Vehicle-mounted DDS adaptive QoS configuration method and system based on width learning

Abstract

The application relates to a vehicle-mounted DDS adaptive QoS configuration method and system based on width learning, which comprises the following steps: based on an original data set, an initial QoS configuration model fusing safety features is constructed through a width learning framework, the initial QoS configuration model is trained through a sparse self-encoding algorithm and in combination with a safety constraint loss function, and a target QoS configuration model is obtained; an initial QoS configuration scheme is generated through the target QoS configuration model, the initial QoS configuration scheme is subjected to compatibility checking and safety verification based on a QoS incompatible configuration item combination library and a safety risk evaluation rule library of the vehicle-mounted DDS, and a safe and usable target QoS configuration file is obtained. Through the application, the automation, intelligence and safety of QoS configuration can be realized, the configuration efficiency, communication adaptability and safety protection capability are improved, and the safety of vehicle-mounted core data transmission is ensured.

Description

Technical Field

[0001] This invention relates to the field of vehicle communication technology, and in particular to a vehicle-mounted DDS adaptive QoS configuration method and system based on width learning. Background Technology

[0002] With the rapid development of autonomous driving technology and intelligent connected vehicles, the communication needs within the vehicle system are becoming increasingly complex. The data streams that need to be transmitted cover various types, including autonomous driving perception data, control command data, and in-vehicle entertainment data. Different types of data streams have significantly different requirements for communication real-time performance, reliability, and bandwidth usage. Among them, core data such as autonomous driving control commands have extremely high requirements for communication security. Once a security incident such as data tampering or unauthorized access occurs, it will directly threaten the driving safety of the vehicle.

[0003] As a highly efficient distributed real-time communication middleware, DDS has been widely used in vehicle communication systems. Its QoS configuration directly determines the quality of communication services and is a core component for adapting to different data stream requirements. Currently, QoS configuration in existing vehicle DDS systems is mostly done manually, requiring technicians to set fixed QoS configuration items for different communication scenarios based on experience. However, this approach has significant drawbacks: First, it has poor adaptability. Vehicle communication scenarios are dynamically changing (e.g., network bandwidth fluctuations and data stream type switching during vehicle operation), and fixed configurations cannot adapt to these changes in real time. Second, it suffers from sluggish response. Manual configuration involves multiple steps such as requirements analysis and parameter tuning, making it difficult to meet the stringent real-time requirements of scenarios like autonomous driving. Third, it relies heavily on professional experience, resulting in low configuration efficiency and susceptibility to human error leading to improper configuration, causing communication failures such as excessive transmission latency and increased packet loss rates. Fourth, it lacks security protection. Existing configuration schemes do not fully consider the security risks of vehicle communication and do not design adaptive configuration strategies for different security levels, making it difficult to resist malicious attacks such as data tampering and unauthorized access, and failing to guarantee the security of core data transmission.

[0004] To address these issues, related technologies have attempted to introduce machine learning techniques to achieve automatic QoS configuration. However, existing solutions often employ complex models such as deep learning, which suffer from long training cycles and slow inference speeds, making them unsuitable for the real-time requirements of in-vehicle systems. Furthermore, existing solutions do not incorporate security enhancement designs, thus failing to resolve the security protection issues of in-vehicle communication. Summary of the Invention

[0005] Therefore, the purpose of this invention is to provide a method and system for adaptive QoS configuration of vehicle-mounted DDS based on width learning, so as to overcome the shortcomings of the prior art.

[0006] To achieve the above objectives, this invention provides a vehicular DDS adaptive QoS configuration method based on width learning, the method comprising: Historical QoS configuration data, corresponding communication performance parameters, communication performance indicators, and security event data of DDS communication in vehicle scenarios are collected to construct an original dataset. The original dataset is then cleaned and standardized to obtain a standardized training dataset. Based on the standardized training dataset, an initial QoS configuration model incorporating security features is constructed using a wide learning framework. The initial QoS configuration model is then trained using a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thus obtaining the target QoS configuration model. Based on real-time vehicle communication requirements and security level requirements, an initial QoS configuration scheme is generated through the target QoS configuration model. The initial QoS configuration scheme is then subjected to compatibility verification and security verification based on the QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle DDS, resulting in a secure and usable target QoS configuration file.

[0007] The beneficial effects of this invention are as follows: By collecting historical QoS configuration data, corresponding communication performance parameters, communication performance indicators, and security event data of DDS communication in vehicle scenarios, an original dataset is constructed. The original dataset is cleaned and standardized to obtain a standardized training dataset. Based on the standardized training dataset, an initial QoS configuration model incorporating security features is constructed using a wide learning framework. The initial QoS configuration model is trained using a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thus obtaining a target QoS configuration model. Then, based on real-time vehicle communication requirements and security level requirements, an initial QoS configuration scheme is generated using the target QoS configuration model. The initial QoS configuration scheme is then subjected to compatibility verification and security verification based on the QoS incompatible configuration item combination library and security risk assessment rule library of vehicle DDS, respectively, to obtain a safe and usable target QoS configuration file. Unlike existing technologies, this invention achieves automation, intelligence, and security of QoS configuration, improves configuration efficiency, communication adaptability, and security protection capabilities, and also ensures the security of core vehicle data transmission.

[0008] Furthermore, the initial QoS configuration model includes an input layer, a feature layer, a reinforcement layer, and an output layer. The feature layer uses ReLU as a non-linear activation function, and the number of nodes n in the feature layer ranges from 100 to 200. The reinforcement layer uses Sigmoid as a non-linear activation function, and the number of nodes m in the reinforcement layer ranges from 200 to 300. The steps of constructing an initial QoS configuration model incorporating security features based on the standardized training dataset using a wide learning framework include: The standardized training dataset is divided into a training set and a validation set according to a preset ratio. The sample features of the training set are received through the input layer. The sample features are mapped into high-dimensional feature vectors through the feature layer. The high-dimensional feature vectors are then subjected to reinforcement learning through the reinforcement layer to obtain reinforcement learning feature vectors. The high-dimensional feature vector and the reinforcement learning feature vector are concatenated into an enhanced feature matrix through the output layer, and the output of the initial QoS configuration model is obtained by solving the pseudo-inverse.

[0009] Furthermore, the expression for the output of the initial QoS configuration model is as follows:

[0010] in, For the input feature matrix, To and The corresponding label matrix, For regularization parameters, To and Identity matrices of the same dimension This is the model weight matrix. This is the enhanced feature matrix output by the model.

[0011] Furthermore, the expression for the safety constraint loss function is as follows:

[0012] in, The output of the safety constraint loss function, For mean square error loss, For safety risk losses, For safety weighting coefficients, The value range is 0.3-0.5. The expression for the safety constraint loss function is as follows:

[0013] in, The output of the safety constraint loss function, For mean square error loss, For safety risk losses, For safety weighting coefficients, The value range is 0.3-0.5.

[0014] Furthermore, after obtaining a secure and usable target QoS profile, the method further includes: The system collects real-time communication performance indicators and real-time security status data of the vehicle-mounted DDS system after adopting the target QoS configuration file, and compares the real-time communication performance indicators and real-time security status data with preset performance thresholds and security thresholds, respectively. When performance fails to meet standards or a security incident occurs, the configuration scheme, corresponding communication data, and security incident details are fed back to the original dataset to obtain the latest standardized training dataset. The target QoS configuration model is incrementally trained based on the latest standardized training dataset to optimize model parameters and security constraints.

