Smart city traffic planning method and system based on big data

By setting up traffic sensors in urban road networks, extracting and analyzing multi-dimensional traffic data, building a traffic panoramic database and simulation model, combining deep learning and digital twin technology, the problem of insufficient big data fusion analysis in the existing technology is solved, and the city is realized Comprehensive, dynamic and intelligent management of the transportation system.

CN118917025AActive Publication Date: 2024-11-08HUIZHI HUATONG INFORMATION TECHNOLOGY (BEIJING) CO LTD

Patent Information

Application Number
CN202411002360.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-08
Estimated Expiration
2044-07-25

AI Technical Summary

Technical Problem

In the existing technology, big data-driven traffic planning analysis has not yet formed a mature theoretical method and technical system, and the lack of integrated analysis of multi-source heterogeneous traffic big data has led to the failure to effectively solve the problem of urban traffic planning.

Method used

Set up various types of traffic sensors in urban road networks, extract key feature parameters of multi-dimensional traffic data through edge computing nodes, combine deep learning algorithms to automatically label and semantic description, generate traffic panoramic database, and build a basic model of multi-level traffic simulation , using digital twin technology and knowledge graph for real-time mapping, combining reinforcement learning and game theory algorithms to generate optimization decisions.

Benefits of technology

It realizes comprehensive perception and dynamic management of urban transportation systems, improves the scientificity and comprehensiveness of planning decisions, forms an optimal planning scheme that takes into account the interests of multiple parties, and enhances the robustness and executability of decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917025A_ABST
    Figure CN118917025A_ABST
Patent Text Reader

Abstract

The invention provides a smart city traffic planning method and system based on big data, and relates to the technical field of traffic planning, and the method comprises the steps: setting a traffic sensor, collecting multi-dimensional traffic data, extracting key feature parameters in the multi-dimensional traffic data, carrying out the automatic marking and semantic description, and generating structured traffic events and abnormal condition data. A traffic panorama database is obtained; the method comprises the following steps: constructing a multi-level traffic simulation basic model, constructing a traffic knowledge graph, updating the traffic knowledge graph, generating a simulation traffic digital twin model, performing simulation scene construction and intelligent identification, automatically deducing and calibrating the model, and generating candidate decisions; extracting a planning mode, constructing an alternative planning scheme set, setting a simulation scene parameter space and performing parameter combination, evaluating the candidate decision, determining a first candidate decision set, evaluating the acceptance of each element in the first candidate decision set and determining an optimal candidate decision, generating a task list and performing task deployment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of traffic planning, and in particular to a smart city traffic planning method and system based on big data. Background Art

[0003] In recent years, the new generation of information technology represented by big data, cloud computing, and artificial intelligence has flourished, providing new ideas and new means to solve the problems of urban traffic planning. Big data technology can realize the rapid collection, storage, and processing of massive and heterogeneous traffic data, making it possible to fully perceive the operation of the traffic system. Scholars at home and abroad have carried out big data analysis and application exploration for traffic planning, mainly focusing on traffic flow prediction, travel demand analysis, traffic induction strategies, etc., and have achieved certain research results. However, in general, big data-driven traffic planning analysis is still in its infancy and has not yet formed a mature theoretical method and technical system. Existing research has focused on the application of single data and lacks the fusion analysis of multi-source heterogeneous traffic big data.

[0004] Therefore, a solution is urgently needed to solve the problems existing in the prior art. Summary of the invention

[0005] The embodiments of the present invention provide a smart city traffic planning method and system based on big data, which can at least solve some of the problems existing in the prior art.

[0006] A first aspect of an embodiment of the present invention provides a smart city traffic planning method based on big data, comprising:

[0007] Various types of traffic sensors are set up in the urban road network. The traffic sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time and combine them to obtain multi-dimensional traffic data. The key feature parameters in the multi-dimensional traffic data are extracted through edge computing nodes. The key feature parameters are automatically annotated and semantically described in combination with a fusion deep learning algorithm to generate structured traffic events and abnormal situation data. A traffic panoramic database is obtained in combination with an association fusion algorithm.

[0008] Based on the traffic panorama database, a multi-level traffic simulation basic model is constructed, and a traffic knowledge graph is constructed by combining ontology and semantic web technology. Key information is extracted and the traffic knowledge graph is updated by combining knowledge extraction technology. Real-time mapping is performed by combining digital twin technology to generate a simulated traffic digital twin model. Simulation scenarios are constructed and intelligently identified through a knowledge reasoning engine, and emergency plans are automatically deduced. The simulated traffic digital twin model is dynamically calibrated by combining a reinforcement learning algorithm and candidate decisions are generated.

[0009] Based on the simulated traffic digital twin model, the planning pattern is extracted by combining knowledge graph and case reasoning technology, a set of alternative planning schemes is constructed by combining optimization algorithms and heuristic search strategies, a simulation scenario parameter space is set and parameters are combined by intelligent sampling algorithms, the candidate decisions are evaluated, and a first candidate decision set is determined by combining multi-criteria decision-making and machine learning sorting algorithms. The acceptability of each element in the first candidate decision set is evaluated by combining game theory and swarm intelligence optimization algorithms, and the optimal candidate decision is determined by combining a pre-set acceptance threshold. A task list is generated according to the optimal candidate decision and the task is deployed through a control center.

[0010] In an optional embodiment,

[0011] Various types of traffic sensors are set up in the urban road network. The traffic sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time and combine them to obtain multi-dimensional traffic data. The key feature parameters in the multi-dimensional traffic data are extracted through edge computing nodes. The key feature parameters are automatically annotated and semantically described in combination with the fusion deep learning algorithm to generate structured traffic events and abnormal situation data. The traffic panoramic database obtained in combination with the association fusion algorithm includes:

[0012] Various types of traffic sensors are set up in the urban road network, wherein the traffic sensors include cameras, microwave radar vehicle detectors, lidars and meteorological sensors. Through the simulated annealing algorithm, the temporal and spatial coverage and data quality of traffic information collection are maximized and the sensor setting cost is minimized to obtain the optimal setting scheme of the sensor network. Sensors are set up based on the optimal setting scheme, and edge computing nodes are deployed in combination with the edge computing mode to collect road traffic status, traffic flow data, vehicle speed and weather environment data. Based on the data fusion theory, the collected data is subjected to temporal and spatial registration and consistency verification through Bayesian inference to obtain the multi-dimensional traffic data.

[0013] For the multidimensional traffic data, feature engineering is performed through pre-set edge computing nodes to extract key feature parameters that reflect traffic status and traffic events, wherein the key feature parameters include average vehicle speed, vehicle volume and traffic density. A multimodal deep learning method is used to perform feature understanding and semantic description of complex traffic scenes, automatically annotate traffic events and abnormal situations, and generate semantic annotation results. A semantic description model is generated through knowledge distillation and model compression algorithms and locally deployed through edge computing nodes. The semantic annotation results are semantically understood through the semantic description model. Based on the semantic understanding results, in combination with natural language processing and knowledge graph technology, traffic professional terms are mapped and associated with daily language to obtain the traffic panoramic database.

[0014] In an optional embodiment,

[0015] Based on data fusion theory, the spatiotemporal registration and consistency check are performed through Bayesian inference as shown in the following formula:

[0016] μ′(t)=μ(t)+K x,z ∑ -1 ({z i (t)}-{h i (μ(t))});

[0017] Among them, μ'(t) represents the posterior mean function, μ(t) represents the prior mean function of the state parameter, K x,z represents the covariance matrix between state parameters and observation data, ∑ represents the spatiotemporal covariance matrix of observation data, z i (t) represents the observation data of the i-th sensor, h i (μ(t)) represents the observation value of the i-th sensor when the prior mean μ(t) of the state parameter is given.

