Signal control system and method based on large model

Abstract

The invention discloses a signal control system and method based on a large model, belongs to the technical field of artificial intelligence, and aims to solve the technical problem of how to optimize traffic signal control so as to improve traffic efficiency, reduce traffic congestion, save human resources and reduce operation cost. Comprising an infrastructure and middleware construction module, a data collecting and processing module, a large model training module, a traffic signal timing module and a monitoring feedback module, through analysis of a large model, traffic flow and trend can be intelligently predicted, effective signal control not only reduces waiting time of vehicles, but also reduces unnecessary acceleration and deceleration, so that the traffic flow and trend can be intelligently predicted. The large model has self-learning and adaptive capabilities, can adjust a signal control strategy according to real-time traffic conditions, and adapts to different traffic modes and emergencies.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and more specifically to a signal control system and method based on large models. Background Art

[0002] With the acceleration of the urbanization process, the pressure on the transportation industry has gradually increased. Traffic signal control is an important means to ensure smooth traffic and improve road capacity. Traditional traffic signal control methods mainly rely on manual experience for adjustment, suffering from problems such as slow adjustment speed, low efficiency, and waste of human resources. With the development of artificial intelligence technology, especially the application of large model technology, new solutions have been provided for traffic signal control.

[0003] Existing traffic signal control methods, such as rule-based control and adaptive control, although can improve traffic efficiency to a certain extent, these methods often rely on manually set rules, are unable to handle complex traffic situations, and are difficult to adapt to real-time changing traffic flows. In addition, traditional traffic signal control methods often cannot make full use of a large amount of historical traffic data, thus unable to achieve precise control.

[0004] How to optimize traffic signal control to improve traffic efficiency, reduce traffic congestion, save human resources, and lower operating costs is a technical problem that needs to be solved. Summary of the Invention

[0005] The technical task of the present invention is to address the above deficiencies and provide a signal control system method based on large models to solve the technical problem of how to optimize traffic signal control to improve traffic efficiency, reduce traffic congestion, save human resources, and lower operating costs.

[0006] In a first aspect, a signal control system based on large models of the present invention includes an infrastructure and middleware construction module, a data collection and processing module, a large model training module, a traffic signal timing module, and a monitoring and feedback module;

[0007] The infrastructure and middleware building blocks are used to perform the following: deploy the infrastructure for monitoring and collecting traffic data, deploy one or more servers as data processing centers, deploy a cloud computing platform for training and deploying large machine learning models, deploy middleware in the data processing center and the cloud computing platform, where the middleware is used to provide data preprocessing, model training, model deployment and optimization, and signal control. Through data preprocessing, perform data preprocessing operations on the collected traffic data and extract features from the preprocessed traffic data to obtain traffic features. Through model training, perform model selection, hyperparameter tuning, model training, and model validation on the large machine learning model. Through model deployment and optimization, deploy the trained large machine learning model to the signal control system, and the signal control is used to adjust the timing of traffic lights according to the prediction results of the trained large machine learning model;

[0008] The data collection and processing module is used to perform the following: collect traffic data based on the infrastructure, and perform data preprocessing on the collected traffic data and feature engineering operations on the preprocessed traffic data based on the middleware, and save the preprocessed traffic data and the extracted traffic features to the database;

[0009] The large model training module is used to perform the following: perform model training and model evaluation on the selected large machine learning model based on the middleware to obtain the trained large machine learning model;

[0010] The traffic signal timing module is used to perform the following: use the traffic features and the preprocessed traffic data as inputs, and perform signal timing optimization through the trained large machine learning model to obtain the optimized signal timing plan;

[0011] The monitoring and feedback module is used to perform the following: deploy the optimized signal timing plan, monitor the real-time traffic data and conduct traffic state evaluation, dynamically adjust the signal timing plan according to the traffic state evaluation results to generate a signal timing adjustment plan, and store the real-time traffic data, traffic state evaluation results, and signal timing adjustment plan in the database, and adjust the large machine learning model based on the real-time traffic data, traffic state evaluation results, and signal timing adjustment plan in the database.

[0012] Preferably, the infrastructure includes cameras, radars, and geomagnetic sensors installed on traffic lights;

[0013] Traffic data includes static data and dynamic data. Static data includes maps and traffic light information, and dynamic data includes vehicle speed, vehicle flow, lane occupancy, accident information, weather forecasts, and road conditions.

[0014] Preferably, the data collection and processing module is used to perform the following operations:

[0015] Determine the data source: Determine the collection channels of traffic data. The collection channels include deployed infrastructure, traffic monitoring devices, in-vehicle devices, and pedestrian mobile phones;

[0016] Data collection: Real-time collect traffic data through the determined collection channels and send the traffic data to the data processing center;

[0017] Data preprocessing: Perform data preprocessing on the collected traffic data through middleware deployed in the data processing center and data processing software or scripts deployed in the data processing center. Through data preprocessing, perform data cleaning, data denoising, and format unification on the traffic data, and perform feature engineering operations on the preprocessed traffic data. Remove invalid, incorrect, or duplicate data through data cleaning, remove random interference in the traffic data through data denoising, unify the formats of traffic data from different sources through format unification, and extract information related to signal control from the traffic data as traffic features through feature engineering operations. The traffic features are used as the input of the large machine learning model;

[0018] Data storage management: Save the preprocessed traffic data and the extracted traffic features to the database, and perform data management on the preprocessed traffic data and the extracted traffic features. Provide data query, data update, data deletion operations, data backup, and data recovery operations through data management.

[0019] Preferably, the large model training module is used to perform the following operations:

[0020] Divide the dataset: Randomly divide the dataset composed of traffic features into a training set, a validation set, and a test set. The training set is used to train the large machine learning model, the validation set is used to adjust the model parameters of the large machine learning model, and the test set is used to evaluate the model performance of the large machine learning model;

[0021] Select a model: Select a suitable large machine learning model based on model evaluation factors. The model evaluation factors include the complexity of the model, training time, training resources, and prediction accuracy;

[0022] Model training: Perform model training on the selected large machine learning model based on the training set and the validation set. During the model training process, adjust the model parameters, and adjust the model parameters based on the optimization algorithm of gradient descent;

[0023] Model evaluation: Perform model evaluation on the trained large machine learning model based on the test set. The evaluation metrics include accuracy, recall rate, and FI value, and adjust the model parameters of the large machine learning model based on the evaluation results.

