Regional and highland dual-link ramp control method under multi-source data fusion

Abstract

The application relates to a ramp control method based on regional and highland dual linkage under multi-source data fusion, and relates to the technical field of intelligent traffic control. The application aims at the problems of the existing ramp control, such as the separation of expressways and ground systems, the isolated lagging strategy, the single data dimension, the lack of behavior induction and the like, adopts multi-source heterogeneous data fusion to construct an integrated traffic state sensing system, identifies the control demand in two dimensions through regional linkage and highland coordination, establishes a rolling optimization model based on a model predictive control algorithm, synchronously outputs ramp release rates, ground signal timing and induction strategies, and realizes accurate instruction issuing and closed-loop feedback self-learning relying on a distributed architecture. The application can effectively relieve ramp congestion and highland traffic mismatch, improve the overall traffic efficiency of a road network, and is suitable for the coordinated control scene of urban expressways and ground road networks.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control technology, specifically to a ramp control method based on multi-source data fusion and regional and high-altitude dual linkage. Background Technology

[0002] With the acceleration of urbanization and the continuous growth of motor vehicle ownership, urban expressway systems are facing enormous pressure. As a key node connecting expressways and surface road networks, the control efficiency of ramps directly affects overall traffic operation. However, existing ramp control methods are increasingly showing their limitations in dealing with dynamic and complex urban traffic. Currently, the mainstream methods can be summarized into three categories: first, signal light control based on fixed rules, completely detached from real-time traffic conditions; second, dynamic switch control triggered by single indicators such as mainline occupancy, which, while possessing some adaptability, still operates in isolation; and third, feedback-based flow regulation based on local detector data (such as the ALINEA algorithm), but only focusing on local flow matching in merging sections. The common drawback of these methods is that their control logic is confined to the expressway system itself, lacking modeling and coordination of the strong coupling relationship between ramps and the surface traffic system, leading to a "high-level disconnect." For example, off-ramp traffic flow easily causes overflow congestion at surface intersections, while the closure of on-ramp may cause queuing back on connecting surface roads, forming a systemic bottleneck.

[0003] Specifically, regarding relevant patented technologies, such as CN114613125A, a hierarchical collaborative control method for multiple ramps on expressways, it calculates ramp merging rates by delineating control sub-zones and integrating local and collaborative adjustment quantities. Although this scheme constructs a hierarchical collaborative framework, its control sub-zone delineation and optimization process still revolves entirely around the expressway mainline status, failing to incorporate the signal status and capacity of ground-level intersections into the collaborative optimization model. Essentially, it remains a form of "self-optimization" within the expressway system, failing to solve the fundamental problem of cross-system (high-level) collaboration. Another patent, CN108053658A, a coordinated control method for multiple ramps on expressways with congestion full-chain management, uses the stability of mainline network traffic flow as an indicator, executing ramp closure and opening through zoning and sequencing. This scheme also centers on the expressway mainline; its control model does not consider the impact of off-ramp release on ground-level intersection saturation and lacks a mechanism to address potential feedback of ground-level congestion to upstream ramps, thus failing to achieve true "high-level two-way" coordination.

[0004] In summary, existing technologies generally suffer from fragmented control systems at the system level, with expressways and ground signal systems operating independently and lacking an integrated linkage mechanism. At the data level, they rely on traditional cross-sectional detection data, resulting in a single data dimension and failing to integrate multi-source information reflecting drivers' actual choices, such as internet trajectories and navigation behavior. This leads to insufficient ability to identify "ramp service targets" and difficulty in implementing differentiated and precise control. At the control strategy level, they are mostly based on rules or local feedback, exhibiting weak foresight, adaptability, and multi-objective balancing capabilities. Furthermore, existing systems lack effective user guidance linkage mechanisms, with control commands being disconnected from drivers' route selection behavior, making it difficult to improve control effectiveness through "flexible guidance." These technical problems collectively hinder further improvements in the overall operational efficiency of urban traffic. Summary of the Invention