[0018] In an optional embodiment,

[0019] Based on the traffic panorama database, a multi-level traffic simulation basic model is constructed, and a traffic knowledge graph is constructed by combining ontology and semantic web technology. Key information is extracted and the traffic knowledge graph is updated by combining knowledge extraction technology. Real-time mapping is performed by combining digital twin technology to generate a simulated traffic digital twin model. Simulation scenarios are constructed and intelligently identified through a knowledge reasoning engine, and emergency plans are automatically deduced. The simulated traffic digital twin model is dynamically calibrated by combining a reinforcement learning algorithm and candidate decisions are generated, including:

[0020] The multi-level traffic simulation basic model is constructed according to the traffic panoramic database, wherein the multi-level traffic simulation basic model includes a micro-simulation model, a macro-simulation model and a meso-simulation model. Traffic participants are modeled by an agent-based modeling method, and the interaction between different traffic participants is simulated in combination with a social force model to obtain the micro-simulation model. Based on the theory of traffic fluid mechanics, the traffic flow is regarded as a continuous medium and a group of partial differential equations for the spatiotemporal evolution of the traffic flow is established. The macro-simulation model is modeled in combination with the compressibility and viscosity of the traffic flow. The road network is abstracted into a service desk and a queuing system by a discrete event simulation method based on queuing theory. The meso-simulation model is modeled in combination with the random arrival of traffic demand and the random distribution of service time. For each traffic area, the macro-simulation model and the micro-simulation model are set through a multi-resolution architecture and the meso-simulation model is used to supplement the blank area to complete model coupling and collaborative solution.

[0021] Define the core concepts, attributes and relationships in the transportation field through ontology engineering methods and form a transportation ontology model layer, build a transportation knowledge description framework based on the transportation ontology model layer, obtain transportation knowledge ontology, and represent transportation entities, attributes and relationships as triples through semantic web technology to form a transportation semantic network and construct a transportation knowledge graph;

[0022] Extract structured, semi-structured and unstructured traffic knowledge from the traffic panorama database through knowledge extraction and knowledge fusion technology, perform semantic matching and knowledge fusion with the existing traffic knowledge graph to update the traffic knowledge graph, and map the real-time data stream in the traffic panorama database with the multi-level traffic simulation basic model in real time in combination with the digital twin technology according to the updated traffic knowledge graph, and generate the simulated traffic digital twin model in combination with the pre-set traffic infrastructure model and traffic environment model;

[0023] Based on the simulated traffic digital twin model, combined with the traffic knowledge graph and the knowledge reasoning engine, a simulation scenario is constructed by generating typical traffic events. Based on the simulation scenario, the emergency plan for traffic events is automatically deduced, and the optimization effects of different decision-making schemes are evaluated. Combined with the reinforcement learning algorithm, the optimal traffic decision-making strategy is learned through interactive exploration of the traffic simulation environment, and the simulated traffic digital twin model is calibrated to generate candidate decisions.

[0024] In an optional embodiment,

[0025] The social force model is combined to simulate the interaction between different traffic participants as shown in the following formula:

[0026]

[0027] Among them, F g represents the net force acting on individual g, represents the driving force of individual g, α gh represents the perception coefficient of individual g to individual h, A represents the first constant parameter, d gh represents the distance between individuals g and h, D represents the expected safe distance between individuals g and h, B represents the second constant parameter, represents the unit vector from individual g to individual h, b represents the boundary, β gb represents the perception coefficient of individual g to boundary b, A b represents the first boundary parameter, d gb represents the distance between individual g and boundary b, D b represents the expected safe distance between individual g and boundary b, B b represents the second boundary parameter, represents the unit vector pointing from individual g to boundary b.

[0028] In an optional embodiment,

[0029] Based on the simulated traffic digital twin model, the planning pattern is extracted by combining knowledge graph and case reasoning technology, a set of alternative planning schemes is constructed by combining optimization algorithm and heuristic search strategy, the simulation scenario parameter space is set and the parameters are combined by intelligent sampling algorithm, the candidate decisions are evaluated, and the first candidate decision set is determined by combining multi-criteria decision and machine learning sorting algorithm, the acceptance of each element in the first candidate decision set is evaluated by combining game theory and swarm intelligence optimization algorithm, and the optimal candidate decision is determined by combining the preset acceptance threshold, and a task list is generated according to the optimal candidate decision and the task is deployed through the control center, including:

[0030] Based on the simulated traffic digital twin model, concepts, entities and relationships in the field of traffic planning are mapped to a low-dimensional vector space through a transfer embedding model to obtain a distributed representation of planning knowledge. Based on the distributed representation, historical traffic planning cases are retrieved, corrected and retained through case similarity calculation and adaptive adjustment in combination with case reasoning technology. Frequent subgraph mining technology is used to determine co-occurring concepts from the traffic planning knowledge graph, extract planning rules and constraints, and obtain the planning model.

[0031] Abstracting the traffic planning problem into a combinatorial optimization problem, determining the decision space through taboo search algorithm and neighborhood search and obtaining the corresponding first alternative planning scheme, solving integer programming and nonlinear programming problems through particle swarm algorithm, generating the second alternative planning scheme, integrating the first alternative planning scheme and the second alternative planning scheme into the set of alternative planning schemes;

[0032] The influence of different parameters on the simulation results is evaluated by the global sensitivity analysis method, and the key parameters are screened. For high-dimensional parameter space, the Latin hypercube sampling algorithm is used to generate sampling points and the parameter combination is modeled as a black box optimization problem. The parameter combination is determined by combining the differential evolution algorithm. For the simulation scenario corresponding to each parameter combination, the adaptability of each alternative planning scheme is determined according to the multi-decision criterion and the machine learning sorting algorithm. The alternative planning schemes are sorted in descending order according to the adaptability, and the top 10% alternative planning schemes are selected as the first candidate decision set;

[0033] For each element in the first candidate decision set, determine the acceptability of each element through the Nash equilibrium strategy, compare the acceptability corresponding to each element with a preset acceptance threshold, and if it is not less than the acceptance threshold, retain the current element, and for each retained element, determine the optimal candidate decision based on information interaction through a swarm intelligence optimization algorithm;

[0034] The optimal candidate decision is decomposed into multiple specific implementation tasks, and the tasks are deployed through a control center according to task relationships and task importance.

[0035] In an optional embodiment,

[0036] The Latin hypercube sampling algorithm generates sampling points as shown in the following formula:

[0037]

[0038] Where X represents the Latin hypercube sampling matrix after random perturbation, k mn It represents the value of the element in the mth row and nth column, s represents the number of factor levels, rand() represents a random number between 0 and 1, and N represents the number of sampling points.

[0039] A second aspect of an embodiment of the present invention provides a smart city traffic planning system based on big data, including:

[0040] The first unit is used to set up various types of traffic sensors in the urban road network, collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time through the traffic sensors and combine them to obtain multi-dimensional traffic data, extract key feature parameters in the multi-dimensional traffic data through edge computing nodes, automatically annotate and semantically describe the key feature parameters in combination with a fusion deep learning algorithm, generate structured traffic events and abnormal situation data, and obtain a traffic panoramic database in combination with an association fusion algorithm;

[0041] The second unit is used to build a multi-level traffic simulation basic model based on the traffic panoramic database, build a traffic knowledge graph in combination with ontology and semantic web technology, extract key information in combination with knowledge extraction technology and update the traffic knowledge graph, and perform real-time mapping in combination with digital twin technology to generate a simulated traffic digital twin model, build simulation scenarios and intelligently identify and automatically deduce emergency plans through a knowledge reasoning engine, and dynamically calibrate the simulated traffic digital twin model in combination with a reinforcement learning algorithm and generate candidate decisions;

[0042] The third unit is used to extract planning patterns based on the simulated traffic digital twin model in combination with knowledge graphs and case-based reasoning technology, construct a set of alternative planning schemes through combinatorial optimization algorithms and heuristic search strategies, set simulation scenario parameter space and combine parameters through intelligent sampling algorithms, evaluate the candidate decisions, determine the first candidate decision set in combination with multi-criteria decision-making and machine learning sorting algorithms, evaluate the acceptability of each element in the first candidate decision set in combination with game theory and swarm intelligence optimization algorithms, and determine the optimal candidate decision in combination with a pre-set acceptance threshold, generate a task list based on the optimal candidate decision, and deploy the tasks through a control center.

