to determine the vehicle terminal from which the experience degradation signal was uploaded and the vehicle terminal from which the data collection instruction was sent.

[0043] S150: The vehicle-mounted system determines the data collection range based on the directional data collection instruction and performs directional data collection within the data collection range.

[0044] In this embodiment, after receiving the directional data collection command from the cloud, the vehicle parses the command to obtain the parsing result. Based on this result, it accurately locates the functional node corresponding to the root cause and the directly upstream related nodes, thereby determining the collection range. Specifically, based on the node information in the command, the operating module of the root cause node in the cockpit model can be located, as well as the directly upstream related nodes that provide input to the root cause node. The collection boundary is then defined, containing only these two types of nodes, without expanding the collection of data from irrelevant modules.

[0045] The technical solution described in this invention collects parameter information generated by user interactions within the cockpit on the vehicle side, determines the corresponding signal quantization score, determines the overall experience score based on the signal quantization score, and triggers an experience degradation signal when the overall experience score meets preset requirements. This signal is then sent to the cloud. The cloud responds to the experience degradation signal, acquires user interaction process information within the cockpit, and determines the root cause contribution causal effect value based on a pre-configured cockpit interaction function dependency topology and interaction process information. Based on the root cause contribution causal effect value, a directional data collection instruction is generated and sent to the vehicle side. Finally, the vehicle side determines the data collection range according to the directional data collection instruction and performs directional data collection within that range. This allows for precise quantification of the user's subjective experience, realizing an intelligent cockpit data collection mechanism driven by experience optimization, dynamically identifying high-value interaction samples, and achieving efficient collaboration and low-redundancy transmission between the vehicle side and the cloud.

[0046] In one embodiment, the directional data acquisition instruction further includes: data format and transmission requirements; correspondingly, the method further includes:

[0047] The vehicle-mounted system organizes and lightweight packages the collected data according to the data format, and uploads it to the cloud according to the transmission requirements.

[0048] In this embodiment, the vehicle-mounted system performs local compliance preprocessing on the collected valid data, completing data standardization and lightweight encapsulation. Specifically, invalid fragmented information in the data is removed, the data format is standardized, and the data is encapsulated according to cloud requirements to ensure the compliance and lightweight nature of data transmission. Finally, the vehicle-mounted system uploads the preprocessed targeted data set to the cloud through a secure channel, completing the entire process of targeted data collection for the cockpit model in this causal-guided experience.

[0049] In one embodiment, Figure 2 This is a flowchart of another method for directional cockpit data acquisition provided by an embodiment of the present invention. This embodiment is a further optimization of any of the above embodiments. This embodiment can be applied to the situation of directional cockpit data acquisition. This method can be executed by the vehicle end and the cloud in directional cockpit data acquisition.

[0050] like Figure 2 As shown, the method includes:

[0051] S210: The vehicle collects parameter information generated by user interaction in the cabin and collects the user's physiological index information through preset sensing devices.

[0052] Among them, physiological indicators include at least facial expression, heart rate, respiratory rate, and blinking frequency.

[0053] In this embodiment, the preset sensing devices may include, but are not limited to, one or more of the following: a DMS driver monitoring camera, an OMS occupant monitoring camera, a ToF 3D depth camera, an infrared night vision camera, a seat-embedded cardiac impact sensor, millimeter-wave radar, and an ultrasonic sensor. One or more of the above-mentioned sensing devices can be used to collect the user's facial expressions, heart rate, respiratory rate, blink frequency, and other physiological characteristics, which can directly reflect the user's internal state such as fatigue, irritability, and discomfort.

[0054] S220 collects user behavior metrics information through vehicle-side log information and CAN bus.

[0055] Among them, the behavioral indicator information includes at least the number of times voice interruption occurred, the number of times manual operation was performed, the number of times instructions were repeated, the time spent on the page, and the number of times operations were undone.

[0056] In this embodiment, behavioral characteristics such as the number of voice interruptions, the number of manual operations, the number of command repetitions, the page dwell time, and the number of operation cancellations can be collected through the vehicle system logs and CAN bus. These behavioral characteristics can directly reflect the smoothness and ease of use of the cockpit interaction.

[0057] S230: Collect feedback signal information from the vehicle-mounted interactive components.

[0058] In this embodiment, the feedback signal information fed back by the vehicle-mounted interactive components can be collected through the vehicle-mounted feedback signal system. This feedback signal information can be the user's explicit subjective evaluation, which can include multiple subjective evaluation levels. For example, the user's explicit subjective evaluation includes six levels: very satisfied, satisfied, neutral, dissatisfied, very dissatisfied, and no feedback. Different levels can correspond to different scores.

[0059] S240. Determine the user's physiological signal quantification score based on physiological indicator information, preset physiological indicator weights, and preset physiological indicator benchmark values.

[0060] The preset physiological indicator weights are the weight values ​​corresponding to each physiological indicator set by the user based on their needs or experience. The sum of the weight values ​​of all physiological indicators is 1. The preset physiological indicator benchmark value is the normal benchmark value of the physiological indicator. This benchmark value can be dynamically generated using a "cloud-based group prior + vehicle-side personalized adaptation" mechanism. Specifically, during the cold start phase, an initial benchmark library can be generated by clustering according to vehicle model and user profile; during the online phase, it can be updated slightly at a fixed ratio based on the measured values ​​under normal user conditions to achieve adaptive adjustment.