[0005] To address the problems of existing ramp control methods, such as the disconnect between expressways and ground-level traffic systems, isolated and lagging control strategies, and the lack of precise and coordinated control capabilities based on multi-source data fusion and user behavior recognition, this invention provides a regional and elevated dual-linkage ramp control method based on multi-source data fusion, comprising:

[0006] S1: Collect multi-source heterogeneous traffic data, perform spatiotemporal standardization processing on the multi-source heterogeneous traffic data, and output a standardized multi-source traffic dataset;

[0007] S2: Based on the standardized multi-source traffic data, construct and update the fusion state matrix in real time. The fusion state matrix is ​​used to represent the traffic operation status of expressways and ground road networks in an integrated manner.

[0008] S3: Based on the fused state matrix, perform regional linkage control requirement identification and high-ground collaborative control requirement identification, and output structured regional linkage control requirements and high-ground collaborative control requirements;

[0009] S4: Taking the regional linkage control requirements and the high-ground coordinated control requirements as inputs, construct an optimization model based on the model predictive control method and solve it, and output the optimal control strategy sequence including ramp release rate, ground signal timing scheme and guidance strategy parameters;

[0010] S5: Encapsulate, distribute and execute the current cycle control instructions in the optimal control strategy sequence through a hierarchical control architecture, and collect status feedback data of instruction execution;

[0011] S6: Based on the state feedback data, evaluate the control effect and dynamically adjust the control model parameters according to the evaluation results to achieve closed-loop rolling updates.

[0012] Furthermore, in S1, the multi-source heterogeneous traffic data includes: roadside detector data, internet trajectory data, traffic signal system data, guidance and navigation platform data, and video AI recognition data; the spatiotemporal standardization processing includes: aligning data from different sources according to a unified sampling frequency and timestamp, and mapping them to a unified urban road topology map according to a geocoding system, with each type of raw data labeled as a standard quintuple data structure:

[0013]

[0014] in, This represents the smallest data unit in a unified format formed after spatiotemporal standardization. Indicates the device from which the data originates. To collect timestamps, For spatial coordinates or road segment numbers, For measured values ​​or indicators, For data types.

[0015] Furthermore, in S2, the formula for constructing the fusion state matrix M(t) is:

[0016]

[0017] in, This represents the state submatrix of the expressway mainline and ramps. This represents the state submatrix of signalized intersections and arterial roads in the ground area. This represents a submatrix of indexes for on / off ramp nodes. This represents the interaction effect matrix of the primary land system.

[0018] Furthermore, in S3, the activation criterion for the regional linkage control requirement is to meet one of the following conditions: the saturation of any key intersection in the region exceeds a preset first threshold, or the fluctuation of the main path traffic delay exceeds a preset second threshold, or there is an overflow phenomenon of vehicle queues spreading from one intersection to the upstream intersection; the high-ground coordinated control requirement identification includes at least one of the following types: ramp flow restriction requirement triggered to alleviate congestion on the main line of the expressway, coordinated control requirement triggered to prevent traffic flow from off-ramp causing congestion at ground intersections, and bidirectional coordinated control requirement triggered to solve the mismatch between the traffic capacity of on-ramp or off-ramp and the connecting ground intersection.

[0019] Furthermore, in S4, the optimization model includes: an equation describing the evolution of the system state and an objective function to be minimized;

[0020] The state evolution equation is:

[0021]

[0022] in, This represents the predicted state vector of the system at time t+1. This represents the current state vector of the system. This represents the control command vector. (⋅) represents the state transition function. Indicates the system disturbance term;

[0023] The objective function is:

[0024]

[0025] in, This indicates the current total system delay. This represents the sum of squares of the changes in the control variables. This indicates penalties for queuing at on-ramps or ground-level intersections that exceeds the tolerance threshold. , , These represent weighting coefficients, reflecting the priorities of different objectives. Describe the objective function. This represents a mathematical operator that optimizes the control variable vector u to minimize the objective function, and H represents the prediction time domain.