[0015] Furthermore, the steps of cleaning and standardizing the original dataset include: The original dataset is processed by deleting invalid samples, imputing missing values, and removing outliers to obtain an intermediate original dataset. The Min-Max normalization method is used to map all sample data in the intermediate original dataset to the interval [0, 1], thereby quantifying and encoding the risk level corresponding to the security event data in the intermediate original dataset.

[0016] Furthermore, the method also includes: The configuration parameters of the target QoS configuration file are encrypted and stored using a symmetric encryption algorithm, and then bound to the identity information of the legitimate nodes of the vehicle-mounted DDS system.

[0017] Furthermore, the method also includes: Collect multiple QoS configuration items of the vehicle-mounted DDS system, verify the compatibility of different combinations of the QoS configuration items through multiple communication tests, record conflicting configuration item pairs and their corresponding conflict behaviors, establish the mapping relationship between the conflicting configuration item pairs and the conflict judgment criteria, and form a library of QoS incompatible configuration item combinations for vehicle-mounted DDS. The security risks of vehicle-mounted DDS communication are identified, assessment indicators and thresholds for these risks are formulated, a mapping relationship between QoS configuration items and security risks is established, and a security risk assessment rule base is formed.

[0018] To achieve the above objectives, the present invention also provides a vehicle-mounted DDS adaptive QoS configuration system based on width learning, used to implement the vehicle-mounted DDS adaptive QoS configuration method based on width learning described above, the system comprising: The module is used to collect historical QoS configuration data, corresponding communication performance parameters, communication performance indicators and security event data of DDS communication in vehicle scenarios, construct the original dataset, and clean and standardize the original dataset to obtain a standardized training dataset. The training module is used to construct an initial QoS configuration model that integrates security features based on the standardized training dataset through a wide learning framework, and to train the initial QoS configuration model through a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thereby obtaining the target QoS configuration model. The verification and validation module is used to generate an initial QoS configuration scheme based on real-time vehicle communication requirements and security level requirements, and through the target QoS configuration model. Based on the QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle DDS, the initial QoS configuration scheme is subjected to compatibility verification and security validation respectively to obtain a safe and usable target QoS configuration file. Attached Figure Description

[0019] Figure 1 This is a flowchart of the vehicle-mounted DDS adaptive QoS configuration method based on width learning according to Embodiment 1 of the present invention; Figure 2 This is an overall flowchart of the vehicle-mounted DDS adaptive QoS configuration method based on width learning according to Embodiment 2 of the present invention; Figure 3 This is a flowchart illustrating the data acquisition and preprocessing process for integrating security dimensions in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the QoS configuration model structure that integrates security features according to Embodiment 2 of the present invention; Figure 5 This is a schematic diagram of the model training and verification process according to Embodiment 2 of the present invention; Figure 6 This is a logical diagram illustrating the compatibility verification and security verification of Embodiment 2 of the present invention; Figure 7 This is a structural block diagram of the in-vehicle DDS adaptive QoS configuration system based on width learning according to Embodiment 3 of the present invention.

[0020] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0022] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0023] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0024] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0025] Example 1 Please see Figure 1 The flowchart below shows the vehicular DDS adaptive QoS configuration method based on width learning in the first embodiment of the present invention. The method includes the following steps: Step S101: Collect historical QoS configuration data, corresponding communication performance parameters, communication performance indicators and security event data of DDS communication in vehicle scenarios, construct the original dataset, clean and standardize the original dataset to obtain the standardized training dataset; The historical QoS configuration data includes configuration items such as reliability level, deadlines, lifetime, and data transmission mode; the communication performance parameters include data transmission rate, data priority, and network bandwidth utilization; the communication performance indicators include transmission delay, packet loss rate, and reliability score (out of 10, automatically scored by the system based on actual communication performance); the security event data includes the occurrence time, impact scope, risk level (1-5, with level 5 being the highest risk) of data tampering events, unauthorized access events, and transmission leakage events, as well as the corresponding QoS configuration.

[0026] Step S102: Based on the standardized training dataset, construct an initial QoS configuration model that integrates security features through a wide learning framework. Train the initial QoS configuration model using a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thereby obtaining the target QoS configuration model. The training dataset was divided into a training set and a validation set in a 7:3 ratio. The Sparse Autoencoder (SAE) algorithm was used to initialize the weight matrix W and the bias term b of the model. The number of iterations of the algorithm was set to 300, the learning rate was 0.02, and the sparsity parameter was set to 0.05 to ensure that the initialized parameters had a good training foundation.

[0027] The standardized training dataset obtained in step S101 is input into the initialized initial QoS configuration model for training. The training objective is to make the error between the predicted performance index and the actual historical performance index corresponding to the QoS configuration scheme output by the model less than a preset threshold, which is 0.03, and to meet the security risk assessment. In addition, a security constraint loss function is introduced during the training process.

[0028] A verification process is set up during training: after every 100 iterations, the prediction accuracy and security risk fit of the model are calculated using the verification set; if the accuracy of the verification set does not improve for 20 consecutive iterations (accuracy fluctuation ≤ 0.5%) and the security risk fit is not optimized, the early stop mechanism is triggered to terminate training and save the current optimal model parameters, thus obtaining the target QoS configuration model after training.

[0029] Step S103: Based on real-time vehicle communication requirements and security level requirements, and by generating an initial QoS configuration scheme through the target QoS configuration model, the initial QoS configuration scheme is subjected to compatibility verification and security verification based on the QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle DDS, to obtain a safe and usable target QoS configuration file.

[0030] Specifically, key parameters are extracted from the real-time vehicle communication requirements. These key parameters include data type (e.g., perception data, control command data, entertainment data), transmission priority (levels 1-5, with level 1 being the highest), real-time requirements (e.g., transmission latency ≤ 50ms), and security level requirements (levels 1-3, with level 1 being the highest security level). The extracted key parameters are converted into input vectors consistent with the training set format and input into the trained target QoS configuration model. The target QoS configuration model outputs an initial QoS configuration scheme through rapid inference. This scheme includes specific parameters for core configuration items such as reliability level, deadlines, and lifetime, as well as security enhancement configuration items.

[0031] Furthermore, the security enhancement configuration items in the security level requirements correspond one-to-one with the security level, specifically: Level 1 security level (core control data transmission scenario) corresponds to AES-256 encryption, two-way authentication, strict access control (only allowing preset legitimate nodes to access), and packet-by-packet verification (checksum using SHA-256); Level 2 security level (awareness data transmission scenario) corresponds to AES-128 encryption, one-way authentication, regular access control (allowing authorized nodes to access), and periodic verification (verifying every 10 data packets); Level 3 security level (entertainment data transmission scenario) corresponds to DES encryption, basic authentication, simplified access control, and critical packet verification (verifying only core data packets).

[0032] It should be noted that the initial QoS configuration scheme is subject to dual verification: it is compared one by one with the QoS incompatible configuration item combination library of the vehicle DDS to complete the compatibility verification; based on the security risk assessment rule library, the security enhancement configuration item parameters, encryption strength, identity authentication strategy, etc. in the scheme are verified to assess whether the security risk level corresponding to the scheme meets the real-time security level requirements.

[0033] Verification result processing: If there are no incompatible configuration items and the security risk level meets the requirements, proceed to the next step; if there are incompatible configuration items or the security risk does not meet the requirements, the conflict information and security risk information are fed back to the target QoS configuration model. The model readjusts the parameters based on the feedback information (focusing on optimizing the matching degree between security enhancement configuration items and core configuration items), generates a new QoS configuration scheme, and performs double verification again until a configuration scheme without incompatibility conflicts and meeting security standards is generated, i.e., the target QoS configuration file.

