A multi-modal fusion urban traffic intelligent agent system construction method

Abstract

The application relates to the field of road traffic control and discloses a multi-modal fusion urban traffic intelligent agent system construction method, which comprises the following steps: acquiring multi-modal traffic data in a target road network and mapping the multi-modal traffic data to intersections, lane groups, phases and directed road segments; correcting the delay of each data source to generate traffic event segments; performing consistency checking on the traffic event segments in the same lane group and the same control period, generating conflict freezing vouchers when conflicts exist, and not directly combining the conflict events into single traffic states; and establishing a demand temporary account according to observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, served demand and unserved demand. Through the traffic event segments, the conflict freezing vouchers, the demand temporary account and the cross-intersection traffic allowance acceptance linkage, the signal timing deviation caused by shielding, delay and downstream overflow is reduced.

Description

Technical Field

[0001] This invention relates to the field of road traffic control technology, specifically to a method for constructing a multimodal fusion urban traffic intelligent system. Background Technology

[0002] With increasing traffic flow at urban road intersections, single roadside detection devices can hardly fully reflect queuing at approach lanes, vehicle arrivals, stop line crossings, and downstream acceptance status. To enable traffic signal control to adjust based on a more complete picture of road operation, it is necessary to process multi-source data such as video, radar, geomagnetic sensors, loop detectors, V2X, bus positioning, or floating car trajectories, and to construct an urban traffic intelligent system suitable for collaborative control of intersection groups.

[0003] In existing technologies, data such as vehicle position, speed, queue length, occupancy rate, number of passing vehicles, and phase status are typically collected through roadside sensing devices and traffic signal controllers. Then, data from different sources are processed through time synchronization, spatial matching, target association, or confidence weighting to form traffic state parameters at the lane group, phase, or road segment level. These traffic state parameters are then input into signal timing control, arterial coordination control, or multi-agent cooperative control models to output commands for green light extension, phase switching, or coordination control.

[0004] While existing technologies can utilize multi-source data to form traffic conditions and participate in signal control, some shortcomings remain. Due to inconsistencies in sampling frequency, detection area, transmission delay, and spatial mapping accuracy among different detection devices, conventional fusion methods can easily weight conflicting data into a single state when video occlusion, radar false detections, loop counting lags, or low-frequency trajectory data updates occur. This weakens or eliminates the demand from vehicles that should have arrived but were not observed. Furthermore, existing cross-intersection coordination relies heavily on upstream release capacity or downstream congestion status for control, lacking mechanisms for freezing and retaining conflicting data, temporarily storing unobserved demands, accepting traffic quotas, and providing post-execution settlement and write-back. This makes it difficult to continuously constrain upstream release responsibility and downstream acceptance capacity. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for constructing a multimodal fusion urban traffic intelligent agent system. This method solves the problem that existing urban traffic control systems struggle to retain, release, and write back unobserved traffic demands when multimodal traffic data is obstructed, delayed, conflicting, or has uncertain cross-intersection acceptance capacity. This results in a lack of continuous and traceable control basis for the construction of traffic intelligent agents and signal control decisions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for constructing a multimodal fusion urban transportation intelligent agent system, comprising the following steps: Acquire multimodal traffic data from the target road network and map it to intersections, lane groups, phases, and directed road segments; Delay correction is performed on each data source to generate traffic event segments; Consistency checks are performed on traffic event segments within the same lane group and the same control cycle. When conflicts exist, conflict freeze certificates are generated, and conflicting events are not directly merged into a single traffic state. Establish a demand temporary ledger based on observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, served demand, and unserved demand. The initial passage quota is generated from the demand temporary storage account, the acceptance request is generated for the initial passage quota released across the intersection, and the final passage quota is obtained according to the acceptance status. The final passage quota generates a signal to control the action, and settlement is performed after execution.

[0007] Preferably, the step of acquiring multimodal traffic data in the target road network and mapping it to intersections, lane groups, phases, and directed road segments includes: establishing a road object configuration table, a data source configuration table, and an intersection association configuration table; The road object configuration table records intersections, lane groups, phases, directed road segments, upstream source relationships, turning relationships, and downstream receiving objects; The data source configuration table records the data source number, data source type, installation location, detection range, corresponding lane group, delay parameters, and initial reliability value. Based on the equipment detection area, lane group spatial range, phase configuration, and steering relationship, data output from at least two of the following data sources—video detection equipment, millimeter-wave radar, geomagnetic detector, coil detector, V2X roadside unit, bus positioning access equipment, and floating car trajectory access equipment—are mapped to the corresponding lane group.

[0008] Preferably, the traffic event segment includes event number, data source number, modality type, lane group number, phase number, downstream receiving object, receiving time window, event time window, spatial location, vehicle quantity range, speed range, queue length, data source reliability, delay parameter, environmental impact factor, and event type. The event time window is obtained by correcting the reception time window according to the delay parameter, and the event types include arrival events, queuing events, stop line passing events, signal phase events, priority traffic events, and abnormal events.

[0009] Preferably, the consistency verification of traffic event segments within the same lane group and the same control cycle includes vehicle quantity range verification, queue length verification, time window verification, spatial attribution verification, and traffic conservation verification. When there are non-overlapping vehicle quantity ranges, queue length conflicts exceeding the preset tolerance, event time window offsets exceeding the preset time tolerance, spatial attribution verification results inconsistent with the original mapped lane group, stop line throughput and queue changes not satisfying the traffic conservation relationship, or upstream expected arrivals inconsistent with local confirmed arrivals, a conflict freeze certificate is determined to exist in the traffic event segment.

[0010] Preferably, the conflict freeze certificate includes a certificate number, lane group number, conflict time window, associated event segment number, vehicle number conflict quantity, queue length conflict quantity, time window offset, conflict type, conflict confidence level, freeze status, freeze requirement quantity, and effective time window. The conflict types include obstruction conflicts, delay conflicts, spatial mismatch conflicts, inconsistency in stop line throughput conflicts, and inconsistency in upstream release throughput conflicts; The freezing demand corresponding to conflict-freezing vouchers that are in the pending status is recorded in the demand temporary storage account. Conflict-freezing vouchers that are in the cancelled status or have been transferred to the abnormal status will no longer participate in the passage quota calculation.

[0011] Preferably, the demand temporary ledger includes lane group number, observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, served demand, abnormal transfer demand, unserved demand, and ledger status. The potential arrival demand is determined by the upstream estimated arrival volume, the locally confirmed number of vehicles, and the stop line throughput, and is used to represent the traffic demand that has been released upstream and should enter the current lane group according to the travel time but has not yet been confirmed by the local traffic event segment; The freeze request is determined by conflict freeze credentials that are in a pending state; The unserved demand is used to reserve traffic demand that has not been released by the final passage quota or has not been actually served through event confirmation.

[0012] Preferably, after establishing the demand temporary ledger, the process further includes constructing a traffic intelligent agent; The traffic intelligent agent includes lane group intelligent agent, phase intelligent agent, corridor intelligent agent and data anomaly intelligent agent; The lane group intelligent agent takes a single lane group as the control responsibility boundary and maintains traffic event fragments, conflict freeze certificates, and demand temporary ledgers. The phase agent takes the set of lane groups controlled by the same phase as the control responsibility boundary and summarizes the demand status of the corresponding lane groups; The corridor intelligent agent takes the upstream intersection, downstream intersection and directional road segment as the control responsibility boundary to handle the acceptance of passage quota and arrival settlement; The data anomaly agent takes the data source or detection area as the control responsibility boundary, outputs reliability correction and anomaly markers, and does not directly output signal control actions to traffic signal controllers.

[0013] Preferably, the step of generating an initial passage quota from the demand reserve account and generating an acceptance request for the initial passage quota released across intersections includes: determining the number of vehicles to be released for a lane group based on the book demand in the demand reserve account, downstream acceptance space, saturation release rate and allocable green light time. When the vehicles corresponding to the quota of vehicles to be released are expected to enter the downstream receiving objects within the expected arrival time window, the upstream corridor intelligent agent generates an acceptance request. The acceptance request includes the acceptance request number, upstream intersection number, downstream intersection number, upstream release lane group number, quota of vehicles to be released, estimated arrival time window, acceptance request credibility level, acceptance status, and acceptance constraint reason.