[0024] Preferably, the traffic signal timing module is used to perform the following operations:

[0025] Determine traffic signal timing parameters: Based on the data stored in the database and the training results of the large machine learning model, determine the basic parameters of signal timing, including cycle length, green light time, red light time, and phase difference;

[0026] Calculate the initial signal timing plan: According to the determined timing parameters, calculate the initial traffic signal timing plan, and use the initial traffic signal timing plan as the reference plan;

[0027] Generate an optimized signal timing plan: Use traffic characteristics and preprocessed traffic data as inputs, and generate an optimized signal timing plan through the trained large machine learning model;

[0028] Scheme evaluation and update: Evaluate the advantages and disadvantages of the optimized signal timing plan and the reference plan. The evaluation indicators include traffic congestion level, average vehicle speed, and number of stops. If the optimized signal timing plan is superior to the reference plan in terms of evaluation indicators, apply the optimized signal timing plan to actual traffic signal control; otherwise, adjust the model training parameters and perform iterative optimization on the large machine learning model.

[0029] Preferably, the monitoring and feedback module is used to perform the following operations:

[0030] Traffic state assessment: Monitor real-time traffic data, evaluate the traffic state by comparing historical traffic data and real-time traffic data based on predetermined evaluation indicators, and the predetermined evaluation indicators include congestion index and traffic flow density;

[0031] Signal control effect assessment: According to real-time traffic data and the traffic state assessment results, evaluate the effect of the current optimized signal timing plan, analyze the advantages and disadvantages of the current signal timing plan, and the evaluation indicators are the same as those used in traffic signal timing;

[0032] Dynamically adjust signal timing: According to the evaluation results of the signal timing plan, dynamically adjust signal timing, and the adjustment methods include fixed cycle control, dynamic green wave control, and adaptive control;

[0033] Data storage and analysis: Store real-time traffic data, traffic state assessment results, and signal timing adjustment plans in the database, and regularly mine and analyze traffic data, traffic state assessment results, and signal timing adjustment plans to train the trained large machine learning model.

[0034] In a second aspect, a signal control method based on a large model of the present invention performs signal timing control through a signal control system based on a large model as described in any item of the first aspect, including the following steps:

[0035] Infrastructure and middleware construction: Deploy infrastructure for monitoring and collecting traffic data, deploy one or more servers as data processing centers, deploy a cloud computing platform for training and deploying large machine learning models, and deploy middleware in the data processing center and the cloud computing platform. The middleware is used to provide data preprocessing, model training, model deployment and optimization, and signal control. Through data preprocessing, perform data preprocessing operations on the collected traffic data and extract features from the preprocessed traffic data to obtain traffic features. Through model training, perform model selection, hyperparameter tuning, model training, and model verification on the large machine learning model. Through model deployment and optimization, deploy the trained large machine learning model to the signal control system. The signal control is used to adjust the timing of traffic lights according to the prediction results of the trained large machine learning model;

[0036] Data collection and processing: Collect traffic data based on the infrastructure, and perform data preprocessing on the collected traffic data based on the middleware and perform feature engineering operations on the preprocessed traffic data, and save the preprocessed traffic data and the extracted traffic features to the database;

[0037] Large model training: Perform model training and model evaluation on the selected large machine learning model based on the middleware to obtain the trained large machine learning model;

[0038] Traffic signal timing: Use traffic features and preprocessed traffic data as inputs, and perform signal timing optimization through the trained large machine learning model to obtain an optimized signal timing plan;

[0039] Monitoring and feedback: Deploy the optimized signal timing plan, monitor real-time traffic data and conduct traffic state evaluation, dynamically adjust the signal timing plan according to the traffic state evaluation results to generate a signal timing adjustment plan, and store the real-time traffic data, traffic state evaluation results, and signal timing adjustment plan in the database, and adjust the large machine learning model based on the real-time traffic data, traffic state evaluation results, and signal timing adjustment plan in the database.

[0040] Preferably, the data collection and processing include the following operations:

[0041] Determine the data source: Determine the collection channels of traffic data, and the collection channels include the deployed infrastructure, traffic monitoring devices, in-vehicle devices, and pedestrian mobile phones;

[0042] Data collection: Real-time collect traffic data through the determined collection channels and send the traffic data to the data processing center;

[0043] Data preprocessing: The collected traffic data is preprocessed through middleware deployed in the data processing center and data processing software or scripts deployed in the data processing center. Through data preprocessing, data cleaning, data denoising, and format unification are performed on the traffic data, and feature engineering operations are carried out on the preprocessed traffic data. Invalid, incorrect, or duplicate data is removed through data cleaning, random interference in the traffic data is removed through data denoising, traffic data from different sources is unified in format through format unification, and information related to signal control is extracted from the traffic data as traffic features through feature engineering operations. The traffic features are used as the input of the large machine learning model;

[0044] Data storage management: The preprocessed traffic data and the extracted traffic features are saved to the database, and data management is performed on the preprocessed traffic data and the extracted traffic features. Through data management, data query, data update, data deletion operations, data backup, and data recovery operations are provided.

[0045] Preferably, the large model training includes the following operations:

[0046] Dataset division: The dataset composed based on traffic features is randomly divided into a training set, a validation set, and a test set. The training set is used to train the large machine learning model, the validation set is used to adjust the model parameters of the large machine learning model, and the test set is used to evaluate the model performance of the large machine learning model;

[0047] Model selection: A suitable large machine learning model is selected based on model evaluation factors. The model evaluation factors include the complexity of the model, training time, training resources, and prediction accuracy;

[0048] Model training: The selected large machine learning model is trained based on the training set and the validation set. During the model training process, the model parameters are adjusted, and the parameter adjustment is based on the optimization algorithm of gradient descent when adjusting the model parameters;

[0049] Model evaluation: The trained large machine learning model is evaluated based on the test set. The evaluation metrics include accuracy, recall rate, and FI value, and the model parameters of the large machine learning model are adjusted based on the evaluation results.

