Vehicle driving track prediction method under highway traffic control based on digital twin tracking simulation

Abstract

The invention relates to the technical field of intelligent high-speed vehicle passing control, and provides a method for predicting a vehicle driving track under highway passing control based on digital twin tracking simulation, and the method comprises the steps: collecting highway digital twin system, target vehicle and environment data in real time, and carrying out the integration to form an input data set; a future driving track of a vehicle is predicted by using a pre-trained LSTM deep neural network model, and an actual track is collected through devices such as a GPS and the like for verification. And when the prediction deviation exceeds the preset range, the model is dynamically adjusted to improve the accuracy. Prediction results and deviation analysis are transmitted to a background control terminal and a driver mobile terminal in real time, and traffic control, accident early warning and driving aid decision making such as front road condition prediction and suggested driving speed providing are supported. The highway traffic efficiency and safety are effectively improved, the driver experience is enhanced, the development of an intelligent traffic system is promoted, the operation cost is reduced, and remarkable social and economic benefits are achieved.

Description

Technical Field

[0001] The present invention relates to the field of highway vehicle traffic control technology, and in particular to a vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation. Background Art

[0002] With the increasing expansion of the highway network and the continuous increase in the number of vehicles, problems such as highway traffic congestion and frequent accidents have gradually become prominent. How to effectively improve the traffic efficiency of road vehicles, ensure driving safety, scientifically and rationally formulate traffic control strategies under different scenarios, and achieve quasi-all-weather traffic on highways has become an urgent problem for road managers to solve.

[0003] Most of the existing highway traffic control strategies rely on empirical predictions based on historical data and simple mathematical models, which are difficult to fully consider real-time traffic conditions and various emergencies. At the same time, traditional vehicle trajectory prediction methods are limited by computing power, real-time and model adaptability, and are difficult to effectively cope with complex and changing traffic environments. In the process of smart highway development, digital twin technology, as an emerging digital simulation technology, has shown great application potential.

[0004] Digital twin technology builds a virtual mirror of the physical world to achieve comprehensive perception, accurate prediction and intelligent decision-making of the physical system, providing a new solution for highway traffic control. In the field of highway traffic control, the vehicle driving trajectory prediction method based on digital twin technology can track the vehicle driving status in real time, combine historical data and real-time traffic information, and accurately predict the vehicle's future driving trajectory, thereby providing a scientific basis for highway traffic control.

[0005] However, the vehicle trajectory prediction method based on digital twin technology is still in the research and exploration stage, and there are many technical problems that need to be solved. For example, how to build a digital twin model that is highly consistent with the physical highway and interacts in real time; how to design an efficient trajectory tracking and prediction algorithm to improve the prediction accuracy and real-time performance; how to achieve the fusion and processing of multi-source data to improve the robustness and reliability of the prediction results, etc.

[0006] Therefore, the present invention proposes a method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation, aiming to achieve real-time tracking and accurate prediction of vehicle driving trajectories by building a digital twin model of the highway, provide a scientific basis for highway traffic control, and improve road traffic efficiency and safety. The proposal of the present invention not only helps to solve the limitations of traditional traffic management methods, but also provides new ideas and technical support for the development and construction of smart highways. Summary of the invention

[0007] The problem solved by the present invention is that, in the current field of highway traffic control, although there are many methods for predicting vehicle driving trajectories, these methods often have problems such as insufficient prediction accuracy, poor real-time performance, and weak robustness when facing complex and changeable traffic environments. Specifically, traditional prediction methods often rely on limited historical data and empirical models, and it is difficult to fully capture and reflect the dynamic characteristics of the traffic system, especially in the case of large traffic volume, complex road conditions, and frequent emergencies. The prediction results often deviate greatly from the actual situation. This patent aims to achieve dynamic prediction of highway vehicle driving trajectories by combining digital twin technology, data clustering analysis, and real-time trajectory prediction models.

[0008] The present invention provides a vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation, comprising:

[0009] Based on digital twin technology, all elements and scenes on the highway are dynamically restored in digital time and space, a digital twin base for the entire domain and all elements is constructed, and a digital twin transportation system with global management, synchronous visualization, and virtual-real interaction is built. Carry out 3D digital twin modeling of highways, complete the three-dimensional static digital twin of all basic environments, structures, equipment and facilities, signs and signboards of highways, and conduct digital twins of dynamic traffic events such as ongoing maintenance and construction on highways. With the support of real-time data, the real-time online vehicle operation simulation of road sections is realized, and the 3D real-time traffic dynamic twin of highways is realized. Through holographic perception of traffic elements, full data fusion is carried out, and real-world information is imported into the twin traffic simulation system. Combined with high-precision maps, real-time data is tracked to locate vehicles and trace, analyze video portraits, and comprehensively study the relationship between vehicles, roads, and the environment, and finally solve traffic problems such as the inability to predict and respond quickly to traffic events.

[0010] Build a digital twin transportation system for highways, integrate multi-source data such as high-precision maps, meteorological environment, and real-time traffic status, upgrade the traditional static highway scene into a visualized dynamic online interactive three-dimensional digital twin highway, and realize real-time and accurate presentation of all elements of road status and vehicle behavior in the entire section and traceability and restoration, thereby realizing real-time management of road sections and road network traffic and service intervention, analysis and judgment of operation status, situation simulation, traffic risk prediction, and analysis and prediction of road traffic operation control effects, meeting the various management needs of highway operation departments under normal and emergency conditions and quasi-all-weather traffic service needs.

[0011] The following are the main steps to build a highway digital twin system:

[0012] Integrate multi-source data such as geographic information systems, video surveillance systems, and traffic flow monitoring systems to comprehensively collect and integrate highway network structures, vehicle information, and traffic events. Use advanced data fusion algorithms to efficiently integrate data from different sensors and IoT devices to improve data accuracy and integrity. Use cloud computing platforms to achieve rapid processing and analysis of large-scale data to meet real-time requirements.

[0013] High-precision 3D modeling technology is used to model highways and their ancillary facilities in 3D, and digital twin models of highways and their ancillary facilities are constructed to achieve accurate replication of the physical world. Through simulation technology, vehicle driving conditions and accident risks in different traffic scenarios are simulated to provide a scientific basis for decision-making.

[0014] Through the Internet of Things technology, real-time data interaction between the digital twin model and the physical highway is achieved to ensure the accuracy and real-time nature of the model data. Through the interconnection and collaborative optimization between devices, the overall efficiency of traffic management is improved.

[0015] By applying the digital twin system of highways, the driving trajectory data of each type of vehicle under different weather environments, different road section structures and traffic control measures can be dynamically collected. Through data clustering, a standardized driving trajectory library of vehicles under different control measures is formed. This library can cover the standard driving modes of vehicles under various common conditions.

[0016] The standardized driving trajectory library is the basis of driving trajectory prediction. The library collects historical traffic data, including vehicle driving trajectories under different time, weather, and traffic flow conditions, and forms a complete and reusable data set after cleaning, denoising, and standardization. In order to maintain the timeliness and representativeness of the data, the system regularly updates the trajectory library, incorporates the latest driving records, and removes outdated or abnormal data. At the same time, data mining technology is used to analyze driving patterns and laws under different conditions, providing rich samples for the training of vehicle trajectory dynamic prediction algorithm models.

[0017] The following are the main steps to build a standardized driving trajectory library:

[0018] Clarify the types of data that need to be collected, including vehicle type classification, vehicle trajectory data, meteorological environment parameters, road structure information and traffic control measures.