[0034] Through the above steps, historical QoS configuration data, corresponding communication performance parameters, communication performance indicators, and security event data of DDS communication in vehicle scenarios are collected to construct an original dataset. The original dataset is cleaned and standardized to obtain a standardized training dataset. Based on the standardized training dataset, an initial QoS configuration model incorporating security features is constructed using a wide learning framework. The initial QoS configuration model is trained using a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thus obtaining a target QoS configuration model. Then, based on real-time vehicle communication requirements and security level requirements, an initial QoS configuration scheme is generated through the target QoS configuration model. The initial QoS configuration scheme is then subjected to compatibility verification and security verification based on the QoS incompatible configuration item combination library and security risk assessment rule library of vehicle DDS, respectively, to obtain a safe and usable target QoS configuration file. Unlike existing technologies, this approach achieves automation, intelligence, and security in QoS configuration, improving configuration efficiency, communication adaptability, and security protection capabilities, while also ensuring the security of core vehicle data transmission.

[0035] Furthermore, the initial QoS configuration model includes an input layer, a feature layer, a reinforcement layer, and an output layer. The feature layer uses ReLU as a non-linear activation function, and the number of nodes n in the feature layer ranges from 100 to 200. The reinforcement layer uses Sigmoid as a non-linear activation function, and the number of nodes m in the reinforcement layer ranges from 200 to 300.

[0036] The input layer receives the standardized training dataset processed in step S101, and selects 10 core input features: data type (encoding: LiDAR data = 1, camera data = 2, fused data = 3, control command data = 4, entertainment data = 5), data transmission rate (normalized), data priority (encoding: levels 1-5 directly mapped to 1-5), network bandwidth utilization (normalized), road condition type (encoding), historical transmission delay (normalized), historical packet loss rate (normalized), historical reliability score (normalized), historical security risk level (normalized), and real-time security level requirements (encoding: levels 1-3 directly mapped to 1-3); the number of input layer nodes is 10.

[0037] Feature Layer: Its core function is to map the low-dimensional features of the input layer, i.e., the core input features (including security features), into a high-dimensional feature vector. ReLU is used as the non-linear activation function (activation function formula: f(x) = max(0, x)). The number of nodes n in the feature layer ranges from 100 to 200, and n can be adaptively adjusted according to the actual dataset size (n is 150-200 when the dataset sample size is greater than 1000; n is 100-150 when the sample size is less than 1000). The output of the feature layer is an n-dimensional high-dimensional feature vector. The expression for the feature layer is:

[0038] in, The output of the feature layer, This is the weight matrix from the input layer to the feature layer (dimension 10×n). This is the feature layer bias term (dimension 1×n). This is the output of the input layer, i.e., the core input features.

[0039] The reinforcement layer further enhances the high-dimensional feature vector output by the feature layer, focusing on strengthening the adaptability of security features. Sigmoid is used as the non-linear activation function (activation function formula: f(x) = 1 / (1+e)). The number of nodes m in the reinforcement layer ranges from 200 to 300, maintaining a ratio of approximately 2:1 with the number of nodes n in the feature layer. The output of the reinforcement layer is an m-dimensional enhanced feature vector. The expression for the reinforcement layer is:

[0040] in, The output of the reinforcement layer, is the weight matrix from the input layer to the reinforcement layer (dimension 10×m), and b is the bias term of the reinforcement layer (dimension 1×m).

[0041] Output layer: The outputs of the feature layer and the enhancement layer are concatenated to form an enhanced feature matrix Y (dimension N×(n+m), where N is the number of samples). The output of the initial QoS configuration model is obtained by solving the pseudo-inverse. The output is a QoS configuration scheme containing security enhancement configuration items.

[0042] Furthermore, the step of constructing an initial QoS configuration model incorporating security features based on the standardized training dataset using a wide learning framework includes: The standardized training dataset is divided into a training set and a validation set according to a preset ratio. The sample features of the training set are received through the input layer. The sample features are mapped into high-dimensional feature vectors through the feature layer. The high-dimensional feature vectors are then subjected to reinforcement learning through the reinforcement layer to obtain reinforcement learning feature vectors. The high-dimensional feature vector and the reinforcement learning feature vector are concatenated into an enhanced feature matrix through the output layer, and the output of the initial QoS configuration model is obtained by solving the pseudo-inverse.

[0043] Furthermore, the expression for the output of the initial QoS configuration model is as follows:

[0044] in, Let be the input feature matrix (the output of the input layer), λ be the regularization parameter (set to 0.001 to prevent overfitting), and I be the regularization parameter. Identity matrices of the same dimension To and The corresponding label matrix, This is the model weight matrix. This is the enhanced feature matrix output by the model.

[0045] Furthermore, the expression for the safety constraint loss function is as follows:

[0046] in, The output of the safety constraint loss function, Mean squared error loss is used to measure the deviation between the model's predicted QoS configuration parameters and historical true values. For security risk losses, it is used to measure the deviation between the model's predicted security configuration parameters and the historical actual security configuration. For safety weighting coefficients, The value ranges from 0.3 to 0.5, and it can be adjusted according to actual security requirements.

[0047] In addition, the mean squared error loss is used to measure the difference between the model-predicted QoS configuration and the actual QoS configuration. The formula is as follows:

[0048] in, For the first The QoS configuration parameters predicted by the model in each sample. For the first Historical real QoS configuration parameters in each sample .

[0049] A stepped weighted penalty loss with safety risk weights and a safety baseline threshold is adopted. A safety baseline threshold is set, and deviations that violate hard safety constraints are penalized squared, resulting in a safety risk loss. The formula is as follows:

[0050] in, For safety risk losses; The total number of samples; The total number of security enhancement configuration items (in this patent, these include four items: encryption algorithm strength, key length, authentication strategy, and access control rules). For the first The risk weights of each security configuration item are set according to the security level of the vehicle-mounted DDS, with core security items having higher weights (e.g., encryption algorithm strength). =5, key length =4, Identity Authentication =3, Access Control =2), reflecting the difference in security priority among different configuration items; For the first The security baseline threshold for each security configuration (such as a key length of no less than 128 bits and an encryption algorithm strength of no less than AES-128) will trigger a quadratic penalty for configurations that are below this threshold, forcing the model to prioritize meeting the hard security constraints. For the first The first sample The model predicts the security configuration parameters; For the first The first sample Historical, real-world security configuration parameters.

[0051] Furthermore, after obtaining a secure and usable target QoS profile, the method further includes: The system collects real-time communication performance indicators and real-time security status data of the vehicle-mounted DDS system after adopting the target QoS configuration file, and compares the real-time communication performance indicators and real-time security status data with preset performance thresholds and security thresholds, respectively. Specifically, the system collects in real time the communication performance indicators (transmission delay, packet loss rate, reliability score) and security status data (whether a security incident has occurred, the type and details of the security incident, and the actual protection effect of the security enhancement configuration items) of the vehicle-mounted DDS system after adopting the target QoS configuration file. The communication performance indicators are compared with preset performance thresholds (such as transmission delay ≤ 50ms, packet loss rate ≤ 2%, reliability score ≥ 8 points) and the security status data is compared with security thresholds (no high-risk security incidents, and the occurrence rate of medium- and low-risk security incidents ≤ 0.1%).