[0050] Preferably, the traffic signal timing includes the following operations:

[0051] Determine traffic signal timing parameters: According to the data stored in the database and the training results of the large machine learning model, the basic parameters of signal timing are determined. The basic parameters include cycle length, green light time, red light time, and phase difference;

[0052] Calculate the initial signal timing plan: According to the determined timing parameters, calculate the initial traffic signal timing plan, and use the initial traffic signal timing plan as the reference plan;

[0053] Generate an optimized signal timing plan: Using traffic characteristics and preprocessed traffic data as inputs, generate an optimized signal timing plan through a trained large-scale machine learning model;

[0054] Scheme evaluation and update: Evaluate the advantages and disadvantages of the optimized signal timing plan and the reference plan. The evaluation indicators include traffic congestion level, average vehicle speed, and number of stops. If the optimized signal timing plan is superior to the reference plan in terms of evaluation indicators, apply the optimized signal timing plan to actual traffic signal control; otherwise, adjust the model training parameters and iteratively optimize the large-scale machine learning model;

[0055] Monitoring and feedback include the following operations:

[0056] Traffic status assessment: Monitor real-time traffic data, and evaluate the traffic status by comparing historical traffic data and real-time traffic data based on predetermined evaluation indicators. The predetermined evaluation indicators include congestion index and traffic flow density;

[0057] Signal control effect assessment: Based on real-time traffic data and the traffic status assessment results, evaluate the effect of the current optimized signal timing plan, analyze the advantages and disadvantages of the current signal timing plan. The evaluation indicators are the same as those used in traffic signal timing;

[0058] Dynamically adjust signal timing: According to the evaluation results of the signal timing plan, dynamically adjust the signal timing. The adjustment methods include fixed-cycle control, dynamic green wave control, and adaptive control;

[0059] Data storage and analysis: Store real-time traffic data, traffic status assessment results, and signal timing adjustment plans in the database, and regularly mine and analyze traffic data, traffic status assessment results, and signal timing adjustment plans to train the trained large-scale machine learning model;

[0060] The signal control system method based on a large model of the present invention has the following advantages:

[0061] 1. Improve traffic efficiency: Through the analysis of the large model, it is possible to intelligently predict traffic flow and trends, thereby optimizing signal control strategies, reducing traffic congestion, and improving road traffic efficiency;

[0062] 2. Reduce energy consumption: Effective signal control not only reduces the waiting time of vehicles but also reduces unnecessary acceleration and deceleration, thereby reducing fuel consumption and exhaust emissions, which helps environmental protection;

[0063] 3. Improve safety: By analyzing and processing traffic data in real time, the large model can promptly identify potential safety risks, such as hidden dangers of traffic accidents, and take measures in advance, such as adjusting the signal timing sequence, to enhance the safety of road participants;

[0064] 4. Intelligent adaptation to changes: The large model has the ability of self - learning and adaptation, and can adjust the signal control strategy according to the real - time traffic conditions, adapting to different traffic patterns and emergencies, such as holiday traffic, construction detours, etc.;

[0065] 5. Reduce maintenance costs: Through intelligent signal control, the need for manual intervention can be reduced, and the maintenance and operation costs can be lowered. At the same time, the predictive maintenance of the system can reduce the losses caused by unexpected failures;

[0066] 6. Data - driven decision - making support: The large model processes and analyzes a large amount of traffic data, providing data support for traffic management and decision - making, and helping traffic management departments make more scientific decisions. Description of the Drawings

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following - described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0068] The present invention will be further described below in conjunction with the drawings.

[0069] Figure 1 It is a flowchart of a signal control method based on a large model for Embodiment 2. Detailed Embodiments

[0070] The following further describes the present invention in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the present invention and be able to implement it. However, the embodiments cited are not intended to limit the present invention. Without conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0071] The embodiments of the present invention provide a signal control system and method based on a large model, which are used to solve the technical problem of how to optimize traffic signal control to improve traffic efficiency, reduce traffic congestion, save human resources, and lower operating costs.

[0072] Embodiment 1:

[0073] A signal control system based on a large model of the present invention includes an infrastructure and middleware construction module, a data collection and processing module, a large model training module, a traffic signal timing module, and a monitoring and feedback module.

[0074] The infrastructure and middleware building blocks are used to perform the following: deploy infrastructure for monitoring and collecting traffic data, deploy one or more servers as data processing centers, deploy a cloud computing platform for training and deploying large machine learning models, and deploy middleware in the data processing center and the cloud computing platform. The middleware is used to provide data preprocessing, model training, model deployment and optimization, and signal control. Through data preprocessing, perform data preprocessing operations on the collected traffic data and extract features from the preprocessed traffic data to obtain traffic features. Through model training, perform model selection, hyperparameter tuning, model training, and model validation on the large machine learning model. Through model deployment and optimization, deploy the trained large machine learning model to the signal control system. The signal control is used to adjust the timing of traffic lights according to the prediction results of the trained large machine learning model.

[0075] In this embodiment, the infrastructure includes cameras, radars, and geomagnetic sensors installed on traffic lights. Traffic data includes static data and dynamic data. Static data includes maps and traffic light information, and dynamic data includes vehicle speed, vehicle flow, lane occupancy, accident information, weather forecasts, and road conditions.

[0076] As a specific implementation of the infrastructure and middleware building blocks, this module is used to perform the following operations:

[0077] (1) Data collection part: used to collect traffic data such as traffic flow, vehicle speed, and road conditions. This can be achieved through devices such as cameras, radars, and geomagnetic sensors installed on traffic lights. These devices can monitor traffic conditions in real time and transmit the data to the data processing center;

[0078] (2) Data processing center: used to process and analyze the collected traffic data. One or more servers can be deployed in the data processing center for storing and processing data. Relevant data processing software or scripts can run on the servers to perform preprocessing, analysis, and mining on the collected data;

[0079] (3) Large model training and deployment platform: used to train and deploy large machine learning models. This can be a cloud computing platform that provides powerful computing resources and storage capabilities. On this platform, large models can be trained using historical data and real-time data and deployed to the actual signal control system according to the training results.