[0061] In this embodiment, the user's physiological signal quantification score can be determined by physiological indicator information, preset physiological indicator weights, and preset physiological indicator benchmark values. Specifically, the physiological indicator information of the object to be evaluated in N dimensions and the corresponding benchmark indicator values ​​are obtained; the weight coefficients corresponding to the physiological indicator information of each dimension are determined; the relative offset of each dimension is calculated, and the relative offset represents the degree of absolute deviation of the current indicator value from the benchmark indicator value. When the degree of deviation reaches or exceeds 100%, the relative offset is set to 1. The relative offset is weighted and summed according to the corresponding weight coefficients to obtain the comprehensive evaluation score.

[0062] More specifically, the formula for calculating the user's physiological signal quantification score is expressed as follows: Where i represents physiological indicator information, with values ​​of 1, 2, 3, and 4, representing facial expression, heart rate, respiratory rate, and blink frequency, respectively. The preset physiological index weights corresponding to i; ; This represents the measured physiological value of the i-th physiological indicator in a vehicle-mounted scenario. This represents the preset physiological indicator baseline value for the i-th indicator. The measured values ​​of the physiological indicators are compared with the baseline values. The degree to which physiological indicators deviate from the normal state can be converted into a standard score.

[0063] S250. Determine the user behavior signal quantification score based on behavior indicator information, preset behavior indicator weights, and preset behavior indicator benchmark values.

[0064] In this embodiment, the preset behavioral indicator weights are the weight values ​​corresponding to each behavioral indicator set by the user based on their needs or experience. The sum of the weight values ​​of all behavioral indicators is 1. The preset behavioral indicator benchmark values ​​can also be called normal behavioral indicator benchmark values. They can also adopt a dynamic generation mechanism of "cloud-based group prior knowledge + vehicle-side personalized adaptation". Specifically, during the cold start phase, an initial benchmark library is generated by clustering according to vehicle model and user profile; during the online phase, the benchmark is updated slightly at a fixed ratio based on the measured values ​​under normal user conditions to achieve adaptive adjustment.

[0065] In this embodiment, the user behavior signal quantification score can be determined by behavioral indicator information, preset behavioral indicator weights, and preset behavioral indicator benchmark values. Specifically, behavioral indicator information of N dimensions of the object to be evaluated and the corresponding benchmark indicator values ​​are obtained; the weight coefficients corresponding to the behavioral indicator information of each dimension are determined; the relative offset of each dimension is calculated, and the relative offset represents the absolute deviation of the current indicator value from the benchmark indicator value. When the deviation reaches or exceeds 100%, the relative offset is set to 1. The relative offset is weighted and summed according to the corresponding weight coefficients to obtain the comprehensive evaluation score.

[0066] More specifically, the formula for calculating the quantization score of user behavior signals is expressed as follows: Where j takes values ​​of 1, 2, 3, 4, and 5, representing the five metrics: number of voice interruptions, number of manual operations, number of command repetitions, page dwell time, and number of operation undoes. The preset behavioral indicator weights corresponding to j; ; This represents the measured value of the j-th behavioral indicator in a vehicle-mounted scenario. This is represented as the preset benchmark value for the j-th indicator. The measured values ​​of the behavioral indicators are compared with the benchmark value. The degree to which behavioral indicators deviate from the normal state can be converted into a standard score.

[0067] S260. Determine the quantitative score of the user feedback signal based on the preset hierarchical mapping rules and feedback signal information.

[0068] In this embodiment, a hierarchical hard mapping rule can be used to quickly convert the subjective evaluation represented by the feedback signal information into an objective score, while eliminating invalid feedback to avoid scoring interference. When the feedback signal information is a first-level evaluation, a first-level score is determined; when the feedback signal information is a second-level evaluation, a second-level score is determined; when the feedback signal information is a third-level evaluation, a third-level score is determined; when the feedback signal information is a fourth-level evaluation, a fourth-level score is determined; and when the feedback signal information is a fifth-level evaluation, a fifth-level score is determined. Specifically, ; By objectifying subjective feedback, it can be used as a direct verification basis for experience ratings.

[0069] S270. The quantitative scores of user physiological signals, user behavior signals, and user feedback signals are respectively used as the corresponding signal quantitative scores.

[0070] In this embodiment, the user's physiological signal quantification score is used as the quantification score of physiological indicator information, the user's behavior signal quantification score is used as the quantification score of behavior indicator information, and the user's feedback signal quantification score is used as the quantification score of feedback signal information.

[0071] S280. Obtain historical parameter information generated by the user's historical interactions, select abnormal parameters and normal parameters from the historical parameter information, and use the abnormal parameters and normal parameters as samples to perform sample labeling to obtain sample labeling results.

[0072] In this embodiment, historical parameter information generated by the user's historical interactions is obtained, and abnormal parameters and normal parameters are selected from the historical parameter information. Abnormal parameters are marked as 1, and normal parameters are marked as 0. The marked abnormal parameters and normal parameters are used as sample marking results.