[0014] Preferably, obtaining the final passage quota based on the acceptance status includes: the downstream corridor agent determining the downstream acceptable amount based on the downstream acceptance space, reserved occupancy capacity, potential arrival demand and frozen demand, and returning the acceptance status to the upstream corridor agent based on the downstream acceptable amount. The acceptance status includes full acceptance, partial acceptance, delayed acceptance, and rejection of acceptance; When the acceptance status is full acceptance, the final passage quota corresponds to the initial passage quota. When the acceptance status is partial acceptance, the final passage quota corresponds to the downstream acceptable volume, and the unaccepted demand is written into the unserved demand of the upstream lane group. When the acceptance status is "acceptance suspended" or "acceptance refused", the release of the corresponding lane group across the intersection is restricted, and the reason for the acceptance constraint is written into the corridor agent log.

[0015] Preferably, the action of generating the signal control from the final passage quota and performing settlement write-back after execution includes: the phase agent summarizing the final passage quota of each lane group under the same phase control to obtain the phase final passage quota; The green light request time is determined based on the final phase passage quota and the phase saturation release rate. Candidate control actions are determined from maintaining the current phase, extending the current phase, shortening the current phase, switching to the next phase, skipping phases with no service demand, and performing conservative control actions; Candidate control actions that do not meet the minimum green light time, maximum green light time, yellow light time, all-red time, pedestrian clearance time, phase conflict relationship, downstream acceptance constraint and acceptance constraint are eliminated, and the retained candidate control actions are converted into signal control actions. Within the travel time window after the signal control action is executed, stop line passing events and downstream arrival events are collected. The actual number of passing vehicles and the actual number of arriving vehicles are matched with the final passage quota, acceptance request and demand temporary account, respectively. Based on the matching results, the freeze status of conflict freeze vouchers, demand temporary account, data source reliability, travel time parameters and saturation release rate are updated.

[0016] This invention provides a method for constructing a multimodal fusion urban transportation intelligent agent system. It has the following beneficial effects: 1. This invention maps multimodal traffic data, such as video detection, millimeter-wave radar, geomagnetic or coil, V2X, bus positioning, and floating car trajectories, to corresponding intersections, lane groups, phases, and directed road segments, and generates traffic event segments containing time windows, spatial locations, vehicle quantity ranges, speed ranges, queue lengths, reliability, and delay parameters. This ensures that data from different sources have a unified data structure and correspondence before entering control calculations, facilitating subsequent consistency verification, responsibility tracing, and control state generation. It also avoids direct impact on signal control actions due to inconsistencies in time references and spatial attributions of different modal data.

[0017] 2. This invention performs quantity range, queue length, time window, spatial attribution, and traffic conservation checks on traffic event segments within the same lane group and control cycle. When conflicts exist, it generates conflict freeze certificates, temporarily storing conflict states caused by occlusion, delay, spatial mismatch, or inconsistency between upstream release volume and local confirmation volume, instead of directly synthesizing conflict data into a single traffic state. At the same time, the conflict freeze certificates are associated with the demand temporary storage account, so that unconfirmed demands with traffic flow constraints can continue to participate in subsequent traffic quota calculations, thereby reducing control judgment bias caused by local perception loss or modal conflict.

[0018] 3. This invention establishes a demand reserve ledger based on observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, and cancelled demand. It then generates the final passage quota by combining downstream acceptable space, acceptance status, and signal safety constraints, thus establishing a correspondence between the release actions of upstream intersections and the acceptance conditions of downstream intersections. After the control actions are executed, the passage quota, acceptance request, conflict freeze certificate, and demand reserve ledger are settled and written back based on the stop line passing event and downstream arrival event. This facilitates the correction of data source reliability, travel time parameters, and saturation release rate, forming a closed loop of traffic agent state generation, quota release, and subsequent correction. Attached Figure Description

[0019] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a schematic diagram illustrating the generation of traffic incident fragments and conflict freeze certificates according to the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0021] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0022] Example: See attached document Figure 1 -Appendix Figure 2 This invention provides a method for constructing a multimodal fusion urban traffic intelligent system, applicable to an urban road traffic control system including signalized intersections, approach lane groups, exit lane groups, roadside sensing devices, traffic signal controllers, edge computing nodes, and a traffic control platform. The roadside sensing devices include at least two types of video detection equipment, millimeter-wave radar, geomagnetic detectors, loop detectors, V2X roadside units, bus positioning access devices, and floating car trajectory access devices.

[0023] Roadside sensing devices are used to collect vehicle location, speed, trajectory, queue length, stop line throughput, vehicle category, or priority traffic requests. Traffic signal controllers provide phase status, phase switching records, and signal control interfaces. Edge computing nodes perform event fragment generation, conflict freezing, demand storage, traffic quota calculation, acceptance judgment, and settlement write-back. The traffic control platform stores road topology, lane group configuration, phase configuration, equipment configuration, and operation logs, and issues control parameters or control commands to the signal control equipment.

[0024] In this invention, the approach directions of intersections are divided into lane groups according to their turning relationships. These lane groups serve as the basis for traffic demand calculation and control responsibility allocation. Directed road segments between adjacent intersections are used as cross-intersection traffic quota redemption objects. Each lane group corresponds to an upstream source relationship and a downstream acceptance relationship; the upstream source relationship is used to calculate the potential demand for vehicles released upstream to reach the current lane group, and the downstream acceptance relationship is used to determine whether vehicles released from the current lane group are constrained by downstream parking space.

[0025] The traffic intelligent agent in this invention is a software functional unit running on an edge computing node or traffic control platform, including a lane group intelligent agent, a phase intelligent agent, a corridor intelligent agent, and a data anomaly intelligent agent. The lane group intelligent agent maintains traffic event fragments, conflict freeze certificates, and demand temporary ledgers for lane groups; the phase intelligent agent summarizes the demands of each lane group under the same phase and generates phase control actions; the corridor intelligent agent handles traffic quota acceptance, reservation occupancy, and arrival settlement between adjacent intersections; the data anomaly intelligent agent handles data source reliability correction, delay anomalies, obstruction anomalies, and conflict transfer, but does not directly generate signal control actions.

[0026] The method of the present invention includes the following steps: S100. Acquire multimodal traffic data in the target road network and map the traffic data to the corresponding intersections, lane groups, phases, and directed road segments.

[0027] S200: Perform delay correction on the reception time of each data source, and generate a traffic event segment in a unified format based on the device detection area, lane group spatial range, phase configuration, and steering relationship. The traffic event segment includes event number, data source number, modality type, lane group number, reception time window, event time window, spatial location, vehicle quantity range, speed range, queue length, data source reliability, delay parameter, environmental impact factor, and event type.

[0028] S300: Perform consistency checks on traffic event segments within the same lane group and the same control cycle. When there are discrepancies such as non-overlapping quantity ranges, inconsistent queue lengths, mismatched time windows, inconsistent spatial attribution, inconsistent stop-line throughput and queue changes, or inconsistent upstream release and local confirmation, a conflict freeze certificate is generated. The conflict freeze certificate includes the certificate number, lane group number, conflict time window, associated event segment number, conflict type, conflict confidence level, and freeze status.

[0029] S400 establishes a demand reserve ledger based on observable queuing demand, potential arrival demand, frozen demand, equivalent demand for priority traffic, and cancelled demand. It then generates an initial passage quota based on the demand reserve ledger, data conflict risk, downstream capacity, saturation release rate, and allocable green light time. For passage quotas that will enter downstream intersections within the expected arrival time window, the upstream corridor agent generates acceptance requests. The downstream corridor agent returns a status of full acceptance, partial acceptance, deferred acceptance, or rejection based on parking capacity, occupied capacity, reserved occupancy, potential arrival demand, and frozen demand. The upstream agent adjusts the final passage quota based on the acceptance status.