[0026] Furthermore, in S5, the hierarchical control architecture includes: a central control decision platform, edge execution agent nodes, and field controllers and guidance equipment terminals; the issued control commands are encapsulated into a strategy package containing a strategy identifier code, an effective timestamp, a ramp instruction sub-table, a signal control sub-table, and a guidance release sub-table; after the field controllers and guidance equipment terminals execute the commands, they collect the actual release rate, the green light execution time of each phase, and the guidance acceptance rate as the status feedback data and send it back.

[0027] Furthermore, in S6, the indicators used to evaluate the control effect include: ramp control effect execution deviation RES. i Mainline mitigation response index (MRC), surface congestion recovery index (IPAR), and incentive strategy acceptance rate (IPAR);

[0028] The dynamic adjustment of the control model parameters specifically involves updating the weighting coefficients in the optimization objective function using a gradient adjustment strategy based on the evaluation results. The update formula is as follows:

[0029]

[0030] in, This represents the updated weighting coefficients. This represents the weighting coefficients before the update. Indicates the learning rate. Describe the objective function Regarding weight The gradient.

[0031] The beneficial effects of this invention are:

[0032] By constructing a dual-linkage control framework of "regional linkage" and "high-ground coordination," urban expressways and ground signal systems are incorporated into a unified control system for the first time, effectively solving the long-standing problem of their separation. By integrating multi-source data (such as internet trajectories and signaling), high-precision panoramic traffic status perception and dynamic behavior recognition of ramp passage value are achieved, making the control strategy more accurate. The use of model predictive control algorithms for rolling optimization significantly improves the strategy's foresight, stability, and multi-objective balance capabilities. Attached Figure Description

[0033] Figure 1 This is a flowchart of the method described in this invention. Detailed Implementation

[0034] The technical solution of the present invention will be further described below with reference to embodiments, but it is not limited thereto. Any modifications or equivalent substitutions to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be covered within the protection scope of the present invention. In the following embodiments, process equipment or devices not specifically specified are all conventional equipment or devices in the art. Unless specifically specified, the technical means used in the embodiments of the present invention are all conventional means well known to those skilled in the art.

[0035] Example 1, combined with Figure 1 This embodiment describes a ramp control method based on multi-source data fusion, involving both regional and elevated traffic flow, comprising:

[0036] S1: Collect multi-source heterogeneous traffic data, perform spatiotemporal standardization processing on the multi-source heterogeneous traffic data, and output a standardized multi-source traffic dataset;

[0037] S2: Based on the standardized multi-source traffic data, construct and update the fusion state matrix in real time. The fusion state matrix is ​​used to represent the traffic operation status of expressways and ground road networks in an integrated manner.

[0038] S3: Based on the fused state matrix, perform regional linkage control requirement identification and high-ground collaborative control requirement identification, and output structured regional linkage control requirements and high-ground collaborative control requirements;

[0039] S4: Taking the regional linkage control requirements and the high-ground coordinated control requirements as inputs, construct an optimization model based on the model predictive control method and solve it, and output the optimal control strategy sequence including ramp release rate, ground signal timing scheme and guidance strategy parameters;

[0040] S5: Encapsulate, distribute and execute the current cycle control instructions in the optimal control strategy sequence through a hierarchical control architecture, and collect status feedback data of instruction execution;

[0041] S6: Based on the state feedback data, evaluate the control effect and dynamically adjust the control model parameters according to the evaluation results to achieve closed-loop rolling updates.

[0042] Furthermore, in S1, the multi-source heterogeneous traffic data includes: roadside detector data, internet trajectory data, traffic signal system data, guidance and navigation platform data, and video AI recognition data; the spatiotemporal standardization processing includes: aligning data from different sources according to a unified sampling frequency and timestamp, and mapping them to a unified urban road topology map according to a geocoding system, with each type of raw data labeled as a standard quintuple data structure:

[0043]

[0044] in, This represents the smallest data unit in a unified format formed after spatiotemporal standardization. Indicates the device from which the data originates. To collect timestamps, For spatial coordinates or road segment numbers, For measured values ​​or indicators, For data types.