[0019] The highway digital twin system combines sensors, cameras, radars and other equipment deployed on the highway to dynamically collect vehicle driving trajectory data in real time.

[0020] Clean the collected raw data to remove errors, redundancy and incomplete information to ensure data quality. Format and standardize data to unify data formats and units for subsequent processing.

[0021] Extract key features from the cleaned data, such as vehicle speed, acceleration, driving direction, lane position, etc., as well as related weather, road section and control measures information.

[0022] Label each driving trajectory to clearly identify the corresponding vehicle type, weather conditions, road structure and traffic control measures.

[0023] The K-means clustering algorithm is used to group vehicles according to the similarities and differences of their driving trajectories to form different driving patterns. Clustering parameters are adjusted to ensure that the clustering results are neither too fragmented nor too general, and can accurately reflect the driving characteristics under different control measures.

[0024] Based on the clustering results, a standardized driving trajectory model is constructed for each vehicle type, each weather condition, each road section structure and each traffic control measure. The commonalities and differences of the standardized trajectories are analyzed to extract key indicators and parameters.

[0025] The standardized driving trajectory model was validated using a data set that did not participate in the clustering analysis to evaluate its accuracy and generalization ability. The model parameters and clustering strategy were adjusted based on the validation results to continuously optimize the model performance.

[0026] The verified standardized driving trajectory models are integrated into a library, including a collection of driving trajectories under different vehicle models, weather conditions, road structure and control measures.

[0027] As new data is continuously added and the traffic environment changes, the standardized driving trajectory library is regularly maintained and updated to ensure that it is consistent with the actual situation.

[0028] In actual applications, by comparing the real-time meteorological environment, road infrastructure, and traffic control measures obtained by the highway digital twin system with the records in the standardized driving trajectory library, the LSTM deep neural network algorithm is used to dynamically predict the driving trajectory of the target vehicle. The model learns the complex mapping relationship between the vehicle's historical driving trajectory and real-time data through multi-level nonlinear transformations.

[0029] The following are the main steps of the prediction algorithm:

[0030] 1) Forget gate setting

[0031] F k =f[W f ×[h k-1 ,x k-1 ]+θ f]

[0032] Among them, f(·) is the activation function of the neural network, h k-1 For the previous unit input, x k-1 is the current input, θ f Indicates the threshold value.

[0033] 2) Input status update

[0034]

[0035] Among them, g(·) is generally selected as a hyperbolic tangent function, and the specific expression is as follows: A newly created candidate quantity.

[0036]

[0037] 3) Update the neural network state

[0038]

[0039] This step completes the update of the candidate quantity. From the above formula, we can see that the historical state S k-1 With f k It can complete the forgetting of some historical information, and then Weighting allows the results to update the status level according to certain rules.

[0040] 4) Network output

[0041]

[0042] After the processing of g(·) in this step, a value in the interval [-1,1] is obtained. k Multiply them together and finally get the neural network output.

[0043] In order to improve the prediction accuracy and generalization ability of the model, a feature vector that can fully reflect the vehicle driving environment, road conditions and traffic control effects is constructed by performing in-depth feature extraction and selection on the input data. These features include but are not limited to weather parameters, road condition indicators, traffic flow data, vehicle status (such as speed, acceleration, steering angle) and historical driving trajectory characteristics. According to the specific needs of the vehicle trajectory prediction task, the number of hidden layers, number of neurons, activation function and other parameters of the LSTM model are adjusted to balance the complexity and computational efficiency of the model. At the same time, the introduction of attention mechanism or gating mechanism can be considered to enhance the model's ability to capture key information. During the model training process, the real-time data features and the records in the standardized driving trajectory library are used as inputs. By adjusting the network parameters, the complex relationship between the vehicle driving trajectory and various influencing factors is learned. According to the real-time updated data, the input of the LSTM model is continuously adjusted to achieve dynamic prediction of the future driving trajectory of the target vehicle. The prediction results not only contain location information, but can also be further refined to dynamic parameters such as speed and acceleration to improve the accuracy and practicality of the prediction.

[0044] Beneficial effects of the present invention:

[0045] The vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation described in the present invention has beneficial effects mainly reflected in the following aspects:

[0046] Through real-time prediction of vehicle driving trajectories, the present invention enables traffic management departments to timely understand road conditions and make advance predictions on potential traffic jams and accident risks, thereby quickly taking control measures, such as adjusting signal light timing, issuing traffic warning information, etc., to effectively avoid traffic jams and accidents and improve the travel efficiency and safety of highways.

[0047] The predicted trajectory and deviation analysis results are transmitted to the driver's mobile terminal to provide the driver with driving assistance information such as road condition prediction and recommended driving speed. This information can help the driver understand the road conditions ahead in advance, make more reasonable driving decisions, avoid dangerous behaviors such as speeding and sudden braking, and improve driving experience and safety.

[0048] When the deviation between the predicted trajectory and the actual trajectory exceeds the preset range, the system will dynamically adjust the model based on the new real-time data and deviation analysis results. This feedback mechanism can continuously optimize the model parameters and structure, improve the prediction accuracy and robustness of the model, and make it better adapt to the vehicle trajectory prediction needs under different road conditions and traffic environments.

[0049] Based on digital twin technology, the present invention accurately simulates vehicle driving and road conditions in a virtual environment, realizes three-dimensional visualization of the real traffic environment and the automatic establishment of maintenance decision-making plans based on simulation analysis results. The application of this technology provides new ideas and methods for the development of intelligent transportation systems, and helps promote the digital transformation and intelligent upgrading of the transportation industry. Through automated trajectory prediction and traffic control decisions, the present invention reduces the possibility of manual intervention and errors, and reduces operating costs. At the same time, by improving traffic efficiency and safety, the present invention also helps to improve the economic and social benefits of highways.

[0050] In summary, the vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation described in the present invention shows significant beneficial effects in improving traffic control efficiency and safety, enhancing driver driving experience and safety, improving model prediction accuracy and robustness, promoting the development of intelligent transportation systems, reducing operating costs and improving economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0052] Figure 1 A schematic flow chart of a method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation provided by the present invention.

[0053] Figure 2 Schematic diagram of the first highway digital twin system provided by the present invention.

[0054] Figure 3 Schematic diagram of the second expressway digital twin system provided by the present invention.

[0055] Figure 4 Schematic diagram of the third expressway digital twin system provided by the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0057] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0058] See also Figures 1 to 4 The present invention provides a method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation, comprising:

[0059] Step S101: Acquire basic geographic data, real-time vehicle driving data, traffic environment data, digital twin model data, historical driving trajectory data, driver behavior data, and traffic control data to build a digital twin model of the highway;

[0060] The basic geographic data obtained include the geographical location of the highway, road direction, intersection location, number of lanes, slope, curvature, etc.

[0061] Collect real-time vehicle driving data such as vehicle location, speed, acceleration, direction, etc., and integrate data from on-board GPS and on-board communication equipment.

[0062] Monitor and collect traffic environment data such as meteorological data (temperature, humidity, wind speed, wind direction, visibility, rainfall, etc.), road condition data (slippery road surface, damaged road surface, construction information, etc.) and real-time traffic condition data (traffic flow, congestion, etc.).

[0063] Step S102: capturing meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data in real time, and pre-processing the meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data;

[0064] Real-time data capture: Through sensors and IoT devices deployed along the highway, meteorological and environmental data, road infrastructure data, traffic control measures and real-time vehicle data are captured in real time.