[0052] When performance fails to meet standards or a security incident occurs, the configuration scheme, corresponding communication data, and security incident details are fed back to the original dataset to obtain the latest standardized training dataset. The target QoS configuration model is incrementally trained based on the latest standardized training dataset to optimize model parameters and security constraints.

[0053] If the performance indicators meet the standards and no security incidents occur, or only low-risk security incidents occur (meeting the security threshold), the current configuration scheme, corresponding communication data, and security status data are stored in an encrypted historical database for subsequent dataset updates. If the performance does not meet the standards or a high-risk security incident occurs (not meeting the security threshold), the configuration scheme, corresponding communication data, security incident details, and reasons for non-compliance are fed back to the original dataset in step S101 to update the training set (focusing on supplementing security incident-related samples). Then, incremental training is performed on the target QoS configuration model to optimize model parameters and security constraints, thereby improving the accuracy and security protection capabilities of subsequent configurations.

[0054] Furthermore, the steps of cleaning and standardizing the original dataset include: The original dataset is processed by deleting invalid samples, imputing missing values, and removing outliers to obtain an intermediate original dataset. The original dataset is cleaned by deleting samples with more than 10% missing values, filling samples with ≤10% missing values ​​(numerical fields) using the mean imputation method, and filling categorical fields (such as data type and security event type) using the mode imputation method. Outliers are removed by the 3σ principle (i.e., when a data value deviates from the mean by more than 3 times the standard deviation, it is judged as an outlier and removed).

[0055] The Min-Max normalization method is used to map all sample data in the intermediate original dataset to the interval [0, 1], thereby quantifying and encoding the risk level corresponding to the security event data in the intermediate original dataset.

[0056] In the preprocessing of security incident data, the risk level is quantified and encoded, and standardized using the Min-Max normalization method along with other feature dimensions. All data are mapped to the [0,1] interval to obtain a training dataset containing the security dimension.

[0057] Furthermore, the method also includes: The configuration parameters of the target QoS configuration file are encrypted and stored using a symmetric encryption algorithm, and then bound to the identity information of the legitimate nodes of the vehicle-mounted DDS system.

[0058] Specifically, the final QoS configuration scheme (target QoS configuration file) is encrypted: a symmetric encryption algorithm (consistent with the encryption algorithm in the security enhancement configuration item) is used to encrypt and store the core parameters in the target QoS configuration file, generating an encrypted QoS configuration file (in XML format); at the same time, the identity information of the legitimate nodes of the vehicle DDS system (such as node MAC address and device serial number) is bound to the configuration file, and an authentication mechanism is set to ensure that only legitimate nodes can read and load the configuration file.

[0059] The encrypted final QoS configuration file is output to the encrypted configuration directory of the vehicle-mounted DDS system. After the vehicle-mounted DDS system completes authentication through the configuration loading module, it automatically reads and loads the configuration file to complete the QoS parameter configuration for the publisher and subscriber. After loading is completed, a "Configuration effective + Security verification passed" notification is triggered.

[0060] Furthermore, the method also includes: Collect multiple QoS configuration items of the vehicle-mounted DDS system, verify the compatibility of different combinations of the QoS configuration items through multiple communication tests, record conflicting configuration item pairs and their corresponding conflict behaviors, establish the mapping relationship between the conflicting configuration item pairs and the conflict judgment criteria, and form a library of QoS incompatible configuration item combinations for vehicle-mounted DDS. This involves collecting common QoS configuration items for vehicle-mounted DDS systems, verifying the compatibility of different configuration item combinations through multiple communication tests, recording conflicting configuration item pairs (such as "reliability level = reliable transmission" and "lifetime ≤ 100ms", "deadlines = 50ms" and "data transmission mode = batch transmission") and their corresponding conflict judgment criteria (such as determining incompatibility when the configuration item combination triggers a transmission delay exceeding 100 ms or a packet loss rate exceeding 5%), establishing a mapping relationship between conflicting configuration item pairs and conflict judgment criteria, and forming a standardized library of incompatible configuration item combinations (supporting subsequent updates and iterations).

[0061] The security risks of vehicle-mounted DDS communication are identified, assessment indicators and thresholds for these risks are formulated, a mapping relationship between QoS configuration items and security risks is established, and a security risk assessment rule base is formed.

[0062] This study identifies typical security risks in vehicle-mounted DDS communication, including data tampering, unauthorized access, security vulnerabilities caused by excessive transmission latency, and insufficient encryption strength. Clear assessment indicators and thresholds are established for each security risk (e.g., insufficient encryption strength: Level 1 security level using AES-128 or lower encryption is considered high risk; unauthorized access: access node failing authentication is considered high risk). A mapping relationship between QoS configuration items and security risks is established, forming a standardized security risk assessment rule base (supporting subsequent updates and iterations).

[0063] The beneficial effects of this invention are as follows: Achieve automated and intelligent QoS configuration: Without manual intervention, the QoS configuration scheme is automatically generated and optimized through a wide learning model. Compared with the traditional manual configuration method, the configuration efficiency is improved by more than 60%, and the labor cost is significantly reduced. At the same time, the model can adapt to the dynamic changes of the vehicle communication scenario in real time to meet the differentiated needs of different data streams.

[0064] A comprehensive security protection system is built to enhance the security of vehicle communication: by introducing security incident data in the data acquisition stage, adding security constraints in the model training stage, designing hierarchical security enhancement configurations in the configuration generation stage, adding security verification in the verification stage, and using encrypted storage and identity binding for configuration files, a closed-loop security protection system is formed throughout the entire process; it can effectively resist malicious attacks such as data tampering, unauthorized access, and transmission leakage, reducing the incidence of security incidents by more than 90%, and especially ensuring the transmission security of core data such as autonomous driving control commands.

[0065] Adapted to dynamic in-vehicle scenarios and meeting real-time requirements: Combining the advantages of wide learning models for fast inference, it can generate an adapted QoS configuration scheme (including security enhancement configuration) within 50ms. Compared with existing deep learning solutions, the inference speed is improved by 80%, meeting the needs of scenarios such as autonomous driving that have stringent requirements for both real-time performance and safety.

[0066] Enhancing communication stability and reliability: Compatibility verification is completed by building a library of incompatible configuration item combinations, effectively avoiding communication failures caused by configuration conflicts; at the same time, configuration accuracy is continuously improved through configuration effect feedback and model optimization mechanisms, reducing the transmission latency of the vehicle DDS system by 35% and the packet loss rate by 40%; the introduction of security enhancement configuration items further ensures the stability of the communication process and avoids communication interruptions caused by security attacks.

[0067] The model is lightweight and easy to deploy, adapting to automotive hardware resources: It adopts a wide learning framework to build the model, which reduces the model parameter size by 70% and the training cycle by 60% compared to deep learning models, making it easy to deploy in automotive embedded systems; at the same time, the safety enhancement mechanism adopts a hierarchical design, which can dynamically adjust the safety configuration parameters according to the performance of automotive hardware, balancing the safety protection effect and hardware resource consumption.

[0068] Security configuration is hierarchically adapted to enhance solution flexibility: Different security enhancement configuration items are designed for different security level requirements, which can flexibly adapt to the transmission scenarios of different types of data streams such as core control data, perception data, and entertainment data. While ensuring the security of core data, it avoids the waste of resources caused by excessive security configuration.

