Highway traffic volume prediction method based on three-elasticity dynamic coupling and related system

Abstract

The invention discloses a highway traffic volume prediction method based on three-elasticity dynamic coupling and a related system. The method comprises the following steps: determining a geographic area and GDP macroscopic factor influence elastic coefficient K1, a microcosmic factor elastic coefficient K2 and an emergency elastic coefficient K3; four-dimensional space-time tensor extraction cross-modal features are constructed by K1, K2 and K3 with satellite remote sensing traffic flow data, a mobile phone signaling OD matrix and toll station license plate recognition data respectively; inputting the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of the dynamic road network game, and outputting multi-dimensional influence factors of the traffic flow; and inputting the GDP macroscopic data, the microcosmic factor data, the emergency data and the four-dimensional space-time tensor data at the current moment into the hybrid coupling prediction model, and predicting a highway traffic prediction result. According to the method, the weights of the three elastic coefficients are redistributed through the hybrid coupling prediction model, so that the method adapts to different traffic scenes and time periods, and the accuracy and reliability of traffic volume prediction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic flow prediction, and in particular to a highway traffic volume prediction method and system based on three-elastic dynamic coupling, and a computing device. Background Art

[0002] With the acceleration of urbanization and rapid economic development, highway traffic volume forecasting is becoming increasingly important in traffic management, urban planning, and emergency response. Highway project traffic volume primarily consists of trend traffic, diverted traffic, and induced traffic, and its magnitude is a key factor influencing the project's financial returns. When acquiring highway projects, investors often raise the issue of traffic volume discounts in feasibility studies (or consulting reports). From planning to completion and opening to traffic, highway projects undergo several phases, including feasibility study preparation, construction drawing design, project bidding, construction, defect resolution, and final acceptance. This typically spans five years or even longer, making accurate forecasts of traffic volume for each year after opening to traffic extremely difficult.

[0003] Traditional traffic flow forecasting methods primarily rely on historical traffic data and simple statistical models (such as time series analysis and regression analysis). While these methods can predict traffic flow to a certain extent, they typically rely on a single data source (such as historical traffic flow data) and ignore other factors that influence traffic flow (such as macroeconomic factors, micro-policy factors, and emergencies). Because these forecasting models or methods are typically static, it is difficult to dynamically adjust weights to adapt to different traffic scenarios and time periods. This makes them less responsive to emergencies (such as accidents and natural disasters), and the forecast results are easily disrupted by these events. This results in low prediction accuracy and instability. Furthermore, because these traditional models are black-box models, they struggle to explain the underlying mechanisms of traffic flow, hindering decision-making by traffic management departments.

[0004] To solve the above problems, the present invention proposes a highway traffic volume prediction method based on three-elastic dynamic coupling. The weights of the three elastic coefficients are redistributed through a hybrid coupling prediction model to adapt to different traffic scenarios and time periods, thereby improving the accuracy and reliability of traffic volume prediction. Summary of the Invention

[0005] In view of the above problems, the present invention provides a highway traffic volume prediction method and system, and computing equipment based on three-elastic dynamic coupling.

[0006] According to one aspect of the present invention, a highway traffic volume prediction method based on three-elastic dynamic coupling is provided, comprising:

[0007] Based on the geographical region elasticity coefficient T1 and the GDP macro-factor elasticity coefficient T2 of the province where the project is located, the geographical region and GDP macro-factor elasticity coefficient K1 is determined; based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the route, urbanization progress, cultural and tourism industry development, and population and enterprise growth along the route, the micro-factor elasticity coefficient K2 is determined; based on the new interchanges, the opening or closure of large factories and mines, roadbed collapse, earthquakes, and major repairs of competitive channels, the emergency elasticity coefficient K3 is determined;

[0008] The GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 are respectively combined with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data to construct a four-dimensional space-time tensor, and the cross-modal features of the four-dimensional space-time tensor are extracted through the tensor decomposition method;

[0009] Inputting the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of a dynamic road network game to output multi-dimensional influencing factors of traffic flow;

[0010] The multi-dimensional influencing factors of the traffic flow are input into the hybrid coupling prediction model as prior knowledge. The hybrid coupling prediction model redistributes the weights of the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 according to the current traffic status;

[0011] The current GDP macro data, micro factor data, emergency data, and four-dimensional space-time tensor data are input into the hybrid coupling prediction model to predict the highway traffic volume within a preset time in the future; wherein the prediction results include traffic flow, speed, and density of different sections.

[0012] In an optional manner, the multi-agent simulation model of the dynamic road network game includes a road network representation layer, an agent decision layer, and a traffic flow simulation layer, which is used to simulate the behavior of vehicles in the road network. Each vehicle acts as an agent and makes decisions based on the surrounding environment and its own goals.

[0013] The road network representation layer models the road network as a graph structure through the GAT neural network. The nodes of the graph structure represent road sections or intersections, and the edges of the graph structure represent the connection relationship between road sections. The GAT neural network learns the topological structure and traffic status of the road network and aggregates traffic information of adjacent road sections to each node.

[0014] The agent decision layer trains the agent through an A2C network, enabling the agent to make decisions based on the road network status and its own goals. The A2C network uses a critic network to evaluate the value of actions. In addition, the agent decision layer divides the road section into multiple cells through a cellular automation, each cell accommodating a vehicle, and simulates the movement of vehicles between cells according to preset rules.

[0015] In an optional manner, the hybrid coupling prediction model includes an input layer, an LSTM network layer, a weight distribution layer, a model coupling layer, a feedback coupling layer and an output layer;

[0016] The input layer is used to input time series data, macro factor data, micro factor data, emergency event data and four-dimensional space-time tensor features;

[0017] The LSTM network layer is used to learn the changing pattern of historical traffic flow;

[0018] The weight distribution layer is used to dynamically adjust the weight of the LSTM network layer according to the current traffic status, wherein the current traffic status includes flow, speed and density;

[0019] The data coupling layer is used to use the multi-dimensional influencing factors of traffic flow output by the multi-agent simulation model as prior knowledge of the hybrid coupling prediction model;

[0020] The feedback coupling layer is used to feed back the prediction results of the hybrid coupling prediction model to the multi-agent simulation model to update the road network status and adjust the strategy of the agent;

[0021] The output layer is used to output the prediction result of traffic flow.