[0073] S290. Based on the quantitative scores of user physiological signals, user behavior signals, and user feedback signals, as well as the sample labeling results and preset dynamic confidence, a comprehensive experience score is determined. If the comprehensive experience score meets the preset requirements, an experience degradation signal is triggered and sent to the cloud.

[0074] In this embodiment, the vehicle-side system weights and fuses the quantified scores of physiological signals, behavioral signals, and feedback signals with a preset dynamic confidence level, introducing outlier penalty correction to calculate a comprehensive experience score. The preset dynamic confidence level includes: a first dynamic confidence level corresponding to the user's quantified physiological signal score, a second dynamic confidence level corresponding to the user's quantified behavioral signal score, and a third dynamic confidence level corresponding to the user's quantified feedback signal score. The first, second, and third confidence levels are dynamically determined based on the real-time operating status of the corresponding signal acquisition devices and the data integrity rate.

[0075] In one embodiment, determining a comprehensive experience score based on user physiological signal quantification scores, user behavior signal quantification scores, user feedback signal quantification scores, sample labeling results, and a preset dynamic confidence level includes: determining a first product result corresponding to the user physiological signal quantification score and a first dynamic confidence level; determining a second product result corresponding to the user behavior signal quantification score and a second dynamic confidence level; determining a third product result corresponding to the user feedback signal quantification score and a third dynamic confidence level; summing the first, second, and third product results to obtain a first summation result; determining a fourth product result corresponding to the sample labeling results and a preset abnormal sample value penalty coefficient; and determining the comprehensive experience score based on the first summation result and the fourth product result.

[0076] Specifically, the formula for calculating the overall experience score can be expressed as follows: In the formula, To calculate the overall experience score, Quantify and score physiological signals Quantify the score of behavioral signals, Quantize the score for the feedback signal; The samples are labeled as follows: 1 for abnormal and 0 for normal. This is the outlier penalty coefficient, used to control the severity of the penalty for outlier samples. , , The dynamic confidence level of physiological, behavioral, and feedback signals is dynamically determined based on the real-time working status and data integrity rate of the corresponding signal acquisition device, with a value range of [0.6-1]. The higher the signal quality, the higher the confidence level. The vehicle-side experience perception layer judges the comprehensive experience score. When the comprehensive experience score is lower than the dynamic threshold (time window length is 3 seconds) for two consecutive time windows, or when the decrease exceeds the preset rate of change threshold within 10 seconds, the causal inference process is triggered.

[0077] The S2100 and cloud respond to experience degradation signals, acquire information about the user's interaction process in the cockpit, and break down the interaction process information into multiple independent functional nodes based on the pre-configured cockpit interaction function dependency topology graph.

[0078] Among them, a functional node represents the smallest functional unit node in the user's cockpit interaction process; each functional node has upstream and downstream dependencies.

[0079] In this embodiment, the cloud responds to experience degradation signals by acquiring information about the user's interaction process within the cockpit. Based on a pre-defined cockpit interaction function dependency topology graph (a directed acyclic graph), the complete interaction process is broken down into the smallest functional unit nodes, such as voice input, ASR recognition, intent understanding, dialogue generation, and TTS output. Directed edges represent the upstream and downstream dependencies between data flow and control flow; node anomalies can only propagate downstream along directed edges and cannot be reversed.

[0080] S2110. Obtain the status identifier corresponding to each functional node; wherein, the status identifier includes normal status identifier and abnormal status identifier.

[0081] In this embodiment, upon receiving a signal indicating a deterioration in user experience, the current complete cockpit interaction process is broken down into independent functional nodes, and the real-time operating status indicators of each functional node are acquired simultaneously. The status indicator is 1 for abnormal and 0 for normal.

[0082] S2120. Determine the coupling degree between functional nodes based on the dependencies between them.

[0083] In this embodiment, the degree of coupling between functional nodes can be determined by the dependencies between them. Specifically, let... The causal propagation weight from node k to node j, also known as the coupling degree, can be assigned a value based on the upstream and downstream link relationships in the functional dependency topology graph and historical intervention log statistics. This coupling degree characterizes the transmission strength of the state deviation of the upstream node to the downstream node. When node k is a direct predecessor dependency of node j, Take the higher value; if there is no dependent path or it is a downstream node, then... This ensures that the calculation of causal contribution strictly follows the data flow. The calculation of causal effect not only considers the anomalies of the node itself, but also the propagation impact of the anomalies along the interaction link.

[0084] S2130. Determine the deviation between the experience score and the benchmark score for each functional node.

[0085] In this embodiment, upon receiving a degradation trigger signal, the current complete cockpit interaction process is broken down into independent functional nodes. The standard experience benchmark score for each functional node is simultaneously acquired, and the deviation between the experience score for each functional node and the benchmark score is calculated. In this embodiment, the experience score for each functional node can be obtained through the node's key performance indicators (such as the "recognition accuracy" of the ASR node and the "synthesis latency" of the TTS node) and a score mapping function. Specifically, the real-time monitoring value of at least one key performance indicator (KPI) of the functional node is acquired; based on a preset experience mapping function, the real-time monitoring value is converted into a normalized real-time experience score; wherein the mapping function includes a piecewise linear mapping. For example, the experience mapping for ASR recognition accuracy is as follows: accuracy ≥ 95%, real-time experience score = 1.0 (full score); accuracy = 85%, real-time experience score = 0.71; accuracy ≤ 30%, real-time experience score = 0 (unacceptable). The standard experience benchmark score in this embodiment can be adaptively set based on experience.