[0045] Specifically, in this step, the system aims to build a multi-level integrated traffic state perception system, targeting expressways, ramps, ground intersections and road sections, and comprehensively accessing multi-source heterogeneous data, including roadside detectors, internet trajectories, signal systems, guidance and navigation platforms, and video AI recognition. By performing time synchronization, spatial mapping and unified encoding processing on all data sources, a standard five-tuple data structure is formed, thereby providing a standardized and scalable data foundation for subsequent data fusion, state analysis and integrated control.

[0046] Furthermore, in S2, the formula for constructing the fusion state matrix M(t) is:

[0047]

[0048] in, This represents the state submatrix of the expressway mainline and ramps. This represents the state submatrix of signalized intersections and arterial roads in the ground area. This represents a submatrix of indexes for on / off ramp nodes. This represents the interaction effect matrix of the primary land system.

[0049] Specifically, in this step, based on the road structure division of the expressway mainline, ramps, ground-level intersections, connecting road sections, and elevated connection points, the system extracts corresponding real-time status indicators for various traffic units: for the mainline, it extracts cross-sectional flow, occupancy rate, speed, and density; for ramps, it extracts queue length, cycle release number, and inflow compression rate; for ground-level intersections, it extracts entrance lane queue length, queue overflow rate, and saturation; for connecting road sections, it extracts unit travel time and congestion level based on speed-density; and for elevated connection points, it extracts the ratio of ramp exit flow to ground-level entrance lane saturation to quantify the matching relationship between release and capacity.

[0050] Subsequently, fusion computing and traffic state matrix construction were carried out, organizing various indicators into a fusion state matrix of expressway-ground road network according to spatial structure; finally, state assessment indicators were output and risk marking was performed, and comprehensive indicators such as mainline overload, off-ramp overflow, and ground intersection backflow were calculated:

[0051] The formula for calculating the overload index of the main line of the expressway is:

[0052]

[0053] in, This indicates the overload index of the expressway mainline. This indicates the section number of the main line of the expressway. Indicates the first Traffic capacity of each main line section Indicates the first The actual flow rate of each main line section, the overloaded section will trigger the compression of the on-ramp;

[0054] The formula for calculating the risk factor of surface overflow at the off-ramp is:

[0055]

[0056] in, This indicates the risk coefficient of ground overflow at the off-ramp. Indicates the effective length of the ground entrance passage. This indicates the queue length of the ground entrance road connected to the off-ramp; a coefficient greater than 0.8 is marked as high risk.

[0057] The formula for calculating the return flow index at ground intersections is:

[0058]

[0059] in, Indicates the return flow index at ground intersections. Indicates the total traffic flow at the intersection. This indicates the number of vehicles returning to the intersection, used to determine whether there is already ramp return traffic.

[0060] The above results will directly serve as the core input for the next stage of regional linkage demand analysis and highland coordinated control decision-making, and can also be updated in real time to support short-cycle prediction models.

[0061] Furthermore, in S3, the activation criterion for the regional linkage control requirement is to meet one of the following conditions: the saturation of any key intersection in the region exceeds a preset first threshold, or the fluctuation of the main path traffic delay exceeds a preset second threshold, or there is an overflow phenomenon of vehicle queues spreading from one intersection to the upstream intersection; the high-ground coordinated control requirement identification includes at least one of the following types: ramp flow restriction requirement triggered to alleviate congestion on the main line of the expressway, coordinated control requirement triggered to prevent traffic flow from off-ramp causing congestion at ground intersections, and bidirectional coordinated control requirement triggered to solve the mismatch between the traffic capacity of on-ramp or off-ramp and the connecting ground intersection.

