Decision-making method for intelligent predictive driving of networked automobile based on machine learning

Abstract

The invention relates to the technical field of predictive driving, in particular to a decision-making method for intelligent predictive driving of a networked automobile based on machine learning. Comprising the following steps: acquiring and preprocessing multi-source data of a networked automobile, real-time data of the Internet of Vehicles and historical driving data, and performing feature extraction on the preprocessed multi-source data of the networked automobile and the preprocessed real-time data of the Internet of Vehicles to obtain preliminary feature data; introducing a multi-layer recursive feature fusion algorithm to perform fusion processing on the preliminary feature data to obtain final fused data; and constructing and training a predictive driving model by using a networked automobile intelligent predictive driving prediction algorithm based on deep cross adaptability, and performing prediction processing on the finally fused data by using the predictive driving model to obtain a driving prediction result and generate a driving decision. The technical problems that processing of multi-source heterogeneous data is not accurate enough in intelligent predictive driving of the networked automobile, the predictive adaptability of driving decisions is poor, and the accuracy rate is low are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of predictive driving, and particularly to a decision-making method for intelligent predictive driving of connected vehicles based on machine learning. Background Art

[0002] With the rapid development of intelligent connected vehicle technology, the degree of vehicle intelligence has been continuously improved, and driving assistance and autonomous driving systems based on in-vehicle sensors and vehicle-to-everything (V2X) have been gradually applied to actual traffic scenarios. Intelligent connected vehicles need to make rapid and accurate driving decisions in complex dynamic traffic environments to ensure driving safety, improve driving comfort, and optimize energy efficiency. However, there are still many technical challenges in current driving decision-making systems in aspects such as data acquisition, information fusion, driving prediction, and real-time decision-making. Connected vehicles rely on various sensors such as lidar, millimeter-wave radar, cameras, and ultrasonic sensors to perceive the surrounding environment in real time, and at the same time, through V2X technology, they interact with other vehicles, infrastructure, and cloud platforms to obtain key data such as traffic signals, road conditions, and weather conditions. These multi-source heterogeneous data have large differences in numerical range, timestamp accuracy, data structure, etc., resulting in problems such as information loss, inconsistency, and redundancy during the data fusion process, which affect the subsequent driving prediction and decision-making accuracy.

[0003] However, the existing decision-making methods for intelligent prediction of connected vehicles have the following technical problems: inaccurate processing of multi-source heterogeneous data in intelligent predictive driving of connected vehicles, poor prediction adaptability and low accuracy of driving decisions. Summary of the Invention

[0004] The present invention provides a decision-making method for intelligent predictive driving of connected vehicles based on machine learning to solve the technical problems of inaccurate processing of multi-source heterogeneous data in intelligent predictive driving of connected vehicles, poor prediction adaptability and low accuracy of driving decisions.

[0005] The decision-making method for intelligent predictive driving of connected vehicles based on machine learning of the present invention specifically includes the following technical solutions:

[0006] The decision-making method for intelligent predictive driving of connected vehicles based on machine learning includes the following steps:

[0007] S1. Obtain and preprocess the multi-source data, V2X real-time data, and historical driving data of the connected vehicle to obtain the preprocessed multi-source data, V2X real-time data, and historical driving data of the connected vehicle; extract features from the preprocessed multi-source data and V2X real-time data of the connected vehicle to obtain preliminary feature data; introduce a multi-layer recursive feature fusion algorithm to perform fusion processing on the preliminary feature data to obtain the finally fused data;

[0008] S2. Using the intelligent predictive driving prediction algorithm for connected vehicles based on deep cross-adaptability, construct and train a predictive driving model based on the preprocessed historical driving data, and use the predictive driving model to perform predictive processing on the finally fused data to obtain driving prediction results, generate driving decisions, and drive according to the driving decisions.

[0009] Preferably, the S1 specifically includes:

[0010] The multi-layer recursive feature fusion algorithm performs multi-level and recursive feature fusion on the preliminary feature data, adopts a multi-level nested structure, and combines non-linear mapping and recursive optimization strategies. By introducing an adaptive weight adjustment mechanism, the preliminary feature data is fused.