[0080] Middleware is various software components and technologies that provide support for implementing the signal control method. These middleware can include the following parts:

[0081] (1) Data preprocessing: Used to preprocess the collected original traffic data, including data cleaning, data transformation, feature extraction, etc. Data preprocessing can be implemented based on programming languages such as Python and Java, and relevant data processing libraries such as NumPy and Pandas are used;

[0082] (2) Model training: Used to train the model. This part can include processes such as model selection, hyperparameter tuning, model training and validation. Training can use deep learning frameworks such as TensorFlow and PyTorch, as well as machine learning libraries such as Scikit-learn;

[0083] (3) Model deployment and optimization: Used to deploy the trained model to the actual signal control system and optimize it according to the actual operation situation. This part can include functions such as model serialization, deserialization, model update and model evaluation. Web frameworks such as Flask and Django can be used to implement the model deployment and optimization part;

[0084] (4) Signal control: Used to automatically adjust the timing of traffic lights according to the prediction results of the machine learning model. This part can generate the optimal traffic signal control strategy according to traffic data and road conditions and implement it by controlling the timing of traffic lights.

[0085] The application of the signal control method based on the large model in the traffic industry requires the construction of a complete set of infrastructure and middleware. The coordinated work of these facilities, components and technologies can achieve intelligent traffic signal control and improve road traffic capacity and safety.

[0086] The data collection and processing module is used to perform the following: collect traffic data based on the infrastructure, perform data preprocessing on the collected traffic data based on the middleware, perform feature engineering operations on the preprocessed traffic data, and save the preprocessed traffic data and the extracted traffic features to the database.

[0087] In this embodiment, the data collection and processing module is used to perform the following operations:

[0088] (1) Determine the data source: Determine the collection channels of traffic data. The collection channels include deployed infrastructure, traffic monitoring devices, in-vehicle devices, and pedestrian mobile phones;

[0089] (2) Data collection: Real-time collect traffic data through the determined collection channels and send the traffic data to the data processing center;

[0090] (3) Data preprocessing: The collected traffic data is preprocessed through middleware deployed in the data processing center and data processing software or scripts deployed in the data processing center. Through data preprocessing, data cleaning, data denoising, and format unification are performed on the traffic data, and feature engineering operations are carried out on the preprocessed traffic data. Invalid, incorrect, or duplicate data is removed through data cleaning, random interference in the traffic data is removed through data denoising, traffic data from different sources is unified in format through format unification, and information related to signal control is extracted from the traffic data as traffic features through feature engineering operations. The traffic features are used as the input to the large machine learning model;

[0091] (4) Data storage and management: The preprocessed traffic data and the extracted traffic features are saved to the database, and data management is performed on the preprocessed traffic data and the extracted traffic features. Through data management, data query, data update, data deletion operations, data backup, and data recovery operations are provided.

[0092] The data collection and processing module is one of the important component modules of the system. Its main purpose is to obtain and process relevant traffic data in order to provide effective input for the signal control method.

[0093] The large model training module is used to perform the following: Based on the middleware, model training and model evaluation are carried out on the selected large machine learning model to obtain the trained large machine learning model. Through the above process of the data collection and processing module, effective data support can be provided for the signal control method based on the large model.

[0094] In this embodiment, the large model training module is used to perform the following operations:

[0095] (1) Divide the dataset: The dataset composed based on traffic features is randomly divided into a training set, a validation set, and a test set. The training set is used to train the large machine learning model, the validation set is used to adjust the model parameters of the large machine learning model, and the test set is used to evaluate the model performance of the large machine learning model;

[0096] (2) Select the model: Based on model evaluation factors, a suitable large machine learning model is selected. The model evaluation factors include the complexity of the model, training time, training resources, and prediction accuracy. Since there are many indicators to be considered in the signal control process and the requirement for model accuracy is relatively high due to safety considerations, theoretically, a large model with a large number of parameters should be used. However, a large model with a large number of parameters requires more training resources and a long training time, which is not conducive to timely adjusting the signal timing of traffic lights. Therefore, a comprehensive consideration should be made, and a large model with a moderate number of parameters, such as Qwen-72B, should be selected;

[0097] (3) Model training: Based on the training set and the validation set, train the selected large-scale machine learning model. During the model training process, adjust the model parameters. When adjusting the model parameters, use optimization algorithms based on gradient descent for parameter adjustment, such as Stochastic Gradient Descent (SGD), Adam, etc.;

[0098] (4) Model evaluation: Based on the test set, evaluate the trained large-scale machine learning model. The evaluation metrics include accuracy, recall rate, and F1 value, and adjust the model parameters of the large-scale machine learning model based on the evaluation results to improve the prediction accuracy.

[0099] The traffic signal timing module is used to perform the following: Take traffic characteristics and preprocessed traffic data as input, and optimize the signal timing through the trained large-scale machine learning model to obtain an optimized signal timing plan.

[0100] In this embodiment, the traffic signal timing module is used to perform the following operations:

[0101] (1) Determine traffic signal timing parameters: According to the data stored in the database and the training results of the large-scale machine learning model, determine the basic parameters of signal timing. The basic parameters include cycle length, green light time, red light time, and phase difference;

[0102] (2) Calculate the initial signal timing plan: According to the determined timing parameters, calculate the initial traffic signal timing plan, and use the initial traffic signal timing plan as the reference plan;

[0103] (3) Generate an optimized signal timing plan: Take traffic characteristics and preprocessed traffic data as input, and generate an optimized signal timing plan through the trained large-scale machine learning model;

[0104] (4) Scheme evaluation and update: Evaluate the advantages and disadvantages of the optimized signal timing plan and the reference plan. The evaluation metrics include traffic congestion degree, average vehicle speed, and number of stops. If the optimized signal timing plan is better than the reference plan in terms of evaluation metrics, then apply the optimized signal timing plan to the actual traffic signal control. Otherwise, adjust the model training parameters and perform iterative optimization on the large-scale machine learning model.