[0065] Data preprocessing: Clean the captured data, remove duplicate records, fill in missing values, and remove outliers. Standardize the data and scale all features to the same scale. Extract features useful for trajectory prediction, such as meteorological parameters, road infrastructure coding, vehicle speed, acceleration, etc. Store the preprocessed data in a data center or cloud computing platform.

[0066] Step S103: extracting features from the preprocessed data to obtain trajectory length features, shape features and speed distribution features, and performing standardization processing on the trajectory data. The standardization processing of the trajectory data includes trajectory alignment, parameterized representation, data compression and metadata annotation. The standardized trajectory data is stored in a database to form a standardized driving trajectory library.

[0067] Extract trajectory length features, shape features (such as curvature, steering angle) and speed distribution features from the preprocessed data. Perform trajectory alignment, parameterization, data compression and metadata annotation on the trajectory data. Store the standardized trajectory data in the database to form a standardized driving trajectory library.

[0068] Step S104: constructing a vehicle driving trajectory prediction model based on the LSTM deep neural network, using the training set data to train the vehicle driving trajectory prediction model, updating the model parameters through forward propagation and back propagation, and obtaining the vehicle driving trajectory prediction model;

[0069] Determine the number of LSTM network layers based on the complexity of the task and the size of the data. Select training and validation sets from the standardized driving trajectory library. Normalize the input and output data.

[0070] Randomly initialize the weights and biases of the LSTM network. Use the training set data to train the LSTM model and update the model parameters through forward propagation and back propagation. Save the trained LSTM model to a file or database.

[0071] Step S105: The real-time collected highway digital twin system data, the target vehicle's driving data, and the similar trajectory data extracted from the standardized driving trajectory library are input into the vehicle driving trajectory prediction model. The model outputs the prediction result of the future driving trajectory of the target vehicle based on the input data, collects the actual driving trajectory of the target vehicle, and compares the deviation between the prediction result of the future driving trajectory of the target vehicle and the actual driving trajectory of the target vehicle. If the deviation is greater than the prediction error range value, the vehicle driving trajectory prediction model is dynamically adjusted according to the new real-time data, and the prediction result is transmitted to the background control terminal and the driver's mobile terminal.

[0072] Collect data from the highway digital twin system and target vehicles in real time. Integrate the real-time data with similar trajectory data extracted from the standardized driving trajectory library to form a complete input data set.

[0073] Preprocess the data collected in real time. Load the pre-trained LSTM model and input the pre-processed input data into the model. The model outputs the prediction result of the future driving trajectory of the target vehicle.

[0074] The actual driving trajectory of the target vehicle is collected, and the deviation between the predicted result and the actual trajectory is compared. If the deviation is greater than the predicted error range, the model is dynamically adjusted according to the new real-time data. The predicted results and deviation analysis are transmitted to the background control terminal and the driver's mobile terminal for traffic control, accident warning and driving assistance.

[0075] Specifically, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation described in the present invention, step S101 includes:

[0076] Step S1011, obtaining geographic location information of the expressway, including road direction, intersection location, number of lanes, slope and curvature, and collecting geographic environment data, such as topography and building distribution;

[0077] Step S1012: Using sensors installed on the highway, the vehicle's location, speed, acceleration, and direction of travel status information are collected in real time, and data from the vehicle's GPS and vehicle-mounted communication equipment are integrated;

[0078] Step S101-3: Real-time monitoring and collection of meteorological data, road condition data, such as slippery, damaged, and construction information of the road surface, and real-time traffic condition data of traffic flow and congestion;

[0079] Step S101-4: using the existing highway digital twin model data, including road models, traffic facility models, and surrounding environment models, as the basis for building a new model, and updating and optimizing the digital twin model based on the real-time collected data;

[0080] Step S101-5: extracting the driving trajectory of the vehicle on the highway from the historical data, including the vehicle starting point, end point, waypoints, driving time, and speed change;

[0081] Step S101-6: Collecting the driver's driving behavior data through the vehicle-mounted device or the driver behavior monitoring system;

[0082] Step S1017: Obtaining traffic control information on the expressway, including speed limits, road closures, and traffic diversion control measures;

[0083] Step S1018: Integrate and fuse the various types of data collected above, and use advanced modeling techniques and algorithms to build a digital twin model of the highway.

[0084] The vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation described in the present invention, in step S101, describes in detail the specific steps of building a highway digital twin model, which are as follows:

[0085] Step S101: Build a digital twin model of the highway

[0086] Step S1011: Obtain basic geographic data and collect geographic location information of highways, including road direction, intersection location, number of lanes, slope, curvature, etc., to provide a basic geographic spatial framework for building a digital twin model.

[0087] Step S1012: The sensors installed on the highway collect the vehicle's location, speed, acceleration, direction and other driving status information in real time, and integrate the data from the vehicle's GPS and vehicle-mounted communication equipment to obtain the vehicle's real-time dynamic information and provide the model with real-time data support for the vehicle's driving status.

[0088] Step S1013: Real-time monitoring and collection of meteorological data (such as temperature, humidity, wind speed, etc.), as well as road condition data (such as slippery, damaged, construction information, etc.) and real-time traffic condition data such as traffic flow and congestion.

[0089] Step S1014: Use the existing highway digital twin model data, including road models, traffic facility models, surrounding environment models, etc., as the basis for building a new model. Use the existing data to accelerate the construction process of the new model and ensure that the model has a certain initial accuracy.

[0090] Step S1015: Extract the driving trajectory of the vehicle on the highway from the historical data, including the vehicle starting point, end point, waypoints, driving time, speed change, etc. Provide rich historical trajectory data for model training and enhance the prediction ability of the model.

[0091] Step S1016: Collect the driver's driving behavior data through the vehicle-mounted equipment or the driver behavior monitoring system, consider the impact of the driver's behavior on the vehicle's driving trajectory, and improve the accuracy and practicality of the prediction model.

[0092] Step S1017: Obtain traffic control information on the expressway, including speed limits, road closures, traffic diversion and other control measures. Incorporating traffic control factors into the model considerations makes the prediction results more consistent with the actual situation.

[0093] Step S1018: Integrate and fuse the various types of data collected above, and use advanced modeling technology and algorithms to build a digital twin model of the highway. A digital twin model that comprehensively and accurately reflects the actual situation of the highway is formed, providing strong support for subsequent vehicle driving trajectory prediction.

[0094] Through the above steps, the present invention successfully constructs a digital twin model of the highway, providing a solid foundation for subsequent vehicle driving trajectory prediction.

[0095] Specifically, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation described in the present invention, step S102 includes:

[0096] Step S1021, by deploying meteorological monitoring stations along the expressway, real-time capture of temperature, humidity, wind speed, wind direction, visibility, and rainfall meteorological parameters, and road infrastructure data capture using pavement status sensors and video surveillance camera equipment to obtain real-time pavement conditions, including slipperiness, potholes and cracks, clarity of traffic signs and markings, and structural safety status data of bridges and tunnels;

[0097] Obtain real-time traffic control measures through the traffic management system, including temporary speed limits, lane closures, and traffic guidance information;

[0098] Through the GPS positioning system and vehicle identification camera equipment, the location, longitude and latitude, speed, acceleration and direction of the target vehicle are obtained in real time;

[0099] Step S1022, checking and deleting duplicate records in the data set, using interpolation to fill in missing values ​​in the data set, identifying and eliminating outliers in the data, including unreasonable speed and acceleration data;

[0100] Step S1023: converting the temperature, speed, and humidity of different dimensions into a unified metric standard, and scaling all feature data to the same scale;

[0101] Step S1024, extracting trajectory prediction features from the standardized data, the trajectory prediction features including temperature and humidity in meteorological parameters, road surface condition codes in road infrastructure data, and speed and acceleration in real-time vehicle data;

[0102] Step S1025: Integrate the feature-extracted meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data, and store the integrated data in a data center or cloud computing platform.