[0086] S2140. Determine the root cause contribution and causal effect values ​​of each functional node based on the status identifier, coupling correlation degree, and deviation.

[0087] In this embodiment, a unified formula can be used to quantitatively assess single-node anomalies and two-node coupled anomalies through state identification, coupling correlation, and deviation. In one embodiment, the formula for calculating the root cause contribution causal effect value is expressed as follows: ;in, This is represented as the root cause contribution causal effect quantification value of the k-th functional node; This represents the status identifier of the k-th functional node, where an abnormal status identifier is 1 and a normal status identifier is 0. This is represented as the status identifier of the j-th functional node; This represents the coupling degree between functional node k and functional node j. When functional node k is a direct predecessor dependency of functional node j, The highest value, 1, is taken when functional node k and functional node j have no dependency, or when functional node k is a downstream functional node of functional node j. The value is 0; This is represented as the deviation between the experience score of the k-th functional node and the benchmark score; This represents the deviation between the experience score of the j-th functional node and the baseline score. When only functional node k is abnormal, because Automatic zeroing, which only calculates the experience loss caused by the node's own anomaly, accurately identifies a single source of failure. This can be understood as follows: one node is normal, one node is abnormal (one has a value of 0, the other a value of 1), and their product is 0. When function k and function node j are both abnormal... When activated, the sum of the absolute values ​​of the deviations between functional nodes k and j is included in the total effect; when functional node k is normal, the comprehensive causal effect value is 0, and the root cause determination and collection instructions for functional node k are not triggered.

[0088] S2150. Sort the root cause contribution and causal effect values ​​corresponding to each functional node according to their numerical values ​​to obtain the sorting results.

[0089] In this embodiment, after calculating the comprehensive causal effect value, the value is directly used as the basis for the root cause contribution, and the roots are sorted by value to select high-priority abnormal roots.

[0090] S2160. Select the high-priority root cause contribution causal effect values ​​from the sorting results as abnormal root causes; among them, the functional nodes corresponding to the abnormal root causes are designated as abnormal root cause nodes.

[0091] In this embodiment, high-priority root cause contribution causal effect values ​​are selected from the sorting results as abnormal root causes; wherein, the functional nodes corresponding to the abnormal root causes are designated as abnormal root cause nodes.

[0092] S2170. Based on the node parameter information of the abnormal root cause node, the associated nodes, and the directional data collection requirements, generate a directional data collection instruction and send it to the corresponding vehicle terminal.

[0093] In this embodiment, a targeted data collection instruction is generated based on the high-priority root cause nodes located and sent to the corresponding vehicle terminal. Specifically, the instruction clearly indicates the root cause functional node number, the data type to be collected, the collection time window length, the data format, and the transmission requirements, and is sent to the vehicle terminal that triggered the experience degradation through the vehicle-cloud security channel.

[0094] S2180: Parse the targeted data collection instructions issued by the cloud to obtain the root cause node of the anomaly and the directly preceding related node corresponding to the root cause node of the anomaly.

[0095] In this embodiment, the vehicle receives and parses the cloud-based directional data collection command, accurately locates the root cause-corresponding functional node and the directly related upstream node, and determines the collection range.

[0096] S2190, Locate the first operating module corresponding to the abnormal root cause node, and the second operating module corresponding to the direct upstream preceding associated node that provides input to the abnormal root cause node, and delineate the collection boundary including the first operating module and the second operating module.

[0097] In this embodiment, the first operating module corresponding to the abnormal root cause node and the second operating module corresponding to the direct upstream preceding associated node that provides input to the abnormal root cause node are located, and the acquisition boundary containing the first operating module and the second operating module is defined. This can be understood as, based on the node information in the instruction, locating the operating module of the root cause node in the cockpit model and the direct upstream preceding node that provides input to the root cause node, defining the acquisition boundary containing only the two types of nodes, and not expanding the acquisition of data from irrelevant modules.

[0098] S2200: The first running data corresponding to the first running module and the second running data corresponding to the second running module in the data acquisition boundary are taken as the data acquisition range, and directional data acquisition is performed within the data acquisition range.

[0099] In this embodiment, the first running data corresponding to the first running module and the second running data corresponding to the second running module within the acquisition boundary are used as the data acquisition range, and directional data acquisition is performed within the data acquisition range. Specifically, the dataset is acquired. Defined as: In the formula, The root cause corresponds to the functional node's running data. This refers to the operational data of the upstream related nodes for the root cause analysis function. The targeted data execution layer strictly limits the collection objects according to the formula, retaining only the data of the root cause node and its upstream related nodes, and not collecting redundant data from irrelevant modules or scenarios. It should be noted that since the experience degradation is mainly caused by the root cause node and its direct input links, collecting only the data of the above-mentioned related nodes is sufficient to fully cover the anomaly propagation path. For example, when the root cause is an anomaly of the intent understanding node, only upstream data such as user input text and ASR recognition results, as well as the output and operational status data of the intent understanding node, are collected, while downstream irrelevant data such as dialogue generation and TTS output are not collected.