[0043] According to a third aspect of the embodiments of the present invention,

[0044] An electronic device is provided, comprising:

[0045] processor;

[0046] a memory for storing processor-executable instructions;

[0047] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0048] A fourth aspect of the embodiments of the present invention is:

[0049] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0050] In the present invention, by deploying multiple types of traffic sensors in the urban road network and using edge computing nodes for data preprocessing and feature extraction, comprehensive perception and data aggregation of the traffic system are achieved, and deep learning algorithms are integrated to automatically label and semantically describe key features, thereby improving the structured degree and interpretability of the data, providing high-quality data support for subsequent modeling analysis and decision optimization. Digital twin technology is introduced to construct a multi-level traffic simulation basic model, which is mapped with the physical world in real time to generate a simulated traffic digital twin model, which can reflect the real-time operating status and dynamic change laws of the traffic system, and provide a realistic virtual environment for traffic simulation, scenario deduction, and emergency decision-making. A combinatorial optimization algorithm and a heuristic search strategy are adopted to perform efficient search and optimization in a set of alternative planning schemes. The parameter space is combined and sampled in combination with an intelligent sampling algorithm, and candidate decisions are evaluated in a simulation environment, thereby improving the scientificity and comprehensiveness of planning decisions. A multi-criteria decision-making method is used to rank and screen candidate schemes, game theory is introduced to analyze the acceptability of alternative schemes, and decision optimization is performed through a swarm intelligence algorithm to form an optimal planning scheme that takes into account the interests of multiple parties and complies with the principle of fairness, thereby enhancing the robustness and executability of the decision. In summary, the present invention achieves the technical effect of comprehensive, dynamic, and intelligent management of urban transportation systems, and has brought significant improvements and optimizations to urban transportation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of a flow chart of a smart city traffic planning method based on big data according to an embodiment of the present invention;

[0052] Figure 2 It is a structural diagram of a smart city traffic planning system based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. 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.

[0054] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0055] Figure 1 FIG. 1 is a flow chart of a smart city traffic planning method based on big data according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0056] S1. Various types of traffic sensors are set up in the urban road network. The traffic sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time and combine them to obtain multi-dimensional traffic data. The key feature parameters in the multi-dimensional traffic data are extracted through edge computing nodes. The key feature parameters are automatically annotated and semantically described in combination with a fusion deep learning algorithm to generate structured traffic events and abnormal situation data. The traffic panoramic database is obtained in combination with an association fusion algorithm;

[0057] The key feature parameters refer to the features or attributes that play a key role in the identification and description of traffic events. In traffic event identification, the key feature parameters may be data such as vehicle speed, vehicle position, acceleration, direction, etc. The automatic labeling refers to the use of machine learning and artificial intelligence technology to automatically add labels or annotations to data. The semantic description refers to the conversion of data and information into a form with semantic understanding capabilities so that computers can understand and process this information. The structured traffic events refer to the organized and hierarchical description and archiving of traffic events so that these events can be easily retrieved, analyzed and utilized. The association fusion algorithm refers to the use of data association analysis technology to associate and fuse information from different data sources. The traffic panorama database is a platform that integrates a large amount of traffic data, including traffic flow, event data, traffic equipment status, etc.

[0058] In an optional embodiment,

[0059] Various types of traffic sensors are set up in the urban road network. The traffic sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time and combine them to obtain multi-dimensional traffic data. The key feature parameters in the multi-dimensional traffic data are extracted through edge computing nodes. The key feature parameters are automatically annotated and semantically described in combination with the fusion deep learning algorithm to generate structured traffic events and abnormal situation data. The traffic panoramic database obtained in combination with the association fusion algorithm includes:

[0060] Various types of traffic sensors are set up in the urban road network, wherein the traffic sensors include cameras, microwave radar vehicle detectors, lidars and meteorological sensors. Through the simulated annealing algorithm, the temporal and spatial coverage and data quality of traffic information collection are maximized and the sensor setting cost is minimized to obtain the optimal setting scheme of the sensor network. Sensors are set up based on the optimal setting scheme, and edge computing nodes are deployed in combination with the edge computing mode to collect road traffic status, traffic flow data, vehicle speed and weather environment data. Based on the data fusion theory, the collected data is subjected to temporal and spatial registration and consistency verification through Bayesian inference to obtain the multi-dimensional traffic data.

[0061] For the multidimensional traffic data, feature engineering is performed through pre-set edge computing nodes to extract key feature parameters that reflect traffic status and traffic events, wherein the key feature parameters include average vehicle speed, vehicle volume and traffic density. A multimodal deep learning method is used to perform feature understanding and semantic description of complex traffic scenes, automatically annotate traffic events and abnormal situations, and generate semantic annotation results. A semantic description model is generated through knowledge distillation and model compression algorithms and locally deployed through edge computing nodes. The semantic annotation results are semantically understood through the semantic description model. Based on the semantic understanding results, in combination with natural language processing and knowledge graph technology, traffic professional terms are mapped and associated with daily language to obtain the traffic panoramic database.

[0062] The spatiotemporal coverage refers to the comprehensiveness and coverage of data collection in space and time. The data quality reflects the performance of the data in terms of accuracy, completeness, consistency, timeliness and credibility. The data fusion theory involves the principles and methods of integrating, merging and analyzing data from different sources and different types. The Bayesian inference is an inference method based on Bayesian statistical theory, which is used to update beliefs about parameters or unknown quantities from observed data. The spatiotemporal registration refers to aligning and matching data from different times and spaces for consistency analysis and integration. The feature engineering refers to the process of extracting relevant features from raw data and preprocessing them. The multimodal deep learning method combines multiple data types (such as images, texts, time series data, etc.) for joint modeling and learning. The model compression algorithm aims to reduce the computing and storage costs of deep learning models while maintaining the performance of the models as much as possible.

[0063] Deploy various types of traffic sensors in the urban road network, including cameras, microwave radar vehicle detectors, lidar and meteorological sensors, define the optimization objective function, consider factors such as spatiotemporal coverage, data quality and setup cost, initialize the parameters of the simulated annealing algorithm, such as initial temperature, cooling rate, termination condition, etc., randomly generate the initial setting scheme of the sensor network, calculate the objective function value of the current setting scheme, perturb the setting scheme according to the current temperature, generate new candidate schemes, calculate the objective function value of the candidate schemes, and compare them with the current scheme, accept the candidate scheme as the new current scheme with a certain probability, reduce the temperature and repeatedly calculate the objective function value until the termination condition is met, and output the optimal sensor network setting scheme;

[0064] According to the optimal setting plan, traffic sensors are deployed in the urban road network. In combination with the edge computing mode, edge computing nodes are deployed near the sensors. The deployed sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data. The collected multi-source heterogeneous data are then subjected to spatiotemporal registration based on data fusion theory and Bayesian inference method. The registered data are then checked for consistency, abnormal and redundant data are eliminated, and the verified data are combined to obtain multi-dimensional traffic data.

[0065] Perform feature engineering on multi-dimensional traffic data on edge computing nodes to extract key feature parameters that reflect traffic status and events, such as average vehicle speed, vehicle volume, and traffic density. Use multimodal deep learning methods to understand and semantically describe complex traffic scenarios, automatically annotate traffic events and abnormal situations, and generate semantic annotation results.