[0069] Example 2 The second embodiment of this invention presents a width-learning-based adaptive QoS configuration method for in-vehicle DDS. This method uses the transmission of core control commands (such as braking and steering commands) of autonomous vehicles as a typical application scenario (this scenario represents the highest level of security, Level 1), providing a detailed and complete implementation. This scenario is a core high-security, high-real-time scenario for in-vehicle DDS systems, with stringent requirements for real-time communication (transmission delay ≤50ms), reliability (packet loss rate ≤2%, reliability score ≥8 points), and security (no data tampering or unauthorized access events). This fully verifies the effectiveness and practicality of the method and security enhancement mechanism of this invention. This embodiment will fully present the entire process of "data acquisition and preprocessing with security dimensions → construction of a QoS configuration model integrating security features → model training → initial configuration generation → compatibility verification and security verification → secure encryption and loading of configuration files → configuration effect feedback and security optimization," as follows: Figure 2 As shown.

[0070] Refine the parameter settings, operational details, and judgment criteria for each step to ensure that technicians in the relevant technical field can fully reproduce the results.

[0071] I. Detailed Description of the Implementation Environment The hardware and software environment of this embodiment is compatible with mainstream in-vehicle intelligent system configurations and meets the protection requirements of Level 1 security. The specific parameters are as follows: 1. Hardware environment: The vehicle communication bus adopts Ethernet (supporting 1000Mbps transmission rate), and the network adapter is a gigabit Ethernet controller (with hardware encryption function); the security authentication module supports two-way identity authentication.

[0072] 2. Software Environment: The operating system is Linux Ubuntu 20.04 LTS (kernel version 5.15.0-78-generic, with security hardening patch). The DDS middleware is Fast DDS 2.9.1 (compliant with DDS 1.4 specification, security plugin enabled). The wide learning network model is implemented based on Python 3.9.16 and TensorFlow 2.10.0 framework. Data processing tools include Pandas 1.5.3 and NumPy 1.24.3. Model training acceleration relies on CUDA 11.7 and cuDNN 8.5.0. The encryption algorithm library uses OpenSSL 3.0.2, supporting AES-256 and SHA-256 algorithms.

[0073] 3. Preset Basic Data: The preset QoS can be automatically configured with 20 items (denoted as T1-T20), including 16 core configuration items (T1-T16, including reliability level, deadline, etc.) and 4 security enhancement configuration items (T17-encryption algorithm type, T18-authentication policy, T19-access control rules, T20-transmission verification mechanism); the in-vehicle DDS QoS incompatibility configuration item combination library pre-stores 20 sets of conflict combinations verified by actual testing; the security risk assessment rule library pre-stores 15 assessment rules, among which the core assessment rules corresponding to security level 1 include: encryption algorithm must be AES-256 or above, two-way authentication must be enabled, access control rules must be in strict mode, and transmission verification must be per packet verification.

[0074] II. Complete Implementation Steps Step 1: Integrate security-dimensional data collection and preprocessing, such as Figure 3 The flowchart for this step is shown below: 1.1 Data Collection Scope and Content: Historical DDS communication data of the target autonomous vehicle over the past 12 months was collected under three typical road conditions: urban roads, highways, and suburban roads. The data primarily covers core control command transmission scenarios, supplemented with some perception data and entertainment data transmission scenario data for comparative training. A total of 2000 samples were constructed for the original dataset. Each sample must contain four core data modules, with specific fields and example values ​​as follows: (1) Communication scenario parameter module: data type (core control command / sensing data / entertainment data), data transmission rate (unit: Mbps), data priority (level 1-5, level 1 is the highest), network bandwidth utilization (unit: %), road condition type (encoding: urban road=1, highway=2, suburban road=3); example value: data type=core control command, transmission rate=5Mbps, priority=level 1, bandwidth utilization=45%, road condition type=2.

[0075] (2) Historical QoS configuration data module: contains specific parameters of 20 configuration items T1-T20; example values: T1=reliable transmission, T2=real-time transmission, T3=40ms, T4=500ms, T17=AES-256, T18=two-way authentication, T19=strict access control, T20=packet-by-packet verification.

[0076] (3) Communication performance index module: transmission delay (unit: ms), packet loss rate (unit: %), reliability score (full score 10 points, calculation formula: reliability score = 10 - (transmission delay / 10) - (packet loss rate × 2), the result is rounded to 1 decimal place); example value: transmission delay = 32ms, packet loss rate = 0.8%, reliability score = 9.2 points.

[0077] (4) Security event data module: security event type (none / data tampering / illegal access / transmission leakage), security risk level (level 1-5), event handling result (resolved / unresolved); Example values: security event type = none, security risk level = level 1, event handling result = none.

[0078] 1.2 Data Cleaning Operation: The strategy adopted is "layered filtering + precise repair + security verification". The specific process is as follows: (1) Invalid sample deletion: Through field integrity verification, delete samples with missing values ​​exceeding 10% (a total of 22 samples were deleted, mainly incomplete data from the early testing phase); missing value percentage determination formula: missing value percentage = (number of missing fields / total number of fields) × 100%.

[0079] (2) Missing value filling: For samples with a missing value ratio of ≤10% (45 samples in total), the mean filling method is used to fill in numerical fields (such as transmission rate, bandwidth utilization, transmission delay, etc.), and the mode filling method is used to fill in categorical fields (such as data type, road condition type, safety event type, etc.). The mean calculation range is limited to a subset of samples with the same road condition, data type, and safety level to ensure the rationality of the filled values.

[0080] (3) Outlier removal: The 3σ principle is used to judge outliers. The specific steps are: calculate the mean μ and standard deviation σ of each numerical field; if the value x of a certain field in the sample satisfies |x-μ|>3σ, it is judged as an outlier; remove samples containing outliers (a total of 33, mainly abnormal data caused by network fluctuations, hardware failures or security attacks).

[0081] (4) Security data verification: The authenticity of security event data is verified by comparing the details of the security event with the encrypted historical logs, and false security event samples (5 in total) are removed to ensure the validity of security dimension data.

[0082] 1.3 Data Standardization: The cleaned dataset was standardized using the Min-Max normalization method, mapping all numerical data to the [0,1] interval. The normalization formula is as follows:

[0083] in, Here, x represents the normalized data, and x represents the original data. The minimum value of this field. This represents the maximum value of the field. The security risk levels (levels 1-5) are quantized and then synchronously normalized. The final training dataset, with 1900 samples, is obtained after standardization.

[0084] Step 2: Constructing a QoS configuration model that integrates security features, such as... Figure 4 The diagram shown is a structural schematic of a QoS configuration model that integrates security features, as detailed below: 2.1 Core Model Structure Design: A QoS configuration model integrating security features is constructed based on a wide learning framework. The model adopts a four-layer structure of "input layer - feature layer - reinforcement layer - output layer". The functions and parameter settings of each layer are as follows: (1) Input layer: Receives the training dataset features after preprocessing in step 1. A total of 10 core input features are selected, namely: data type (encoding: core control command = 1, perception data = 2, entertainment data = 3), data transmission rate (after normalization), data priority (encoding: levels 1-5 are directly mapped to 1-5), network bandwidth utilization rate (after normalization), road condition type (after encoding), historical transmission delay (after normalization), historical packet loss rate (after normalization), historical reliability score (after normalization), historical security risk level (after normalization), and real-time security level requirements (encoding: levels 1-3 are directly mapped to 1-3); the number of nodes in the input layer is 10.

[0085] (2) Feature Layer: ReLU is used as the non-linear activation function, and the number of mapping layer nodes n=180 (based on the sample size of 1900 in this scenario, the ratio of sample size to number of nodes is approximately 10:1 to ensure a balance between model fitting ability and generalization ability). The output of the feature layer is a 180-dimensional high-dimensional feature vector, calculated as follows:

[0086] in, The output of the feature layer, This is the weight matrix from the input layer to the feature layer (dimension 10×180). This is the feature layer bias term (dimension 1×180). This is the output of the input layer.