[0100] The technical solution described in this invention establishes a direct causal relationship between subjective experience quantification indicators and data triggering mechanisms, enabling targeted collection of closed-loop data only when experience deteriorates, thus filtering low-value redundant data at the source. This significantly reduces cloud storage and computing load, ensuring that model iteration direction is always highly aligned with users' real experience improvement needs. By designing a causal effect quantification formula, the complete interaction process is broken down into the smallest functional nodes. Through dynamic calculation of anomaly identification and node coupling degree, it automatically distinguishes between single faults and multi-node collaborative anomalies and accurately sorts them. False associations can be eliminated without pre-setting complex rules, directly outputting high-priority root causes, providing strong causal samples for targeted model iteration, significantly shortening the optimization cycle and improving diagnostic accuracy. By automatically adjusting trigger thresholds and collection strategies based on real-time interaction status, historical benchmarks, and multimodal signal quality, it effectively overcomes the adaptive bottleneck of "rules lagging behind experience evolution," significantly enhancing the generalization ability for long-tail scenarios and differentiated interactions, and greatly reducing rule maintenance and manual annotation costs. By strictly limiting the collection scope to only "root cause node operation data + directly related data", redundant crawling of unrelated upstream and downstream modules is eliminated; combined with vehicle-side local compliance filtering, invalid fragment removal and lightweight encapsulation technology, the data packet size is greatly compressed under the premise of strictly meeting privacy and security requirements, minimizing vehicle-to-cloud transmission bandwidth, and building a high-quality data foundation for continuous model iteration.

[0101] In one embodiment, to facilitate a better understanding of the cockpit data directional acquisition method, Figure 3 This is a flowchart illustrating another method for targeted cockpit data collection provided in an embodiment of the present invention. By constructing a subjective experience objectification and standardization system, it achieves accurate identification of experience degradation and triggers data collection; by introducing a vehicle-cloud collaborative causal reasoning mechanism, it deeply locates the root causes of user experience degradation, and then implements targeted collection of high-value samples based on the contribution of the root causes, ultimately constructing a data collection method and system with user experience as the core driver, low redundancy transmission, and adaptive evolution capabilities.

[0102] like Figure 3 As shown, the specific methods for directional cockpit data acquisition include:

[0103] a1. The vehicle-side experience perception layer synchronously collects the user's physiological signals, behavioral signals, and feedback signals within the cabin, and performs quantification and scoring for each signal. Specifically, the user's physiological indicators are collected through the in-cabin sensing devices, the user's behavioral indicators are collected through the vehicle system logs and CAN bus, and the feedback signals from the vehicle's in-vehicle interaction components are collected through the user feedback system. The corresponding quantification scores are calculated using the physiological indicators, behavioral signals, and feedback signals, respectively, to obtain the physiological signal quantification score and the user feedback signal quantification score.

[0104] a2. The vehicle-side combines the quantitative scores of physiological signals, behavioral signals, and feedback signals with dynamic confidence to perform weighted fusion, introduces outlier penalty correction, and calculates the comprehensive experience score. When the comprehensive experience score is lower than the dynamic threshold for two consecutive time windows (the time window length is 3 seconds), or the decrease exceeds the preset rate of change threshold within 10 seconds, the causal inference process is triggered.

[0105] a3. After receiving the experience degradation trigger signal, the cloud-based causal reasoning layer breaks down the current cockpit's complete interaction process into independent functional nodes, and simultaneously obtains the real-time operating status and standard experience benchmark score of each node. After calculating the comprehensive causal effect value, it directly uses this value as the basis for the root cause contribution, sorts them by value size, and filters out high-priority abnormal root causes.

[0106] a4. The vehicle receives and parses the cloud-based directional data collection instructions, accurately locates the root cause-corresponding functional nodes and directly related upstream nodes, and determines the collection range.

[0107] a5. The vehicle-side performs local compliance preprocessing on the collected valid data, completing data regularization and lightweight packaging.

[0108] In one embodiment, Figure 4 This is a schematic diagram of a cockpit data orientation acquisition system provided in an embodiment of the present invention. Figure 4 As shown, the system includes a vehicle-side module 410 and a cloud-side module 420; wherein, the vehicle-side module 410 includes: a signal quantization and determination module 4101, a degraded signal transmission module 4102, and a directional data acquisition module 4103; the cloud-side module 420 includes: a root cause contribution determination module 4201 and an acquisition command issuance module 4202;

[0109] The signal quantization score determination module 4101 is used to collect parameter information generated by user interaction in the cockpit and determine the signal quantization score corresponding to the parameter information; wherein the parameter information includes at least user physiological signals, user behavior signals and user feedback signals.

[0110] The degradation signal sending module 4102 is used to determine the overall experience score based on the signal quantization score, and when the overall experience score meets the preset requirements, to trigger an experience degradation signal and send the experience degradation signal to the cloud 420.

[0111] The root cause contribution determination module 4201 responds to the experience degradation signal by acquiring the user's interaction process information in the cockpit and determining the root cause contribution causal effect value based on the pre-configured cockpit interaction function dependency topology and the interaction process information.