[0030] S500 generates signal control actions based on the final passage quota, satisfying minimum green, maximum green, yellow light, all red, pedestrian clearance, phase conflict, downstream acceptance, and acceptance constraints. After the control actions are executed, stop line passage events and downstream arrival events are collected within the corresponding travel time window. The passage quota, acceptance request, conflict freeze certificate, and demand temporary account are settled, and the data source reliability, travel time parameters, and saturation release rate are updated.

[0031] The technical implementation details of the above steps are explained in detail below with reference to specific embodiments: Before acquiring multimodal traffic data, the system establishes a road object configuration table, a data source configuration table, and an intersection association configuration table. The road object configuration table records at least the intersection, lane group, phase, directed road segment, upstream source relationship, turning relationship, and downstream receiving object; the data source configuration table records at least the data source number, data source type, installation location, detection range, corresponding lane group, delay parameter, and initial reliability value.

[0032] The target road network is represented as: In the formula, Represents the target road network. Represents the set of intersections. This represents a set of directed road segments. A directed road segment indicates the driving connection relationship between vehicles at adjacent intersections.

[0033] The set of lane groups at an intersection is represented as: In the formula, Indicates an intersection The set of lane groups, Indicates an intersection The Each lane group Indicates an intersection The number of lane groups. Lane groups are divided according to the direction of entry, lane attributes, and steering relationship. Vehicles in the same lane group have the same or can be processed together by signal control.

[0034] The control relationship between phases and lane groups is determined by a phase configuration table. The phase configuration table records the controlled lane groups corresponding to each phase, used to map signal phase records to lane groups. The set of lane groups controlled by phase is represented as follows: In the formula, Indicates phase Controlled lane group set, Indicates phase With lane group Control relationship; when At that time, the lane group Phase Control; when At that time, the lane group Unaffected by phase control.

[0035] The upstream source lane group set of the current lane group is represented as: In the formula, Indicates the ability to direct traffic to lane groups The upstream lane group of the transport vehicles is assembled. Indicates the upstream lane group, Indicates the upstream lane group With lane group The source relationship is determined by road topology, inlet channelization, and steering configuration, and is used to subsequently calculate the potential arrival demand corresponding to upstream released vehicles.

[0036] The number of source vehicles mapped from upstream release vehicles to the current lane group is represented as: In the formula, Indicates the upstream lane group Mapped to lane group The number of vehicles originating from, Indicates the upstream lane group The number of vehicles released Indicates the upstream lane group To the lane group Steering ratio, This indicates the source relationship between the two. The steering ratio can be determined by historical steering flow, channelization configuration, or detector statistics. The specific statistical methods are well-known in the field and will not be elaborated here.

[0037] The correspondence between lane groups and downstream receiving objects is expressed as follows: In the formula, Indicates lane group The set of downstream receiving objects that the vehicle enters after it is released. This indicates the downstream exit channel, road section, or queuing storage area. Indicates lane group With downstream recipients The mapping relationship is used for subsequent calculations of downstream acceptable space and serves as input for cross-intersection traffic quota acceptance.

[0038] The data source collection is represented as: In the formula, Represents a collection of data sources. Indicates the first One data source, Indicates the number of data sources. Data sources include at least two of the following: video detection equipment, millimeter-wave radar, geomagnetic detectors, coil detectors, traffic signals, V2X roadside units, bus positioning access equipment, and floating car trajectory access equipment.

[0039] The raw records of the data source are represented as follows: In the formula, Indicates data source The One original record, Indicates the data source number. Indicates the data source type. Indicates the system's receiving time. Indicates spatial location, Indicates speed information, Indicates the number of vehicles or trigger count. Indicates the queue length. Represents signal phase information. Indicates the target number or vehicle number. This indicates the vehicle category or priority attribute. Fields not provided by different data sources are set to null values ​​and proceed to subsequent processing according to the data source type.

[0040] The system completes the original record access verification according to the data source configuration table. For device protocol parsing, timestamp reading, and communication status determination, those skilled in the art can complete these tasks based on the device interface protocol; their implementation is well-known technology and will not be elaborated upon here.

[0041] Data latency is represented as: In the formula, Indicates data source Data latency, Indicates the time when the system receives data. This indicates the time when the event corresponding to the data occurred. For data sources that cannot directly provide the time of event occurrence, the system uses configured latency parameters or the most recent valid latency estimate to determine the data latency. The corrected event time is represented as: In the formula, Indicates data source The timing of the correction event.

[0042] After time correction, the system maps the original records to lane groups, phases, and downstream receiving objects. Point-based data determines lane groups based on the inclusion relationship between the detection location and the spatial range of the lane group; trajectory-based data determines lane groups based on the number of trajectory points belonging to each lane group's spatial range; signal data determines associated lane groups according to the phase configuration table; upstream release data is mapped to the current lane group based on source relationship and steering ratio. The above spatial matching is a standard implementation in traffic detection data processing, and its algorithm formula is not given separately here.

[0043] The mapping result of the original record is represented as follows: In the formula, Represents the original record The mapping results Indicates the time of the correction event. Indicates the mapping of lane groups. Indicates the mapped phase. This indicates the downstream receiving object of the mapping.

[0044] When an original record can be mapped to a unique lane group, the system uses this mapping result as input for generating traffic event segments. When an original record corresponds to multiple candidate lane groups, the system determines the final lane group based on spatial matching results, trajectory direction, phase control relationship, and data source installation direction. When an original record cannot be mapped to any lane group, the system marks it as a spatially mismatched record and it does not participate in the traffic quota calculation for the current control cycle. Through the above processing, the system associates multimodal traffic data with intersections, lane groups, phases, and downstream acceptance objects, providing a data foundation for subsequent traffic event segment generation, conflict freeze certificate generation, demand temporary account establishment, and traffic quota acceptance.

[0045] After acquiring multimodal traffic data and mapping road objects, the system processes the mapping results in control cycles, converting raw records from different data sources into traffic event fragments in a unified format. These traffic event fragments represent traffic state records formed within the same lane group, the same detection area, or the same phase relationship within a corresponding time window.

[0046] The traffic incident segment is represented as follows: In the formula, This represents a segment of a traffic incident. Indicates the event number. Indicates the data source number. Indicates the data source type. Indicates the lane group number. Indicates the phase number, Indicates the downstream receiving object. Indicates the start time of the reception time window. Indicates the end time of the receiving time window. Indicates the start time of the event time window. Indicates the end time of the event time window. Indicates spatial location, Indicates a range of vehicle numbers. Indicates the speed range. Indicates the queue length. Indicates the reliability of the data source. Indicates data latency. Indicates environmental impact factors, Indicates the event type.

[0047] The vehicle quantity range and speed range are represented as follows: In the formula, This indicates the lower bound of the number of vehicles. This indicates the upper limit of the number of vehicles. Indicates the lower bound of velocity. This indicates the upper bound of the speed. Fields not provided by different data sources do not participate in the calculation of the corresponding event type and are retained as null or default values ​​in the event fragment; this processing is used to ensure that video, radar, coil, geomagnetic, signal, and trajectory data can all be included in the same event fragment format.

[0048] Event types include arrival events, queuing events, stop line crossing events, signal phase events, priority traffic events, and abnormal events. Video detection data generates arrival events or queuing events based on vehicle targets, trajectory directions, stop line positions, and low-speed status within the detection area; radar data generates arrival events or low-speed target events based on target distance, speed, heading, and target number; coil or geomagnetic data generates stop line crossing events based on trigger time, occupancy status, and passage count; signal data generates signal phase events based on phase number, phase status, and phase switching records; V2X, bus positioning, or floating car trajectory data generates arrival events or priority traffic events based on vehicle position, speed, heading, vehicle type, and vehicle number. Radar target tracking, coil trigger counting, and video target detection can be performed using existing traffic detection processing methods, which are well-known technologies in the field and will not be elaborated upon here.