[0087] (3) Enhancement Layer: Sigmoid is used as the non-linear activation function, and the number of nodes in the enhancement layer is m=360 (maintaining a 2:1 ratio with the number of nodes in the feature layer to ensure the feature enhancement effect, focusing on enhancing the adaptability of security features); the output of the enhancement layer is a 360-dimensional enhanced feature vector, calculated as follows:

[0088] in, The output of the reinforcement layer, is the weight matrix from the input layer to the reinforcement layer (dimension 10×360), and b is the bias term of the reinforcement layer (dimension 1×360).

[0089] (4) Output layer: The outputs of the feature layer and the enhancement layer are concatenated to form an enhanced feature matrix X (dimension 1900×540). The model output is obtained by pseudo-inverse solving. The output is the optimal parameter combination of 20 QoS configuration items (T1-T20); Output layer The calculation formula is:

[0090] in, The input feature matrix; To and The corresponding tag matrix (dimension 1900×20, composed of historical QoS configuration data). This is the regularization parameter (set to 0.001 to prevent model overfitting). To and An identity matrix of the same dimension (dimension 540×540). This is the model weight matrix; This is the enhanced feature matrix output by the model.

[0091] 2.2 Model Initialization Configuration: The Sparse Autoencoder (SAE) algorithm is used to initialize the model's weight matrix W and bias term b. The number of hidden layer nodes of the Sparse Autoencoder network is set to 200, the number of iterations is 300, the learning rate is 0.02, and the sparsity parameter is set to 0.05 to ensure that the initialized parameters have a good training foundation.

[0092] Step 3: Model training and validation, such as Figure 5 The diagram shown below illustrates the flowchart for QoS configuration model training and validation. 3.1 Dataset Partitioning: The training dataset obtained in Step 1 is randomly divided into a training set and a validation set in a 7:3 ratio. The training set contains 1330 samples (1900 × 70%), used for model parameter learning; the validation set contains 570 samples (1900 × 30%), used for real-time verification of model accuracy and security risk adaptability to avoid overfitting. It should be noted that the training sample data table for the QoS configuration model is shown below:

[0093] 3.2 Training Parameter Settings: The total number of training iterations was set to 800, with an initial learning rate of 0.03. A dynamic learning rate adjustment strategy was adopted: During the early training phase (iterations 1-300), the learning rate was maintained at 0.03. If the rate of decrease in the training set loss value was less than 0.001 per iteration, the learning rate was increased to 0.04. During the middle training phase (iterations 301-600), the learning rate was adjusted to 0.02. During the later training phase (iterations 601-800), the learning rate was adjusted to 0.01 to ensure stable convergence of the model in the later training phase. The total training loss function was adopted as follows:

[0094] in, The output of the safety constraint loss function, For mean square error loss, For safety risk losses, The safety weighting coefficient is α = 0.4.

[0095] 3.3 Validation and Early Stopping Mechanism: After every 100 iterations, the model's prediction accuracy and security risk fit are calculated using the validation set. The accuracy criterion is: if ≥90% of the 20 predicted QoS configuration parameters have an error ≤5% compared to the optimal configuration parameter (the configuration with the highest reliability score and lowest security risk in the corresponding scenario in historical data), then the prediction is considered accurate. The security risk fit criterion is: the model's predicted security enhancement configuration items meet the corresponding security level requirements, and the error between the predicted security risk level and the actual security risk level is ≤1 level. When the validation set accuracy shows no improvement (accuracy fluctuation ≤0.5%) and the security risk fit shows no optimization after 20 consecutive iterations, the early stopping mechanism is triggered, training is terminated, and the current optimal model parameters are saved. In this embodiment, when the model is trained to 580 iterations, the accuracy of the validation set reaches 97.1% and the safety risk fit reaches 96.8%. Furthermore, the accuracy and safety risk fit remain stable for the subsequent 20 iterations, triggering the early stop mechanism. Training is stopped and the model file (named "QoS_security_config_model.h5") is encrypted and saved to the vehicle system's security model library (path: / home / vehicle / secure_dds_model / ).

[0096] Step 4: Generate the initial QoS configuration scheme, as follows: 4.1 Real-time Communication Requirement Acquisition and Analysis: The vehicle-mounted DDS system acquires real-time core control command transmission requirements through an encrypted requirement acquisition module. Specific requirement information includes: data type = core control command, transmission priority = level 1, real-time requirement = transmission delay ≤ 50ms, reliability requirement = packet loss rate ≤ 2%, security level requirement = level 1 (highest security level), and application scenario = highway (road condition type = 2). The requirement analysis module converts these requirements into a model-recognizable input feature vector. The converted input vector is: [1, 0.5, 1, 0.45, 2, 0.32, 0.08, 0.92, 0.2, 1] (corresponding to the normalized values ​​of the 10 core input features).

[0097] 4.2 Model Inference and Initial Configuration Generation: The transformed input feature vector is input into the QoS configuration model trained in step 3. The model outputs the initial QoS configuration scheme through fast inference (inference time is 35ms). The specific configuration parameters are as follows: T1=Reliable transmission, T2=Real-time transmission, T3=40ms, T4=500ms, T5=Keep_Last(20), T6=10MB, T17=AES-256, T18=Two-way authentication, T19=Strict access control (only three legal nodes, namely vehicle controller, brake actuator and steering actuator, are allowed to access), T20=Packet verification (SHA-256), and the remaining configuration items are set according to the optimal default value of security level 1.

[0098] Step 5: Compatibility verification, security verification, and reconfiguration, such as... Figure 6 The diagram shown illustrates the logic of compatibility and security verification for the initial QoS configuration scheme. 5.1 Dual Verification Process: The preset QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle-mounted DDS are called, and dual verification is performed by "item-by-item comparison + simulation verification": the configuration item combination in the initial configuration scheme is extracted item by item and accurately matched with the conflicting combination in the incompatible combination library to complete the compatibility verification; according to the assessment rules of level 1 security in the security risk assessment rule library, the security enhancement configuration items (T17-T20) in the initial scheme are verified item by item, and at the same time, a simulated transmission test is performed (test duration 30s, simulating data tampering and illegal access attacks) to verify the security protection effect of the scheme.

[0099] 5.2 Verification Results and Reconfiguration: Compatibility comparison showed that none of the configuration item combinations in the initial QoS configuration scheme matched any conflicting combinations in the incompatible combination library, thus the compatibility verification passed. Security verification showed that all security enhancement configuration items in the initial scheme met Level 1 security requirements (T17 for AES-256, T18 for two-way authentication, etc.). Simulated transmission tests successfully resisted data tampering and unauthorized access attacks, with no security events occurring, thus the security verification passed. Therefore, the initial configuration scheme is deemed compatible and secure, and reconfiguration is unnecessary.

[0100] Step 6: Secure Encryption and Loading of Configuration Files 6.1 Configuration File Encryption: The initial QoS configuration scheme is securely encrypted using the AES-256 algorithm to encrypt core parameters in the configuration file (such as T3=40ms, T17=AES-256, T19=strict access control rules, etc.), generating the encrypted final QoS configuration file (in XML format, named final_secure_QoS_config.xml). Simultaneously, this configuration file is bound to the identity information of the vehicle controller (MAC address: 00:1B:44:11:3A:B7), brake actuator, and steering actuator, and an authentication mechanism is set up.