[0011] Preferably, the S1 specifically includes:

[0012] During the implementation process of the multi-layer recursive feature fusion algorithm, the preliminary feature data is standardized to obtain the standardized feature data, and the standardized feature data is used as the input to enter the recursive feature transformation stage, and then the finally recursively transformed feature data is obtained. The specific implementation formula is:

[0013]

[0014] Among them, is the feature data after the (i + 1)-th layer of recursive feature transformation; is the feature data after the i-th layer of recursive feature transformation, W i is the weight of the recursive feature transformation at the i-th layer of recursion; b i is the bias term at the i-th layer of recursion; α is the adaptive balance coefficient; γ is the frequency adjustment factor of the sine non-linear transformation; X norm is the standardized feature data.

[0015] Preferably, the S1 specifically includes:

[0016] During the implementation process of the multi-layer recursive feature fusion algorithm, after completing the recursive feature transformation, the finally recursively transformed feature data enters the orthogonal feature fusion stage to process the finally recursively transformed feature data to obtain an orthogonal projection matrix; the orthogonal projection matrix is used to perform a redundancy removal operation on the finally recursively transformed feature data to obtain the preliminary non-redundant feature data.

[0017] Preferably, the S1 specifically includes:

[0018] During the implementation process of the multi-layer recursive feature fusion algorithm, after the orthogonal feature fusion, the preliminary non-redundant feature data enters the adaptive weight fusion stage for processing. The specific implementation formula is:

[0019]

[0020] Among them, is the weight of the -th initially non-redundant feature data in the final fusion result; β is the weight smoothing factor; represents the second norm of the -th initially non-redundant feature data; represents the second norm of the j-th initially non-redundant feature data; is the total number of initially non-redundant feature data.

[0021] Preferably, the S1 specifically includes:

[0022] In the implementation process of the multi-layer recursive feature fusion algorithm, after weight calculation, the initially non-redundant feature data are weighted and summed to obtain the fused feature data, and the fused feature data are subjected to feature data dimensionality reduction processing to obtain the finally fused data.

[0023] Preferably, the S2 specifically includes:

[0024] In the implementation process of the intelligent predictive driving prediction algorithm for connected vehicles based on deep cross-adaptability, based on the acquired historical driving data, the corresponding environmental perception data and vehicle networking data are acquired, and after preprocessing the environmental perception data and vehicle networking data, they are merged with the preprocessed historical driving data to obtain the comprehensively preprocessed data.

[0025] Preferably, the S2 specifically includes:

[0026] After completing the data preprocessing, a predictive driving model is constructed by using multi-layer recursive cross-sensing modeling, and this modeling process is divided into a local perception layer, a global trend layer, and a cross-feedback layer.

[0027] Preferably, the S2 specifically includes:

[0028] In the process of training the predictive driving model, a hierarchical game strategy is adopted to infer and optimize the driving decision-making strategy of the predictive driving model; after completing the training of the predictive driving model, the finally fused data is processed to generate a driving prediction result.

[0029] The beneficial effects of the technical solution of the present invention are:

[0030] 1. By adopting the multi-layer recursive feature fusion algorithm, deep optimization and adaptive adjustment are carried out at the feature level, which can effectively solve the problems of inconsistency in scale, modality, time synchronization, etc. of multi-source data, improve the accuracy and robustness of data fusion, and provide high-quality input for the predictive driving model.

[0031] 2. Through a hierarchical modeling strategy, combined with a local perception layer, a global trend layer, and a cross-feedback layer, short-term and long-term driving behavior characteristics can be comprehensively captured. The local perception layer quickly extracts the immediate driving behaviors of the vehicle within a short time window, such as acceleration, braking, lane changing, etc., while the global trend layer can monitor the influence of driver habits and the external environment in the long term, thereby enhancing the generalization ability of the predictive driving model in different scenarios and improving the stability and adaptability of driving prediction results.

[0032] 3. By adopting a multi-level recursive cross-update strategy, the prediction strategy can be quickly adjusted when the real-time traffic environment changes, reducing short-term prediction errors and avoiding wrong driving behaviors caused by sudden environmental changes. In complex traffic scenarios, such as sudden traffic accidents, temporary road closures, etc., the driving prediction results can be quickly corrected, providing accurate real-time decision support for the driving system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart of the decision-making method for intelligent predictive driving of connected vehicles based on machine learning according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0036] The following specifically describes the specific solution of the decision-making method for intelligent predictive driving of connected vehicles based on machine learning provided by the present invention in conjunction with the accompanying drawings.