[0049] The event time window is determined based on the reception time window and data latency. Specifically, the event time window is obtained by subtracting the corresponding data source latency from the reception time window. This event time window is used to determine the control cycle to which the event segment belongs, and is not directly used as the basis for multimodal data alignment based on the system reception time.

[0050] The system performs consistency checks on event segments within the same lane group and the same control cycle. Consistency checks include vehicle quantity interval checks, queue length checks, time window checks, spatial attribution checks, and traffic conservation checks. Vehicle quantity interval checks are used to determine whether there is overlap in the descriptions of vehicle quantities for the same lane group by different modalities.

[0051] Let the ranges of vehicle numbers for the two event segments be: In the formula, This indicates the range of vehicle numbers in the first event segment. This represents the lower bound of the number of vehicles in the first event segment. This represents the upper bound of the number of vehicles in the first event segment. This indicates the range of vehicle numbers in the second event segment. This represents the lower bound of the number of vehicles in the second event segment. This represents the upper bound of the number of vehicles in the second event segment. A vehicle number conflict is determined to exist when the following formula is satisfied: The number of vehicle conflicts is expressed as follows: In the formula, This indicates the number of vehicle conflicts. , , , The meaning is the same as above.

[0052] Queue length verification determines whether there is a queue length conflict based on the difference in queue length between different event segments and the preset tolerance. Time window verification determines whether there is a time window conflict based on whether the event time windows of two event segments have effective overlap. Spatial attribution verification verifies the lane group attribution of event segments based on the lane group spatial range, detection area, trajectory direction, and phase configuration; its point matching and trajectory matching are well-known techniques in traffic detection data processing and will not be elaborated here.

[0053] Traffic conservation checks are used to determine the consistency between upstream inputs, stop-line outputs, and queue changes. Upstream expected arrivals are determined by the number of vehicles released upstream, turning ratios, source relationships, and travel time shifts; stop-line throughput is determined by stop-line passage events; and queue changes are determined by the difference between the observable queue numbers in the current control cycle and the previous control cycle. The traffic conservation residual is expressed as: In the formula, This represents the traffic conservation residual. This indicates the expected arrival volume from upstream. Indicates the amount of time the stop line has passed. This indicates the change in the queue.

[0054] When vehicle quantity conflicts, queue length conflicts, time window conflicts, spatial attribution conflicts, or traffic conservation residuals meet the corresponding conflict conditions, the system outputs a conflict judgment result. The conflict judgment result includes at least the conflict type, associated event number, lane group number, conflict time window, and corresponding conflict quantity, and serves as the input for subsequently generating conflict freeze certificates; this output does not overwrite the original traffic event segment.

[0055] After generating and verifying traffic event fragments, the system does not directly merge conflicting event fragments into a single traffic state value. Instead, it generates a conflict freeze certificate based on the conflict determination result. The conflict freeze certificate records the source of the conflict, related events, conflict time window, conflict amount, freeze requirement, and processing status, and serves as input for the requirement temporary ledger and reliability write-back.

[0056] The conflict freeze certificate is represented as follows: In the formula, Indicates a conflict-freezing certificate. Indicates the voucher number. Indicates the lane group number. Indicates the conflict time window. Indicates the number of the first associated event segment. Indicates the number of the second related event segment. This indicates the number of vehicle conflicts. Indicates the number of queue length conflicts. Indicates the time window offset. Indicates the type of conflict. Indicates the confidence level of the conflict. Indicates a frozen state. This indicates the amount of frozen demand. This indicates the valid time window of the certificate.

[0057] The conflict time window is represented as: In the formula, Indicates the start time of the conflict time window. This indicates the end time of the conflict time window. The conflict time window is determined by the event time windows of the related event segments; for two related event segments, the conflict time window takes the coverage area of ​​both event time windows.

[0058] S310. The system reads the conflict judgment result output by the consistency check. The conflict judgment result includes lane group number, associated event segment number, conflict type candidate, conflict time window, vehicle number conflict amount, queue length conflict amount, and time window offset. The system reads the corresponding traffic event segment according to the associated event segment number and retains the original event segment fields. The conflict freeze certificate is saved as an independent data object and does not overwrite the original traffic event segment.

[0059] S320. The system determines the triggering conditions for conflict freeze certificates. When there are non-overlapping vehicle quantity ranges, queue length conflict exceeding the preset tolerance, event time window offset exceeding the preset time tolerance, spatial attribution verification results inconsistent with the original mapped lane group, stop line throughput and queue change not satisfying the traffic conservation relationship, or upstream expected arrival volume inconsistent with local confirmed volume, the system generates a conflict freeze certificate.

[0060] S330, The system determines the conflict type. The conflict type is represented as follows: In the formula, Indicates occlusion-type conflict. Indicates a delayed conflict. This indicates a spatial mismatch conflict. This indicates a conflict caused by inconsistent passage volume at the stop line. This indicates a conflict caused by inconsistent upstream release volumes.

[0061] A conflict is classified as occlusion-type when the number of vehicles in a video event segment is lower than that of radar, coil, or geomagnetic event segments, and the video event segment contains occlusion markers, strong backlight markers, rain / fog markers, or target loss markers. A conflict is classified as delay-type when the event time window is offset and the data delay of at least one data source exceeds the corresponding tolerance. A conflict is classified as spatial mismatch-type when the spatial location, trajectory direction, or detection area of ​​the event segment is inconsistent with the mapped lane group. A conflict is classified as stop line throughput inconsistency-type when the stop line throughput, queue change, and upstream expected arrival do not satisfy the conservation relationship. A conflict is classified as upstream release inconsistency-type when upstream released vehicles should enter the current lane group after travel time shift, but no corresponding arrival event or queue event has formed locally.

[0062] S340, The system calculates the conflict trust level. The conflict trust level is expressed as: In the formula, Indicates the confidence level of the conflict. This indicates the reliability of the data source corresponding to the first associated event fragment. This indicates the reliability of the data source corresponding to the second associated event fragment. Indicates the conflict type weight. Indicates the time window offset. This represents the environmental impact factor. When a conflict is triggered by traffic conservation relationships or upstream release relationships and there are no two directly related event segments,... The overall reliability of locally confirmed event fragments is taken. Take the combined reliability of the upstream release event or the stop line pass event.

[0063] S350, the system maps conflict freeze credentials to freeze requests. For vehicle quantity conflicts, the freeze request is represented as: In the formula, This indicates the amount of frozen demand. Indicates the confidence level of the conflict. This indicates a higher estimate of the number of vehicles involved in the conflict. This indicates a lower estimate of the number of vehicles involved in the conflict.

[0064] For conflicts arising from inconsistent upstream release volumes, the frozen demand is represented as follows: In the formula, This indicates the amount of frozen demand. Indicates the confidence level of the conflict. This indicates the expected arrival volume from upstream. This indicates the number of locally confirmed vehicles. This indicates the amount of time the stop line has passed.

[0065] S360, System settings are frozen. The frozen state is indicated as follows: In the formula, Indicates a pending status. This indicates that the data has been released to the demand staging account. This indicates that the transaction has been cancelled. This indicates that the system has entered an abnormal state.

[0066] When a conflict-freezing certificate is generated, it is set to a pending status. If a subsequent event confirms that the vehicle corresponding to the freeze request exists but service has not yet been completed, the status is updated to "Released to the Request Temporary Account." If a stop-line pass event or downstream arrival event has already covered the freeze request, the status is updated to "Write-off." If no stop-line pass event, downstream arrival event, or local confirmation event is received within the valid time window, the status is updated to "Transferred to Abnormal Status."

[0067] The frozen state transition relationship is represented as follows: In the formula, , , , The meaning is the same as above.

[0068] S370. The system establishes a relationship between conflict-freezing vouchers and the demand staging ledger. Conflict-freezing vouchers in the pending state are included in frozen demands; conflict-freezing vouchers that have been released to the demand staging ledger are transferred to pending service demands and are no longer included in frozen demands; conflict-freezing vouchers that have been cancelled or transferred to an abnormal state are no longer included in demand calculation.