[0066] Apply knowledge distillation and model compression algorithms to generate a lightweight semantic description model, deploy the semantic description model to the edge computing node, realize localized semantic understanding, perform semantic understanding on the semantic annotation results through the deployed semantic description model, extract high-level semantic information of the traffic scene, combine natural language processing and knowledge graph technology, map and associate traffic professional terms with daily language, and obtain the traffic panoramic database.

[0067] In this embodiment, a simulated annealing algorithm is used to optimize the setting scheme of the sensor network. While ensuring the spatiotemporal coverage and data quality, the sensor setting cost is minimized, the economy and feasibility of the sensor network are improved, and an edge computing mode is introduced to deploy edge computing nodes near the sensors to realize localized processing and analysis of data, reduce the computing pressure of the central server, and improve the efficiency and real-time performance of data processing. A multimodal deep learning method is used to understand the features and semantically describe complex traffic scenes, automatically annotate traffic events and abnormal situations, realize intelligent judgment and early warning of traffic status, map and associate traffic professional terms with daily language, build a semantically rich traffic panoramic database, and support high-level semantic query and analysis. In summary, this embodiment provides data support and intelligent means, which is helpful to improve the level of refinement and scientific decision-making ability of urban traffic management.

[0068] In an optional embodiment,

[0069] Based on data fusion theory, the spatiotemporal registration and consistency check are performed through Bayesian inference as shown in the following formula:

[0070] μ′(t)=μ(t)+K x,z ∑ -1 ({z i (t)}-{h i (u(t))});

[0071] Among them, μ'(t) represents the posterior mean function, μ(t) represents the prior mean function of the state parameter, K x,z represents the covariance matrix between state parameters and observation data, ∑ represents the spatiotemporal covariance matrix of observation data, z i (t) represents the observation data of the i-th sensor, h i (μ(t)) represents the observation value of the i-th sensor when the prior mean μ(t) of the state parameter is given.

[0072] In this embodiment, by considering the spatiotemporal correlation between state parameters and observation data, the observation data of different sensors are mapped to the same spatiotemporal reference system to achieve spatiotemporal alignment of data, which helps to eliminate the spatiotemporal deviations between data and improve the spatiotemporal consistency of data. By introducing the spatiotemporal covariance matrix of the observation data, the observation noise can be effectively suppressed, and the robustness and reliability of the data fusion results can be improved. By introducing the prior mean function and the posterior mean function of the state parameters, it can adapt to changes in traffic conditions and achieve adaptability and real-time performance. In summary, this embodiment can effectively solve the problems of spatiotemporal alignment, noise suppression, information completion, dynamic updating, uncertainty quantification, etc. of traffic sensor data, improve the accuracy, reliability and adaptability of data fusion, and provide a high-quality data foundation for subsequent traffic state estimation, prediction and decision-making, which is of great significance for improving the performance and efficiency of intelligent transportation systems.

[0073] S2. Based on the traffic panorama database, a multi-level traffic simulation basic model is constructed, and a traffic knowledge graph is constructed by combining ontology and semantic web technology. Key information is extracted and the traffic knowledge graph is updated by combining knowledge extraction technology. Real-time mapping is performed by combining digital twin technology to generate a simulated traffic digital twin model. Simulation scenarios are constructed and intelligently identified through a knowledge reasoning engine, and emergency plans are automatically deduced. The simulated traffic digital twin model is dynamically calibrated by combining a reinforcement learning algorithm and candidate decisions are generated.

[0074] The multi-level traffic simulation basic model refers to a model for modeling and simulating the traffic system at different levels and granularities. The ontology is a theoretical basis for knowledge representation and management, which is used to describe the structured form of domain knowledge. The semantic web technology refers to the use of semantic description language and semantic reasoning methods to construct and organize information resources on the Internet. The knowledge extraction technology is a technology for automatically extracting structured knowledge or information from unstructured or semi-structured data. The digital twin technology refers to a technology that simulates, optimizes and predicts by establishing a virtual model of a physical system or process and reflecting the state and behavior of the physical system in real time. The knowledge reasoning engine refers to a system or engine that uses logical reasoning, rule reasoning, statistical reasoning and other technologies to infer and interpret knowledge. The reinforcement learning algorithm is a type of machine learning algorithm that achieves the goal of maximizing the expected cumulative reward through interactive learning between the intelligent agent and the environment.

[0075] In an optional embodiment,

[0076] Based on the traffic panorama database, a multi-level traffic simulation basic model is constructed, and a traffic knowledge graph is constructed by combining ontology and semantic web technology. Key information is extracted and the traffic knowledge graph is updated by combining knowledge extraction technology. Real-time mapping is performed by combining digital twin technology to generate a simulated traffic digital twin model. Simulation scenarios are constructed and intelligently identified through a knowledge reasoning engine, and emergency plans are automatically deduced. The simulated traffic digital twin model is dynamically calibrated by combining a reinforcement learning algorithm and candidate decisions are generated, including:

[0077] The multi-level traffic simulation basic model is constructed according to the traffic panoramic database, wherein the multi-level traffic simulation basic model includes a micro-simulation model, a macro-simulation model and a meso-simulation model. Traffic participants are modeled by an agent-based modeling method, and the interaction between different traffic participants is simulated in combination with a social force model to obtain the micro-simulation model. Based on the theory of traffic fluid mechanics, the traffic flow is regarded as a continuous medium and a group of partial differential equations for the spatiotemporal evolution of the traffic flow is established. The macro-simulation model is modeled in combination with the compressibility and viscosity of the traffic flow. The road network is abstracted into a service desk and a queuing system by a discrete event simulation method based on queuing theory. The meso-simulation model is modeled in combination with the random arrival of traffic demand and the random distribution of service time. For each traffic area, the macro-simulation model and the micro-simulation model are set through a multi-resolution architecture and the meso-simulation model is used to supplement the blank area to complete model coupling and collaborative solution.

[0078] Define the core concepts, attributes and relationships in the transportation field through ontology engineering methods and form a transportation ontology model layer, build a transportation knowledge description framework based on the transportation ontology model layer, obtain transportation knowledge ontology, and represent transportation entities, attributes and relationships as triples through semantic web technology to form a transportation semantic network and construct a transportation knowledge graph;

[0079] Extract structured, semi-structured and unstructured traffic knowledge from the traffic panorama database through knowledge extraction and knowledge fusion technology, perform semantic matching and knowledge fusion with the existing traffic knowledge graph to update the traffic knowledge graph, and map the real-time data stream in the traffic panorama database with the multi-level traffic simulation basic model in real time in combination with the digital twin technology according to the updated traffic knowledge graph, and generate the simulated traffic digital twin model in combination with the pre-set traffic infrastructure model and traffic environment model;

[0080] Based on the simulated traffic digital twin model, combined with the traffic knowledge graph and the knowledge reasoning engine, a simulation scenario is constructed by generating typical traffic events. Based on the simulation scenario, the emergency plan for traffic events is automatically deduced, and the optimization effects of different decision-making schemes are evaluated. Combined with the reinforcement learning algorithm, the optimal traffic decision-making strategy is learned through interactive exploration of the traffic simulation environment, and the simulated traffic digital twin model is calibrated to generate candidate decisions.