[0037] Referring to the attached Figure 1 , which shows a flowchart of the decision-making method for intelligent predictive driving of connected vehicles based on machine learning provided by an embodiment of the present invention. The method includes the following steps:

[0038] S1. Obtain and preprocess the multi-source data, vehicle networking real-time data, and historical driving data of the connected vehicle to obtain the preprocessed multi-source data, vehicle networking real-time data, and historical driving data of the connected vehicle; extract features from the preprocessed multi-source data and vehicle networking real-time data of the connected vehicle to obtain preliminary feature data; introduce a multi-layer recursive feature fusion algorithm to perform fusion processing on the preliminary feature data to obtain the finally fused data;

[0039] Obtain the multi-source data of the connected vehicle from data acquisition devices such as lidar (LiDAR), millimeter-wave radar, cameras, ultrasonic sensors, etc., including vehicle dynamic data such as sensor data, vehicle speed, acceleration, steering angle, tire pressure, accelerator pedal position, brake state, etc., and position data such as GPS positioning information, longitude, latitude, and elevation of the vehicle; obtain the vehicle networking real-time data in the weather and traffic management system through existing integrated API interfaces, including traffic signal information, traffic flow, road conditions information, road closure, accident alerts, weather data, construction information, temporary road closures, information interaction between vehicles (such as vehicle speed, position), and communication between vehicles and roadside units (RSUs) (such as traffic signal status, real-time road conditions), etc.; obtain historical driving data from existing databases, including driving habit data such as hard acceleration, hard braking, lane-changing behavior, etc., environmental data such as the weather, road surface conditions, traffic density, etc. at that time, and driving trajectory data such as the driving path of the vehicle on historical roads.

[0040] Perform preprocessing such as data cleaning, denoising, normalization, and standardization on the obtained multi-source data, vehicle networking real-time data, and historical driving data of the connected vehicle to obtain the preprocessed multi-source data, vehicle networking real-time data, and historical driving data of the connected vehicle. The technical means adopted in the preprocessing process are well-known to those skilled in the art and will not be elaborated here.

[0041] Further, use existing feature engineering techniques to extract features from the preprocessed multi-source data and vehicle networking real-time data of the connected vehicle to obtain time feature data such as time period, day-night conversion, etc., spatial feature data such as distance to obstacles, lane deviation, etc., and driving behavior feature data such as sudden changes in acceleration, frequency of braking behavior, etc., and use the obtained time feature data, spatial feature data, and driving behavior feature data as preliminary feature data. Introduce a multi-layer recursive feature fusion algorithm to perform fusion processing on the preliminary feature data. The multi-layer recursive feature fusion algorithm performs multi-level and recursive feature fusion on the preliminary feature data, adopts a multi-level nested structure, combines non-linear mapping and recursive optimization strategies, and realizes the gradual optimization fusion of the preliminary feature data by introducing an adaptive weight adjustment mechanism to ensure the optimal mapping of the preliminary feature data after feature extraction in a high-dimensional space. The specific implementation process is as follows:

[0042] Represent the preliminary feature data as: where, X t represents the time feature data, j s represents the spatial feature data, X d represents the driving behavior feature data, n is the number of feature data, and m is the dimension of the feature data. Since there are differences in the numerical scales and distributions of feature data from different sources, in order to ensure the stability of subsequent calculations, it is necessary to perform standardization processing on the preliminary feature data to obtain the standardized feature data X norm , and use the standardized feature data as the input to enter the recursive feature transformation stage. The technical goal of this stage is to extract more discriminative deep features from the interaction relationships between the preliminary feature data by constructing a multi-layer non-linear mapping structure. In a recursive form, the output of each layer of feature transformation is used as the input of the next layer to form an iteratively optimized feature representation. The mathematical expression of the recursive feature transformation is as follows:

[0043]