[0069] The lane group freeze requirement is expressed as follows: In the formula, This indicates the need to freeze lane groups. This indicates the number of conflict freeze certificates associated with this lane group. Indicates the first The required amount of freeze corresponding to each conflicting freeze certificate. Indicates the first The frozen status of a conflicting frozen certificate. Indicates an indicator function; takes the value when the condition is true. Otherwise take .

[0070] After the above processing, the conflict freeze certificate retains the conflicts between multimodal event fragments as settleable data objects, and provides the pending confirmation requirements to the requirement staging account in the form of frozen requirements.

[0071] After generating the conflict freeze certificate, the system establishes a demand temporary storage account according to the lane group, and constructs lane group intelligent agents, phase intelligent agents, corridor intelligent agents, and data anomaly intelligent agents with the demand temporary storage account, traffic event fragments, conflict freeze certificates, and road object configurations as inputs.

[0072] The demand stub is used to record traffic demand that lane groups need to retain for quota calculation and subsequent settlement during the control period. This demand includes locally confirmed queuing demand, potential arrival demand that has not yet been locally confirmed after upstream release, frozen demand formed by conflict freeze certificates, and equivalent demand formed by priority traffic objects.

[0073] S410. The system establishes a temporary demand ledger for lane groups. The temporary demand ledger is represented as follows: In the formula, This indicates a temporary storage account for demand. Indicates the lane group number. This indicates observable queuing demand. Indicates potential arrival demand. This indicates a freeze on demand. Indicates priority traffic equivalent demand. This indicates that the service request has been fulfilled. Indicates an abnormal transfer request. Indicates unserved demand. This indicates the status of the account. Among them, "Serviced Demand" indicates demand that has been confirmed as completed by the stop line through an event or downstream arrival event; "Abnormal Transfer Demand" indicates demand that is no longer processed as normal traffic demand after exceeding the effective time window; and "Unserviced Demand" is used to pass on to the next control cycle.

[0074] The current book requirement for lane group to participate in quota calculation is expressed as follows: In the formula, Indicates book requirements, Indicates unserved demand. This indicates observable queuing demand. Indicates potential arrival demand. This indicates a freeze on demand. Indicates priority traffic equivalent demand. This indicates that the service request has been fulfilled. This indicates an abnormal transfer request. A separate field for "written-out requests" is no longer used in this formula to avoid duplicate deductions from already served requests.

[0075] Observable queuing demand is formed by queuing events, arrival events, and local confirmation events. The system reads traffic event segments within the same lane group that meet reliability requirements and determines observable queuing demand based on vehicle quantity ranges, queue lengths, and stop line positions. For video queue recognition, radar low-speed target recognition, and loop occupancy status statistics, those skilled in the art can use existing traffic detection and processing methods, which are well-known technologies in the field and will not be elaborated upon here.

[0076] S420. The system calculates potential arrival demand. Potential arrival demand represents traffic demand that has been released upstream and should enter the current lane group according to travel time, but has not yet been confirmed by local event segments. Potential arrival demand is represented as follows: In the formula, Indicates potential arrival demand. This indicates the expected arrival volume from upstream. This indicates the number of locally confirmed vehicles. The stop line throughput is shown. The upstream expected arrival volume is determined by the number of vehicles released upstream, source relationships, turning ratios, and travel time shift results; the locally confirmed vehicle count is determined by arrival events, queuing events, and vehicle targets that meet reliability requirements; the stop line throughput is determined by stop line passage events.

[0077] S430. The system generates a freeze request based on the conflicting freeze credentials. The freeze request is expressed as follows: In the formula, This indicates a freeze on demand. This indicates the number of conflict freeze certificates associated with the lane group. Indicates the first The required amount of freeze corresponding to each conflicting freeze certificate. Indicates the first The frozen status of a conflicting frozen certificate. Indicates a pending status. Indicates an indicator function; takes the value when the condition is true. Otherwise take Conflicting frozen vouchers that have been released to the demand staging account will have their demands converted to unserved demands and will no longer be counted as frozen demands; conflicting frozen vouchers that have been cancelled or transferred to an abnormal state will not be included in the calculation of frozen demands.

[0078] S440, The system generates the equivalent demand for priority traffic. The equivalent demand for priority traffic is expressed as: In the formula, Indicates priority traffic equivalent demand. This indicates the number of priority traffic objects associated with the current lane group. Indicates the first The weight of each priority traffic object Indicates the first An indicator value indicating whether a priority traffic object is within the effective control range. The effective control range is determined by vehicle position, direction of travel, estimated arrival time window, and phase relationship.

[0079] S450. The system periodically updates the demand temporary ledger. The recursive relationship is represented as follows: In the formula, This indicates the unserved demand for the next control cycle. This represents the book demand after participating in quota calculation for the current control period. If serviced demand and abnormally transferred demand have already been deducted during the settlement phase, the next control period will only inherit the demand for incomplete services and will not deduct it again.

[0080] S460. The system constructs a traffic intelligent agent. The traffic intelligent agent is represented as follows: In the formula, Represents a traffic intelligent agent. Indicates the boundaries of control responsibility. Represents the set of observable objects. Represents a set of states. Represents a set of executable actions. Represents the set of security constraints. This represents the set of settlement logs.

[0081] When a lane group has a corresponding traffic event fragment or demand ledger, a lane group agent is constructed. The lane group agent takes a single lane group as the control responsibility boundary. Its observable objects include the event fragments, conflict freeze certificates, and demand ledger of that lane group. Its state set includes observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, unserved demand, and conflict state. Its action set is used to output lane group traffic quota requests.

[0082] When a phase controls at least one lane group, a phase agent is constructed. The phase agent takes the set of lane groups controlled by that phase as the control responsibility boundary, reads the demand state of the lane group agents, and generates phase-level control actions in subsequent steps based on minimum green, maximum green, yellow light, all red, pedestrian clearance, and phase conflict constraints.

[0083] When a directed road segment exists between two intersections, and vehicles released from the upstream lane group enter the downstream receiving area, a corridor agent is constructed. The corridor agent uses the upstream intersection, the downstream intersection, and the directed road segment as control responsibility boundaries, and reads the initial upstream traffic quota, downstream receiving space, reservation occupancy, and estimated arrival time window for subsequent traffic quota acceptance and arrival settlement.

[0084] When a data source is involved in conflict-freezing credentials, spatial mismatch records occur, latency is abnormal, or reliability falls below the availability boundary, a data anomaly agent is constructed. This data anomaly agent uses the data source or detection area as the control responsibility boundary, outputting reliability corrections, anomaly markers, and recovery status, without directly outputting control actions to the signal controller.

[0085] The reference relationship between lane group agents is represented as follows: In the formula, This represents the lane group intelligent agent. Indicates lane group Demand temporary storage account, Indicates lane group A collection of traffic incident clips Indicates lane group The set of conflict-freezing certificates.

[0086] After the above processing, the system incorporates lane group demand, conflict freeze certificate, phase control relationship and downstream acceptance relationship into the corresponding traffic intelligent agent, providing state input for traffic quota calculation, cross-intersection acceptance and signal control action generation.

[0087] After establishing a demand reserve ledger and constructing traffic agents, the lane group agent provides the lane group demand status, the corridor agent provides the downstream acceptance status, and the phase agent provides signal control constraints. The system generates an initial traffic quota for each lane group. When a vehicle corresponding to this traffic quota is expected to enter the downstream intersection within the estimated arrival time window, the system executes cross-intersection acceptance through the corridor agent and adjusts the final traffic quota based on the acceptance status.

[0088] S510, the system reads the lane group's book demand, downstream acceptance space, and phase constraints. Book demand is determined by the demand buffer. Downstream acceptance space limits the number of vehicles that the upstream lane group can release during the current control cycle.

[0089] The downstream receiving space is represented as: In the formula, Indicates downstream acceptance space. This indicates the maximum available storage capacity for downstream vehicles. This indicates that the downstream devices are currently occupying the capacity. This indicates the reserved capacity that has been accepted but has not yet arrived. The maximum available parking capacity can be determined by the downstream exit lane length, the number of lanes, and the average vehicle occupancy length. Its calculation is a standard procedure in traffic engineering and will not be elaborated here.