[0112] The data acquisition instruction sending module 4202 generates a directional data acquisition instruction based on the root cause contribution causal effect value and sends it to the vehicle terminal 410.

[0113] The directional data acquisition module 4103 determines the data acquisition range according to the directional data acquisition instruction, and performs directional data acquisition within the data acquisition range.

[0114] In one embodiment, the preset requirements include: the overall experience score being below a preset threshold for a continuous preset time window, or the overall experience score decreasing by a rate of change exceeding a preset threshold within a preset time period.

[0115] In one embodiment, the signal quantization and determination module 4101 includes:

[0116] A physiological indicator determination unit is used to collect the user's physiological indicator information through a preset sensing device; wherein, the physiological indicator information includes at least facial expression, heart rate, respiratory rate, and blink frequency;

[0117] The behavior indicator determination unit is used to collect the user's behavior indicator information through the vehicle's log information and the CAN bus; wherein, the behavior indicator information includes at least the number of voice interruptions, the number of manual operations, the number of command repetitions, the page dwell time, and the number of operation cancellations;

[0118] The feedback signal determination unit is used to collect feedback signal information from the in-vehicle interaction component at the vehicle end;

[0119] The physiological signal quantification score determination unit is used to determine the user's physiological signal quantification score based on the physiological indicator information, preset physiological indicator weights, and preset physiological indicator benchmark values.

[0120] The behavior signal quantification score determination unit is used to determine the user behavior signal quantification score based on behavior indicator information, preset behavior indicator weights, and preset behavior indicator benchmark values.

[0121] The feedback signal quantization score determination unit is used to determine the user feedback signal quantization score based on a preset hierarchical mapping rule and the feedback signal information.

[0122] The signal quantization score determination unit is used to take the user's physiological signal quantization score, the user's behavior signal quantization score, and the user's feedback signal quantization score as the corresponding signal quantization scores.

[0123] In one embodiment, the degraded signal transmission module 4102 includes:

[0124] The sample labeling unit is used to obtain historical parameter information generated by the user's historical interactions, select abnormal parameters and normal parameters from the historical parameter information, and use the abnormal parameters and normal parameters as samples to perform sample labeling to obtain sample labeling results;

[0125] The comprehensive experience score determination unit is used to determine the comprehensive experience score based on the user's physiological signal quantification score, the user's behavior signal quantification score, the user's feedback signal quantification score, the sample labeling result, and a preset dynamic confidence level.

[0126] In one embodiment, the preset dynamic confidence level includes: a first dynamic confidence level corresponding to the user's physiological signal quantification score, a second dynamic confidence level corresponding to the user's behavioral signal quantification score, and a third dynamic confidence level corresponding to the user's feedback signal quantification score.

[0127] The first confidence level, the second confidence level, and the third confidence level are dynamically determined based on the real-time working status of the corresponding signal acquisition device and the data integrity rate.

[0128] In one embodiment, the comprehensive experience score determination unit is specifically used for:

[0129] Determine the first product result of the user's physiological signal quantization score and the first dynamic confidence level;

[0130] Determine the second product result corresponding to the quantized score of the user behavior signal and the second dynamic confidence level;

[0131] Determine the third product result corresponding to the quantized score of the user feedback signal and the third dynamic confidence level;

[0132] The first product result, the second product result, and the third product result are summed to obtain the first summation result;

[0133] Determine the fourth product result corresponding to the sample labeling result and the preset abnormal sample value penalty coefficient;

[0134] The overall experience score is determined based on the first summation result and the fourth product result.

[0135] In one embodiment, the root cause contribution determination module 4201 includes:

[0136] The node splitting unit is used to split the interaction process information into multiple independent functional nodes based on the pre-configured cockpit interaction function dependency topology graph; wherein, the functional node represents a minimum functional unit node of the user in the cockpit interaction process; each functional node has upstream and downstream dependencies.

[0137] The identifier acquisition unit is used to acquire the status identifier corresponding to each of the functional nodes; wherein, the status identifier includes a normal status identifier and an abnormal status identifier;

[0138] The coupling correlation degree determination unit is used to determine the coupling correlation degree between each of the functional nodes based on the dependency relationship between each of the functional nodes.

[0139] The deviation determination unit is used to determine the deviation between the experience score and the benchmark score for each of the aforementioned functional nodes.

[0140] The root cause effect value determination unit is used to determine the root cause contribution causal effect value corresponding to each of the functional nodes based on the state identifier, the coupling correlation degree, and the deviation.

[0141] In one embodiment, the formula for calculating the root cause contribution causal effect value is expressed as follows: ;in, This is represented as the root cause contribution causal effect quantification value of the k-th functional node; This represents the status identifier of the k-th functional node, where an abnormal status identifier is 1 and a normal status identifier is 0. This is represented as the status identifier of the j-th functional node; This represents the coupling degree between functional node k and functional node j. When functional node k is a direct predecessor dependency of functional node j, The highest value, 1, is taken when functional node k and functional node j have no dependency, or when functional node k is a downstream functional node of functional node j. The value is 0; This is represented as the deviation between the experience score of the k-th functional node and the benchmark score; This represents the deviation between the experience score of the j-th functional node and the benchmark score. because Automatic zeroing, which only calculates the experience loss caused by the node's own anomaly, accurately identifies a single source of failure. This can be understood as follows: one node is normal, one node is abnormal (one has a value of 0, the other a value of 1), and their product is 0. When function k and function node j are both abnormal... When activated, the sum of the absolute values ​​of the deviations between functional nodes k and j is included in the total effect; when functional node k is normal, the comprehensive causal effect value is 0, and the root cause determination and collection instructions for functional node k are not triggered.