[0022] In an optional manner, the calculation formula for the elasticity coefficient K1 of the geographical region and GDP macro factors is:

[0023]

[0024] Among them, α is the spatial diffusion coefficient, which is used to describe the propagation and attenuation characteristics of geographical factors in space; β is the multi-factor coupling strength, which quantifies the comprehensive impact of the interaction between the geographical elasticity coefficient 1 and the GDP elasticity coefficient T2 on K1; γ is the Lagrange multiplier, which is used to introduce traffic flow conservation constraints in the optimization process; L is the loss function considering traffic flow conservation, L = ∫ Ω (ρu 2 +p)dω, ρ is the traffic flow density, u is the velocity field, and p is the pressure term; Ω is the road network topology domain, which represents the spatial range of the combined effect of geographical area and GDP factors; is the Laplace operator, which represents the diffusion behavior of K1 in space; dω represents the integral differential element in the region Ω.

[0025] In an optional manner, the calculation formula of the micro-factor elasticity coefficient K2 is:

[0026]

[0027] Among them, ω i is the weight coefficient of the i-th micro-factor; P i ∈R 3×3 , is a three-dimensional matrix of policy intensity-implementation scope-duration; E i ∈R 3×1 , is the economy-population-enterprise synergy vector; λ is the time decay coefficient; Δt i is the time interval from the release of micro factors to the current moment; σ j is the standard deviation coefficient of the j-th competitive factor; C j is the covariance matrix of urbanization rate-population migration speed-enterprise opening and closing status; ||.|| F is the Frobenius norm; ∈ is the smoothing term.

[0028] In an optional manner, the calculation formula of the emergency elasticity coefficient K3 is:

[0029]

[0030] in, is the composite impact field of five types of emergencies, α k is the intensity coefficient of the k-th event, is the space-time Gaussian kernel function; D(x)=diag(d road ,d geo ,d eco ) is the spatial heterogeneity diffusion tensor, d road ,d geo ,d eco are road network density correction term, terrain relief, and regional economic resilience respectively; Ω is the integral domain; is the divergence operator; dVdt is the differential of volume and time, which means integration over the entire space-time domain; x is the spatial coordinate; t is the time coordinate; x k is the spatial position associated with the kth event; t k is the time associated with the kth event.

[0031] In an optional manner, the data coupling layer uses a tensor train decomposition method to perform tensor fusion on the four-dimensional space-time tensor and the elastic coefficient. The calculation formula for tensor fusion is:

[0032]

[0033] Among them, T fuse is the fused tensor, representing the comprehensive traffic status information integrating spatial, temporal features and elastic coefficients; TT decompose is the tensor chain decomposition function; is the tensor outer product; T spatial is the spatial tensor; T temporal is the time tensor; K1, K2, and K3 are the elasticity coefficients 1, 2, and 3 of the emergency respectively;

[0034] The feedback coupling layer measures the difference between the predicted flow and the simulated flow. The calculation formula of the difference measurement is:

[0035]

[0036] When the difference metric of road section i is greater than the error threshold, the road network status is updated. The update formula of the road network status is:

[0037] S i (t+1) =(1-α)S i (t) +α·MLP([S i (t) ,δ i,t ])

[0038] Among them, S i (t+1) is the road network status of section i at time t+1; S i (t) is the road network status of section i at time t; To predict traffic; is the simulated traffic; α is the update step size; MLP is the multi-layer perceptron corrector.

[0039] In an optional manner, the agent decision layer uses the A2C-MA algorithm to maintain the policy network of each agent, wherein the policy network includes a short-term decision network and a long-term planning network;

[0040] The short-term decision network outputs the acceleration / deceleration probability according to the current road state, and the long-term planning network generates a path selection strategy according to the road network level target. The loss function of the strategy network is:

[0041] L total =ηL A2C +(1-η)L MA +λ||Θ||2 2

[0042] Where η is the dynamic weight coefficient; L MAis the multi-agent collaborative loss; λ is the L2 regularization coefficient; ||Θ||2 2 is the square of the L2 norm of the model parameter set Θ.

[0043] According to another aspect of the present invention, a highway traffic volume prediction system based on three-elastic dynamic coupling is provided, comprising:

[0044] The elasticity coefficient determination module is used to determine the elasticity coefficient K1 of the geographical region and GDP macro factors based on the geographical region elasticity coefficient T1 and the GDP macro factor elasticity coefficient T2 of the project province; the micro factor elasticity coefficient K2 is determined based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the route, urbanization progress, cultural and tourism industry development, and population and enterprise growth along the route; and the emergency elasticity coefficient K3 is determined based on the addition of new interchanges, the opening or closure of large factories and mines, roadbed collapse, earthquakes, and major repairs of competitive channels.

[0045] A cross-modal feature extraction module is used to construct a four-dimensional space-time tensor by combining the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data, and extract cross-modal features of the four-dimensional space-time tensor through a tensor decomposition method;

[0046] A multi-factor impact analysis module is used to input the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of a dynamic road network game and output multi-dimensional influencing factors of traffic flow;

[0047] A weight redistribution module is used to input the multi-dimensional influencing factors of the traffic flow as prior knowledge into the hybrid coupling prediction model, and the hybrid coupling prediction model redistributes the weights of the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 according to the current traffic status;

[0048] The traffic volume prediction module is used to input the current GDP macro data, micro factor data, emergency data, and four-dimensional space-time tensor data into the hybrid coupling prediction model to predict the highway traffic volume within a preset time in the future; wherein the prediction results include traffic flow, speed and density of different sections.

[0049] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0050] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned highway traffic volume prediction method based on three elastic dynamic coupling.