[0090] S520, system calculation data conflict risk and downstream restriction risk. Data conflict risk is determined by conflict-freezing credentials in a pending state: In the formula, This indicates the risk of data conflict. This indicates the number of conflict freeze certificates associated with the lane group. Indicates the first The conflict trust level of a conflict-frozen credential. Indicates the first The frozen status of a conflicting frozen certificate. Indicates a pending status. Indicates an indicator function; takes the value when the condition is true. Otherwise take .

[0091] Downstream constraint risk is used to indicate the degree to which insufficient downstream acceptance capacity restricts the release of quotas, and it is expressed as follows: In the formula, This indicates the risk of downstream restrictions. Indicates book requirements, This indicates the downstream receiving space.

[0092] Demand pressure is expressed as: In the formula, This indicates demand pressure. Indicates book requirements, Indicates potential arrival demand. This indicates a freeze on demand. Indicates priority traffic equivalent demand. This indicates the risk of data conflict. This indicates the risk of downstream restrictions. , , , , , This indicates the corresponding weight. Demand pressure is used for lane group or phase resource sorting, and is not directly used as the number of vehicles released.

[0093] S530, The system generates the initial traffic quota for lane groups. The initial traffic quota is expressed as: In the formula, Indicates the initial passage quota. Indicates book requirements, Indicates downstream acceptance space. This indicates the saturation release rate of the lane group. This indicates the green light time allowed to be allocated to this lane group within the current control cycle. The saturation release rate can be determined by data from lane group attributes, number of lanes, vehicle composition, and historical stop lines. Its configuration or estimation is well-known in the art and will not be elaborated here.

[0094] S540. The system determines the trigger conditions for cross-intersection acceptance. Cross-intersection acceptance is triggered when a lane group has an initial passage quota, a mapping relationship exists between the lane group and the downstream receiving object, and the expected arrival time window overlaps with the control-related time window of the downstream intersection. When inter-intersection communication is unavailable, the system does not generate a successful acceptance result and executes temporary suspension of acceptance or conservative release according to the anomaly and downgrade handling rules.

[0095] The estimated arrival time window is represented as: In the formula, Indicates the expected arrival time window. Indicates the start time of the current control cycle. This indicates the shortest travel time from the upstream intersection to the downstream intersection. This indicates the longest travel time from the upstream intersection to the downstream intersection. The shortest and longest travel times can be determined by the road segment length and speed range.

[0096] S550, the upstream corridor intelligent agent generates an acceptance request. The acceptance request is expressed as: In the formula, Indicates a request for acceptance. Indicates the acceptance request number, Indicates the upstream intersection number. Indicates the downstream intersection number. Indicates the upstream release lane group number, This indicates the number of vehicles to be released. Indicates the expected arrival time window. Indicates the credibility level of the acceptance request. Indicates the acceptance status. This indicates the reason for the acceptance constraint.

[0097] In the formula, This indicates the number of vehicles to be released. This indicates the initial passage quota.

[0098] The credibility level of an acceptance request is represented as follows: In the formula, Indicates the credibility level of the acceptance request. Indicates the overall reliability of lane group data. This indicates the risk of data conflict. This represents the segment travel time validity coefficient. The overall reliability of lane group data is obtained by weighting the reliability of the data sources involved in the state judgment of that lane group; the segment travel time validity coefficient is determined by the most recent release to settlement result or the segment speed stability.

[0099] S560, downstream corridor intelligent agents calculate the redeemable amount. The downstream redeemable amount is expressed as: In the formula, This indicates the amount that can be accepted by downstream customers. Indicates downstream acceptance space. This indicates the amount of risk occupancy.

[0100] Risk occupancy is expressed as: In the formula, Indicates the amount of risk occupancy. Indicates the number of lane groups associated with the downstream receiving object. Indicates the first Potential arrival demand for a group of associated lanes Indicates the first Freezing requirements for each associated lane group Indicates the first The occupancy mapping coefficient of the associated lane group for the downstream receiving object.

[0101] S570, the downstream corridor smart agent determines the acceptance status. The acceptance status is represented as follows: In the formula, This indicates full acceptance. This indicates partial acceptance. This indicates a temporary suspension of acceptance. This indicates a refusal to accept the payment.

[0102] The number of vehicles accepted is expressed as follows: In the formula, Indicates the number of vehicles accepted. This indicates the number of vehicles to be released. This indicates the amount that can be accepted by downstream customers. Indicates the acceptance status. This indicates full acceptance. This indicates partial acceptance. This indicates a temporary suspension of acceptance. This indicates a refusal to accept the payment.

[0103] The downstream corridor intelligent agent will include the number of accepted vehicles in the reserved capacity and record the corresponding acceptance request number and estimated arrival time window. The reserved capacity will be released after the vehicle arrives and is settled; reserved capacity that has not been confirmed for arrival after the estimated arrival time window will be released or transferred to the abnormal record according to the abnormal and downgrade handling rules.

[0104] In the formula, This indicates the capacity used by the updated reservations. This indicates the capacity used by reservations before the update. This indicates the number of vehicles accepted for payment.

[0105] S580, Upstream Corridor Agents Revise Final Passage Quota. The final passage quota is expressed as: In the formula, Indicates the final passage quota. Indicates the initial passage quota. This indicates the number of vehicles accepted for payment.

[0106] Unaccepted demand is represented as: In the formula, Indicates unaccepted demand. This indicates the number of vehicles to be released. This indicates the number of vehicles that accepted the payment. When all vehicles accepted the payment, the final passage quota equals the initial passage quota. When some vehicles accepted the payment, the final passage quota equals the number of vehicles that accepted the payment, and the unaccepted requests are recorded as unserved requests in the upstream lane group. When acceptance is postponed or rejected, the final passage quota is set to zero, and the reason for the acceptance constraint is recorded in the corridor agent log.

[0107] After generating traffic quotas and accepting them at intersections, the system uses the final traffic quota as input to generate signal control actions by the phase agent. These signal control actions are constrained by phase structure, safety time, pedestrian clearance time, phase conflict relationships, downstream acceptance constraints, and acceptance status.

[0108] The phase agent summarizes the final passage quotas for lane groups under the same phase control based on the phase configuration table, and obtains the final phase quota: In the formula, This indicates the final passage quota for the phase. This represents the set of lane groups for phase control. Indicates the lane group number. Indicates lane group The final passage quota. For lane groups whose acceptance status is suspended or refused, their final passage quota is zero; for lane groups whose acceptance status is partially accepted, only the passage quota corresponding to the number of accepted vehicles enters the phase's final passage quota.

[0109] The phase agent determines the green light request time based on the final phase passage quota and the phase saturation release rate: In the formula, Indicates requesting green light time. This indicates the final passage quota for the phase. Indicates phase saturation release rate. This represents the minimum saturation release rate protection value. The phase saturation release rate can be synthesized from the lane group saturation release rate, and its configuration or modification is well-known in the art and will not be elaborated here.

[0110] The phase agent formulates candidate control actions based on the current phase state, phase sequence, and signal interface. Candidate control actions include maintaining the current phase, extending the current phase, shortening the current phase, switching to the next phase, skipping phases with no service demand, and executing conservative control actions. When maintaining or extending the current phase, the green light duration is determined based on the current phase's elapsed time, maximum green light time, and requested green light time. When switching to the next phase, the green light duration is determined based on the new phase's minimum green light time, maximum green light time, and requested green light time.

[0111] The set of candidate control actions is represented as follows: In the formula, Represents the set of candidate control actions. Indicates the first One candidate control action, This indicates the number of candidate control actions.