[0081] The social force model is used to describe the interactions and behavioral patterns between individuals (such as vehicles, pedestrians, etc.), and is derived from theories in sociology. In the field of transportation, the social force model is used to simulate the interactive behaviors between traffic participants, such as rear-end collisions between vehicles and clustering behaviors between pedestrians. The traffic fluid mechanics theory is a scientific theory that studies traffic flow behavior, which is similar to the fluid mechanics theory in physics, but is used to describe and analyze the movement and interactive behaviors of participants such as vehicles and pedestrians in the traffic system. The discrete event simulation method based on queuing theory is a simulation method that analyzes the operation of the system by establishing a queuing model and simulating the queuing behavior of individuals in the system and the order of occurrence of events. The ontology model layer refers to the hierarchical structure or organization of the ontology in the knowledge representation. The typical traffic events refer to common or important event types in the traffic system, such as traffic accidents, traffic congestion, traffic signal failures, etc.

[0082] Based on the traffic panorama database, a multi-level traffic simulation basic model including micro-simulation model, macro-simulation model and meso-simulation model is constructed. The intelligent agent-based modeling method is used to model traffic participants (such as vehicles, pedestrians, etc.). The social force model is combined to simulate the interaction between different traffic participants to construct a micro-simulation model. Based on the traffic fluid mechanics theory, the traffic flow is regarded as a continuous medium, and a group of partial differential equations describing the spatiotemporal evolution of the traffic flow is established. The macro-simulation model is modeled in combination with the compressibility and viscosity characteristics of the traffic flow. The discrete event simulation method based on queuing theory is used to abstract the road network into a service desk and a queuing system. The meso-simulation model is modeled considering the random arrival of traffic demand and the random distribution of service time. For each traffic area, a macro-simulation model and a micro-simulation model are set up through a multi-resolution architecture, and the meso-simulation model is used to supplement the blank area to achieve coupling and collaborative solution between models.

[0083] Adopting ontology engineering method, define the core concepts, attributes and relationships in the field of transportation, form the transportation ontology model layer, build the transportation knowledge description framework based on the transportation ontology model layer, obtain the transportation knowledge ontology, apply semantic web technology, represent transportation entities, attributes and relationships as triples, form transportation semantic network, build transportation knowledge graph, and represent the knowledge structure and semantic relationship in the field of transportation;

[0084] Use knowledge extraction and knowledge fusion technology to extract structured, semi-structured and unstructured traffic knowledge from the traffic panorama database, perform semantic matching and knowledge fusion on the extracted traffic knowledge and the existing traffic knowledge graph, update and expand the traffic knowledge graph based on the matching and fusion results, and map the real-time data stream in the traffic panorama database with the multi-level traffic simulation basic model in real time based on the updated traffic knowledge graph in combination with the digital twin technology, combine the pre-set traffic infrastructure model and traffic environment model, and generate a simulated traffic digital twin model to realize the digital representation and simulation of the real traffic system;

[0085] Based on the simulated traffic digital twin model, combined with the traffic knowledge graph and knowledge reasoning engine, by generating typical traffic incidents, constructing simulation scenarios, and automatically deducing emergency plans for traffic incidents based on the simulation scenarios, the optimization effects of different decision-making schemes are evaluated at the same time. Combined with the reinforcement learning algorithm, the optimal traffic decision-making strategy is learned through interactive exploration of the traffic simulation environment, the simulated traffic digital twin model is calibrated, and candidate decision schemes are generated.

[0086] In this embodiment, by constructing traffic simulation models at the micro, macro and meso levels and realizing coupling and collaborative solving between models, the traffic system can be simulated in detail at different scales and granularities, providing comprehensive and accurate traffic status analysis and prediction capabilities. By constructing a traffic knowledge ontology and knowledge graph, the core concepts, attributes and relationships in the traffic field are represented and associated in a structured manner to form a knowledge base rich in semantic information, which provides a solid knowledge foundation for the understanding, integration and application of traffic data. Through knowledge extraction and knowledge fusion technology, traffic knowledge is continuously extracted and updated from the traffic panoramic database and mapped with the simulation model in real time. and calibration, to achieve dynamic updating of knowledge-driven traffic simulation models, improve the accuracy and timeliness of simulation models, based on the simulated traffic digital twin model, combined with knowledge reasoning and reinforcement learning algorithms, to achieve simulation deduction of typical traffic events and automatic evaluation of emergency plans, through interactive exploration of the traffic simulation environment, learn the optimal traffic decision-making strategy, generate intelligent candidate decision-making solutions, and provide scientific and efficient decision-making support for traffic managers. In summary, this embodiment constructs a traffic simulation and decision optimization system with complete functions, high performance, intelligent and flexible, which has good technical innovation and practical value, and is of great significance to promoting the development and application of smart transportation.

[0087] In an optional embodiment,

[0088] The social force model is combined to simulate the interaction between different traffic participants as shown in the following formula:

[0089]

[0090] Among them, F g represents the net force acting on individual g, represents the driving force of individual g, α gh represents the perception coefficient of individual g to individual h, A represents the first constant parameter, d gh represents the distance between individuals g and h, D represents the expected safe distance between individuals g and h, B represents the second constant parameter, represents the unit vector from individual g to individual h, b represents the boundary, β gb represents the perception coefficient of individual g to boundary b, A b represents the first boundary parameter, d gb represents the distance between individual g and boundary b, D b represents the expected safe distance between individual g and boundary b, B b represents the second boundary parameter, represents the unit vector pointing from individual g to boundary b.

[0091] In this embodiment, the interaction forces between individuals and between individuals and boundaries are taken into account. By introducing parameters such as driving force, perception coefficient, and safe distance, the behavioral characteristics of different traffic participants can be finely characterized, such as following, avoiding, and pulling over, making the micro-simulation model closer to the real traffic scene. Through the summation term in the formula, the social force model can calculate the comprehensive force exerted on an individual by other individuals and boundaries, and dynamically update the individual's motion state. It can effectively simulate the real-time interaction and response between traffic participants, improve the authenticity and accuracy of the micro-simulation, and transform the motion decisions of traffic participants into force analysis and kinematic calculation problems, giving the micro-simulation model a clear physical meaning and explanation, and enhancing the rationality and credibility of the simulation results. In summary, this embodiment can significantly improve the modeling accuracy of traffic participants' behaviors and the reliability of simulation results, laying an important foundation for building a more realistic and intelligent traffic simulation system.

[0092] S3. Based on the simulated traffic digital twin model, the planning pattern is extracted by combining knowledge graph and case reasoning technology, a set of alternative planning schemes is constructed through combinatorial optimization algorithms and heuristic search strategies, the simulation scenario parameter space is set and the parameters are combined through intelligent sampling algorithms, the candidate decisions are evaluated, and the first candidate decision set is determined by combining multi-criteria decision-making and machine learning sorting algorithms. The acceptance of each element in the first candidate decision set is evaluated by combining game theory and swarm intelligence optimization algorithms, and the optimal candidate decision is determined by combining a pre-set acceptance threshold, and a task list is generated according to the optimal candidate decision, and task deployment is performed through the control center.

[0093] The case-based reasoning technology is an inference method based on past experience cases, which solves new problems through analogy, comparison and analogy. The combinatorial optimization algorithm focuses on solving combinatorial optimization problems, that is, finding the optimal solution under limited resources and constraints. The heuristic search strategy is a method for finding solutions. It evaluates and guides the search direction through heuristic functions, and continuously adjusts the current solution to find a better solution. The intelligent sampling algorithm is a class of algorithms for efficiently generating data samples, usually based on model generation methods or optimization search methods to better explore data distribution and generate representative sample sets. The multi-criteria decision-making involves decision-making problems that consider multiple decision criteria or goals at the same time. The machine learning sorting algorithm aims to sort items according to the characteristics and context of the data. The game theory studies the strategic interactions and influences between decision-makers, as well as the mathematical theories and models for reaching the optimal decision under given rules. The swarm intelligence optimization algorithm uses group wisdom and collective behavior principles, such as ant colony algorithms, particle swarm algorithms, etc., to solve complex optimization problems and simulate the collaborative behavior of groups in natural systems.