[0103] The vehicle trajectory prediction method under highway traffic control based on digital twin tracking simulation described in the present invention describes in detail the real-time data capture and preprocessing process in step S102. The following are the specific steps:

[0104] Step S1021: Meteorological environment data capture: Meteorological parameters such as temperature, humidity, wind speed, wind direction, visibility, and rainfall are captured in real time through meteorological monitoring stations deployed along the highway.

[0105] Road infrastructure data capture: Use road condition sensors, video surveillance cameras and other equipment to obtain real-time road condition data, including the slipperiness, potholes and cracks of the road, the clarity of traffic signs and markings, and the structural safety of bridges and tunnels.

[0106] Obtaining traffic control measures: Obtain real-time traffic control measures through the traffic management system, such as temporary speed limits, lane closures, traffic guidance information, etc.

[0107] Real-time vehicle data capture: Through GPS positioning systems, vehicle identification cameras and other equipment, the target vehicle's location (latitude and longitude), speed, acceleration, direction and other driving data are obtained in real time.

[0108] Step S1022: Delete duplicate records: Check whether there are duplicate records in the data set and delete them.

[0109] Missing value processing: For missing values ​​in the data set, interpolation method is used to fill them in to ensure the integrity of the data.

[0110] Outlier processing: Identify and remove outliers in the data, such as unreasonable speed and acceleration data, to improve data accuracy.

[0111] Step S1023: Convert data of different dimensions such as temperature, speed, humidity, etc. into a unified metric standard to facilitate subsequent data processing and analysis.

[0112] Scale all feature data to the same scale to avoid some features having too large a weight in the model.

[0113] Step S1024: Trajectory prediction feature extraction: Extract features useful for trajectory prediction from the standardized data, such as temperature and humidity in meteorological parameters, road surface condition coding in road infrastructure data, and speed and acceleration in real-time vehicle data.

[0114] Step S1025: Integrate the feature-extracted meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data to form a complete data set. Store the integrated data in a data center or cloud computing platform to facilitate subsequent data analysis and model training.

[0115] Through the above steps, the present invention successfully realizes the capture and preprocessing of real-time data along the expressway, providing high-quality data support for subsequent vehicle driving trajectory prediction.

[0116] Specifically, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation described in the present invention, step S103 includes:

[0117] Step S1031, by calculating the sum of distances between all points in the trajectory, the total length of each trajectory is obtained as the basic feature of the trajectory, and the shape feature of the trajectory is extracted. The shape feature of the trajectory includes the curvature and the turning angle of the trajectory;

[0118] Step S103.2: aligning the starting points and end points of different trajectories into a unified reference coordinate system, converting the trajectory data into a parameterized form, and adding metadata annotations to each trajectory data, wherein the metadata annotations added to each trajectory data include the collection time, location, vehicle type, and driver information of the trajectory;

[0119] Step S1033: Integrate the standardized trajectory data into the database to form a standardized driving trajectory library.

[0120] The vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation described in the present invention extracts, aligns and standardizes the trajectory features, and constructs a driving trajectory library in step S103. The following is a detailed description of the specific steps:

[0121] Step S1031: Trajectory feature extraction

[0122] Track length calculation: The total length of each track is obtained by calculating the sum of the distances between all points in the track. This length is the basic feature of the track and reflects the distance traveled by the vehicle.

[0123] Shape feature extraction: Extract the shape features of the trajectory, including the curvature and steering angle of the trajectory. The curvature describes the degree of bending of the trajectory at a certain point, while the steering angle reflects the change in the direction of the vehicle. These features are crucial for understanding the vehicle's driving pattern and predicting future driving trajectories.

[0124] Step S103.2: Track alignment: Align the start and end points of different tracks to a unified reference coordinate system. This step ensures the spatial consistency of the track data, facilitating subsequent data analysis and comparison.

[0125] Parametric transformation: Convert trajectory data into parametric form to facilitate mathematical processing and modeling.

[0126] Metadata annotation: metadata annotation is added to each track data, including the track collection time, location, vehicle type, and driver information. These metadata provide contextual information of the track data, which helps to better understand the track characteristics and perform data analysis.

[0127] Step S1033: Integrate the standardized trajectory data into the database. This step ensures the orderly storage and efficient access of the trajectory data.

[0128] Driving trajectory library formation: By integrating and storing standardized trajectory data, a standardized driving trajectory library is formed. This library provides rich data support for subsequent vehicle driving trajectory prediction, which helps to improve the accuracy and reliability of the prediction.

[0129] Through the above steps, the present invention successfully extracts useful features from the original trajectory data and constructs a standardized driving trajectory library. These works provide a solid foundation for subsequent vehicle driving trajectory prediction.

[0130] Specifically, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation described in the present invention, step S104 includes:

[0131] Step S1041: Determine the number of layers of the LSTM network according to the complexity of the task and the data size;

[0132] Step S1042: from the standardized driving trajectory library, select a part of the data as a training set and another part as a validation set, divide the trajectory data in the training set into an input sequence and a corresponding output sequence, the input sequence includes the driving data of the vehicle for a period of time in the past, and the output sequence includes the driving data of the vehicle for a period of time in the future, and perform normalization processing on the input data and the output data;

[0133] Step S104.3: Randomly initialize the weights and biases of the LSTM network, select a loss function to evaluate the prediction error of the model according to the prediction target, select an optimization algorithm to update the parameters of the model to minimize the loss function, input the input data into the LSTM network, calculate the output and loss, calculate the gradient according to the loss function, and update the parameters of the model through the back propagation algorithm, repeat the forward propagation and back propagation process until the predetermined number of training rounds is reached or the loss function converges;

[0134] Step S1044: Save the trained LSTM model to a file or database.

[0135] The vehicle trajectory prediction method for highway traffic control based on digital twin tracking simulation described in the present invention describes in detail the process of model training using a long short-term memory network (LSTM) in step S104. The following is an in-depth analysis of the specific steps:

[0136] Step S1041: Task complexity analysis: According to the complexity of the vehicle trajectory prediction task and the data scale, the depth and width of the required LSTM network are evaluated. The depth (number of layers) determines the length of the long-term dependencies that the network can capture, while the width (number of neurons in each layer) affects the representation ability of the network.

[0137] Determine the number of network layers: Select the appropriate number of LSTM network layers through experiments or experience to balance the model's expressiveness and training efficiency.

[0138] Step S1042: Dataset division: Randomly select a portion of the data from the standardized driving trajectory library as a training set and another portion as a validation set. The training set is used to train the model, and the validation set is used to evaluate the performance of the model and adjust the hyperparameters.

[0139] Sequence segmentation and normalization: The trajectory data in the training set is segmented into input sequences and corresponding output sequences.

[0140] The input sequence contains the vehicle's driving data (such as position, speed, acceleration, etc.) over a period of time.

[0141] The output sequence contains the vehicle's driving data for a period of time in the future (prediction target).