[0105] The traffic signal timing module is the core module of the system. The results of model training directly control the traffic signals through the traffic signal timing module. Through the above method, the characteristics of traffic data and the training results of the large model can be fully utilized to achieve intelligent traffic signal timing, thereby alleviating traffic congestion and improving road traffic efficiency.

[0106] The monitoring feedback module is used to perform the following: deploy optimized signal timing plans, monitor real-time traffic data and conduct traffic status assessments, dynamically adjust signal timing plans based on traffic status assessment results, generate signal timing adjustment plans, and store real-time traffic data, traffic status assessment results, and signal timing adjustment plans in a database; adjust large machine learning models based on real-time traffic data, traffic status assessment results, and signal timing adjustment plans in the database.

[0107] In this embodiment, the monitoring feedback module is used to perform the following operations:

[0108] (1) Traffic status assessment: monitor real-time traffic data, compare historical traffic data with real-time traffic data, and assess traffic status based on predetermined assessment indicators, including congestion index and traffic density;

[0109] (2) Signal control effect evaluation: Based on real-time traffic data and traffic status evaluation results, the effect of the current optimized signal timing scheme is evaluated, and the advantages and disadvantages of the current signal timing scheme are analyzed. The evaluation indicators are consistent with the evaluation indicators used in traffic signal timing.

[0110] (3) Dynamically adjust signal timing: According to the evaluation results of the signal timing scheme, the signal timing is dynamically adjusted. The adjustment methods include fixed cycle control, dynamic green wave control and adaptive control;

[0111] (4) Data storage and analysis: Real-time traffic data, traffic status assessment results, and signal timing adjustment plans are stored in the database, and traffic data, traffic status assessment results, and signal timing adjustment plans are regularly mined and analyzed to train the trained large-scale machine learning model.

[0112] The main purpose of the monitoring and feedback module is to monitor and evaluate the effect of the traffic signal control method, so as to dynamically adjust the traffic signal timing plan and improve traffic flow and safety. Through the above methods, the monitoring and feedback module can monitor the effect of the traffic signal control method in real time, adjust the signal timing in a targeted manner, and continuously optimize the signal control strategy.

[0113] The system of this embodiment adopts large model technology, and can automatically learn and optimize control strategies through training of a large amount of historical traffic data, so as to achieve more accurate and efficient traffic signal control. This system can not only improve traffic efficiency and reduce traffic congestion, but also save human resources and reduce operating costs.

[0114] Embodiment 2:

[0115] The present invention relates to a signal control method based on a large model, which realizes signal timing control through the system disclosed in Embodiment 1. The method includes five steps: infrastructure and middleware construction, data collection and processing, large model training, traffic signal timing, and monitoring and feedback.

[0116] Step S100 Infrastructure and middleware construction: Deploy infrastructure for monitoring and collecting traffic data, deploy one or more servers as a data processing center, deploy a cloud computing platform for training and deploying large machine learning models, and deploy middleware in the data processing center and the cloud computing platform. The middleware is used to provide data preprocessing, model training, model deployment and optimization, and signal control. Through data preprocessing, perform data preprocessing operations on the collected traffic data and extract features from the preprocessed traffic data to obtain traffic features. Through model training, perform model selection, hyperparameter tuning, model training, and model verification on the large machine learning model. Through model deployment and optimization, deploy the trained large machine learning model to the signal control system. The signal control is used to adjust the timing of traffic lights according to the prediction results of the trained large machine learning model.

[0117] In this embodiment, the infrastructure includes cameras, radars, and geomagnetic sensors installed on traffic lights. Traffic data includes static data and dynamic data. Static data includes maps and traffic light information, and dynamic data includes vehicle speed, vehicle flow, lane occupancy, accident information, weather forecast, and road condition information.

[0118] As a specific implementation of infrastructure and middleware construction, this step is used to perform the following operations:

[0119] (1) Data collection part: Collect traffic data such as traffic flow, vehicle speed, and road conditions. This can be achieved through devices such as cameras, radars, and geomagnetic sensors installed on traffic lights. These devices can monitor traffic conditions in real time and transmit the data to the data processing center;

[0120] (2) Data processing center: Process and analyze the collected traffic data. The data processing center can deploy one or more servers to store and process data. Relevant data processing software or scripts can run on the servers to preprocess, analyze, and mine the collected data;

[0121] (3) Large model training and deployment platform: Train and deploy large machine learning models. This can be a cloud computing platform that provides powerful computing resources and storage capabilities. On this platform, large models can be trained using historical data and real-time data and deployed to the actual signal control system according to the training results.

[0122] Middleware refers to various software components and technologies that support the implementation of signal control methods. These middleware can include the following parts:

[0123] (1) Data preprocessing: Preprocess the collected raw traffic data, including data cleaning, data transformation, feature extraction, etc. Data preprocessing can be implemented based on programming languages such as Python and Java, and relevant data processing libraries such as NumPy and Pandas are used;

[0124] (2) Model training: Include processes such as model selection, hyperparameter tuning, model training and validation. Training can use deep learning frameworks such as TensorFlow and PyTorch, as well as machine learning libraries such as Scikit-learn;

[0125] (3) Model deployment and optimization: Deploy the trained model to the actual signal control system and optimize it according to the actual operation situation. This part can include functions such as model serialization, deserialization, model update and model evaluation. Web frameworks such as Flask and Django can be used to implement the model deployment and optimization part;

[0126] (4) Signal control: Automatically adjust the timing of traffic lights according to the prediction results of the machine learning model. This part can generate the optimal traffic signal control strategy based on traffic data and road conditions and implement it by controlling the timing of traffic lights.

[0127] The application of the signal control method based on large models in the transportation industry requires the construction of a complete set of infrastructure and middleware. The coordinated work of these facilities, components and technologies can achieve intelligent traffic signal control and improve road traffic capacity and safety.