[0142] In one embodiment, the acquisition command issuing module 4202 includes:

[0143] The effect value sorting unit is used to sort the root cause contribution causal effect values ​​corresponding to each of the functional nodes according to their numerical values ​​to obtain the sorting results.

[0144] A filtering unit is used to filter out high-priority root cause contribution causal effect values ​​from the sorting results as abnormal root causes; wherein, the functional node corresponding to the abnormal root cause is an abnormal root cause node.

[0145] The instruction issuing unit is used to generate a targeted data collection instruction based on the node parameter information of the abnormal root cause node, associated nodes, and targeted data collection requirements, and then issue it to the corresponding vehicle terminal.

[0146] In one embodiment, the directional data acquisition instruction includes the root cause node number, data acquisition type, data acquisition time window length, and associated nodes of the root cause node;

[0147] Correspondingly, the directional data acquisition module 4103 includes:

[0148] The parsing unit is used to parse the targeted data collection instructions sent from the cloud to obtain the root cause node of the anomaly and the direct preceding associated node corresponding to the root cause node of the anomaly.

[0149] The positioning unit is used to locate the first operating module corresponding to the abnormal root cause node and the second operating module corresponding to the direct upstream preceding associated node that provides input to the abnormal root cause node, and to delineate the collection boundary including the first operating module and the second operating module.

[0150] The data acquisition range determination unit is used to take the first running data corresponding to the first running module and the second running data corresponding to the second running module in the acquisition boundary as the data acquisition range.

[0151] In one embodiment, the directional data acquisition instruction further includes: data format and transmission requirements; correspondingly, the system further includes:

[0152] The data upload module is used by the vehicle to organize and lightweight encapsulate the directionally collected data information according to the data format, and upload it to the cloud according to the transmission requirements.

[0153] The cockpit data orientation acquisition system provided in this embodiment of the invention can execute the cockpit data orientation acquisition method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0154] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0155] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for directional acquisition of cockpit data, characterized in that, An application to a cockpit data orientation acquisition system, the system comprising a vehicle-side terminal and a cloud-based terminal; the method comprising: The vehicle-mounted device collects parameter information generated by user interactions within the cockpit and determines the signal quantization score corresponding to the parameter information. The vehicle determines a comprehensive experience score based on the signal quantization score, and if the comprehensive experience score meets the preset requirements, it triggers an experience degradation signal and sends the experience degradation signal to the cloud. In response to the experience degradation signal, the cloud acquires information about the user's interaction process in the cabin and determines the root cause contribution causal effect value based on the pre-configured cabin interaction function dependency topology and the interaction process information. The cloud platform generates a targeted data collection instruction based on the root cause contribution causal effect value and sends it to the vehicle terminal. The vehicle-mounted device determines the data collection range according to the directional data collection instruction, and performs directional data collection within the data collection range.

2. The method according to claim 1, characterized in that, The preset requirements include: the overall experience score is below a preset threshold for a continuous preset time window, or the overall experience score decreases by more than a preset rate of change threshold within a preset time period.

3. The method according to claim 1, characterized in that, The vehicle-side determines the signal quantization score corresponding to the parameter information, including: The user's physiological indicators are collected through preset sensing devices; The user's behavioral metrics are collected through the vehicle's log information and the CAN bus. Collect feedback signal information from the in-vehicle interaction components at the vehicle end; The user's physiological signal quantification score is determined based on the physiological indicator information, preset physiological indicator weights, and preset physiological indicator benchmark values. The user behavior signal quantification score is determined based on behavioral indicator information, preset behavioral indicator weights, and preset behavioral indicator benchmark values. The user feedback signal quantization score is determined based on the preset hierarchical mapping rules and the feedback signal information. The user's physiological signal quantification score, the user's behavioral signal quantification score, and the user's feedback signal quantification score are respectively used as the corresponding signal quantification scores.

4. The method according to claim 1, characterized in that, The vehicle-side determines a comprehensive experience score based on the signal quantization score, including: Obtain historical parameter information generated by the user's historical interactions, select abnormal parameters and normal parameters from the historical parameter information, and use the abnormal parameters and normal parameters as samples to perform sample labeling to obtain sample labeling results; The comprehensive experience score is determined based on the user's physiological signal quantification score, the user's behavioral signal quantification score, the user's feedback signal quantification score, the sample labeling results, and the preset dynamic confidence level.

5. The method according to claim 4, characterized in that, The preset dynamic confidence level includes: a first dynamic confidence level corresponding to the user's physiological signal quantification score, a second dynamic confidence level corresponding to the user's behavioral signal quantification score, and a third dynamic confidence level corresponding to the user's feedback signal quantification score. The first confidence level, the second confidence level, and the third confidence level are dynamically determined based on the real-time working status of the corresponding signal acquisition device and the data integrity rate.