[0142] The input data and output data are normalized to eliminate the dimensional differences between different features and improve the convergence speed and prediction accuracy of the model.

[0143] Step S104.3: Parameter initialization: Randomly initialize the weights and biases of the LSTM network as the starting point for model training. According to the prediction target (such as position prediction, speed prediction, etc.), select a suitable loss function to evaluate the prediction error of the model. Select an optimization algorithm (such as Adam, SGD, etc.) to update the parameters of the model to minimize the loss function.

[0144] Input data is fed into the LSTM network and forward propagation is performed to calculate output and loss. Gradients are calculated based on the loss function, and the model parameters are updated through the back-propagation algorithm. The forward and back-propagation processes are repeated until the predetermined number of training rounds is reached or the loss function converges.

[0145] Model validation and adjustment: During the training process, the validation set is used regularly to evaluate the performance of the model. Hyperparameters (such as learning rate, number of network layers, etc.) are adjusted based on the validation results to improve the generalization ability of the model.

[0146] Step S1044: Model saving: Save the trained LSTM model to a file or database for subsequent loading and prediction. The saved model should contain information such as network structure, parameter weights and biases to ensure that the model state can be accurately restored.

[0147] Through the above steps, the present invention successfully uses the LSTM network to model and train the vehicle's driving trajectory, providing an effective model support for subsequent trajectory prediction.

[0148] Specifically, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation described in the present invention, step S105 includes:

[0149] Step S1051: collect data from the highway digital twin system and the target vehicle in real time, including vehicle position, speed, acceleration, direction, weather conditions, road conditions, and traffic control measures, and integrate the real-time data with similar trajectory data extracted from the standardized driving trajectory library to form a complete input data set;

[0150] Step S1052: pre-processing the data collected in real time, loading a pre-trained vehicle driving trajectory prediction model based on the LSTM deep neural network, inputting the pre-processed input data into the model, and the model outputting a prediction result of the future driving trajectory of the target vehicle based on the input data;

[0151] Step S1053: Collect the actual driving trajectory of the target vehicle through GPS, radar, and camera sensor equipment.

[0152] Step S1051: Collect data in real time from the highway digital twin system, which integrates multiple sensors and data analysis tools to provide detailed information on weather conditions, road conditions, and traffic control measures. Collect data in real time from sensors and communication devices installed on the target vehicle, including driving parameters such as the vehicle's location, speed, acceleration, and direction.

[0153] Data integration: The data collected in real time from the digital twin system and the target vehicle are integrated with similar trajectory data extracted from the standardized driving trajectory library. Similar trajectory data refers to historical trajectory data that is similar to the current driving environment, vehicle status, etc., which can provide valuable reference for current trajectory prediction. Through data integration, a complete input data set is formed, which contains all the key information required for accurate trajectory prediction.

[0154] Step S1052: Data preprocessing: Preprocess the data collected in real time, including data cleaning, denoising, normalization and other operations to ensure the quality and consistency of the data. According to the requirements of the LSTM model, the input data is format converted and feature extracted to meet the input requirements of the model.

[0155] Model loading and prediction: Load the pre-trained vehicle trajectory prediction model based on LSTM deep neural network. The model has mastered the laws and patterns of vehicle trajectories through learning a large amount of historical data.

[0156] The preprocessed input data is input into the model, and the model calculates and outputs the prediction results of the future driving trajectory of the target vehicle based on the input data. The prediction results may include parameters such as the position, speed, acceleration and direction of the vehicle in the future.

[0157] Step S1053: The actual driving trajectory data of the target vehicle is collected in real time through sensor devices such as GPS, radar, and camera. These data will be used to verify the prediction results of the model. The driving trajectory predicted by the model is compared and analyzed with the actual collected driving trajectory to evaluate the prediction accuracy and reliability of the model. If it is found that there is a large deviation between the prediction result and the actual situation, the model can be further tuned and trained to improve its prediction performance.

[0158] At the same time, actual driving trajectory data can also be added to the standardized driving trajectory library as new training samples to enrich the training data and improve the generalization ability of the model.

[0159] Through the above steps, the present invention realizes the real-time prediction and verification of vehicle driving trajectories under highway traffic control, providing important decision-making support for traffic management departments and drivers.

[0160] Specifically, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation described in the present invention, step S1053 includes:

[0161] Calculate the deviation between the predicted trajectory and the actual trajectory, and compare the calculated deviation with the preset prediction error range value. If the deviation is greater than the prediction error range value, dynamically adjust the vehicle driving trajectory prediction model based on the new real-time data and the deviation analysis results;

[0162] The predicted trajectory and deviation analysis are transmitted to the background control terminal and the driver's mobile terminal. The background control terminal uses the prediction results to perform traffic control and accident warning, and the driver's mobile terminal provides the driver with driving assistance information, including the road condition prediction ahead and the recommended driving speed.

[0163] In step S1053, the method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation of the present invention further refines the verification, feedback and application process of the predicted trajectory. The following is a detailed description of this step:

[0164] Step S1053: Calculate the deviation between the predicted trajectory and the actual trajectory. The deviation can be measured by a variety of indicators, such as position deviation, speed deviation, acceleration deviation, etc. Compare the calculated deviation with the preset prediction error range value. The prediction error range value is set according to actual needs and historical experience, and is used to judge the accuracy of the prediction results.

[0165] Dynamic model adjustment: If the deviation is greater than the prediction error range, it means that the current model's prediction performance is poor and needs to be adjusted. Dynamically adjust the vehicle trajectory prediction model based on new real-time data and deviation analysis results. Adjustment methods may include adjusting model parameters, adding new training data, improving model structure, etc. Through continuous iteration and optimization, the prediction accuracy and robustness of the model can be improved.

[0166] The predicted trajectory and deviation analysis results are transmitted to the background control terminal and the driver's mobile terminal in real time. The background control terminal can understand the vehicle's driving status and predicted trajectory in real time, and provide decision support for traffic control and accident warning. The driver's mobile terminal can provide the driver with real-time driving assistance information to help him better cope with changes in road conditions ahead.

[0167] Driving assistance information provision: Driving assistance information includes road condition prediction and recommended driving speed. Road condition prediction can help drivers understand the road conditions, traffic flow and other information ahead in advance, so as to make more reasonable driving decisions. The recommended driving speed is calculated based on the current road conditions and predicted trajectory, aiming to help drivers maintain a reasonable driving speed and ensure driving safety.

[0168] The backend control terminal uses the prediction results to control traffic, such as adjusting the timing of traffic lights and publishing traffic congestion information, in order to optimize traffic flow and reduce congestion. At the same time, the backend control terminal can also issue accident warnings based on the predicted trajectory and real-time data, timely identify potential traffic accident risks and take corresponding preventive measures.

[0169] Through the implementation of step S1053, the present invention not only realizes the accurate prediction and verification of the vehicle's driving trajectory, but also effectively applies the prediction results to traffic control and driving assistance, providing strong support for improving the traffic efficiency and safety of highways.

[0170] The present invention provides a vehicle driving trajectory prediction method under highway traffic control based on digital twin tracking simulation, comprising:

[0171] The highway digital twin system is initialized and constructed, including data acquisition module, data analysis module, and virtual simulation module.