[0128] Step S200 Data collection and processing: Collect traffic data based on the infrastructure, perform data preprocessing on the collected traffic data based on the middleware, and perform feature engineering operations on the preprocessed traffic data. Save the preprocessed traffic data and the extracted traffic features to the database.

[0129] In this embodiment, the data collection and processing steps perform the following operations:

[0130] (1) Determine the data source: Determine the collection channels of traffic data. The collection channels include the deployed infrastructure, traffic monitoring devices, in-vehicle devices, and pedestrian mobile phones;

[0131] (2) Data collection: Real-time collect traffic data through the determined collection channels and send the traffic data to the data processing center;

[0132] (3) Data preprocessing: The collected traffic data is preprocessed through middleware deployed in the data processing center and data processing software or scripts deployed in the data processing center. Through data preprocessing, data cleaning, data denoising, and format unification are performed on the traffic data, and feature engineering operations are carried out on the preprocessed traffic data. Invalid, incorrect, or duplicate data is removed through data cleaning, random interference in the traffic data is removed through data denoising, traffic data from different sources is unified in format through format unification, and information related to signal control is extracted from the traffic data as traffic features through feature engineering operations. The traffic features serve as the input to the large machine learning model;

[0133] (4) Data storage and management: The preprocessed traffic data and the extracted traffic features are saved to the database, and data management is performed on the preprocessed traffic data and the extracted traffic features. Through data management, data query, data update, data deletion operations, data backup, and data recovery operations are provided.

[0134] Step S300 Large model training: Based on the middleware, model training and model evaluation are performed on the selected large machine learning model to obtain the trained large machine learning model.

[0135] The large model training in this embodiment includes the following operations:

[0136] (1) Dataset division: The dataset composed based on traffic features is randomly divided into a training set, a validation set, and a test set. The training set is used to train the large machine learning model, the validation set is used to adjust the model parameters of the large machine learning model, and the test set is used to evaluate the model performance of the large machine learning model;

[0137] (2) Model selection: A suitable large machine learning model is selected based on model evaluation factors. The model evaluation factors include model complexity, training time, training resources, and prediction accuracy. Since many indicators need to be considered during signal control and a high requirement for model accuracy is imposed due to safety factors, theoretically, a large model with a large number of parameters should be used. However, a large model with a large number of parameters requires more training resources and a long training time, which is not conducive to timely adjusting the signal timing of traffic lights. Therefore, a comprehensive consideration should be made, and a large model with a moderate number of parameters, such as Qwen-72B, should be selected;

[0138] (3) Model training: Based on the training set and the validation set, model training is performed on the selected large machine learning model. During the model training process, the model parameters are adjusted. When adjusting the model parameters, parameter adjustment is performed based on optimization algorithms such as gradient descent, such as Stochastic Gradient Descent (SGD), Adam, etc.;

[0139] (4) Model evaluation: Based on the test set, model evaluation is carried out on the trained large-scale machine learning model. The evaluation metrics include accuracy, recall rate, and FI value, and the model parameters of the large-scale machine learning model are adjusted based on the evaluation results to improve prediction accuracy.

[0140] Step S400 Traffic signal timing: Using traffic characteristics and preprocessed traffic data as input, signal timing optimization is carried out through the trained large-scale machine learning model to obtain the optimized signal timing plan.

[0141] In this embodiment, traffic signal timing includes the following operations:

[0142] (1) Determine traffic signal timing parameters: According to the data stored in the database and the training results of the large-scale machine learning model, the basic parameters of signal timing are determined. The basic parameters include cycle length, green light time, red light time, and phase difference;

[0143] (2) Calculate the initial signal timing plan: According to the determined timing parameters, calculate the initial traffic signal timing plan, and use the initial traffic signal timing plan as the reference plan;

[0144] (3) Generate the optimized signal timing plan: Using traffic characteristics and preprocessed traffic data as input, generate the optimized signal timing plan through the trained large-scale machine learning model;

[0145] (4) Scheme evaluation and update: Evaluate the advantages and disadvantages of the optimized signal timing plan and the reference plan. The evaluation metrics include traffic congestion level, average vehicle speed, and number of stops. If the optimized signal timing plan is superior to the reference plan in terms of evaluation metrics, the optimized signal timing plan is applied to the actual traffic signal control. Otherwise, adjust the model training parameters and perform iterative optimization on the large-scale machine learning model.

[0146] Step S500 Monitoring and feedback: Deploy the optimized signal timing plan, monitor the real-time traffic data and conduct traffic state evaluation, dynamically adjust the signal timing plan according to the traffic state evaluation results, generate the signal timing adjustment plan, and store the real-time traffic data, traffic state evaluation results, and signal timing adjustment plan in the database. Adjust the large-scale machine learning model based on the real-time traffic data, traffic state evaluation results, and signal timing adjustment plan in the database.

[0147] In this embodiment, monitoring and feedback includes the following operations:

[0148] (1) Traffic state evaluation: Monitor the real-time traffic data, evaluate the traffic state by comparing the historical traffic data and the real-time traffic data based on the predetermined evaluation metrics. The predetermined evaluation metrics include congestion index and traffic flow density;

[0149] (2) Signal control effect evaluation: Based on the real-time traffic data and the traffic status evaluation results, evaluate the effect of the current optimized signal timing plan, analyze the advantages and disadvantages of the current signal timing plan, and the evaluation indicators are the same as those used in traffic signal timing;

[0150] (3) Dynamically adjust signal timing: According to the evaluation results of the signal timing plan, dynamically adjust the signal timing, and the adjustment methods include fixed-cycle control, dynamic green wave control, and adaptive control;

[0151] (4) Data storage and analysis: Store the real-time traffic data, traffic status evaluation results, and signal timing adjustment plan in the database, and regularly mine and analyze the traffic data, traffic status evaluation results, and signal timing adjustment plan to train the trained large machine learning model.

[0152] The method of this embodiment can automatically learn and optimize the control strategy through the training of a large amount of historical traffic data, and achieve more accurate and efficient traffic signal control. This system can not only improve traffic efficiency, reduce traffic congestion, but also save human resources and reduce operating costs.