[0051] According to the solution provided by the present invention, the geographical region and GDP macro-factor elasticity coefficient K1 is determined based on the geographical region elasticity coefficient T1 and the GDP macro-factor elasticity coefficient T2 of the province where the project is located; the micro-factor elasticity coefficient K2 is determined based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the line, urbanization progress, cultural and tourism industry development, population and enterprise growth along the line; the emergency elasticity coefficient K3 is determined by adding new interconnections, the opening or closing of large factories and mines, roadbed collapse, earthquakes and major repairs of competitive channels; the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 are respectively combined with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data to construct a four-dimensional space-time tensor, and through A tensor decomposition method extracts cross-modal features from the four-dimensional space-time tensor; the cross-modal features of the four-dimensional space-time tensor are input into a multi-agent simulation model of a dynamic road network game to output multi-dimensional influencing factors of traffic flow; these multi-dimensional influencing factors of traffic flow are input as prior knowledge into a hybrid coupling prediction model, which redistributes weights for the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 based on the current traffic state; current GDP macro-factor data, micro-factor data, emergency data, and four-dimensional space-time tensor data are input into the hybrid coupling prediction model to predict highway traffic volume within a preset timeframe; the prediction results include traffic flow, speed, and density for different road sections. The present invention redistributes the weights of the three elasticity coefficients through the hybrid coupling prediction model to adapt to different traffic scenarios and time periods, thereby improving the accuracy and reliability of traffic volume predictions. Specifically, through the dynamic road network game multi-agent simulation model, the behavior of vehicles in the road network is simulated, and decisions are made based on the surrounding environment and their own goals, more realistically reflecting actual traffic conditions. A four-dimensional spatiotemporal tensor is constructed by combining satellite remote sensing traffic flow data, mobile phone signaling OD matrices, and toll booth license plate recognition data. Tensor decomposition is used to extract cross-modal features, better capturing the spatiotemporal variations of traffic flow. The multi-dimensional influencing factors of traffic flow output by a multi-agent simulation model are used as prior knowledge, and weights are dynamically adjusted based on current traffic conditions, improving the adaptability and accuracy of the prediction model. Prediction results are fed back to the multi-agent simulation model through a feedback coupling layer to update the road network state and adjust the agent strategies, forming a closed-loop control loop, further improving prediction reliability. The A2C-MA algorithm is used to maintain the short-term decision-making network and long-term planning network of each agent, allowing decisions to be made at different time scales, improving decision adaptability.

[0052] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0054] Figure 1 A schematic flow chart of a highway traffic volume prediction method based on three-elastic dynamic coupling according to an embodiment of the present invention is shown;

[0055] Figure 2 A schematic diagram of a framework of a highway traffic volume prediction system based on three elastic dynamic coupling according to an embodiment of the present invention is shown;

[0056] Figure 3 A schematic structural diagram of a computing device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0057] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0058] Figure 1 FIG. 1 shows a flow chart of a highway traffic volume prediction method based on three elastic dynamic coupling according to an embodiment of the present invention. Specifically, Figure 1 As shown, the following steps are included:

[0059] Step S101: Determine the elasticity coefficient K1 of the geographical region and GDP macro factors based on the geographical region elasticity coefficient T1 and the GDP macro factor elasticity coefficient T2 of the province where the project is located; determine the micro factor elasticity coefficient K2 based on the automobile industry support policy, high-speed travel exemption policy, overall economic development plan along the route, urbanization progress, cultural and tourism industry development, and population and enterprise growth along the route; and determine the emergency elasticity coefficient K3 based on the addition of new interchanges, the opening or closure of large factories and mines, roadbed collapse, earthquakes, and major repairs of competitive channels.

[0060] In this embodiment, the elasticity coefficient can be used to adjust and supplement macro, micro, and emergency factors based on actual conditions, adapting to the characteristics of different regions and highways, and better assessing potential risk factors for highways. Compared to traditional methods that only use OD surveys and GDP to derive traffic growth rates, this method considers geographic region, macroeconomic GDP factors, automotive industry support policies, highway travel exemption policies, economic development along the route, urbanization progress, cultural and tourism industry development, population and business growth along the route, and emergencies. It fully utilizes operational traffic volume results and effectively considers the actual annual natural vehicle growth rate. The prediction process is closer to reality and has higher practical operability.

[0061] Specifically, for the geographic elasticity coefficient, geographic data (such as topography, climate characteristics, and hydrological conditions) for the project province was collected through GIS data, remote sensing imagery, and historical meteorological data. For the GDP macroeconomic elasticity coefficient, macroeconomic data such as GDP, per capita GDP, industrial structure, and investment status for the project province or region were collected from economic research institutions and other sources. For automotive industry support policies, national and local government-issued documents related to the automotive industry, including subsidies, tax incentives, and technical standards, were collected. Highway toll exemption policies, such as holiday toll-free travel and green channel policies, were also collected. Economic development plans, industrial layout, and investment attraction information for the regions along the route were also collected. Data on population growth, land development, and infrastructure construction were also collected for cities along the route, as well as data on visitor numbers and revenue for tourist attractions, hotels, and the catering and cultural tourism industries along the route. Data on changes in population and the number of businesses in cities and villages along the route were also collected. The location, timing, and status of connecting roads for newly added highway interchanges were recorded. Information on the start, stop, and relocation of large factories and mines along the route was collected. Information on the time, location, and impact range of natural disasters such as roadbed collapses and earthquakes was also collected. The system also collects information on overhaul plans, schedules, and sections of other competing expressways. It then calculates the macro-factor elasticity coefficient, K1, for geographic region and GDP, and the micro-factor elasticity coefficient, K2. A three-dimensional matrix of policy intensity, implementation scope, and duration, as well as an economic-demographic-enterprise synergy vector, is constructed and populated based on the collected data. The emergency elasticity coefficient, K3, is calculated and assigned different values ​​based on the type of emergency.

[0062] In an optional manner, the calculation formula for the elasticity coefficient K1 of the geographical region and GDP macro factors is:

[0063]

[0064] Among them, α is the spatial diffusion coefficient, which is used to describe the propagation and attenuation characteristics of geographical factors in space; β is the multi-factor coupling strength, which quantifies the comprehensive impact of the interaction between the geographical elasticity coefficient T1 and the GDP elasticity coefficient T2 on K1; γ is the Lagrange multiplier, which is used to introduce traffic flow conservation constraints in the optimization process; L is the loss function considering traffic flow conservation, L = ∫ Ω (ρu 2 +p)dω, ρ is the traffic flow density, u is the velocity field, and p is the pressure term; Ω is the road network topology domain, which represents the spatial range of the combined effect of geographical area and GDP factors; is the Laplace operator, which represents the diffusion behavior of K1 in space; dω represents the integral differential element in the region Ω.

[0065] In this embodiment, the interaction between geographical factors and GDP macro factors is quantified by coupling terms to more accurately capture the impact of the macroeconomic environment on traffic flow. Taking into account the propagation and attenuation characteristics of geographical factors in space, it can simulate the impact of macroeconomic and geographical factors on traffic flow in surrounding areas, which is more in line with actual conditions. The loss function L takes into account the constraints of traffic flow conservation in the calculation of K1, improving the physical rationality of the prediction. The loss function L includes traffic flow density, velocity field, and pressure terms, and the calculation process is consistent with actual traffic patterns.