[0172] Data collection is the cornerstone of building a digital twin system. This module collects multi-dimensional data such as traffic flow, speed, vehicle type, road conditions (such as slippery, waterlogged), and meteorological conditions (such as temperature, humidity, visibility) in real time through various sensors deployed along the highway (such as traffic flow monitors, vehicle identification cameras, weather stations, road condition sensors, etc.). At the same time, combined with GPS positioning data and vehicle driving trajectories, a comprehensive highway operation data network is formed. The data collection module also needs to ensure the accuracy, real-time and integrity of the data to provide a reliable basis for subsequent analysis.

[0173] The collected data is transmitted to the data center through a high-speed, low-latency network, and a distributed storage architecture is used to ensure data security and scalability. The cloud computing platform is used to achieve rapid processing and efficient management of massive data, providing strong data support for the data analysis module.

[0174] The data analysis module first performs pre-processing such as cleaning, denoising, and format conversion on the received raw data to ensure data quality. Subsequently, feature extraction and dimensionality reduction are performed to extract the key factors that have the greatest impact on the operation of the highway. Using big data analysis and machine learning algorithms, the processed data is deeply mined. Including but not limited to: traffic flow prediction, congestion warning, accident risk assessment, road condition optimization suggestions, etc. Through real-time analysis, the system can promptly identify potential problems and automatically or semi-automatically propose solutions to provide decision support for managers.

[0175] The data analysis results are displayed through advanced visualization tools such as dashboards, heat maps, and three-dimensional models, which intuitively show the operating status, problem areas, and optimization effects of the highway, making it easier for managers to understand and respond quickly.

[0176] Based on high-precision maps and real-time data, a digital twin model of the highway is constructed. The model not only contains geometric information of road infrastructure, but also integrates dynamic data such as traffic flow and environmental factors to achieve a comprehensive mapping of the physical world of the highway. Virtual reality (VR) and augmented reality (AR) technologies are used to simulate and emulate various scenes on the highway, including changes in traffic flow under different weather conditions, emergency response to emergencies, and the implementation effect of traffic control measures.

[0177] The virtual simulation module supports interactive operations between users and digital twin models, such as adjusting parameters, modifying strategies, etc. The system updates the simulation results in real time according to the user's operations and gives feedback.

[0178] Data collection is the first and most important step in the entire vehicle trajectory prediction process. The highway digital twin system can capture multiple types of data in real time, including but not limited to:

[0179] Meteorological environment data: temperature, humidity, wind speed, wind direction, visibility, rainfall, etc. These data have a direct impact on the vehicle's driving speed and safety.

[0180] Road infrastructure data: road surface conditions (such as slipperiness, potholes, cracks), clarity of traffic signs and markings, structural safety of bridges and tunnels, etc. These data reflect the road's capacity and safety.

[0181] Traffic control measures: temporary speed limits, lane closures, traffic guidance information, etc. These measures directly affect the vehicle's driving path and speed.

[0182] Real-time vehicle data: including the target vehicle’s location (latitude and longitude), speed, acceleration, direction, etc. These data are the direct basis for predicting the driving trajectory.

[0183] In addition, it is necessary to obtain historical driving trajectory data from the standardized driving trajectory library. These data record the driving path and speed changes of vehicles under different time, weather, traffic flow, road section structure, and traffic control strategy conditions, providing rich samples for the training of the LSTM model.

[0184] The collected raw data often has problems such as noise, missing values, and outliers, and a series of preprocessing operations are required to ensure the quality and usability of the data. The preprocessing steps include:

[0185] Data cleaning: Remove duplicate data, fill missing values, and identify and eliminate outliers by using interpolation, mean filling, mode filling and other methods.

[0186] Data standardization: Since the dimensions and ranges of different features may vary greatly, directly using them for model training may lead to problems such as slow convergence or gradient vanishing / explosion. Therefore, the data needs to be standardized to scale all features to the same scale.

[0187] Feature extraction and construction: Extract or construct features that are useful for the prediction target from the original data. For example, you can extract features such as temperature and humidity from meteorological environment data; extract features such as road condition coding and traffic sign status from road infrastructure data; extract features such as speed, acceleration, and direction from real-time vehicle data. In addition, you can also construct new features based on the dynamic real-time prediction of new requirements based on vehicle trajectory, such as average speed and acceleration change rate within a time window.

[0188] Data partitioning: Divide the preprocessed data set into training set, validation set and test set. Usually, the training set is used for model training, the validation set is used for model parameter adjustment and to prevent overfitting, and the test set is used to evaluate the final performance of the model.

[0189] Vehicle trajectory data usually contains rich information, such as position, speed, acceleration, direction, etc. In order to apply clustering algorithms more effectively, it is necessary to extract features that have an important impact on trajectory classification from the original data. These features may include trajectory length, shape, speed distribution, turning frequency, etc. Feature selection is to select the most representative feature subset from the extracted features to reduce the amount of calculation and improve the clustering effect.

[0190] Based on preprocessing and feature extraction, clustering algorithm is applied to classify trajectory data. Clustering algorithm is an unsupervised learning method that can group similar data points into multiple clusters. After clustering, the trajectory data in each cluster needs to be standardized to form a standardized driving trajectory database. Standardization includes but is not limited to the following aspects:

[0191] Trajectory alignment: Trajectory data within the same cluster are aligned in time and space to facilitate subsequent comparison and analysis.

[0192] Parametric representation: Convert trajectory data into a series of parametric representations, such as polynomial curves, Bezier curves, etc., for easy storage and query.

[0193] Data compression: While maintaining the main features of the trajectory, the data is compressed to reduce the requirements for storage space and transmission bandwidth.

[0194] Metadata annotation: Add metadata annotations to each standardized trajectory, such as timestamp, weather conditions, road type, etc., to facilitate subsequent data analysis and application.

[0195] Finally, the standardized trajectory data is stored in the database to form a standardized driving trajectory database. The database should support efficient data retrieval, query and update operations to meet the needs of different application scenarios. At the same time, with the continuous collection of new data and the update of old data, the database also needs to be regularly maintained and optimized to ensure the accuracy and timeliness of the data.

[0196] Construction of vehicle trajectory prediction model;

[0197] When applying the LSTM deep neural network algorithm to dynamically predict the target vehicle's trajectory, the following aspects need to be considered:

[0198] Input layer: The input layer receives the preprocessed feature vector sequence. Each feature vector contains multiple feature values ​​(such as temperature, humidity, road structure, control strategy, vehicle speed, acceleration, etc.), and the length of the sequence can be set according to actual needs.

[0199] LSTM layer: The LSTM layer is the core part of the model, responsible for capturing long-term dependencies in sequence data. Multiple LSTM layers can be set as needed to enhance the expressiveness of the model. Each LSTM layer contains a certain number of neurons, which transmit information through a gating mechanism.

[0200] Optional layers: After the LSTM layer, you can add other types of layers to further optimize the model performance. For example, you can add a Dropout layer to prevent overfitting; you can add a Dense (fully connected) layer to extract high-level features and output prediction results; you can also add a BatchNormalization layer to speed up the training process and improve model stability.

[0201] Output layer: The output layer is designed according to the dynamic prediction requirements of vehicle trajectories. In the prediction method of the present invention, the output layer should contain prediction values ​​related to the target trajectory (such as longitude, latitude, speed, etc.). If the trajectory of multiple time steps is predicted, the output layer needs to adjust its structure accordingly to output the prediction values ​​of multiple time steps.