[0062] Specifically, this step analyzes and determines whether to activate regional coordinated control and high-level coordinated control strategies from the perspectives of macro-level road network collaborative efficiency and micro-level bottleneck mitigation, forming a logical framework for control requirements, serving as an intermediary link connecting the perception layer and the decision-making layer. Control requirements analysis is divided into two parallel sub-tasks: regional coordinated control requirements identification and high-level coordinated control requirements identification. Regional coordinated control is used to coordinate signal timing at ground-level intersections and on-ramp traffic strategies to alleviate regional structural congestion. The system extracts ground-level intersection signal indicators based on the fused state matrix, forming an intersection-level saturation matrix.

[0063]

[0064] in, Represents the intersection-level saturation matrix. This represents an n-row, 1-column matrix of real number column vectors.

[0065] And by identifying the main origin-destination (OD) paths through trajectory clustering, control is initiated based on three criteria:

[0066] Region saturation threshold criterion:

[0067]

[0068] in, This represents the actual arrival flow at key intersection i. This represents the actual arrival flow at key intersection i. This represents the regional saturation threshold (typical value 0.85).

[0069] Criterion for decreased path reliability:

[0070]

[0071] in, Indicates the OD path travel delay. This indicates the average travel delay along the OD path. Indicates the path reliability threshold. This represents the variance of travel delay along the OD path.

[0072] And the criteria for cross-intersection coordinated flow interference. A regional coordinated control task item is generated when any one of the conditions is met.

[0073] The high-ground coordinated control demand identification is based on the extraction of core coupling indicators from the interaction submatrix, including off-ramp traffic flow, queue length, ground intersection saturation, and the ratio of on-ramp queue length to available entrance lane length. Control demands are triggered according to three types of models: mainline protection, ground overflow protection, and bidirectional matching coordination. After demand identification, the ramps and associated ground intersections are divided into linkage node groups. Finally, the S3 stage outputs two types of structured control demands: regional linkage demand and high-ground coordinated demand, which serve as inputs to the model predictive control solver to complete strategy scheduling and trigger the closed loop of the scheduling cycle.

[0074] Furthermore, in S4, the optimization model includes: an equation describing the evolution of the system state and an objective function to be minimized;

[0075] The state evolution equation is:

[0076]

[0077] in, This represents the predicted state vector of the system at time t+1. This represents the current state vector of the system. This represents the control command vector. (⋅) represents the state transition function. Indicates the system disturbance term;

[0078] The objective function is:

[0079]

[0080] in, This indicates the current total system delay. This represents the sum of squares of the changes in the control variables. This indicates penalties for queuing at on-ramps or ground-level intersections that exceeds the tolerance threshold. , , These represent weighting coefficients, reflecting the priorities of different objectives. Describe the objective function. This represents a mathematical operator that optimizes the control variable vector u to minimize the objective function, and H represents the prediction time domain.

[0081] Specifically, this step constructs a state-space model of an integrated urban expressway network. State variables include the mainline traffic flow status, ramp queue status, and surface intersection traffic status. Control variables include ramp release rates, off-ramp flow restriction status indicators, and surface signal timing parameters. The prediction time domain spans three to five control cycles. Within each cycle, the future state trajectory is predicted based on the previous state and control variables. Simultaneously, state constraints, control variable constraints, system coupling constraints, and feasible region limitations are set. A sequential quadratic programming or interior-point optimizer is used for rolling optimization, outputting the optimal control sequence for multiple future cycles. Only the control action for the current cycle is executed. All solution processes are completed within 30 to 60 seconds. The final output includes the ramp release rate decision vector, surface signal timing scheme, off-ramp flow restriction activation indicator, and guidance strategy parameters, achieving a closed-loop operation encompassing data perception, state assessment, strategy optimization, and control issuance.

[0082] Furthermore, in S5, the hierarchical control architecture includes: a central control decision platform, edge execution agent nodes, and field controllers and guidance equipment terminals; the issued control commands are encapsulated into a strategy package containing a strategy identifier code, an effective timestamp, a ramp instruction sub-table, a signal control sub-table, and a guidance release sub-table; after the field controllers and guidance equipment terminals execute the commands, they collect the actual release rate, the green light execution time of each phase, and the guidance acceptance rate as the status feedback data and send it back.