[0066] For example, a project is located in Province A, a province characterized by hilly terrain and rapid GDP growth. Analysis of historical data for the province reveals that the geographic elasticity coefficient, T1, is primarily influenced by cultivated land area, population density, and transportation network density, resulting in a calculated T1 value of 0.7. Analysis of GDP data for the province reveals that GDP growth rate significantly influences traffic flow, resulting in a calculated T2 value of 1.2. Based on the province's geographic characteristics, a spatial diffusion coefficient of 0.2 is determined. Based on the relationship between T1 and T2, a multi-factor coupling strength of 0.5 is determined. Based on the strength of the traffic flow conservation constraint, a Lagrange multiplier of 0.1 is determined. The road network topology domain is divided into 1000 small grids, and the finite difference method is used to solve the formula to obtain the K1 value for each grid. Ultimately, the elasticity coefficient, K1, of the geographic and GDP macroeconomic factors affecting highway traffic volume forecasts for the province is determined.

[0067] In an optional manner, the calculation formula of the micro-factor elasticity coefficient K2 is:

[0068]

[0069] Among them, ω i is the weight coefficient of the i-th micro-factor; P i ∈R3×3 , is a three-dimensional matrix of policy intensity-implementation scope-duration; E i ∈R 3×1 , is the economy-population-enterprise synergy vector; λ is the time decay coefficient; Δt i is the time interval from the release of micro factors to the current moment; σ j is the standard deviation coefficient of the j-th competitive factor; C j is the covariance matrix of urbanization rate-population migration speed-enterprise opening and closing status; ||.|| F is the Frobenius norm; ∈ is the smoothing term.

[0070] In this embodiment, both positive promoting factors (such as policies supporting the automobile industry) and negative competitive factors (such as the opening of other roads) are included, and different micro factors are distinguished by weight coefficients. Abstract concepts such as policies, economy, population, and enterprises are converted into quantifiable indicators, and the intensity, scope of implementation, and duration of policies are taken into account, so that the impact of micro factors can be further accurately evaluated. For example, the following micro factors of a certain highway are known: i = 1: automobile industry support policy (high policy intensity, wide implementation scope, and long duration); i = 2: economic development plan along the route (medium planning intensity, covering some areas, and long duration); j = 1: opening of another parallel highway (large impact scope and strong competitiveness);

[0071] P1=[[0.8,0.7,0.9],[0.7,0.8,0.8],[0.9,0.8,0.7]](policy intensity 0.8, implementation scope 0.7, duration 0.9); E1=[[0.06],[0.03],[0.05]](GDP growth 6%, population growth 3%, enterprise growth 5%); P2=[[0.6,0.5,0.7],[0.5,0.6,0.6],[0.7,0.6,0. 5]] (planning intensity 0.6, partial coverage 0.5, duration 0.7); E2 = [[0.04], [0.02], [0.04]] (GDP growth 4%, population growth 2%, enterprise growth 4%); C1 = [[0.2, 0.1, 0.05], [0.1, 0.3, 0.1], [0.05, 0.1, 0.4]] (covariance matrix of urbanization rate 0.2, population migration speed 0.3, and enterprise opening and closing status 0.4). Substituting into the above formula, we get K2 = 0.236 / 0.089 ≈ 2.65, indicating that micro-factors have a significant impact on the traffic volume of this highway.

[0072] In an optional manner, the calculation formula of the emergency elasticity coefficient K3 is:

[0073]

[0074] in, is the composite impact field of five types of emergencies, α k is the intensity coefficient of the k-th event, is the space-time Gaussian kernel function; D(x)=diag(d road ,d geo ,d eco ) is the spatial heterogeneity diffusion tensor, d road ,d geo ,d eco are road network density correction term, terrain relief, and regional economic resilience respectively; Ω is the integral domain; is the divergence operator; dVdt is the differential of volume and time, which means integration over the entire space-time domain; x is the spatial coordinate; t is the time coordinate; x k is the spatial position associated with the kth event; t k is the time associated with the kth event.

[0075] In this implementation, the spatiotemporal Gaussian kernel function captures the spatial and temporal impact and intensity of emergencies, more accurately reflecting their dynamic impact on traffic flow. The emergency resilience coefficient considers multiple factors, including road network density, topographical relief, and regional economic resilience, enabling more accurate traffic flow predictions across diverse geographic and economic contexts.

[0076] For example, a major traffic accident occurred on a highway, resulting in the closure of part of the road. The time, location, impact range, road network density, terrain relief, and regional economic resilience of the accident are known. According to the formula Calculate the impact of the accident on traffic flow. Use D(x)=diag(d road ,d geo ,d eco ) calculates the impact of road network density, terrain relief, and regional economic resilience. K3 is calculated using integral and divergence operators, and the emergency elasticity coefficient is effectively utilized to improve the accuracy of highway traffic volume forecasts, thereby better addressing the impact of emergencies on traffic flow.

[0077] In step S102, the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 are respectively combined with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data to construct a four-dimensional space-time tensor, and the cross-modal features of the four-dimensional space-time tensor are extracted through the tensor decomposition method.

[0078] In this embodiment, tensor decomposition extracts the potential correlation and influence patterns between different modes (time, space, macro factors, micro factors, and emergency factors) in the four-dimensional space-time tensor. For example, predict the traffic flow of a specific section of a highway in the next hour. It is known that the GDP growth rate of the province where the highway is located is 5%, and the geographical area elasticity coefficient is 0.8. The formula is used to calculate K1 = 0.7. Micro factors: There is a newly developed tourist attraction along the route, and the government has introduced tourism promotion policies, and K2 = 0.9 is calculated. Emergency: A minor traffic accident occurred on a section of the highway, and K3 = 0.6 was calculated. Collect satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data for the past week. A four-dimensional space-time tensor was constructed, consisting of the following: time (24 hours per day over the past week); space (all road sections on the expressway); elasticity coefficients (K1, K2, and K3); and traffic data (satellite remote sensing traffic flow, mobile phone signaling outbound volume, and toll booth traffic flow). All of this data was then populated into the four-dimensional tensor and decomposed into four factor matrices using CP decomposition. The correlations between the time dimension and the elasticity coefficient dimension were extracted to characterize the daily variations in traffic flow due to macro, micro, and emergency factors. The correlations between the space dimension and the traffic data dimension were extracted to characterize the traffic flow distribution across different road sections. The extracted features were then fed into a multi-agent simulation model for dynamic road network game simulations to simulate road network traffic flow. The extracted features and simulation results were then fed into a hybrid coupling prediction model to predict traffic flow for the road section within the next hour. This demonstrates that integrating macro, micro, and emergency factors with multi-source traffic data into the four-dimensional space-time tensor and extracting cross-modal features through tensor decomposition allows for a more comprehensive capture of the factors influencing traffic flow.