[0153] The above has introduced in detail the signal control system and method based on a large model provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A signal control system based on a large model, characterized in that: It includes infrastructure and middleware building modules, data collection and processing modules, large model training modules, traffic signal timing modules, and monitoring feedback modules; The infrastructure and middleware building module is used to perform the following: deploy infrastructure for monitoring and collecting traffic data, deploy one or more servers as data processing centers, deploy a cloud computing platform for training and deploying large-scale machine learning models, and deploy middleware in the data processing center and the cloud computing platform. The middleware is used to provide data preprocessing, model training, model deployment and optimization, and signal control. Through data preprocessing, the collected traffic data is subjected to data preprocessing operations, and feature extraction is performed on the preprocessed traffic data to obtain traffic features. Through model training, model selection, hyperparameter tuning, model training, and model verification are performed on the large-scale machine learning model. Through model deployment and optimization, the trained large-scale machine learning model is deployed to the signal control system. The signal control is used to adjust the timing of traffic lights according to the prediction results of the trained large-scale machine learning model. The data collection and processing module is used to perform the following: collect traffic data based on the infrastructure, perform data preprocessing on the collected traffic data based on the middleware, perform feature engineering operations on the preprocessed traffic data, and save the preprocessed traffic data and the extracted traffic features to the database; The large model training module is used to perform the following: perform model training and model evaluation on the selected large machine learning model based on the middleware to obtain a trained large machine learning model; The traffic signal timing module is used to perform the following: using traffic characteristics and pre-processed traffic data as input, optimizing signal timing through a trained large-scale machine learning model, and obtaining an optimized signal timing plan; The monitoring feedback module is used to perform the following: deploy optimized signal timing plans, monitor real-time traffic data and conduct traffic status assessments, dynamically adjust signal timing plans based on traffic status assessment results, generate signal timing adjustment plans, and store real-time traffic data, traffic status assessment results, and signal timing adjustment plans in a database; adjust large machine learning models based on real-time traffic data, traffic status assessment results, and signal timing adjustment plans in the database.

2. The signal control system based on a large model according to claim 1, characterized in that: The infrastructure includes cameras, radars and geomagnetic sensors mounted on traffic lights; Traffic data includes static data and dynamic data. Static data includes map and traffic light information, and dynamic data includes vehicle speed, vehicle flow, lane occupancy, accident information, weather forecast and road condition information.

3. The signal control system based on a large model according to claim 1, characterized in that: The data collection and processing module is used to perform the following operations: Determine the data source: Determine the collection channel of traffic data, which includes the deployed infrastructure and traffic monitoring equipment, vehicle-mounted equipment and pedestrians’ mobile phones; Data collection: collect traffic data in real time through the specified collection channels and send the traffic data to the data processing center; Data preprocessing: The collected traffic data is preprocessed through the middleware and data processing software or scripts deployed in the data processing center. The traffic data is cleaned, denoised and formatted through data preprocessing, and feature engineering operations are performed on the preprocessed traffic data. Invalid, erroneous or duplicate data is removed through data cleaning, random interference in traffic data is removed through data denoising, and traffic data from different sources are unified through format unification. Information related to signal control is extracted from the traffic data as traffic features through feature engineering operations, and traffic features are used as input for large machine learning models. Data storage management: Save the preprocessed traffic data and the extracted traffic features into the database, and perform data management on the preprocessed traffic data and the extracted traffic features, and provide data query, data update, data deletion, data backup and data recovery operations through data management.

4. The signal control system based on a large model according to claim 1, characterized in that: The large model training module is used to perform the following operations: Divide the data set: randomly divide the data set composed based on traffic characteristics into training set, validation set and test set. The training set is used to train the large machine learning model, the validation set is used to adjust the model parameters of the large machine learning model, and the test set is used to evaluate the model performance of the large machine learning model. Model selection: Select an appropriate large-scale machine learning model based on model evaluation factors, including model complexity, training time, training resources, and prediction accuracy; Model training: Perform model training on the selected large-scale machine learning model based on the training set and the validation set. During the model training process, adjust the model parameters. When adjusting the model parameters, perform parameter adjustment based on the gradient descent optimization algorithm. Model evaluation: The trained large-scale machine learning model is evaluated based on the test set. The evaluation indicators include accuracy, recall rate and FI value. The model parameters of the large-scale machine learning model are adjusted based on the evaluation results.

5. The signal control system based on a large model according to claim 1, characterized in that: The traffic signal timing module is used to perform the following operations: Determine traffic signal timing parameters: Based on the data stored in the database and the training results of the large machine learning model, determine the basic parameters of signal timing, including cycle length, green light time, red light time and phase difference; Calculate the initial signal timing plan: Calculate the initial traffic signal timing plan based on the determined timing parameters, and use the initial traffic signal timing plan as the benchmark plan; Generate optimized signal timing plans: Using traffic characteristics and pre-processed traffic data as input, the optimized signal timing plans are generated through the trained large-scale machine learning model; Scheme evaluation update: Evaluate the advantages and disadvantages of the optimized signal timing scheme and the benchmark scheme. The evaluation indicators include traffic congestion level, average vehicle speed and number of stops. If the optimized signal timing scheme is better than the benchmark scheme in terms of evaluation indicators, the optimized signal timing scheme will be applied to the actual traffic signal control. Otherwise, adjust the model training parameters and iteratively optimize the large machine learning model.

6. The signal control system based on a large model according to claim 1, characterized in that: The monitoring feedback module is used to perform the following operations: Traffic status assessment: monitor real-time traffic data, evaluate traffic status based on predetermined assessment indicators by comparing historical traffic data with real-time traffic data, including congestion index and traffic density; Signal control effect evaluation: Based on real-time traffic data and traffic status evaluation results, the effect of the current optimized signal timing scheme is evaluated, and the advantages and disadvantages of the current signal timing scheme are analyzed. The evaluation indicators are consistent with the evaluation indicators used in traffic signal timing; Dynamically adjust signal timing: According to the evaluation results of the signal timing scheme, the signal timing is dynamically adjusted. The adjustment methods include fixed cycle control, dynamic green wave control and adaptive control; Data storage and analysis: Real-time traffic data, traffic status assessment results, and signal timing adjustment plans are stored in the database, and traffic data, traffic status assessment results, and signal timing adjustment plans are regularly mined and analyzed to train the trained large-scale machine learning model.