[0083] Specifically, this step adopts a distributed control architecture, with a three-tiered collaborative approach involving a central control decision platform, edge execution agent nodes, and field controllers and guidance equipment terminals to distribute policies. The central control decision platform deploys a dual-linkage model predictive control optimizer, a prediction module, and a policy scheduling scheduler, periodically generating policy instruction packages. Edge execution agent nodes possess intermediate caching, protocol conversion, and regional scheduling functions, responsible for boundary orchestration of local control modules. Field controllers and guidance equipment terminals, including ramp signal control boxes, ground intersection signals, guidance screens, information publishing terminals, and navigation plugins, are responsible for executing control commands and collecting status feedback. The system encapsulates control policies into policy packages containing policy identifiers, effective period numbers, timestamps, ramp instruction sub-tables, signal control sub-tables, guidance publishing sub-tables, and anomaly handling policy tables, which are then encrypted before being sent downwards. Control commands are driven by a unified clock, completing a calculation and distribution every 30 or 60 seconds, sequentially going through five stages: command generation, data relay, command activation, execution monitoring, and anomaly policy replacement, forming a stable and reliable execution process. The system is designed with a unified interface standard, compatible with various signal control systems, ramp control systems, guidance information systems, and urban traffic sensing platforms. Data semantic conversion, timestamp alignment, and format standardization are achieved through interface adapters. To ensure real-time execution and stability, the system employs time synchronization, transmission delay detection, command acknowledgment, redundant control channels, and fault-tolerant mechanisms to ensure that control commands take effect synchronously, transmission is reliable, and basic control strategies are automatically activated in case of faults, guaranteeing minimum safe passage capacity at core intersections and ramps.

[0084] Furthermore, in S6, the indicators used to evaluate the control effect include: ramp control effect execution deviation RES. i Mainline mitigation response index (MRC), surface congestion recovery index (IPAR), and incentive strategy acceptance rate (IPAR);

[0085] The dynamic adjustment of the control model parameters specifically involves updating the weighting coefficients in the optimization objective function using a gradient adjustment strategy based on the evaluation results. The update formula is as follows:

[0086]

[0087] in, This represents the updated weighting coefficients. This represents the weighting coefficients before the update. Indicates the learning rate. Describe the objective function Regarding weight The gradient.

[0088] Specifically, this step involves real-time data collection of ramp control execution status, ground intersection signal response, guidance strategy acceptance rate, and system event logs. A comprehensive score is then calculated based on the ramp control effect, mainline mitigation response, ground congestion recovery, and the effectiveness of the guidance strategy. Model parameters are dynamically adjusted using batch comparison and gradient adjustment methods based on the score results, and a strategy memory library is established to enable rapid reuse of historically high-scoring strategies. Continuous low-score warnings, guidance acceptance rate monitoring, and ground-level adverse effect identification provide early warnings of strategy degradation, forming a closed-loop rolling update mechanism encompassing state awareness, strategy generation, execution feedback, effect evaluation, and parameter optimization.

Claims

1. A ramp control method under regional and highland dual linkage of multi-source data fusion, characterized in that, include: S1: Collect multi-source heterogeneous traffic data, perform spatiotemporal standardization processing on the multi-source heterogeneous traffic data, and output a standardized multi-source traffic dataset; S2: Based on the standardized multi-source traffic data, construct and update the fusion state matrix in real time. The fusion state matrix is ​​used to represent the traffic operation status of expressways and ground road networks in an integrated manner. S3: Based on the fused state matrix, perform regional linkage control requirement identification and high-ground collaborative control requirement identification, and output structured regional linkage control requirements and high-ground collaborative control requirements; S4: Taking the regional linkage control requirements and the high-ground coordinated control requirements as inputs, construct an optimization model based on the model predictive control method and solve it, and output the optimal control strategy sequence including ramp release rate, ground signal timing scheme and guidance strategy parameters; S5: Encapsulate, distribute and execute the current cycle control instructions in the optimal control strategy sequence through a hierarchical control architecture, and collect status feedback data of instruction execution; S6: Based on the state feedback data, evaluate the control effect and dynamically adjust the control model parameters according to the evaluation results to achieve closed-loop rolling updates.