[0094] In an optional embodiment,

[0095] Based on the simulated traffic digital twin model, the planning pattern is extracted by combining knowledge graph and case reasoning technology, a set of alternative planning schemes is constructed by combining optimization algorithm and heuristic search strategy, the simulation scenario parameter space is set and the parameters are combined by intelligent sampling algorithm, the candidate decisions are evaluated, and the first candidate decision set is determined by combining multi-criteria decision and machine learning sorting algorithm, the acceptance of each element in the first candidate decision set is evaluated by combining game theory and swarm intelligence optimization algorithm, and the optimal candidate decision is determined by combining the preset acceptance threshold, and a task list is generated according to the optimal candidate decision and the task is deployed through the control center, including:

[0096] Based on the simulated traffic digital twin model, concepts, entities and relationships in the field of traffic planning are mapped to a low-dimensional vector space through a transfer embedding model to obtain a distributed representation of planning knowledge. Based on the distributed representation, historical traffic planning cases are retrieved, corrected and retained through case similarity calculation and adaptive adjustment in combination with case reasoning technology. Frequent subgraph mining technology is used to determine co-occurring concepts from the traffic planning knowledge graph, extract planning rules and constraints, and obtain the planning model.

[0097] Abstracting the traffic planning problem into a combinatorial optimization problem, determining the decision space through taboo search algorithm and neighborhood search and obtaining the corresponding first alternative planning scheme, solving integer programming and nonlinear programming problems through particle swarm algorithm, generating the second alternative planning scheme, integrating the first alternative planning scheme and the second alternative planning scheme into the set of alternative planning schemes;

[0098] The influence of different parameters on the simulation results is evaluated by the global sensitivity analysis method, and the key parameters are screened. For high-dimensional parameter space, the Latin hypercube sampling algorithm is used to generate sampling points and the parameter combination is modeled as a black box optimization problem. The parameter combination is determined by combining the differential evolution algorithm. For the simulation scenario corresponding to each parameter combination, the adaptability of each alternative planning scheme is determined according to the multi-decision criterion and the machine learning sorting algorithm. The alternative planning schemes are sorted in descending order according to the adaptability, and the top 10% alternative planning schemes are selected as the first candidate decision set;

[0099] For each element in the first candidate decision set, determine the acceptability of each element through the Nash equilibrium strategy, compare the acceptability corresponding to each element with a preset acceptance threshold, and if it is not less than the acceptance threshold, retain the current element, and for each retained element, determine the optimal candidate decision based on information interaction through a swarm intelligence optimization algorithm;

[0100] The optimal candidate decision is decomposed into multiple specific implementation tasks, and the tasks are deployed through a control center according to task relationships and task importance.

[0101] The transfer embedding model is a technology for learning relationships between entities. It represents entities and relationships through embedding vectors to capture the semantic associations between them. The frequent subgraph mining technology is used to discover subgraph patterns that frequently appear in graph datasets. The co-occurrence concept refers to two or more concepts or events that appear at the same time in a given context. The taboo search algorithm is a heuristic search method for solving optimization problems. It avoids local optimal solutions in the search process by maintaining a taboo table so as to more comprehensively explore the solution space. The global sensitivity analysis method is used to evaluate the responsiveness of the model output to changes in input parameters. The Latin hypercube sampling algorithm is a method for designing experiments and sampling parameter space to ensure uniform distribution and efficiency of samples. The Nash equilibrium strategy refers to a solution in game theory, which means that under the strategies of multiple participants, each participant cannot improve its own interests by changing the strategy alone.

[0102] Through the transfer embedding model, the concepts, entities and relationships in the field of transportation planning are mapped to a low-dimensional vector space to obtain a distributed representation of planning knowledge and form a vectorized representation of planning knowledge. Based on the distributed representation of planning knowledge, historical transportation planning cases are retrieved, revised and retained through case similarity calculation and adaptive adjustment. Combined with frequent subgraph mining technology, co-occurring concepts are determined from the transportation planning knowledge graph, and planning rules and constraints are extracted to form a transportation planning model.

[0103] Abstract the traffic planning problem into a combinatorial optimization problem, determine the decision space and obtain the corresponding first alternative planning scheme through tabu search algorithm and neighborhood search, integrate the first alternative planning scheme and the second alternative planning scheme to obtain a set of alternative planning schemes;

[0104] A global sensitivity analysis method is used to evaluate the influence of different parameters on the simulation results and screen out key parameters. For high-dimensional parameter space, sampling points are generated by the Latin hypercube sampling algorithm. The parameter combination is modeled as a black box optimization problem. The optimal parameter combination is determined by combining the differential evolution algorithm. For each simulation scenario corresponding to each parameter combination, the adaptability of each alternative planning scheme is determined based on multiple decision criteria and machine learning sorting algorithms. The alternative planning schemes are sorted in descending order according to their adaptability, and the alternative planning schemes in the top 10% are selected as the first candidate decision set.

[0105] For each element in the first candidate decision set, its acceptability is determined by the Nash equilibrium strategy, the acceptability of each element is compared with the preset acceptance threshold, and the elements whose acceptability is not less than the threshold are retained. For each retained element, the optimal candidate decision is determined based on information interaction through the swarm intelligence optimization algorithm;

[0106] The optimal candidate decision is broken down into multiple specific implementation tasks. Tasks are deployed through the control center based on task relationships and task importance. Each execution entity carries out corresponding traffic planning implementation work according to task allocation and scheduling.

[0107] In this embodiment, by constructing a distributed representation of planning knowledge, combining case-based reasoning and frequent subgraph mining technology, making full use of historical planning experience and domain knowledge, extracting effective planning patterns and constraints, and laying the foundation for generating reasonable and feasible traffic planning schemes, introducing simulation scenario analysis and sensitivity analysis technology, evaluating the adaptability of schemes for different parameter combinations, so that planning decisions can dynamically adapt to complex and changing traffic environments, using intelligent optimization algorithms to automatically evaluate and screen alternative schemes, combining Nash equilibrium game strategies to analyze the acceptability of schemes, and further optimizing schemes through group intelligence, ultimately obtaining the optimal candidate decision, greatly improving The efficiency and quality of planning decisions are improved by integrating planning decisions with simulated traffic digital twin models, realizing real-time testing and dynamic adjustment of planning schemes in a virtual environment, supporting refined management and real-time optimization control of traffic systems, and improving the effectiveness and flexibility of traffic planning. In summary, this embodiment has constructed an integrated, automated, and intelligent planning decision-making system, which has demonstrated significant technical effects in improving the efficiency and quality of planning decisions, supporting multi-objective trade-offs and balanced decisions, and enabling refined traffic management. It provides innovative solutions for urban traffic planning and management in the era of smart transportation, and is of great significance to promoting the progress of traffic planning theory and practice.

[0108] In an optional embodiment,

[0109] The Latin hypercube sampling algorithm generates sampling points as shown in the following formula:

[0110]

[0111] Where X represents the Latin hypercube sampling matrix after random perturbation, k mn It represents the value of the element in the mth row and nth column, s represents the number of factor levels, rand() represents a random number between 0 and 1, and N represents the number of sampling points.

[0112] In this embodiment, by dividing the value range of each factor into multiple levels and ensuring that each level appears only once, the uniform distribution of sampling points in each factor dimension is guaranteed. At the same time, while ensuring the sampling quality, the number of sampling points is significantly reduced. The sampling points generated by Latin hypercube sampling have multiple different levels of value combinations in each factor dimension, which can better describe the uncertainty distribution of the parameters. The sampling points generated by Latin hypercube sampling are evenly distributed and cover different areas of the parameter space. They can be used as the initial population or candidate solution set of the intelligent optimization algorithm. The high-quality initial solution helps to accelerate the convergence speed of the optimization algorithm, improve the optimization effect, and reduce the risk of the algorithm falling into the local optimum. In summary, this embodiment can obtain representative sampling points in the parameter space at a low computing cost, lay a solid data foundation for subsequent steps, and has broad application prospects in the field of parameter sampling and optimization design.