[0079] Step S103: input the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of a dynamic road network game, and output multi-dimensional influencing factors of traffic flow.

[0080] In this embodiment, a multi-agent simulation model based on dynamic road network games simulates vehicle behavior in a road network. Each vehicle is treated as an agent and interactions between vehicles are considered, enabling them to make decisions based on their surroundings and their own goals, thereby generating game-playing behaviors. This allows for more accurate predictions of traffic flow trends, taking into account both competitive and collaborative relationships between vehicles, and thus more realistically reflecting actual traffic conditions. Furthermore, the system not only predicts traffic flow but also outputs the multi-dimensional influencing factors of traffic flow, such as vehicle routing preferences, congestion propagation patterns, and the impact of emergencies. This helps traffic managers understand traffic phenomena and develop more effective traffic management strategies.

[0081] In an optional manner, the multi-agent simulation model of the dynamic road network game includes a road network representation layer, an agent decision layer, and a traffic flow simulation layer, which is used to simulate the behavior of vehicles in the road network. Each vehicle acts as an agent and makes decisions based on the surrounding environment and its own goals.

[0082] The road network representation layer models the road network as a graph structure through the GAT neural network. The nodes of the graph structure represent road sections or intersections, and the edges of the graph structure represent the connection relationship between road sections. The GAT neural network learns the topological structure and traffic status of the road network and aggregates traffic information of adjacent road sections to each node.

[0083] The agent decision layer trains the agent through an A2C network, enabling the agent to make decisions based on the road network status and its own goals. The A2C network uses a critic network to evaluate the value of actions. In addition, the agent decision layer divides the road section into multiple cells through a cellular automation, each cell accommodating a vehicle, and simulates the movement of vehicles between cells according to preset rules.

[0084] In this embodiment, the road network is modeled as a graph structure using the Graph Attention Network (GAT), capturing the network's topology and traffic status, and adaptively aggregating traffic information from adjacent road sections. The Advantage Actor-Critic (A2C) network combines policy gradients and value functions to help intelligent agents make better decisions in complex traffic environments under uncertainty. Cellular automata divide road sections into multiple cells, each housing a vehicle, allowing for detailed simulation of vehicle movement within the section and improving the accuracy of traffic flow simulation.

[0085] For example, a highway network consists of multiple road sections and intersections. The network representation layer models the highway network as a graph, with nodes representing road sections or intersections and edges representing connections between sections. The GAT neural network learns the network's topology and traffic status and aggregates traffic information from adjacent sections to each node. The agent decision layer uses an A2C network to train agents to make decisions based on the network status and their own goals (agents decide whether to accelerate or decelerate based on the traffic flow and speed on the current section). Cellular automata divide the road section into multiple cells, each housing a vehicle, and simulate vehicle movement between cells according to preset rules. The traffic flow simulation layer treats each vehicle as an agent, making decisions based on the surrounding environment and its own goals, simulating vehicle behavior within the network. Multi-agent collaboration dynamically adapts to varying traffic conditions.

[0086] In an optional manner, the agent decision layer uses the A2C-MA algorithm to maintain the policy network of each agent, wherein the policy network includes a short-term decision network and a long-term planning network;

[0087] The short-term decision network outputs the acceleration / deceleration probability according to the current road state, and the long-term planning network generates a path selection strategy according to the road network level target. The loss function of the strategy network is:

[0088] L total =ηL A2C +(1-η)L MA +λ||Θ||2 2

[0089] Where η is the dynamic weight coefficient; L MA is the multi-agent collaborative loss; λ is the L2 regularization coefficient; ||Θ||2 2 is the square of the L2 norm of the model parameter set Θ.

[0090] In this embodiment, the short-term decision network outputs the acceleration / deceleration probability based on the current road section status, and the long-term planning network generates a path selection strategy based on the road network-level goals. For example, multiple intelligent agents (vehicles) in a highway network make decisions between different road sections. The short-term decision network inputs the traffic flow, speed and density data of the current road section into the short-term decision network, and the network outputs the acceleration / deceleration probability. If the traffic flow of the current road section is large, a higher deceleration probability is output. The long-term planning network is used to input road network-level traffic data to generate a path selection strategy. If the global traffic flow is large, a relatively idle path is selected to reduce overall traffic delays. The weight coefficient is dynamically adjusted according to the current traffic status to balance the weights of short-term decisions and long-term planning. Multiple intelligent agents share information and make collaborative decisions through collaborative losses.

[0091] In step S104, the multi-dimensional influencing factors of the traffic flow are input into the hybrid coupling prediction model as prior knowledge. The hybrid coupling prediction model redistributes the weights of the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 according to the current traffic status.

[0092] In this embodiment, the multi-dimensional influencing factors of traffic flow output by the multi-agent simulation model are used as prior knowledge to not only predict but also explain the reasons behind the prediction results. The hybrid coupling prediction model includes an input layer, an LSTM network layer, a weight distribution layer, a model coupling layer, a feedback coupling layer, and an output layer;

[0093] The input layer is used to input time series data, macro factor data, micro factor data, emergency event data and four-dimensional space-time tensor features;

[0094] The LSTM network layer is used to learn the changing pattern of historical traffic flow;

[0095] The weight distribution layer is used to dynamically adjust the weight of the LSTM network layer according to the current traffic status, wherein the current traffic status includes flow, speed and density;

[0096] The data coupling layer is used to use the multi-dimensional influencing factors of traffic flow output by the multi-agent simulation model as prior knowledge of the hybrid coupling prediction model;

[0097] The feedback coupling layer is used to feed back the prediction results of the hybrid coupling prediction model to the multi-agent simulation model to update the road network status and adjust the strategy of the agent;

[0098] The output layer is used to output the prediction result of traffic flow.