[0044] where, is the feature data after the (i + 1)-th layer of recursive feature transformation; is the feature data after the i-th layer of recursive feature transformation, W i is the weight during the i-th layer of recursive feature transformation, representing the influence weight of each feature data on the final fusion result, controlling the contributions of different types of preliminary feature data (time feature data, spatial feature data, driving behavior feature data) in the high-dimensional space, and determined by the expert experience method; b i is the bias term during the i-th layer of recursion, determined by the experimental method; α is the adaptive balance coefficient, determining the influence weight of the sine non-linear component in the recursive feature transformation on the finally recursively transformed feature data, used to control the importance of the non-linear component in the feature fusion process, with a value range of [0, 1] to adapt to the requirements of different scenarios, specifically set according to the specific scenario and not limited here; γ is the frequency adjustment factor of the sine non-linear transformation, used to control the non-linear change degree of the standardized feature data in the high-dimensional space and improve the discrimination ability of the standardized feature data, determined by the expert experience method;

[0045] The above recursive operations continuously update the feature representation until the convergence condition preset according to the expert experience method is reached, such as reaching the preset number of layers according to the expert experience method, so as to obtain the finally recursively transformed feature data X trans .

[0046] After completing the recursive feature transformation, use the finally recursively transformed feature data X transEnter the orthogonal feature fusion stage. The technical goal of this stage is to eliminate the redundant information between the feature data after the final recursive feature transformation, ensure the independence of the feature data after different final recursive feature transformations during the fusion process, and thus improve the effectiveness of data fusion. This step uses the mathematical method of orthogonal projection to process the feature data after the final recursive feature transformation and calculates the orthogonal projection matrix:

[0047]

[0048] where P is the orthogonal projection matrix; T is the transpose of the matrix;

[0049] Then, use the orthogonal projection matrix P to perform a redundancy removal operation on the feature data after the final recursive feature transformation to obtain the initially redundancy-free feature data X orth :

[0050] X orth = X trans - PX trans

[0051] After orthogonal feature fusion, the obtained initially redundancy-free feature data X orth , still has a certain degree of feature redundancy. Therefore, further operations need to be carried out in the adaptive weight fusion stage. The goal of this stage is to assign appropriate weights to different dimensions of the initially redundancy-free feature data to ensure that the contribution of each feature data to the final decision is optimal. The calculation method of adaptive weight fusion is based on the exponential function of the feature norm, and the specific implementation is as follows:

[0052]

[0053] where is the weight of the th initially redundancy-free feature data in the final fusion result, reflecting the relative importance of this feature data in the decision-making; β is the weight smoothing factor, used to control the sensitivity of the exponential function to different feature norms. The larger the value, the more sensitive the weight is to the change of the feature norm. When the value is small, the weight distribution tends to be balanced, and it is determined according to the expert experience method; represents the second norm of the th initially redundancy-free feature data; represents the second norm of the jth initially redundancy-free feature data; is the total number of initially redundancy-free feature data;

[0054] After the weight calculation, perform a weighted sum on the initially redundancy-free feature data to obtain the fused feature data F fuse d :

[0055]

[0056] Further, the existing principal component analysis method is used to perform feature data dimensionality reduction processing on the fused feature data to obtain the finally fused data;

[0057] S2. Using the intelligent predictive driving prediction algorithm for connected vehicles based on deep cross-adaptability, based on the preprocessed historical driving data, construct and train a predictive driving model, and use the predictive driving model to perform prediction processing on the finally fused data to obtain a driving prediction result, generate a driving decision, and drive according to the driving decision.

[0058] Using the intelligent predictive driving prediction algorithm for connected vehicles based on deep cross-adaptability, based on the preprocessed historical driving data, construct and train a predictive driving model. The specific process is as follows: First, on the premise of the obtained historical driving data, obtain the corresponding environmental perception data, including high-dimensional unstructured data such as lidar point clouds, camera images, and meteorological data, and vehicle networking data, including the surrounding vehicle status of V2V and traffic signal information of V2I. After preprocessing the environmental perception data and vehicle networking data, merge them with the preprocessed historical driving data to obtain comprehensively preprocessed data;

[0059] After completing the data preprocessing, use a multi-layer recursive cross-sensing model to construct a predictive driving model. This modeling process is divided into a local perception layer, a global trend layer, and a cross-feedback layer. The technical purpose of the local perception layer is to analyze the immediate driving behavior of the vehicle through a short-time window and use gated temporal convolution to capture driver behavior patterns, such as short-term features like sudden acceleration and emergency braking. The specific implementation method is to perform time window slicing on the comprehensively preprocessed data, input it into the gated temporal convolution network, use the gated unit for feature extraction to obtain the local perception result, and adjust the short-term decision weight through an adaptive learning rate.