[0202] When building a vehicle trajectory dynamic prediction algorithm model based on the LSTM deep neural network, a series of parameters need to be configured to ensure that the model can be trained and run normally. These parameters include but are not limited to:

[0203] Learning rate: The learning rate determines the step size of the model parameter update during the training process. A learning rate that is too small will cause the training process to be slow or even stagnant; a learning rate that is too large may cause the model to fail to converge to the optimal solution or even diverge. Therefore, the learning rate needs to be adjusted according to the actual situation.

[0204] Optimizer: The optimizer is used to update the parameters of the model to minimize the loss function. Common optimizers include SGD (stochastic gradient descent), Adam (adaptive moment estimation), RMSprop, etc. Each optimizer has its own characteristics and applicable scenarios, and you need to choose the appropriate optimizer based on the actual situation.

[0205] Loss function: The loss function is used to evaluate the degree of difference between the model prediction value and the true value. In the prediction method of the present invention, since the goal is to predict the vehicle's driving trajectory (including information such as position and speed), a regression loss function such as mean square error (MSE) or root mean square error (RMSE) can be selected as an evaluation indicator.

[0206] Epochs: Epochs refers to the number of times the entire training set is traversed. Too many epochs may lead to overfitting of the model; too few epochs may lead to underfitting of the model. Therefore, it is necessary to select the appropriate number of epochs according to the actual situation.

[0207] Batch Size: Batch size refers to the number of samples input into the model at each iteration. A smaller batch size can improve the generalization ability of the model but increase the training time; a larger batch size can reduce the training time but may reduce the generalization ability of the model. Therefore, it is necessary to choose an appropriate batch size based on the actual situation.

[0208] In the model training phase, the training set data needs to be used to train the LSTM model. The training process includes two steps: forward propagation and back propagation:

[0209] Forward propagation: Input the samples in the training set into the model, obtain the predicted value through layer-by-layer calculation, and compare it with the true value to calculate the loss function value.

[0210] Back propagation: Calculate the gradient according to the loss function value, and propagate the gradient back to each layer through the chain rule, updating the parameters of each layer to reduce the loss function value.

[0211] During the training process, the model needs to be validated using the validation set data to evaluate the performance of the model and prevent overfitting. If the performance of the model on the validation set begins to decline (i.e., overfitting occurs), it is necessary to stop training in advance or use other regularization methods (such as Dropout, L1 / L2 regularization, etc.) to prevent overfitting.

[0212] After the model training is completed, the test set data needs to be used to evaluate the model to verify its performance. Evaluation indicators usually include commonly used indicators for classification tasks such as accuracy, recall, F1 score, and commonly used indicators for regression tasks such as mean square error (MSE) and root mean square error (RMSE). In the prediction method of the present invention, since the goal is to predict the vehicle's driving trajectory (belonging to the regression task), MSE or RMSE can be used as the evaluation indicator.

[0213] In addition to quantitative evaluation, qualitative evaluation can also be performed, that is, observing the degree of agreement between the trajectory predicted by the model and the actual trajectory. This helps to more intuitively understand the prediction performance of the model and discover potential problems.

[0214] After obtaining preliminary training results, the model may need to be optimized and adjusted to improve its performance. Further combined with feedback data from actual applications, the trajectory prediction model is continuously optimized to adapt to different traffic control requirements and environmental changes. Optimization and adjustment methods include but are not limited to:

[0215] Adjust the model structure: Try increasing or decreasing the number of LSTM layers, changing the number of hidden units, adding or removing other types of layers, etc. to explore the impact of different structures on model performance.

[0216] Tune hyperparameters: Try different combinations of learning rate, optimizer, loss function, training rounds, and batch size to find the optimal hyperparameter configuration.

[0217] Feature engineering: further mining and constructing useful features to improve the prediction accuracy of the model.

[0218] The real-time collected highway digital twin system data, the target vehicle's driving data, and similar trajectory data extracted from the standardized driving trajectory library are input into the trained LSTM model. The model outputs the prediction result of the target vehicle's future driving trajectory based on the input data.

[0219] Compare the deviation between the prediction result and the actual driving trajectory. If the deviation is large, dynamically adjust the model according to the new real-time data to improve the accuracy of the prediction. According to the prediction results and actual road conditions, timely adjust the traffic control measures to optimize traffic flow and improve road traffic efficiency.

[0220] Consider using map visualization tools to display the predicted vehicle trajectories on a map to visually demonstrate the prediction effect. At the same time, compare the actual driving trajectory with the predicted trajectory to evaluate the prediction accuracy and precision of the model. Establish a feedback mechanism to collect feedback from users (such as drivers, traffic management departments, etc.) on the prediction results to continuously improve and optimize the model.

[0221] The present invention proposes a method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation. Aiming at the problems existing in the current field of highway traffic control, the present invention effectively solves the problems of insufficient prediction accuracy, poor real-time performance, and weak robustness through the following technical means:

[0222] The present invention first collects and integrates basic geographic data, real-time vehicle driving data, traffic environment data, historical driving trajectory data, driver behavior data and traffic control data, and uses these data to build a digital twin model of the highway. This model can fully reflect the physical environment, traffic flow status and dynamic changes of the highway, providing a solid foundation for subsequent trajectory prediction.

[0223] Through the deployment of various sensors and IoT devices along the highway, meteorological and environmental data, road infrastructure data, traffic control measures and real-time vehicle data are captured in real time. These data are the key to reflecting the dynamic characteristics of the traffic system. The captured data is pre-processed by cleaning, denoising, interpolation to fill missing values, and outlier removal to ensure the accuracy and completeness of the data. At the same time, the data is standardized and all features are scaled to the same scale to facilitate subsequent model processing.

[0224] :Extract trajectory length features, shape features, speed distribution features, etc. from the preprocessed data, and perform standardization on the trajectory data, including trajectory alignment, parameterized representation, data compression, and metadata annotation. These processing steps make the trajectory data more standardized and easier to process. The standardized trajectory data is stored in the database to form a standardized driving trajectory library. The library covers vehicle driving trajectories under different time, weather, and traffic flow conditions, providing rich sample data for subsequent trajectory prediction.

[0225] A deep neural network model based on LSTM (Long Short-Term Memory Network) is constructed for vehicle trajectory prediction. LSTM network can capture long-term dependencies in time series data and is suitable for processing data with time characteristics such as traffic flow. The LSTM model is trained using training set data, and the model parameters are updated through forward propagation and back propagation. During the training process, the model continuously learns the complex mapping relationship between the vehicle's historical trajectory and real-time data to improve the prediction accuracy.

[0226] The real-time collected highway digital twin system data, the target vehicle's driving data, and similar trajectory data extracted from the standardized driving trajectory library are input into the LSTM model. The model outputs the prediction result of the target vehicle's future driving trajectory based on the input data. The actual driving trajectory of the target vehicle is collected and the deviation between the prediction result and the actual trajectory is compared. If the deviation is greater than the prediction error range value, the model is dynamically adjusted according to the new real-time data to improve the accuracy of the prediction.

[0227] The prediction results and deviation analysis are transmitted to the backend control terminal and the driver's mobile terminal. The backend control terminal uses the prediction results to support decision-making such as traffic control and accident warning; the driver's mobile terminal provides the driver with driving assistance information, such as road condition prediction ahead and recommended driving speed, to improve driving safety and traffic efficiency.

[0228] In summary, the present invention effectively solves the problems of insufficient prediction accuracy, poor real-time performance, and weak robustness of traditional vehicle driving trajectory prediction methods in the face of complex and changeable traffic environments through the steps of constructing a digital twin model of a highway, real-time data capture and preprocessing, building a standardized driving trajectory library, building and training a deep neural network model based on LSTM, real-time prediction and dynamic adjustment, and result transmission and application.