[0099] In an optional manner, the data coupling layer uses a tensor train decomposition method to perform tensor fusion on the four-dimensional space-time tensor and the elastic coefficient. The calculation formula for tensor fusion is:

[0100]

[0101] Among them, T fuse is the fused tensor, representing the comprehensive traffic status information integrating spatial, temporal features and elastic coefficients; TT decompose is the tensor chain decomposition function; is the tensor outer product; T spatial is the spatial tensor; T temporal is the time tensor; K1, K2, and K3 are the elasticity coefficients 1, 2, and 3 of the emergency respectively;

[0102] The feedback coupling layer measures the difference between the predicted flow and the simulated flow. The calculation formula of the difference measurement is:

[0103]

[0104] When the difference metric of road section i is greater than the error threshold, the road network status is updated. The update formula of the road network status is:

[0105] S i (t+1) =(1-α)S i (t) +α·MLP([S i (t) ,δ i,t ])

[0106] Among them, S i (t+1) is the road network status of section i at time t+1; Si (t) is the road network of section i at time t; To predict traffic; is the simulated traffic; α is the update step size; MLP is the multi-layer perceptron corrector.

[0107] In this embodiment, the tensor train decomposition method integrates spatial and temporal features and elasticity coefficients, capturing the complex relationships between multi-dimensional traffic data. The fused tensor is generated by using a tensor chain decomposition function and a tensor outer product, making the prediction results more reliable. The predicted and simulated traffic flows are then measured, and the road network status is dynamically updated based on the difference, enabling adaptive model adjustment.

[0108] Step S105, inputting the current GDP macro data, micro factor data, emergency data, and four-dimensional space-time tensor data into the hybrid coupling prediction model to predict the highway traffic volume within a preset time in the future; wherein the prediction results include traffic flow, speed, and density of different road sections.

[0109] In this embodiment, the three-elasticity coefficient method leverages operational traffic volume and effectively considers the actual annual natural growth rate of vehicles. This results in a more realistic forecast and greater operability. For example, when comparing traffic volume and toll revenue predicted using the three-elasticity coefficient method with the actual project traffic volume and toll revenue, the cumulative error over the past four years has been less than 1%, while the error in feasibility studies is around 20%. Therefore, the traffic volume predicted by the new model demonstrates considerable predictive superiority and is more consistent with actual conditions. The three-elasticity coefficient method provides a new perspective and a novel prediction path for feasibility studies of traffic volume, offering broad application prospects in planning, post-evaluation, feasibility studies, and project financial benefit evaluation. As shown in Table 1, the actual cumulative operating toll revenue of a certain operating company from 2021 to 2024 was 247.75 million yuan. As can be seen from Table 1, the model predicts revenue of 247.496 million yuan, with an error of (247.750 - 247.496) / 247.750 = 0.1%.

[0110] Table 1 Calculation of fee income after 4 years of operation after acquisition

[0111]

[0112]

[0113] According to the solution provided by the present invention, the geographical region and GDP macro-factor elasticity coefficient K1 is determined based on the geographical region elasticity coefficient T1 and the GDP macro-factor elasticity coefficient T2 of the province where the project is located; the micro-factor elasticity coefficient K2 is determined based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the line, urbanization progress, cultural and tourism industry development, population and enterprise growth along the line; the emergency elasticity coefficient K3 is determined by adding new interconnections, the opening or closing of large factories and mines, roadbed collapse, earthquakes and major repairs of competitive channels; the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 are respectively combined with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data to construct a four-dimensional space-time tensor, and through A tensor decomposition method extracts cross-modal features from the four-dimensional space-time tensor; the cross-modal features of the four-dimensional space-time tensor are input into a multi-agent simulation model of a dynamic road network game to output multi-dimensional influencing factors of traffic flow; these multi-dimensional influencing factors of traffic flow are input as prior knowledge into a hybrid coupling prediction model, which redistributes weights for the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 based on the current traffic state; current GDP macro-factor data, micro-factor data, emergency data, and four-dimensional space-time tensor data are input into the hybrid coupling prediction model to predict highway traffic volume within a preset timeframe; the prediction results include traffic flow, speed, and density for different road sections. The present invention redistributes the weights of the three elasticity coefficients through the hybrid coupling prediction model to adapt to different traffic scenarios and time periods, thereby improving the accuracy and reliability of traffic volume predictions. Specifically, through the dynamic road network game multi-agent simulation model, the behavior of vehicles in the road network is simulated, and decisions are made based on the surrounding environment and their own goals, more realistically reflecting actual traffic conditions. A four-dimensional spatiotemporal tensor is constructed by combining satellite remote sensing traffic flow data, mobile phone signaling OD matrices, and toll booth license plate recognition data. Tensor decomposition is used to extract cross-modal features, better capturing the spatiotemporal variations of traffic flow. The multi-dimensional influencing factors of traffic flow output by a multi-agent simulation model are used as prior knowledge, and weights are dynamically adjusted based on current traffic conditions, improving the adaptability and accuracy of the prediction model. Prediction results are fed back to the multi-agent simulation model through a feedback coupling layer to update the road network state and adjust the agent strategies, forming a closed-loop control loop, further improving prediction reliability. The A2C-MA algorithm is used to maintain the short-term decision-making network and long-term planning network of each agent, allowing decisions to be made at different time scales, improving decision adaptability.

[0114] Figure 2The schematic diagram of the framework of the highway traffic volume prediction system based on three elastic dynamic coupling according to an embodiment of the present invention is shown. The highway traffic volume prediction system based on three elastic dynamic coupling includes:

[0115] Elasticity coefficient determination module 210 is used to determine the elasticity coefficient K1 of the geographical region and GDP macro factors based on the geographical region elasticity coefficient T1 and the GDP macro factor elasticity coefficient T2 of the province where the project is located; determine the micro factor elasticity coefficient K2 based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the route, urbanization progress, cultural and tourism industry development, and population and enterprise growth along the route; and determine the emergency elasticity coefficient K3 based on the newly added interchanges, the opening or closure of large factories and mines, roadbed collapse, earthquakes, and major repairs of competitive channels;

[0116] A cross-modal feature extraction module 220 is configured to construct a four-dimensional space-time tensor by combining the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data, and extract cross-modal features from the four-dimensional space-time tensor using a tensor decomposition method.

[0117] A multi-factor impact analysis module 230 is configured to input the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of a dynamic road network game and output multi-dimensional influencing factors of traffic flow;

[0118] A weight redistribution module 240 is configured to input the multi-dimensional influencing factors of the traffic flow as prior knowledge into a hybrid coupling prediction model, wherein the hybrid coupling prediction model redistributes the weights of the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 according to the current traffic status;

[0119] Traffic volume prediction module 250 is used to input the current GDP macro data, micro factor data, emergency data, and four-dimensional space-time tensor data into the hybrid coupling prediction model to predict the highway traffic volume within a preset time in the future; wherein the prediction results include traffic flow, speed and density of different road sections.