7. A signal control method based on a large model, characterized in that: The signal timing control is performed by a signal control system based on a large model as described in any one of claims 1 to 6, comprising the following steps: Infrastructure and middleware construction: Deploy infrastructure for monitoring and collecting traffic data, deploy one or more servers as data processing centers, deploy cloud computing platforms for training and deploying large-scale machine learning models, and deploy middleware in the data processing center and cloud computing platform. The middleware is used to provide data preprocessing, model training, model deployment and optimization, and signal control. Through data preprocessing, the collected traffic data is preprocessed and features are extracted from the preprocessed traffic data to obtain traffic features. Through model training, model selection, hyperparameter tuning, model training and model verification are performed on large-scale machine learning models. Through model deployment and optimization, the trained large-scale machine learning models are deployed to the signal control system. Signal control is used to adjust the timing of traffic lights according to the prediction results of the trained large-scale machine learning model. Data collection and processing: Collect traffic data based on the infrastructure, preprocess the collected traffic data based on the middleware, perform feature engineering operations on the preprocessed traffic data, and save the preprocessed traffic data and extracted traffic features to the database; Large model training: Based on the middleware, the selected large machine learning model is trained and evaluated to obtain a large machine learning model after training; Traffic signal timing: Using traffic characteristics and pre-processed traffic data as input, the signal timing is optimized through a trained large-scale machine learning model to obtain an optimized signal timing plan; Monitoring feedback: Deploy optimized signal timing plans, monitor real-time traffic data and conduct traffic status assessments, dynamically adjust signal timing plans based on traffic status assessment results, generate signal timing adjustment plans, and store real-time traffic data, traffic status assessment results, and signal timing adjustment plans in a database. Adjust large-scale machine learning models based on real-time traffic data, traffic status assessment results, and signal timing adjustment plans in the database.

8. The signal control method based on a large model according to claim 7, characterized in that: Data collection and processing includes the following operations: Determine the data source: Determine the collection channel of traffic data, which includes the deployed infrastructure and traffic monitoring equipment, vehicle-mounted equipment and pedestrians’ mobile phones; Data collection: collect traffic data in real time through the specified collection channels and send the traffic data to the data processing center; Data preprocessing: The collected traffic data is preprocessed through the middleware and data processing software or scripts deployed in the data processing center. The traffic data is cleaned, denoised and formatted through data preprocessing, and feature engineering operations are performed on the preprocessed traffic data. Invalid, erroneous or duplicate data is removed through data cleaning, random interference in traffic data is removed through data denoising, and traffic data from different sources are unified through format unification. Information related to signal control is extracted from the traffic data as traffic features through feature engineering operations, and traffic features are used as input for large machine learning models. Data storage management: Save the preprocessed traffic data and the extracted traffic features into the database, and perform data management on the preprocessed traffic data and the extracted traffic features, and provide data query, data update, data deletion, data backup and data recovery operations through data management.

9. The signal control method based on large model according to claim 7, characterized in that: Large model training includes the following operations: Divide the data set: randomly divide the data set composed based on traffic characteristics into training set, validation set and test set. The training set is used to train the large machine learning model, the validation set is used to adjust the model parameters of the large machine learning model, and the test set is used to evaluate the model performance of the large machine learning model. Model selection: Select an appropriate large-scale machine learning model based on model evaluation factors, including model complexity, training time, training resources, and prediction accuracy; Model training: Perform model training on the selected large-scale machine learning model based on the training set and the validation set. During the model training process, adjust the model parameters. When adjusting the model parameters, perform parameter adjustment based on the gradient descent optimization algorithm. Model evaluation: The trained large-scale machine learning model is evaluated based on the test set. The evaluation indicators include accuracy, recall rate and FI value. The model parameters of the large-scale machine learning model are adjusted based on the evaluation results.

10. The signal control method based on large model according to claim 7, characterized in that: Traffic signal timing includes the following operations: Determine traffic signal timing parameters: Based on the data stored in the database and the training results of the large machine learning model, determine the basic parameters of signal timing, including cycle length, green light time, red light time and phase difference; Calculate the initial signal timing plan: Calculate the initial traffic signal timing plan based on the determined timing parameters, and use the initial traffic signal timing plan as the benchmark plan; Generate optimized signal timing plans: Using traffic characteristics and pre-processed traffic data as input, the optimized signal timing plans are generated through the trained large-scale machine learning model; Scheme evaluation and update: Evaluate the advantages and disadvantages of the optimized signal timing scheme and the benchmark scheme. The evaluation indicators include traffic congestion level, average vehicle speed and number of stops. If the optimized signal timing scheme is better than the benchmark scheme in terms of evaluation indicators, the optimized signal timing scheme will be applied to the actual traffic signal control. Otherwise, adjust the model training parameters and iteratively optimize the large machine learning model. Monitoring feedback includes the following operations: Traffic status assessment: monitor real-time traffic data, evaluate traffic status based on predetermined assessment indicators by comparing historical traffic data with real-time traffic data, including congestion index and traffic density; Signal control effect evaluation: Based on real-time traffic data and traffic status evaluation results, the effect of the current optimized signal timing scheme is evaluated, and the advantages and disadvantages of the current signal timing scheme are analyzed. The evaluation indicators are consistent with the evaluation indicators used in traffic signal timing; Dynamically adjust signal timing: According to the evaluation results of the signal timing scheme, the signal timing is dynamically adjusted. The adjustment methods include fixed cycle control, dynamic green wave control and adaptive control; Data storage and analysis: Real-time traffic data, traffic status assessment results, and signal timing adjustment plans are stored in the database, and traffic data, traffic status assessment results, and signal timing adjustment plans are regularly mined and analyzed to train the trained large-scale machine learning model.

Images