[0113] Figure 2 FIG. 1 is a schematic diagram of the structure of a smart city traffic planning system based on big data according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0114] The first unit is used to set up various types of traffic sensors in the urban road network, collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time through the traffic sensors and combine them to obtain multi-dimensional traffic data, extract key feature parameters in the multi-dimensional traffic data through edge computing nodes, automatically annotate and semantically describe the key feature parameters in combination with a fusion deep learning algorithm, generate structured traffic events and abnormal situation data, and obtain a traffic panoramic database in combination with an association fusion algorithm;

[0115] The second unit is used to build a multi-level traffic simulation basic model based on the traffic panoramic database, build a traffic knowledge graph in combination with ontology and semantic web technology, extract key information in combination with knowledge extraction technology and update the traffic knowledge graph, and perform real-time mapping in combination with digital twin technology to generate a simulated traffic digital twin model, build simulation scenarios and intelligently identify and automatically deduce emergency plans through a knowledge reasoning engine, and dynamically calibrate the simulated traffic digital twin model in combination with a reinforcement learning algorithm and generate candidate decisions;

[0116] The third unit is used to extract planning patterns based on the simulated traffic digital twin model in combination with knowledge graphs and case-based reasoning technology, construct a set of alternative planning schemes through combinatorial optimization algorithms and heuristic search strategies, set simulation scenario parameter space and combine parameters through intelligent sampling algorithms, evaluate the candidate decisions, determine the first candidate decision set in combination with multi-criteria decision-making and machine learning sorting algorithms, evaluate the acceptability of each element in the first candidate decision set in combination with game theory and swarm intelligence optimization algorithms, and determine the optimal candidate decision in combination with a pre-set acceptance threshold, generate a task list based on the optimal candidate decision, and deploy the tasks through a control center.

[0117] According to a third aspect of the embodiments of the present invention,

[0118] An electronic device is provided, comprising:

[0119] processor;

[0120] a memory for storing processor-executable instructions;

[0121] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0122] A fourth aspect of the embodiments of the present invention is:

[0123] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0124] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. Smart city traffic planning method based on big data, It is characterized in that include: Various types of traffic sensors are set up in the urban road network. The traffic sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time and combine them to obtain multi-dimensional traffic data. The key feature parameters in the multi-dimensional traffic data are extracted through edge computing nodes. The key feature parameters are automatically annotated and semantically described in combination with a fusion deep learning algorithm to generate structured traffic events and abnormal situation data. A traffic panoramic database is obtained in combination with an association fusion algorithm. Based on the traffic panorama database, a multi-level traffic simulation basic model is constructed, and a traffic knowledge graph is constructed by combining ontology and semantic web technology. Key information is extracted and the traffic knowledge graph is updated by combining knowledge extraction technology. Real-time mapping is performed by combining digital twin technology to generate a simulated traffic digital twin model. Simulation scenarios are constructed and intelligently identified through a knowledge reasoning engine, and emergency plans are automatically deduced. The simulated traffic digital twin model is dynamically calibrated by combining a reinforcement learning algorithm and candidate decisions are generated. Based on the simulated traffic digital twin model, the planning pattern is extracted by combining knowledge graph and case reasoning technology, a set of alternative planning schemes is constructed by combining optimization algorithm and heuristic search strategy, the simulation scenario parameter space is set and the parameters are combined by intelligent sampling algorithm, the candidate decisions are evaluated, and the first candidate decision set is determined by combining multi-criteria decision-making and machine learning sorting algorithm, the acceptance of each element in the first candidate decision set is evaluated by combining game theory and swarm intelligence optimization algorithm, and the optimal candidate decision is determined by combining the preset acceptance threshold, and a task list is generated according to the optimal candidate decision and the task is deployed through the control center; Based on the traffic panorama database, a multi-level traffic simulation basic model is constructed, and a traffic knowledge graph is constructed by combining ontology and semantic web technology. Key information is extracted and the traffic knowledge graph is updated by combining knowledge extraction technology. Real-time mapping is performed by combining digital twin technology to generate a simulated traffic digital twin model. Simulation scenarios are constructed and intelligently identified through a knowledge reasoning engine, and emergency plans are automatically deduced. The simulated traffic digital twin model is dynamically calibrated by combining a reinforcement learning algorithm and candidate decisions are generated, including: The multi-level traffic simulation basic model is constructed according to the traffic panoramic database, wherein the multi-level traffic simulation basic model includes a micro-simulation model, a macro-simulation model and a meso-simulation model. Traffic participants are modeled by an agent-based modeling method, and the interaction between different traffic participants is simulated in combination with a social force model to obtain the micro-simulation model. Based on the theory of traffic fluid mechanics, the traffic flow is regarded as a continuous medium and a group of partial differential equations for the spatiotemporal evolution of the traffic flow is established. The macro-simulation model is modeled in combination with the compressibility and viscosity of the traffic flow. The road network is abstracted into a service desk and a queuing system by a discrete event simulation method based on queuing theory. The meso-simulation model is modeled in combination with the random arrival of traffic demand and the random distribution of service time. For each traffic area, the macro-simulation model and the micro-simulation model are set through a multi-resolution architecture and the meso-simulation model is used to supplement the blank area to complete model coupling and collaborative solution. Define the core concepts, attributes and relationships in the transportation field through ontology engineering methods and form a transportation ontology model layer, build a transportation knowledge description framework based on the transportation ontology model layer, obtain transportation knowledge ontology, and represent transportation entities, attributes and relationships as triples through semantic web technology to form a transportation semantic network and construct a transportation knowledge graph; Extract structured, semi-structured and unstructured traffic knowledge from the traffic panorama database through knowledge extraction and knowledge fusion technology, perform semantic matching and knowledge fusion with the existing traffic knowledge graph to update the traffic knowledge graph, and map the real-time data stream in the traffic panorama database with the multi-level traffic simulation basic model in real time in combination with the digital twin technology according to the updated traffic knowledge graph, and generate the simulated traffic digital twin model in combination with the pre-set traffic infrastructure model and traffic environment model; Based on the simulated traffic digital twin model, combined with the traffic knowledge graph and the knowledge reasoning engine, a simulation scenario is constructed by generating typical traffic events. Based on the simulation scenario, the emergency plan for traffic events is automatically deduced, and the optimization effects of different decision-making schemes are evaluated. Combined with the reinforcement learning algorithm, the optimal traffic decision-making strategy is learned through interactive exploration of the traffic simulation environment, and the simulated traffic digital twin model is calibrated to generate candidate decisions.

2. The method according to claim 1, It is characterized in that Various types of traffic sensors are set up in the urban road network. The traffic sensors collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time and combine them to obtain multi-dimensional traffic data. The key feature parameters in the multi-dimensional traffic data are extracted through edge computing nodes. The key feature parameters are automatically annotated and semantically described in combination with the fusion deep learning algorithm to generate structured traffic events and abnormal situation data. The traffic panoramic database obtained in combination with the association fusion algorithm includes: Various types of traffic sensors are set up in the urban road network, wherein the traffic sensors include cameras, microwave radar vehicle detectors, lidars and meteorological sensors. Through the simulated annealing algorithm, the temporal and spatial coverage and data quality of traffic information collection are maximized and the sensor setting cost is minimized to obtain the optimal setting scheme of the sensor network. Sensors are set up based on the optimal setting scheme, and edge computing nodes are deployed in combination with the edge computing mode to collect road traffic status, traffic flow data, vehicle speed and weather environment data. Based on the data fusion theory, the collected data is subjected to temporal and spatial registration and consistency verification through Bayesian inference to obtain the multi-dimensional traffic data. For the multidimensional traffic data, feature engineering is performed through pre-set edge computing nodes to extract key feature parameters that reflect traffic status and traffic events, wherein the key feature parameters include average vehicle speed, vehicle volume and traffic density. A multimodal deep learning method is used to perform feature understanding and semantic description of complex traffic scenes, automatically annotate traffic events and abnormal situations, and generate semantic annotation results. A semantic description model is generated through knowledge distillation and model compression algorithms and locally deployed through edge computing nodes. The semantic annotation results are semantically understood through the semantic description model. Based on the semantic understanding results, in combination with natural language processing and knowledge graph technology, traffic professional terms are mapped and associated with daily language to obtain the traffic panoramic database.