[0120] Figure 3 The schematic diagram of the structure of the computing device embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0121] like Figure 3As shown, the computing device may include: a processor 302 , a communications interface 304 , a memory 306 , and a communication bus 308 .

[0122] Processor 302, communication interface 304, and memory 306 communicate with each other via communication bus 308. Communication interface 304 is used to communicate with other devices, such as client devices or other server network elements. Processor 302 is used to execute program 310, which specifically performs the steps described in the aforementioned embodiment of the highway traffic volume prediction method based on triple-elastic dynamic coupling.

[0123] Specifically, the program 310 may include program codes, which include computer operation instructions.

[0124] Processor 302 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in a computing device may be processors of the same type, such as one or more CPUs, or processors of different types, such as one or more CPUs and one or more ASICs.

[0125] The memory 306 is used to store the program 310. The memory 306 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0126] According to the solution provided by the present invention, the geographical region and GDP macro-factor elasticity coefficient K1 is determined based on the geographical region elasticity coefficient T1 and the GDP macro-factor elasticity coefficient T2 of the province where the project is located; the micro-factor elasticity coefficient K2 is determined based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the line, urbanization progress, cultural and tourism industry development, population and enterprise growth along the line; the emergency elasticity coefficient K3 is determined by adding new interconnections, the opening or closing of large factories and mines, roadbed collapse, earthquakes and major repairs of competitive channels; the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 are respectively combined with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data to construct a four-dimensional space-time tensor, and through A tensor decomposition method extracts cross-modal features from the four-dimensional space-time tensor; the cross-modal features of the four-dimensional space-time tensor are input into a multi-agent simulation model of a dynamic road network game to output multi-dimensional influencing factors of traffic flow; these multi-dimensional influencing factors of traffic flow are input as prior knowledge into a hybrid coupling prediction model, which redistributes weights for the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 based on the current traffic state; current GDP macro-factor data, micro-factor data, emergency data, and four-dimensional space-time tensor data are input into the hybrid coupling prediction model to predict highway traffic volume within a preset timeframe; the prediction results include traffic flow, speed, and density for different road sections. The present invention redistributes the weights of the three elasticity coefficients through the hybrid coupling prediction model to adapt to different traffic scenarios and time periods, thereby improving the accuracy and reliability of traffic volume predictions. Specifically, through the dynamic road network game multi-agent simulation model, the behavior of vehicles in the road network is simulated, and decisions are made based on the surrounding environment and their own goals, more realistically reflecting actual traffic conditions. A four-dimensional spatiotemporal tensor is constructed by combining satellite remote sensing traffic flow data, mobile phone signaling OD matrices, and toll booth license plate recognition data. Tensor decomposition is used to extract cross-modal features, better capturing the spatiotemporal variations of traffic flow. The multi-dimensional influencing factors of traffic flow output by a multi-agent simulation model are used as prior knowledge, and weights are dynamically adjusted based on current traffic conditions, improving the adaptability and accuracy of the prediction model. Prediction results are fed back to the multi-agent simulation model through a feedback coupling layer to update the road network state and adjust the agent strategies, forming a closed-loop control loop, further improving prediction reliability. The A2C-MA algorithm is used to maintain the short-term decision-making network and long-term planning network of each agent, allowing decisions to be made at different time scales, improving decision adaptability.

[0127] Those skilled in the art will appreciate that modules in the devices of the embodiments may be adaptively modified and deployed in one or more devices different from the embodiments. Modules, units, or components in the embodiments may be combined into a single module, unit, or component, and furthermore, they may be divided into multiple submodules, subunits, or subcomponents. All features disclosed in this specification (including the accompanying claims, abstract, and drawings), as well as all processes or units of any method or device disclosed therein, may be combined in any combination, except where at least some of such features and / or processes or units are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that provides the same, equivalent, or similar purpose. Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the present invention and to form different embodiments. For example, in the claims below, any of the claimed embodiments may be used in any combination. The present invention may be implemented using hardware comprising a number of different elements and using a suitably programmed computer. In a unit claim that lists several systems, several of these systems may be embodied by the same hardware item. Unless otherwise specified, the steps in the above embodiments should not be understood as limiting the order of execution.

Claims

1. A highway traffic volume prediction method based on three-elastic dynamic coupling, characterized by: include: Based on the geographical region elasticity coefficient T1 and the GDP macro-factor elasticity coefficient T2 of the province where the project is located, the geographical region and GDP macro-factor elasticity coefficient K1 is determined; based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the route, urbanization progress, cultural and tourism industry development, and population and enterprise growth along the route, the micro-factor elasticity coefficient K2 is determined; based on the new interchanges, the opening or closure of large factories and mines, roadbed collapse, earthquakes, and major repairs of competitive channels, the emergency elasticity coefficient K3 is determined; The GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 are respectively combined with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data to construct a four-dimensional space-time tensor, and the cross-modal features of the four-dimensional space-time tensor are extracted through the tensor decomposition method; Inputting the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of a dynamic road network game to output multi-dimensional influencing factors of traffic flow; The multi-dimensional influencing factors of the traffic flow are input into the hybrid coupling prediction model as prior knowledge. The hybrid coupling prediction model redistributes the weights of the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 according to the current traffic status; The current GDP macro data, micro factor data, emergency data, and four-dimensional space-time tensor data are input into the hybrid coupling prediction model to predict the highway traffic volume within a preset time in the future; wherein the prediction results include traffic flow, speed, and density of different sections.

2. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 1 is characterized in that: The multi-agent simulation model of the dynamic road network game includes a road network representation layer, an agent decision layer, and a traffic flow simulation layer, which is used to simulate the behavior of vehicles in the road network. Each vehicle acts as an agent and makes decisions based on the surrounding environment and its own goals. The road network representation layer models the road network as a graph structure through the GAT neural network. The nodes of the graph structure represent road sections or intersections, and the edges of the graph structure represent the connection relationship between road sections. The GAT neural network learns the topological structure and traffic status of the road network and aggregates traffic information of adjacent road sections to each node. The agent decision layer trains the agent through an A2C network, enabling the agent to make decisions based on the road network status and its own goals. The A2C network uses a critic network to evaluate the value of actions. In addition, the agent decision layer divides the road section into multiple cells through a cellular automation, each cell accommodating a vehicle, and simulates the movement of vehicles between cells according to preset rules.

3. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 1 is characterized in that: The hybrid coupling prediction model includes an input layer, an LSTM network layer, a weight distribution layer, a model coupling layer, a feedback coupling layer and an output layer; The input layer is used to input time series data, macro factor data, micro factor data, emergency event data and four-dimensional space-time tensor features; The LSTM network layer is used to learn the changing pattern of historical traffic flow; The weight distribution layer is used to dynamically adjust the weight of the LSTM network layer according to the current traffic status, wherein the current traffic status includes flow, speed and density; The data coupling layer is used to use the multi-dimensional influencing factors of traffic flow output by the multi-agent simulation model as prior knowledge of the hybrid coupling prediction model; The feedback coupling layer is used to feed back the prediction results of the hybrid coupling prediction model to the multi-agent simulation model to update the road network status and adjust the strategy of the agent; The output layer is used to output the prediction result of traffic flow.

4. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 1 is characterized in that: The calculation formula for the elasticity coefficient K1 of the geographical region and GDP macroeconomic factors is: Among them, α is the spatial diffusion coefficient, which is used to describe the propagation and attenuation characteristics of geographical factors in space; β is the multi-factor coupling strength, which quantifies the comprehensive impact of the interaction between the geographical elasticity coefficient T1 and the GDP elasticity coefficient T2 on K1; γ is the Lagrange multiplier, which is used to introduce traffic flow conservation constraints in the optimization process; L is the loss function considering traffic flow conservation, L = ∫ Ω (ρu 2 +p)dω, ρ is the traffic flow density, u is the velocity field, and p is the pressure term; Ω is the road network topology domain, which represents the spatial range of the combined effect of geographical area and GDP factors; is the Laplace operator, which represents the diffusion behavior of K1 in space; dω represents the integral differential element in the region Ω.

5. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 1 is characterized in that: The calculation formula of the micro-factor elasticity coefficient K2 is: Among them, ω i is the weight coefficient of the i-th micro-factor; P i ∈R 3×3 , is a three-dimensional matrix of policy intensity-implementation scope-duration; E i ∈R 3×1 , is the economy-population-enterprise synergy vector; λ is the time decay coefficient; Δt i is the time interval from the release of micro factors to the current moment; σ j is the standard deviation coefficient of the j-th competitive factor; C j is the covariance matrix of urbanization rate-population migration speed-enterprise opening and closing status; ||.|| F is the Frobenius norm; ∈ is the smoothing term.

6. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 1 is characterized in that: The calculation formula of the emergency elasticity coefficient K3 is: in, is the composite impact field of five types of emergencies, α k is the intensity coefficient of the k-th event, is the space-time Gaussian kernel function; D(x)=diag(d road ,d geo ,d eco ) is the spatial heterogeneity diffusion tensor, d road ,d geo ,d eco are road network density correction term, terrain relief, and regional economic resilience respectively; Ω is the integral domain; is the divergence operator; dVdt is the differential of volume and time, which means integration over the entire space-time domain; x is the spatial coordinate; t is the time coordinate; x k is the spatial position associated with the kth event; t k is the time associated with the kth event.

7. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 3 is characterized in that: The data coupling layer uses the tensor train decomposition method to perform tensor fusion on the four-dimensional space-time tensor and the elastic coefficient. The calculation formula of tensor fusion is: Among them, T fuse is the fused tensor, representing the comprehensive traffic status information integrating spatial, temporal features and elastic coefficients; TT decompose is the tensor chain decomposition function; is the tensor outer product; T spatial is the spatial tensor; T temporal is the time tensor; K1, K2, and K3 are the elasticity coefficients 1, 2, and 3 of the emergency respectively; The feedback coupling layer measures the difference between the predicted flow and the simulated flow. The calculation formula of the difference measurement is: When the difference metric of road section i is greater than the error threshold, the road network status is updated. The update formula of the road network status is: S i (t+1) =(1-α)S i (t) +α·MLP([S i (t) ,δ i,t ]) Among them, S i (t+1) is the road network status of section i at time t+1; S i (t) is the road network status of section i at time t; To predict traffic; is the simulated traffic; α is the update step size; MLP is the multi-layer perceptron corrector.

8. The highway traffic volume prediction method based on three-elastic dynamic coupling according to claim 2 is characterized in that: The agent decision layer uses the A2C-MA algorithm to maintain the strategy network of each agent, and the strategy network includes a short-term decision network and a long-term planning network; The short-term decision network outputs the acceleration / deceleration probability according to the current road state, and the long-term planning network generates a path selection strategy according to the road network level target. The loss function of the strategy network is: L total =ηL A2C +(1-n)L MA +λ||Θ||2 2 Where η is the dynamic weight coefficient; L MA is the multi-agent collaborative loss; λ is the L2 regularization coefficient; ||Θ||2 2 is the square of the L2 norm of the model parameter set Θ.

9. A highway traffic volume prediction system based on three elastic dynamic coupling, characterized by: include: The elasticity coefficient determination module is used to determine the elasticity coefficient K1 of the geographical region and GDP macro factors based on the geographical region elasticity coefficient T1 and the GDP macro factor elasticity coefficient T2 of the project province; the micro factor elasticity coefficient K2 is determined based on the automobile industry support policy, high-speed travel exemption policy, the overall economic development plan along the route, urbanization progress, cultural and tourism industry development, and population and enterprise growth along the route; and the emergency elasticity coefficient K3 is determined based on the addition of new interchanges, the opening or closure of large factories and mines, roadbed collapse, earthquakes, and major repairs of competitive channels. A cross-modal feature extraction module is used to construct a four-dimensional space-time tensor by combining the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 with satellite remote sensing traffic flow data, mobile phone signaling OD matrix, and toll station license plate recognition data, and extract cross-modal features of the four-dimensional space-time tensor through a tensor decomposition method; A multi-factor impact analysis module is used to input the cross-modal features of the four-dimensional space-time tensor into a multi-agent simulation model of a dynamic road network game and output multi-dimensional influencing factors of traffic flow; A weight redistribution module is used to input the multi-dimensional influencing factors of the traffic flow as prior knowledge into the hybrid coupling prediction model, and the hybrid coupling prediction model redistributes the weights of the GDP macro-factor elasticity coefficient K1, the micro-factor elasticity coefficient K2, and the emergency elasticity coefficient K3 according to the current traffic status; The traffic volume prediction module is used to input the current GDP macro data, micro factor data, emergency data, and four-dimensional space-time tensor data into the hybrid coupling prediction model to predict the highway traffic volume within a preset time in the future; wherein the prediction results include traffic flow, speed and density of different sections.

10. A computing device comprising: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned highway traffic volume prediction method based on three elastic dynamic coupling.

Images