[0101] 6.2 Configuration Loading: The encrypted final QoS configuration file is output to the encrypted configuration directory of the vehicle DDS system (path: / home / vehicle / secure_dds_config / ). The configuration loading module of the vehicle DDS system first authenticates the loading node. After successful authentication (in this embodiment, the vehicle controller loads the configuration file), it automatically reads and decrypts the configuration file, completing the QoS parameter configuration loading for the publisher (core control command output module) and subscribers (brake actuators, steering actuators). After loading is complete, a "Configuration effective + security verification passed" notification is triggered to the vehicle security monitoring center.

[0102] Step 7: Configuration feedback and security optimization 7.1 Performance Data Collection: After the configuration takes effect, the vehicle safety monitoring module starts real-time monitoring and continuously collects communication performance indicators and safety status data of the core control command transmission. The collection period is 1 second, and the collection is continuous for 60 minutes, collecting a total of 3600 sets of data. The average value is calculated as the final performance indicator.

[0103] 7.2 Performance Verification: The final performance indicators are: transmission latency = 31.5ms (≤50ms, meets requirements), packet loss rate = 0.7% (≤2%, meets requirements), and reliability score = 9.3 points (≥8 points, meets requirements). Security status data shows that no security events (data tampering, unauthorized access, etc.) occurred during the 60-minute monitoring period, and the security protection effect met the standards. Compared with the traditional manual configuration method, the configuration efficiency is improved by 75% (the average time for manual configuration is 25 minutes, while the configuration time of this invention is 35ms), the transmission latency is reduced by 38% (the average transmission latency for manual configuration is 50.8ms), and the packet loss rate is reduced by 44% (the average packet loss rate for manual configuration is 1.25%). Furthermore, traditional manual configuration lacks security enhancement mechanisms and cannot resist security attacks; the security advantages of this invention are significant.

[0104] 7.3 Model Optimization: Since the configuration effect met the requirements and the security protection was effective, the communication scenario parameters, the final QoS configuration scheme, communication performance indicators and security status data were encrypted and added to the original dataset as new samples (sample number: 1901) to update the training dataset. Subsequently, when the cumulative number of new samples reaches 500, the QoS configuration model will be incrementally trained to further optimize the model parameters and security constraints, and improve the model's adaptability to complex security scenarios.

[0105] III. Key Explanations of the Implementation Examples The core value of this embodiment is that it fully reproduces the entire technical process of the present invention (including the safety enhancement mechanism), refines the parameter settings, operating procedures and judgment criteria of each link, and focuses on the highest level 1 safety scenario to verify the effectiveness of the safety enhancement mechanism. Those skilled in the art can implement the technical solution of the present invention in the same or similar vehicle DDS system based on the detailed description of this embodiment, which is especially suitable for core data transmission scenarios of autonomous driving with strict safety requirements.

[0106] Key technological highlights: As can be clearly seen from this embodiment, the wide learning model adopted by this invention has the advantage of fast inference, which can meet the real-time requirements of core control command transmission for autonomous driving; the model design that integrates safety features and the full-process security protection mechanism can effectively resist security attacks and ensure the security of core data transmission; the effect feedback and incremental training mechanism realize the continuous optimization of the model, fully demonstrating the core advantages of "automation, intelligence and safety" of this invention.

[0107] Boundary condition description: Although this embodiment focuses on the core control command transmission scenario of Level 1 safety level, the technical solution of the present invention can be adapted to other vehicle DDS communication scenarios such as perception data and entertainment data of Level 2 and Level 3 safety levels by adjusting the safety level requirements and corresponding safety enhancement configuration parameters, and has good versatility and flexibility.

[0108] Example 3 Please see Figure 7 This is a structural block diagram of the vehicle-mounted DDS adaptive QoS configuration system based on width learning in the third embodiment of the present invention, used to implement the vehicle-mounted DDS adaptive QoS configuration method based on width learning described in the first embodiment. The system includes: Module 10 is used to collect historical QoS configuration data, corresponding communication performance parameters, communication performance indicators and security event data of DDS communication in vehicle scenarios, construct the original dataset, and clean and standardize the original dataset to obtain a standardized training dataset. Training module 20 is used to construct an initial QoS configuration model that integrates security features based on the standardized training dataset through a wide learning framework, and to train the initial QoS configuration model through a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thereby obtaining the target QoS configuration model. The verification and validation module 30 is used to generate an initial QoS configuration scheme based on real-time vehicle communication requirements and security level requirements, and through the target QoS configuration model. Based on the QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle DDS, the initial QoS configuration scheme is subjected to compatibility verification and security validation respectively to obtain a safe and usable target QoS configuration file.

[0109] In practical implementation, historical QoS configuration data, corresponding communication performance parameters, communication performance indicators, and security event data of DDS communication in vehicle scenarios are collected to construct an original dataset. The original dataset is cleaned and standardized to obtain a standardized training dataset. Based on the standardized training dataset, an initial QoS configuration model incorporating security features is constructed using a wide learning framework. The initial QoS configuration model is trained using a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thus obtaining a target QoS configuration model. Then, based on real-time vehicle communication requirements and security level requirements, an initial QoS configuration scheme is generated through the target QoS configuration model. The initial QoS configuration scheme is then subjected to compatibility verification and security verification based on the QoS incompatible configuration item combination library and security risk assessment rule library of vehicle DDS, respectively, to obtain a safe and usable target QoS configuration file. Unlike existing technologies, this approach achieves automation, intelligence, and security in QoS configuration, improving configuration efficiency, communication adaptability, and security protection capabilities, while also ensuring the security of core vehicle data transmission.

[0110] Furthermore, the initial QoS configuration model includes an input layer, a feature layer, a reinforcement layer, and an output layer. The feature layer uses ReLU as a non-linear activation function, and the number of nodes n in the feature layer ranges from 100 to 200. The reinforcement layer uses Sigmoid as a non-linear activation function, and the number of nodes m in the reinforcement layer ranges from 200 to 300. The training module 20 includes: The standardized training dataset is divided into a training set and a validation set according to a preset ratio. The sample features of the training set are received through the input layer. The sample features are mapped into high-dimensional feature vectors through the feature layer. The high-dimensional feature vectors are then subjected to reinforcement learning through the reinforcement layer to obtain reinforcement learning feature vectors. The high-dimensional feature vector and the reinforcement learning feature vector are concatenated into an enhanced feature matrix through the output layer, and the output of the initial QoS configuration model is obtained by solving the pseudo-inverse.

[0111] Furthermore, the expression for the output of the initial QoS configuration model is as follows:

[0112]

[0113] in, For the input feature matrix, To and The corresponding label matrix, For regularization parameters, To and Identity matrices of the same dimension This is the model weight matrix. This is the enhanced feature matrix output by the model.

[0114] Furthermore, the expression for the safety constraint loss function is as follows:

[0115] in, The output of the safety constraint loss function, For mean square error loss, For safety risk losses, For safety weighting coefficients, The value range is 0.3-0.5.

[0116] Furthermore, after the verification and validation module 30, the system further includes: The acquisition module is used to acquire real-time communication performance indicators and real-time security status data of the vehicle-mounted DDS system after adopting the target QoS configuration file, and compare the real-time communication performance indicators and real-time security status data with preset performance thresholds and security thresholds respectively. The module is used to feed back the configuration scheme, corresponding communication data, and security event details to the original dataset when performance is not up to standard or a security event occurs, so as to obtain the latest standardized training dataset. The incremental training module is used to incrementally train the target QoS configuration model based on the latest standardized training dataset in order to optimize the model parameters and security constraints.