[0060] After local perception, enter the global trend layer. The technical purpose of the global trend layer is to establish a long-term driving behavior model, identify the driver's long-term driving preferences and environmental influence trends, and use a hierarchical temporal recurrent neural network. The implementation process includes: First, define a long-term observation window, associate the local perception result with the historical driving data, then construct a long-time dependence through a recurrent neural network, use the gated unit to capture the long-term behavior changes of the driver, and finally use a weighted attention mechanism to assign higher weights to high-frequency behaviors to enhance the prediction ability of the predictive driving model.

[0061] After the global trend layer, it enters the cross-feedback layer. The technical purpose of the cross-feedback layer is to introduce a feedback correction mechanism between the local and global trends, and adjust the local and global weights through a multi-level recursive cross-update strategy. The implementation steps of the multi-level recursive cross-update strategy include: First, calculate the dynamic weight difference between the local prediction error and the global trend error. Subsequently, establish an optimization objective for the loss function to reduce the interference of short-term decisions on long-term strategies. Finally, use the dynamic weight adjustment strategy to optimize the driving prediction results, so that the driving decision can balance short-term accuracy and long-term stability.

[0062] During the process of training the predictive driving model, a hierarchical game strategy reasoning is adopted to optimize the driving decision-making strategy of the predictive driving model. The technical purpose is to improve the intelligence and cooperation of driving decisions and prevent global decision biases caused by local optima. The specific implementation method is to first layer the driving scenario, dividing it into a short-term local game layer and a long-term global game layer. The short-term local game layer is responsible for immediate driving operations, such as lane changes and overtaking. Based on the game theory Nash equilibrium principle, the optimal solution is selected from multiple possible decision-making options. The long-term global game layer models the long-term driving strategy, uses existing deep reinforcement learning methods, and combines the driver's behavior preferences to gradually optimize the long-term driving decision. The specific implementation is that in the initial state, environmental feedback data is collected through a random strategy, and an experience replay buffer pool is used to store the historical state-action-reward sequence. Subsequently, the long-term driving strategy is updated using a deep Q network, and the existing value iteration method is used to select the action plan with the highest long-term benefit. The innovation of this hierarchical strategy lies in the interaction of driving decisions at different time scales, avoiding the limitations of traditional single-time-domain methods.

[0063] After the training of the predictive driving model is completed, the finally fused data is processed to generate driving prediction results. These include short-term behavior prediction results, mid-term driving strategies, and long-term path planning results. The short-term behavior prediction results mainly focus on driving actions in the next 1-3 seconds, such as accelerating, braking, and lane changing. The mid-term driving strategy focuses on the driving plan in the next 3-10 seconds. Combining the prediction results of the global trend layer, considering environmental variables such as road conditions and traffic flow, the optimal driving lane and speed control plan are determined. The long-term path planning results are based on long-term predictions of more than 10 seconds, combining historical driving data and existing traffic flow models to generate a path recommendation plan with the optimal energy consumption and the shortest travel distance.

[0064] To ensure the reliability and sustainable optimization of the predictive driving model, a real-time feedback closed-loop mechanism is introduced to obtain the actual driving data of the vehicle, compare it with the driving prediction results, calculate the deviation, and adjust the model parameters. The specific implementation methods include, first, during the actual driving process of the vehicle, data such as acceleration, vehicle speed, and path deviation are collected in real time, and the existing weighted moving average method is used to smooth the data. Subsequently, the error between the processed data and the model prediction value is calculated, and the existing Bayesian online update method is used to fine-tune the model parameters. The goal of the real-time feedback closed-loop is to enable the predictive driving model to adapt to the dynamically changing traffic environment and continuously optimize its decision-making ability. By introducing an incremental learning mechanism, new driving data can be continuously absorbed without losing historical experience, and the robustness and adaptability of the predictive driving model can be maintained by adjusting the neural network structure.

[0065] In summary, the decision-making method for intelligent predictive driving of connected vehicles based on machine learning is completed.

[0066] The sequence of the invention embodiments is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments.