[0112] Candidate control actions must satisfy signal safety constraints, downstream acceptance constraints, and acceptance constraints. Signal safety constraints include minimum green light time, maximum green light time, yellow light time, all-red time, pedestrian clearance time, and phase conflict relationships. Downstream acceptance constraints exclude actions that release more vehicles than the downstream acceptance space can accommodate. Acceptance constraints exclude cross-intersection release actions that have not obtained valid acceptance. For actions that maintain the current phase, the system checks whether the current phase exceeds the maximum green light time; for phase switching actions, the system checks whether the current phase has met the minimum green light time, and also checks the yellow light, all-red time, and pedestrian clearance time.

[0113] The set of executable actions is represented as: In the formula, Represents a set of executable actions. Indicates candidate control actions, Represents the set of candidate control actions. Indicates candidate control actions Does it meet signal security constraints? Indicates candidate control actions Is there any downstream acceptance exceeding the limit? Indicates candidate control actions Whether the acceptance constraints are met. The phase conflict relationship, yellow light time, all-red time and pedestrian clearance time can be configured according to the traffic signal controller parameters, which is a well-known technology in this field and will not be elaborated here.

[0114] When the set of executable actions is not empty, the system determines the final signal control action based on the final phase passage quota, risk correction amount, and action switching cost: In the formula, Indicates the final signal control action, Indicates an action that can be performed. Represents a set of executable actions. This indicates the final passage quota for the phase. This indicates the risk adjustment amount. Indicates candidate control actions Action switching cost, , , This indicates the corresponding weight. The risk adjustment amount is formed by the data conflict risk and downstream restriction risk of the lane group corresponding to the phase. It is used to select a more conservative action when multiple actions meet the quota and security constraints, and does not replace the acceptance constraint and downstream acceptance constraint.

[0115] When the set of executable actions is empty, the system does not output control actions that violate signal safety constraints, and executes the conservative control actions preset by the traffic signal controller. Conservative control actions include maintaining the current phase until the minimum green light time is met before switching, executing a fixed timing scheme, or executing a local inductive control scheme. For fixed timing and local inductive control, those skilled in the art can implement these based on existing traffic signal controller functions; these are well-known technologies in the field and will not be elaborated upon here.

[0116] The final signal control action includes the phase number, action type, green light duration, yellow light duration, all-red light duration, control activation time, and associated quota number. If the signal does not support direct dynamic phase control, the system will convert the final signal control action into timing parameters or phase adjustment suggestions that the signal can receive.

[0117] Control action records are represented as follows: In the formula, Indicates control action recording, Indicates the final signal control action, This indicates the final passage quota for the phase. This represents the final set of lane group passage quotas. This represents the set of related acceptance request numbers. This indicates the time window for the execution of control actions. This record is used for subsequent settlement to verify the stop-line throughput, acceptance requests, and the release of reserved occupancy.

[0118] After the signal control action is completed, the system reads the phase execution log, stop line passage event, downstream arrival event, acceptance request record, and conflict freeze certificate status to settle the executed passage quota. If the control action is affected by manual intervention, signal protection, pedestrian clearance constraints, or communication abnormalities, the actual execution result recorded by the signal will be used for settlement.

[0119] The actual effective green light time for a phase is expressed as follows: In the formula, Indicates lane group During the control cycle The actual effective green light time within the period, Indicates phase The actual start time of clearance. Indicates phase The actual time of release is when the clearance is completed. This indicates the start-up loss, clearing loss, and unavailability time caused by signal protection. The calibration of start-up loss and clearing loss can be determined by those skilled in the art based on conventional methods for traffic signal timing, and will not be elaborated upon here.

[0120] The expected service volume of a lane group is expressed as: In the formula, Indicates the expected service volume. Indicates the final passage quota. This represents the temporary demand held before quota calculation is executed. Indicates the saturation release rate. Indicates the actual effective green light time. Indicates the downstream allowable capacity. This indicates rounding down. When there is a cross-intersection acceptance request, the downstream allowable acceptance quantity is the number of accepted vehicles; when there is no cross-intersection acceptance request, the downstream allowable acceptance quantity is the remaining downstream acceptable space.

[0121] The system reads passing event segments formed by coils, geomagnetic sensors, video crossings, radar tracks, or V2X tracks within the stop line crossing time window, and deduplicates them according to vehicle identification, track continuity, lane group, passing direction, and time slice. The actual release amount is expressed as: In the formula, This indicates the actual amount released. This represents the set of vehicle objects that have completed matching. This represents the set of time slices used when a single-bike identifier cannot be generated. Indicates lane group In the A collection of event fragments within a time slice. This indicates the reliability of the data source corresponding to the event fragment. Indicates the number of vehicles in the event segment. This indicates taking the nearest integer. For vehicle detection, trajectory association, and coordinate transformation, those skilled in the art can use existing traffic perception processing methods, which will not be elaborated upon here.

[0122] For transit quotas accepted across intersections, the system determines the downstream arrival time window based on the upstream release time and the travel time window: In the formula, Indicates lane group The corresponding settlement time window for the vehicle arriving at the downstream intersection. Indicates the upstream release reference time. Indicates the shortest travel time. Indicates the longest travel time. Indicates the tolerance at the front of the arrival time window. This indicates the tolerance at the end of the arrival time window.

[0123] The confirmed arrival volume downstream is expressed as follows: In the formula, This indicates the confirmed arrival quantity downstream. This represents the set of arrival event segments corresponding to lane group ggg within the downstream arrival time window. This indicates that the upstream has confirmed the release of vehicle objects or time slice records. This function represents the matching confirmation based on direction mapping, time window, vehicle object, and lane group relationship.

[0124] Acceptance deviation is expressed as: In the formula, Indicates the relative deviation of acceptance. This indicates the confirmed arrival quantity downstream. This indicates the number of vehicles that accepted the payment. The number of cancelled acceptances is determined by records of unexecuted phases, manual intervention, signal protection interruption, or cancellation of acceptance requests, and is not inferred from subsequent test results.

[0125] The downstream reservation occupancy will be updated after settlement as follows: In the formula, This indicates the updated reservation occupancy rate. This indicates the amount of reservations made before settlement. This indicates the confirmed arrival quantity downstream. This indicates the amount of cancelled acceptances. Reservations that remain unconfirmed beyond the arrival time window are transferred to an exception log and then released.

[0126] The deviation in the release of the passage quota is expressed as: In the formula, This indicates the relative deviation in release. This indicates the actual amount released. This indicates the expected service volume.

[0127] The demand already served is expressed as follows: In the formula, This represents the amount of serviced demand deducted from the demand buffer. This represents the temporary demand held before quota calculation is executed. This indicates the actual release amount. The stop line event is used to deduct from the demand reserve account, while the downstream arrival event is used for acceptance settlement and travel time write-back. The two do not deduct the same demand repeatedly.

[0128] The system cancels conflicting frozen vouchers. The confirmation quantity corresponding to the frozen voucher is represented as follows: In the formula, This indicates the number of confirmed vehicles within the scope associated with the frozen certificate. This indicates that the stop line passes through the event set within the associated time window. This represents the set of arrival events within the associated time window. This indicates conflict-freezing certificates pending verification.

[0129] The status of frozen vouchers is updated according to the confirmation results to be cancelled, partially cancelled, pending cancellation, or abnormal transfer. Cancellation of frozen vouchers only changes the voucher status and does not increase the need for temporary storage; the remaining frozen amount that has not been confirmed after exceeding the maximum valid travel time is converted into an abnormal event.

[0130] The write-back of the temporary storage account is represented as follows: In the formula, This indicates the demand reserve for the next control cycle. This represents the temporary demand held before quota calculation is executed. This represents the newly observed demand, potential arrival demand, and priority transportation equivalent demand during the implementation period. This indicates the amount of demand that has been served. This indicates the amount of demand transferred to the abnormal event. Unresolved frozen demands remain in the demand staging account and will not be added again during the write-back.

[0131] Data source reliability write-back is represented as: In the formula, Indicates the reliability of the updated data source. Indicates the current reliability. This indicates the contribution amount confirmed by the stopped line through the event or downstream of the event. This represents the attribution penalty amount corresponding to missed detections, overestimations, delays, spatial mismatches, or unresolved conflicts. This indicates the confirmed contribution correction factor. This represents the attribution penalty correction coefficient. Indicates the lower limit of reliability. Indicates the upper limit of reliability. This represents the interval truncation function.