Claims

1. A method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation, characterized in that: include: Step S101: Acquire basic geographic data, real-time vehicle driving data, traffic environment data, digital twin model data, historical driving trajectory data, driver behavior data, and traffic control data to build a digital twin model of the highway; Step S102: capturing meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data in real time, and pre-processing the meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data; Step S103: extracting features from the preprocessed data to obtain trajectory length features, shape features and speed distribution features, and performing standardization processing on the trajectory data. The standardization processing of the trajectory data includes trajectory alignment, parameterized representation, data compression and metadata annotation. The standardized trajectory data is stored in a database to form a standardized driving trajectory library. Step S104: constructing a vehicle driving trajectory prediction model based on the LSTM deep neural network, using the training set data to train the vehicle driving trajectory prediction model, updating the model parameters through forward propagation and back propagation, and obtaining the vehicle driving trajectory prediction model; Step S105: The real-time collected highway digital twin system data, the target vehicle's driving data, and the similar trajectory data extracted from the standardized driving trajectory library are input into the vehicle driving trajectory prediction model. The model outputs the prediction result of the future driving trajectory of the target vehicle based on the input data, collects the actual driving trajectory of the target vehicle, and compares the deviation between the prediction result of the future driving trajectory of the target vehicle and the actual driving trajectory of the target vehicle. If the deviation is greater than the prediction error range value, the vehicle driving trajectory prediction model is dynamically adjusted according to the new real-time data, and the prediction result is transmitted to the background control terminal and the driver's mobile terminal.

2. The method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation according to claim 1 is characterized in that: The step S101 includes: Step S1011, obtaining geographic location information of the expressway, including road direction, intersection location, number of lanes, slope and curvature, and collecting geographic environment data, such as topography and building distribution; Step S1012: Using sensors installed on the highway, the vehicle's location, speed, acceleration, and direction of travel status information are collected in real time, and data from the vehicle's GPS and vehicle-mounted communication equipment are integrated; Step S101-3: Real-time monitoring and collection of meteorological data, road condition data, such as slippery, damaged, and construction information of the road surface, and real-time traffic condition data of traffic flow and congestion; Step S101-4: using the existing highway digital twin model data, including road models, traffic facility models, and surrounding environment models, as the basis for building a new model, and updating and optimizing the digital twin model based on the real-time collected data; Step S101-5: extracting the driving trajectory of the vehicle on the highway from the historical data, including the vehicle starting point, end point, waypoints, driving time, and speed change; Step S101-6: Collecting the driver's driving behavior data through the vehicle-mounted device or the driver behavior monitoring system; Step S1017: Obtaining traffic control information on the expressway, including speed limits, road closures, and traffic diversion control measures; Step S1018: Integrate and fuse the various types of data collected above, and use advanced modeling techniques and algorithms to build a digital twin model of the highway.

3. The method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation according to claim 1 is characterized in that: The step S102 includes: Step S1021, by deploying meteorological monitoring stations along the expressway, real-time capture of temperature, humidity, wind speed, wind direction, visibility, and rainfall meteorological parameters, and road infrastructure data capture using pavement status sensors and video surveillance camera equipment to obtain real-time pavement conditions, including slipperiness, potholes and cracks, clarity of traffic signs and markings, and structural safety status data of bridges and tunnels; Obtain real-time traffic control measures through the traffic management system, including temporary speed limits, lane closures, and traffic guidance information; Through the GPS positioning system and vehicle identification camera equipment, the location, longitude and latitude, speed, acceleration and direction of the target vehicle are obtained in real time; Step S1022, checking and deleting duplicate records in the data set, using interpolation to fill in missing values ​​in the data set, identifying and eliminating outliers in the data, including unreasonable speed and acceleration data; Step S1023: converting the temperature, speed, and humidity of different dimensions into a unified metric standard, and scaling all feature data to the same scale; Step S1024, extracting trajectory prediction features from the standardized data, the trajectory prediction features including temperature and humidity in meteorological parameters, road surface condition codes in road infrastructure data, and speed and acceleration in real-time vehicle data; Step S1025: Integrate the feature-extracted meteorological environment data, road infrastructure data, traffic control measures and real-time vehicle data, and store the integrated data in a data center or cloud computing platform.

4. The method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation according to claim 1, characterized in that: The step S103 includes: Step S1031, by calculating the sum of distances between all points in the trajectory, the total length of each trajectory is obtained as the basic feature of the trajectory, and the shape feature of the trajectory is extracted. The shape feature of the trajectory includes the curvature and the turning angle of the trajectory; Step S103.2: aligning the starting points and end points of different trajectories into a unified reference coordinate system, converting the trajectory data into a parameterized form, and adding metadata annotations to each trajectory data, wherein the metadata annotations added to each trajectory data include the collection time, location, vehicle type, and driver information of the trajectory; Step S1033: Integrate the standardized trajectory data into the database to form a standardized driving trajectory library.

5. The method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation according to claim 1, characterized in that: The step S104 includes: Step S1041: Determine the number of layers of the LSTM network according to the complexity of the task and the data size; Step S1042: from the standardized driving trajectory library, select a part of the data as a training set and another part as a validation set, divide the trajectory data in the training set into an input sequence and a corresponding output sequence, the input sequence includes the driving data of the vehicle for a period of time in the past, and the output sequence includes the driving data of the vehicle for a period of time in the future, and perform normalization processing on the input data and the output data; Step S104.3: Randomly initialize the weights and biases of the LSTM network, select a loss function to evaluate the prediction error of the model according to the prediction target, select an optimization algorithm to update the parameters of the model to minimize the loss function, input the input data into the LSTM network, calculate the output and loss, calculate the gradient according to the loss function, and update the parameters of the model through the back propagation algorithm, repeat the forward propagation and back propagation process until the predetermined number of training rounds is reached or the loss function converges; Step S1044: Save the trained LSTM model to a file or database.

6. The method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation according to claim 1, characterized in that: The step S105 includes: Step S1051: collect data from the highway digital twin system and the target vehicle in real time, including vehicle position, speed, acceleration, direction, weather conditions, road conditions, and traffic control measures, and integrate the real-time data with similar trajectory data extracted from the standardized driving trajectory library to form a complete input data set; Step S1052: pre-processing the data collected in real time, loading a pre-trained vehicle driving trajectory prediction model based on the LSTM deep neural network, inputting the pre-processed input data into the model, and the model outputting a prediction result of the future driving trajectory of the target vehicle based on the input data; Step S1053: Collect the actual driving trajectory of the target vehicle through GPS, radar, and camera sensor equipment.

7. A method for predicting vehicle driving trajectories under highway traffic control based on digital twin tracking simulation as claimed in claim 6, characterized in that: The step S1053 includes: Calculate the deviation between the predicted trajectory and the actual trajectory, and compare the calculated deviation with the preset prediction error range value. If the deviation is greater than the prediction error range value, dynamically adjust the vehicle driving trajectory prediction model based on the new real-time data and the deviation analysis results; The predicted trajectory and deviation analysis are transmitted to the background control terminal and the driver's mobile terminal. The background control terminal uses the prediction results to perform traffic control and accident warning, and the driver's mobile terminal provides the driver with driving assistance information, including the road condition prediction ahead and the recommended driving speed.

Images