[0117] Furthermore, the building module 10 includes: The cleaning unit is used to perform invalid sample deletion, missing value imputation, and outlier removal on the original dataset to obtain an intermediate original dataset. The standardized unit is used to map all sample data in the intermediate original dataset to the [0, 1] interval using the Min-Max normalization method, and to quantify and encode the risk level corresponding to the security event data in the intermediate original dataset.

[0118] Furthermore, the system also includes: The encryption module is used to encrypt and store the configuration parameters of the target QoS configuration file using a symmetric encryption algorithm, and bind the identity information of the legitimate nodes of the vehicle-mounted DDS system.

[0119] Furthermore, the system also includes: The first forming module is used to collect multiple QoS configuration items of the vehicle-mounted DDS system, verify the compatibility of different combinations of the QoS configuration items through multiple communication tests, record conflicting configuration item pairs and their corresponding conflict behaviors, establish a mapping relationship between the conflicting configuration item pairs and the conflict judgment criteria, and form a library of QoS incompatible configuration item combinations for the vehicle-mounted DDS. The second forming module is used to identify the security risk points of vehicle-mounted DDS communication, formulate the evaluation indicators and thresholds for the security risk points, establish the mapping relationship between the QoS configuration items and security risks, and form a security risk assessment rule base.

[0120] Example 4 In the fourth embodiment of the present invention, based on the same inventive concept, the present invention proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the width-learning-based vehicle DDS adaptive QoS configuration method of the above embodiments. The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means containing storage, communication, propagation, or transmission programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0121] The memory may include a large-capacity storage device for data or instructions. For example, and not limitingly, the memory may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory may include removable or non-removable (or fixed) media. Where appropriate, the memory may be internal or external to the data processing device. In a particular embodiment, the memory is non-volatile memory. In a particular embodiment, the memory includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0122] Example 5 In the fifth embodiment of the present invention, based on the same inventive concept, the present invention proposes a terminal, the terminal comprising: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute the width learning-based vehicle DDS adaptive QoS configuration method of the above embodiment.

[0123] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0124] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0125] Without causing conflict, those skilled in the art can freely combine and use the above-mentioned additional technical features.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for adaptive QoS configuration of in-vehicle DDS based on width learning, characterized in that, The method includes: Historical QoS configuration data, corresponding communication performance parameters, communication performance indicators, and security event data of DDS communication in vehicle scenarios are collected to construct an original dataset. The original dataset is then cleaned and standardized to obtain a standardized training dataset. Based on the standardized training dataset, an initial QoS configuration model incorporating security features is constructed using a wide learning framework. The initial QoS configuration model is then trained using a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thus obtaining the target QoS configuration model. Based on real-time vehicle communication requirements and security level requirements, an initial QoS configuration scheme is generated through the target QoS configuration model. The initial QoS configuration scheme is then subjected to compatibility verification and security verification based on the QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle DDS, resulting in a secure and usable target QoS configuration file.

2. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 1, characterized in that, The initial QoS configuration model includes an input layer, a feature layer, a reinforcement layer, and an output layer. The feature layer uses ReLU as a non-linear activation function, and the number of nodes n in the feature layer ranges from 100 to 200. The reinforcement layer uses Sigmoid as a non-linear activation function, and the number of nodes m in the reinforcement layer ranges from 200 to 300. The steps of constructing an initial QoS configuration model incorporating security features based on the standardized training dataset using a wide learning framework include: The standardized training dataset is divided into a training set and a validation set according to a preset ratio. The sample features of the training set are received through the input layer. The sample features are mapped into high-dimensional feature vectors through the feature layer. The high-dimensional feature vectors are then subjected to reinforcement learning through the reinforcement layer to obtain reinforcement learning feature vectors. The high-dimensional feature vector and the reinforcement learning feature vector are concatenated into an enhanced feature matrix through the output layer, and the output of the initial QoS configuration model is obtained by solving the pseudo-inverse.

3. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 2, characterized in that, The expression for the output of the initial QoS configuration model is as follows: in, For the input feature matrix, To and The corresponding label matrix, For regularization parameters, To and Identity matrices of the same dimension This is the model weight matrix. This is the enhanced feature matrix output by the model.

4. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 1, characterized in that, The expression for the safety constraint loss function is as follows: in, The output of the safety constraint loss function, For mean square error loss, For safety risk losses, For safety weighting coefficients, The value range is 0.3-0.

5.

5. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 1, characterized in that, After obtaining a secure and usable target QoS profile, the method further includes: The system collects real-time communication performance indicators and real-time security status data of the vehicle-mounted DDS system after adopting the target QoS configuration file, and compares the real-time communication performance indicators and real-time security status data with preset performance thresholds and security thresholds, respectively. When performance fails to meet standards or a security incident occurs, the configuration scheme, corresponding communication data, and security incident details are fed back to the original dataset to obtain the latest standardized training dataset. The target QoS configuration model is incrementally trained based on the latest standardized training dataset to optimize model parameters and security constraints.

6. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 1, characterized in that, The steps for cleaning and standardizing the original dataset include: The original dataset is processed by deleting invalid samples, imputing missing values, and removing outliers to obtain an intermediate original dataset. The Min-Max normalization method is used to map all sample data in the intermediate original dataset to the interval [0, 1], thereby quantifying and encoding the risk level corresponding to the security event data in the intermediate original dataset.

7. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 1, characterized in that, The method further includes: The configuration parameters of the target QoS configuration file are encrypted and stored using a symmetric encryption algorithm, and then bound to the identity information of the legitimate nodes of the vehicle-mounted DDS system.

8. The vehicular DDS adaptive QoS configuration method based on width learning according to claim 1, characterized in that, The method further includes: Collect multiple QoS configuration items of the vehicle-mounted DDS system, verify the compatibility of different combinations of the QoS configuration items through multiple communication tests, record conflicting configuration item pairs and their corresponding conflict behaviors, establish the mapping relationship between the conflicting configuration item pairs and the conflict judgment criteria, and form a library of QoS incompatible configuration item combinations for vehicle-mounted DDS. The security risks of vehicle-mounted DDS communication are identified, assessment indicators and thresholds for these risks are formulated, a mapping relationship between QoS configuration items and security risks is established, and a security risk assessment rule base is formed.

9. A vehicle-mounted DDS adaptive QoS configuration system based on width learning, used to implement the vehicle-mounted DDS adaptive QoS configuration method based on width learning as described in any one of claims 1-8, characterized in that, The system includes: The module is used to collect historical QoS configuration data, corresponding communication performance parameters, communication performance indicators and security event data of DDS communication in vehicle scenarios, construct the original dataset, and clean and standardize the original dataset to obtain a standardized training dataset. The training module is used to construct an initial QoS configuration model that integrates security features based on the standardized training dataset through a wide learning framework, and to train the initial QoS configuration model through a sparse autoencoder algorithm combined with a security constraint loss function until the model prediction error is less than a preset threshold and meets the security risk assessment requirements, thereby obtaining the target QoS configuration model. The verification and validation module is used to generate an initial QoS configuration scheme based on real-time vehicle communication requirements and security level requirements, and through the target QoS configuration model. Based on the QoS incompatible configuration item combination library and security risk assessment rule library of the vehicle DDS, the initial QoS configuration scheme is subjected to compatibility verification and security validation respectively to obtain a safe and usable target QoS configuration file.

Images