[0068] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present invention and should all be included in the protection scope of the present invention.

Claims

1. A decision-making method for intelligent predictive driving of connected vehicles based on machine learning, characterized in that: The following steps are involved: S1. Acquire and preprocess multi-source data, real-time data of the Internet of Vehicles, and historical driving data of connected vehicles to obtain preprocessed multi-source data, real-time data of the Internet of Vehicles, and historical driving data of connected vehicles; Extract features from pre-processed multi-source data of connected cars and real-time data of Internet of Vehicles to obtain preliminary feature data; introduce a multi-layer recursive feature fusion algorithm to fuse the preliminary feature data to obtain the final fused data; S2. Use the connected car intelligent predictive driving prediction algorithm based on deep cross adaptability to build and train a predictive driving model based on pre-processed historical driving data, and use the predictive driving model to predict and process the final fused data to obtain driving prediction results, generate driving decisions, and drive according to the driving decisions.

2. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 1, characterized in that: The S1 specifically includes: The multi-layer recursive feature fusion algorithm performs multi-level, recursive feature fusion on the preliminary feature data, adopts a multi-level nested structure, and combines nonlinear mapping with a recursive optimization strategy, and introduces an adaptive weight adjustment mechanism to perform fusion processing on the preliminary feature data.

3. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 2 is characterized in that: The S1 specifically includes: In the implementation process of the multi-layer recursive feature fusion algorithm, the preliminary feature data is standardized to obtain the standardized feature data, and the standardized feature data is used as input to enter the recursive feature transformation stage to obtain the final recursive feature transformed feature data. The specific implementation formula is: in, It is the feature data after the i+1th layer recursive feature transformation; is the feature data after the recursive feature transformation of the i-th layer, W i is the weight of the recursive feature transformation at the i-th layer; b i is the bias term for the i-th layer recursion; α is the adaptive balance coefficient; γ is the frequency adjustment factor of the sinusoidal nonlinear transformation; X norm It is the standardized feature data.

4. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 3 is characterized in that: The S1 specifically includes: In the implementation process of the multi-layer recursive feature fusion algorithm, after completing the recursive feature transformation, the feature data after the final recursive feature transformation enters the orthogonal feature fusion stage, and the feature data after the final recursive feature transformation is processed to obtain an orthogonal projection matrix; the orthogonal projection matrix is ​​used to perform redundancy removal operations on the feature data after the final recursive feature transformation to obtain preliminary non-redundant feature data.

5. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 4 is characterized in that: The S1 specifically includes: In the implementation process of the multi-layer recursive feature fusion algorithm, after the orthogonal feature fusion, the initial non-redundant feature data enters the adaptive weight fusion stage for processing. The specific implementation formula is: in, It is The weight of the initial non-redundant feature data in the final fusion result; β is the weight smoothing factor; Representative The bi-norm of the initial non-redundant feature data; Represents the bi-norm of the jth initial non-redundant feature data; is the total number of initial non-redundant feature data.

6. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 5 is characterized in that: The S1 specifically includes: In the implementation process of the multi-layer recursive feature fusion algorithm, after the weight calculation, the initial non-redundant feature data is weighted summed to obtain the fused feature data, and the fused feature data is subjected to feature data dimensionality reduction processing to obtain the final fused data.

7. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 1, characterized in that: The S2 specifically includes: In the process of implementing the intelligent predictive driving prediction algorithm for connected vehicles based on deep cross adaptability, corresponding environmental perception data and Internet of Vehicles data are obtained based on the acquired historical driving data, and after preprocessing the environmental perception data and Internet of Vehicles data, they are merged with the preprocessed historical driving data to obtain comprehensive preprocessed data.

8. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 7, characterized in that: The S2 specifically includes: After completing data preprocessing, a predictive driving model is constructed using multi-layer recursive cross-perception modeling. The modeling process is divided into a local perception layer, a global trend layer, and a cross-feedback layer.

9. The decision-making method for intelligent predictive driving of a connected vehicle based on machine learning according to claim 8, characterized in that: The S2 specifically includes: In the process of training the predictive driving model, a hierarchical game strategy is used to infer and optimize the driving decision-making strategy of the predictive driving model; after completing the training of the predictive driving model, the final fused data is processed to generate the driving prediction results.

Images