6. The method according to claim 5, characterized in that, The determination of the comprehensive experience score based on the user's physiological signal quantification score, the user's behavioral signal quantification score, the user's feedback signal quantification score, the sample labeling results, and the preset dynamic confidence level includes: Determine the first product result of the user's physiological signal quantization score and the first dynamic confidence level; Determine the second product result corresponding to the quantized score of the user behavior signal and the second dynamic confidence level; Determine the third product result corresponding to the quantized score of the user feedback signal and the third dynamic confidence level; The first product result, the second product result, and the third product result are summed to obtain the first summation result; Determine the fourth product result corresponding to the sample labeling result and the preset abnormal sample value penalty coefficient; The overall experience score is determined based on the first summation result and the fourth product result.

7. The method according to claim 1, characterized in that, The cloud-based determination of root cause contribution causal effect values ​​based on a pre-configured cockpit interaction function dependency topology and the interaction process information includes: Based on the pre-configured cockpit interaction function dependency topology graph, the interaction process information is divided into multiple independent functional nodes; wherein, each functional node represents a minimum functional unit node of the user in the cockpit interaction process; each functional node has upstream and downstream dependencies. Obtain the status identifier corresponding to each of the aforementioned functional nodes; wherein, the status identifier includes a normal status identifier and an abnormal status identifier; The coupling degree between the functional nodes is determined based on the dependencies between them. Determine the deviation between the experience score and the benchmark score for each of the aforementioned functional nodes; The root cause contribution causal effect value of each functional node is determined based on the status identifier, the coupling correlation degree, and the deviation.

8. The method according to claim 7, characterized in that, The formula for calculating the causal effect value of the root cause contribution is expressed as follows: ;in, This is represented as the root cause contribution causal effect quantification value of the k-th functional node; This represents the status identifier of the k-th functional node, where an abnormal status identifier is 1 and a normal status identifier is 0. This is represented as the status identifier of the j-th functional node; This represents the coupling degree between functional node k and functional node j. When functional node k is a direct predecessor dependency of functional node j, The highest value, 1, is taken when functional node k and functional node j have no dependency, or when functional node k is a downstream functional node of functional node j. The value is 0; This is represented as the deviation between the experience score of the k-th functional node and the benchmark score; This represents the deviation between the experience score of the j-th functional node and the benchmark score.

9. The method according to claim 1, characterized in that, The cloud platform generates a targeted data collection instruction based on the root cause contribution causal effect value and sends it to the vehicle terminal, including: The root cause contribution causal effect values ​​corresponding to each of the aforementioned functional nodes are sorted according to their numerical values ​​to obtain the sorting results. From the sorting results, high-priority root cause contribution causal effect values ​​are selected as abnormal root causes; wherein, the functional nodes corresponding to the abnormal root causes are designated as abnormal root cause nodes. Based on the node parameter information, associated nodes, and directional data collection requirements of the abnormal root cause node, a directional data collection instruction is generated and sent to the corresponding vehicle terminal.

10. The method according to claim 1, characterized in that, The targeted data acquisition instruction includes the abnormal root cause node number, data acquisition type, data acquisition time window length, and associated nodes of the abnormal root cause node; Accordingly, the vehicle-mounted device determines the data collection range according to the directional data collection instruction, including: The targeted data collection instructions sent from the cloud are analyzed to obtain the root cause node of the anomaly and the directly preceding associated node of the root cause node of the anomaly. Locate the first operating module corresponding to the abnormal root cause node, and the second operating module corresponding to the direct upstream preceding associated node that provides input to the abnormal root cause node, and define the collection boundary that includes the first operating module and the second operating module. The first running data corresponding to the first running module and the second running data corresponding to the second running module within the acquisition boundary are taken as the data acquisition range.

11. The method according to claim 1, characterized in that, The directional data acquisition instruction also includes: data format and transmission requirements; correspondingly, the method also includes: The vehicle-mounted device organizes and lightweight packages the directionally collected data information according to the data format, and uploads it to the cloud according to the transmission requirements.

12. A cockpit data orientation acquisition system, characterized in that, The system is applied to the cockpit data directional acquisition method according to claims 1-10; the system includes a vehicle-side terminal and a cloud-based terminal. The vehicle-side component includes: a signal quantization and determination module, a degraded signal transmission module, and a directional data acquisition module; the cloud-side component includes: a root cause contribution determination module and an acquisition command issuance module. The signal quantization score determination module is used to collect parameter information generated by user interaction in the cockpit and determine the signal quantization score corresponding to the parameter information; wherein the parameter information includes at least user physiological signals, user behavioral signals and user feedback signals; The degradation signal sending module is used to determine the overall experience score based on the signal quantization score, and when the overall experience score meets the preset requirements, to trigger an experience degradation signal and send the experience degradation signal to the cloud. The root cause contribution determination module responds to the experience degradation signal by acquiring user interaction process information in the cabin and determining the root cause contribution causal effect value based on the pre-configured cabin interaction function dependency topology and the interaction process information. The data acquisition instruction issuing module generates a targeted data acquisition instruction based on the root cause contribution causal effect value and issues it to the vehicle terminal. The directional data acquisition module determines the data acquisition range according to the directional data acquisition instruction, and performs directional data acquisition within the data acquisition range.