2. The ramp control method of claim 1, wherein, In S1, the multi-source heterogeneous traffic data includes: roadside detector data, internet trajectory data, traffic signal system data, guidance and navigation platform data, and video AI recognition data; the spatiotemporal standardization processing includes: aligning data from different sources according to a unified sampling frequency and timestamp, and mapping them to a unified urban road topology map according to a geocoding system, with each type of raw data labeled as a standard quintuple data structure: in, This represents the smallest data unit in a unified format formed after spatiotemporal standardization. Indicates the device from which the data originates. To collect timestamps, For spatial coordinates or road segment numbers, For measured values ​​or indicators, For data types.

3. The ramp control method of claim 1, wherein, In S2, the formula for constructing the fusion state matrix M(t) is: in, This represents the state submatrix of the expressway mainline and ramps. This represents the state submatrix of signalized intersections and arterial roads in the ground area. This represents a submatrix of indexes for on / off ramp nodes. This represents the interaction effect matrix of the primary land system.

4. The ramp control method based on multi-source data fusion and regional and high-altitude dual linkage according to claim 1, characterized in that, In S3, the activation criterion for the regional linkage control requirement is to meet one of the following conditions: the saturation of any key intersection in the region exceeds a preset first threshold, or the fluctuation of the main path traffic delay exceeds a preset second threshold, or there is an overflow phenomenon of vehicle queues spreading from one intersection to the upstream intersection; the high-ground coordinated control requirement identification includes at least one of the following types: ramp flow restriction requirement triggered to alleviate congestion on the main line of the expressway, coordinated control requirement triggered to prevent traffic flow from off-ramp causing congestion at ground intersections, and bidirectional coordinated control requirement triggered to solve the mismatch between the traffic capacity of on-ramp or off-ramp and the connecting ground intersection.

5. The ramp control method of claim 1, wherein, In S4, the optimization model includes: an equation describing the evolution of the system state and an objective function to be minimized; The state evolution equation is: wherein, denotes the predicted system state vector at time t+1, denotes the current system state vector, denotes the control command vector, (·) denotes the state transition function, denotes the system disturbance term; The objective function is: in, This indicates the current total system delay. This represents the sum of squares of the changes in the control variables. This indicates penalties for queuing at on-ramps or ground-level intersections that exceeds the tolerance threshold. , , These represent weighting coefficients, reflecting the priorities of different objectives. Describe the objective function. This represents a mathematical operator that optimizes the control variable vector u to minimize the objective function, and H represents the prediction time domain.

6. The ramp control method based on multi-source data fusion and regional and high-altitude dual linkage according to claim 1, characterized in that, In S5, the hierarchical control architecture includes: a central control decision platform, edge execution agent nodes, and field controllers and guidance equipment terminals; the issued control commands are encapsulated into a strategy package containing a strategy identifier code, an effective timestamp, a ramp instruction sub-table, a signal control sub-table, and a guidance release sub-table; after the field controllers and guidance equipment terminals execute the commands, they collect the actual release rate, the green light execution time of each phase, and the guidance acceptance rate as the status feedback data and send it back.

7. The ramp control method of claim 1, wherein, In S6, the indicators used for evaluating the control effect include: ramp metering effect execution deviation RES i , main line relief response indicator MRC, and ground congestion recovery indicator and induction strategy acceptance rate IPAR. The dynamic adjustment control model parameter is specifically updating the weighting coefficient in the optimization objective function according to the evaluation result by using a gradient adjustment strategy, and the updating formula is: wherein, denotes the updated weight coefficient, denotes the weight coefficient before update, denotes the learning rate, denotes the objective function with respect to the weight of the gradient.

Images