3. The method according to claim 2, It is characterized in that Based on data fusion theory, the spatiotemporal registration and consistency check are performed through Bayesian inference as shown in the following formula: μ′(t)=μ(t)+K x,z S -1 (z i (t)}-{h i (μ(t))}); Among them, μ'(t) represents the posterior mean function, μ(t) represents the prior mean function of the state parameter, K x,z represents the covariance matrix between state parameters and observation data, ∑ represents the spatiotemporal covariance matrix of observation data, z i (t) represents the observation data of the i-th sensor, h i (μ(t)) represents the observation value of the i-th sensor when the prior mean μ(t) of the state parameter is given.

4. The method according to claim 1, It is characterized in that The social force model is combined to simulate the interaction between different traffic participants as shown in the following formula: Among them, F g represents the net force acting on individual g, represents the driving force of individual g, α gh represents the perception coefficient of individual g to individual h, A represents the first constant parameter, d gh represents the distance between individuals g and h, D represents the expected safe distance between individuals g and h, B represents the second constant parameter, represents the unit vector from individual g to individual h, b represents the boundary, β gb represents the perception coefficient of individual g to boundary b, A b represents the first boundary parameter, d gb represents the distance between individual g and boundary b, D b represents the expected safe distance between individual g and boundary b, B b represents the second boundary parameter, represents the unit vector pointing from individual g to boundary b.

5. The method according to claim 1, It is characterized in that Based on the simulated traffic digital twin model, the planning pattern is extracted by combining knowledge graph and case reasoning technology, a set of alternative planning schemes is constructed by combining optimization algorithm and heuristic search strategy, the simulation scenario parameter space is set and the parameters are combined by intelligent sampling algorithm, the candidate decisions are evaluated, and the first candidate decision set is determined by combining multi-criteria decision and machine learning sorting algorithm, the acceptance of each element in the first candidate decision set is evaluated by combining game theory and swarm intelligence optimization algorithm, and the optimal candidate decision is determined by combining the preset acceptance threshold, and a task list is generated according to the optimal candidate decision and the task is deployed through the control center, including: Based on the simulated traffic digital twin model, concepts, entities and relationships in the field of traffic planning are mapped to a low-dimensional vector space through a transfer embedding model to obtain a distributed representation of planning knowledge. Based on the distributed representation, historical traffic planning cases are retrieved, corrected and retained through case similarity calculation and adaptive adjustment in combination with case reasoning technology. Frequent subgraph mining technology is used to determine co-occurring concepts from the traffic planning knowledge graph, extract planning rules and constraints, and obtain the planning model. Abstracting the traffic planning problem into a combinatorial optimization problem, determining the decision space through the taboo search algorithm and the neighborhood search and obtaining the corresponding first alternative planning scheme, solving the integer programming and nonlinear programming problems through the particle swarm algorithm, generating the second alternative planning scheme, and integrating the first alternative planning scheme and the second alternative planning scheme into the set of alternative planning schemes; The influence of different parameters on the simulation results is evaluated by the global sensitivity analysis method, and the key parameters are screened. For high-dimensional parameter space, the Latin hypercube sampling algorithm is used to generate sampling points and the parameter combination is modeled as a black box optimization problem. The parameter combination is determined by combining the differential evolution algorithm. For the simulation scenario corresponding to each parameter combination, the adaptability of each alternative planning scheme is determined according to the multi-decision criterion and the machine learning sorting algorithm. The alternative planning schemes are sorted in descending order according to the adaptability, and the top 10% alternative planning schemes are selected as the first candidate decision set; For each element in the first candidate decision set, determine the acceptability of each element through the Nash equilibrium strategy, compare the acceptability corresponding to each element with a preset acceptance threshold, and if it is not less than the acceptance threshold, retain the current element, and for each retained element, determine the optimal candidate decision based on information interaction through a swarm intelligence optimization algorithm; The optimal candidate decision is decomposed into multiple specific implementation tasks, and the tasks are deployed through a control center according to task relationships and task importance.

6. The method according to claim 5, It is characterized in that The Latin hypercube sampling algorithm generates sampling points as shown in the following formula: Where X represents the Latin hypercube sampling matrix after random perturbation, k mn It represents the value of the element in the mth row and nth column, s represents the number of factor levels, rand() represents a random number between 0 and 1, and N represents the number of sampling points.

7. A smart city traffic planning system based on big data, used to implement the method described in any one of claims 1 to 6. It is characterized in that include: The first unit is used to set up various types of traffic sensors in the urban road network, collect road traffic status, traffic flow data, vehicle speed and weather environment data in real time through the traffic sensors and combine them to obtain multi-dimensional traffic data, extract key feature parameters in the multi-dimensional traffic data through edge computing nodes, automatically annotate and semantically describe the key feature parameters in combination with a fusion deep learning algorithm, generate structured traffic events and abnormal situation data, and obtain a traffic panoramic database in combination with an association fusion algorithm; The second unit is used to build a multi-level traffic simulation basic model based on the traffic panoramic database, build a traffic knowledge graph in combination with ontology and semantic web technology, extract key information in combination with knowledge extraction technology and update the traffic knowledge graph, and perform real-time mapping in combination with digital twin technology to generate a simulated traffic digital twin model, build simulation scenarios and intelligently identify and automatically deduce emergency plans through a knowledge reasoning engine, and dynamically calibrate the simulated traffic digital twin model in combination with a reinforcement learning algorithm and generate candidate decisions; The third unit is used to extract planning patterns based on the simulated traffic digital twin model in combination with knowledge graphs and case-based reasoning technology, construct a set of alternative planning schemes through combinatorial optimization algorithms and heuristic search strategies, set simulation scenario parameter space and combine parameters through intelligent sampling algorithms, evaluate the candidate decisions, determine the first candidate decision set in combination with multi-criteria decision-making and machine learning sorting algorithms, evaluate the acceptability of each element in the first candidate decision set in combination with game theory and swarm intelligence optimization algorithms, and determine the optimal candidate decision in combination with a pre-set acceptance threshold, generate a task list based on the optimal candidate decision, and deploy the tasks through a control center.

8. An electronic device, It is characterized in that include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, It is characterized in that When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Citation Information

Patent Citations

  • Smart city decision-making system and method based on digital twin

    CN113609656A

  • Urban road online microscopic simulation method and system based on digital twinborn technology

    CN113704956A

  • Smart city management method and system based on digital twinning

    CN117057656A

  • Urban planning management and control system based on digital twinborn technology

    CN118297775A

Cited By

  • Digital twin city construction and operation optimization method based on cloud computing and big data technology

    CN119129085A

  • Intelligent auxiliary method and device for command system

    CN119151276A

  • Shunt induction control method and system for expressway unmanned toll station

    CN119476660A

  • Intelligent traffic data processing method and system based on edge calculation

    CN119479316A

  • Sunken urban global governance comprehensive management and control method and platform

    CN119558535A