[0132] The matched upstream release events and downstream arrival events form a travel time sample set, and the central travel time is written back as follows: In the formula, Indicates the updated central travel time. Indicates the current central travel time. Indicates the travel time update coefficient. Represents the travel time sample set, This indicates taking the median.

[0133] In the absence of downstream congestion, manual intervention, signal protection shutdown, or stop line detection anomalies, the saturation release rate is rewritten as follows: In the formula, This represents the updated saturation release rate. This indicates the current saturation release rate. This represents the saturation release rate update coefficient. This indicates the actual amount released. Indicates the actual effective green light time. This indicates the minimum effective green light time. If there are abnormal situations that affect the judgment of the release capacity, the saturation release rate will remain unchanged.

[0134] The system generates an agent settlement record. This record includes at least the actual control action, final passage quota, actual effective green light time, expected service volume, actual release volume, served demand volume, acceptance deviation, release deviation, frozen voucher status, demand temporary ledger update result, data source reliability update result, travel time window update result, and deviation reason identifier. Subsequent control cycles read this settlement record to generate new conflicting frozen vouchers, demand temporary ledgers, passage quotas, and cross-intersection acceptance requests.

[0135] Within each control cycle, the system performs anomaly checks on data sources, traffic event fragments, conflict freeze certificates, demand temporary ledgers, cross-intersection acceptance requests, and signal controller execution feedback. Anomalies include equipment offline, timestamp interruption, excessive observation delay, inconsistent spatial calibration, modal observation conflicts, downstream non-acceptance, and inconsistencies between the actual actions executed by the signal controller and the control actions. For online detection of the interfaces of cameras, radar, geomagnetic sensors, coils, and signal controllers, those skilled in the art can achieve this through heartbeat messages, device status codes, timestamp continuity, and interface return codes. The specific communication protocols are well-known in the field and will not be elaborated upon here.

[0136] The system calculates the health of the data source: In the formula, Indicates the health status of the data source; Indicates the reliability of the data source; This represents the normalized consistent residual, with values ​​restricted to a range of 100. ; Indicates observation delay; Indicates the delay reference value; Indicates online status; Indicates consistency of spatial mapping; , , , , Let represent the weights, and the sum of the weights is . When the health level is below the anomaly threshold, the data source does not directly generate releasable demands; when the health level is between the anomaly threshold and the degradation threshold, its event fragments retain the number, time window, spatial location, vehicle quantity range, and delay parameters, and are only used for conflict freeze vouchers, anomaly identification, and reduction calculation of the demand temporary ledger.

[0137] The downgrade factor is: In the formula, Indicates the downgrade factor; Indicates the abnormal threshold; This indicates the downgrade threshold. The number of vehicles after the reduction is: In the formula, This indicates the number of vehicles after the reduction; This indicates the number of vehicles corresponding to the original event segment. This processing does not delete abnormal data, but rather limits its direct impact on traffic quotas.

[0138] In abnormal situations, only the releaseable portion of the demand temporary account participates in quota calculation: In the formula, This represents the temporary demand that can be used in quota calculation; This indicates that the original demand is temporarily stored in the ledger. Indicates the downstream available space; This indicates the pending service demand that can continue to be maintained. The downgraded passage quota is: In the formula, This indicates a downgrade of the passage quota; This indicates a normal passage quota; Indicates the saturation release rate; This indicates the safe green light time that satisfies the constraints of minimum green, maximum green, yellow light, all red, and pedestrian clearance.

[0139] When an acceptance anomaly occurs at a cross-road intersection, the upstream release quota is adjusted according to the acceptance status: In the formula, This indicates the cross-border release quota after acceptance adjustment; This indicates that the original plan was to release the quota; Indicates the acceptance downgrade coefficient; Indicates the downstream available space; Indicates the number of vehicles to be released; This represents the retention factor when acceptance is temporarily suspended; This indicates the acceptance status. The unreleased portion remains in the upstream demand temporary account and is not considered as served demand.

[0140] When the actual execution result of the signal is inconsistent with the control action, the system deducts the demand based on the actual effective green light time: In the formula, This indicates the service demand calculated based on the actual execution results; This indicates the actual effective green light time. Such deviations are written into the signal execution anomaly log and are not intended to reduce the reliability of video, radar, or coil data sources. When the anomaly is resolved, the data source must continuously meet the recovery conditions, and reliability is restored incrementally in single-cycle increments to avoid frequent switching between abnormal and normal states.

[0141] In one specific embodiment, to illustrate that the method of the present invention can complete data fusion, conflict freezing, demand storage, traffic quota acceptance and signal control action generation based on real-time traffic parameters, the following uses adjacent intersections and their corresponding lane groups in the target road network as examples to specifically describe the processing process within a control cycle.

[0142] Taking the eastbound through lane group G1 between the first intersection I1 and the downstream second intersection I2 in a certain city as an example, the method of the present invention will be described. The directional road section between the first intersection I1 and the second intersection I2 is 320m long. G1 includes two through lanes, and the downstream receiving object D1 is the queuing storage area at the westbound entrance of the second intersection I2. The effective storage length of D1 is 120m, and the number of lanes is 2. Calculated based on an average vehicle occupancy length of 7.5m, the maximum available storage capacity of D1 is 32 vehicles.

[0143] During the control period from 07:32:00 to 07:32:30 on May 18, 2026, the system accessed the following real-time data sources: Video detection device V1, installed approximately 80m before the stop line at the east entrance of I1, with an initial reliability value of 0.82 and an average delay of 0.18s; millimeter-wave radar R1, installed on the roadside at the east entrance of I1, with an initial reliability value of 0.90 and an average delay of 0.11s; loop detector L1, located at the stop line of G1, with an initial reliability value of 0.95 and an average delay of 0.03s; and traffic signal SIG1, with an initial reliability value of 0.98 and an average delay of 0.02s. The environmental impact factor for this period was 0.12, indicating a slight obstruction effect on video detection under light rain conditions.

[0144] The raw data received by the system from 07:32:00 to 07:32:30 is as follows: Video detection device V1 detected the number of vehicles in G1 within the range of [10, 12] vehicles, the speed range of [0, 4] km / h, and the queue length of 86m within the receiving time window from 07:32:05.180 to 07:32:10.180, and generated a queue event E1; millimeter-wave radar R1 detected vehicles in G1 within the receiving time window from 07:32:05.110 to 07:32:10.110. The number of vehicles is in the range of [14, 15] vehicles, the speed range is [0, 5] km / h, the queue length is 105m, and a queueing event E2 is generated; the loop detector L1 detects 3 vehicles passing through the stop line between 07:32:00 and 07:32:30, and generates a stop line passing event E3; the upstream lane group G0 released 12 vehicles in the previous control cycle, the steering ratio from G0 to G1 is 0.75, the travel time window is 38s to 52s, so the upstream expected arrival number mapped to G1 is 9 vehicles.

[0145] The system corrects the event time window based on the data source delay. For example, the receiving time window of video detection device V1 is from 07:32:05.180 to 07:32:10.180, with a delay of 0.18s. Therefore, the corrected event time window is from 07:32:05.000 to 07:32:10.000. The receiving time window of millimeter-wave radar R1 is from 07:32:05.110 to 07:32:10.110, with a delay of 0.11s. Therefore, the corrected event time window is from 07:32:05.000 to 07:32:10.000. After correction, E1 and E2 belong to the same lane group G1 and the same event time window.

[0146] The system performs a vehicle quantity consistency check on E1 and E2. The vehicle quantity range for E1 is [10, 12], and the vehicle quantity range for E2 is [14, 15]. Since there is no overlap, a vehicle quantity conflict is determined. Because video detection device V1 has an occlusion marker in this period, and the number of video detections is lower than the number of radar detections, the system classifies this conflict as an occlusion-type conflict. Based on V1 reliability (0.82), R1 reliability (0.90), occlusion-type conflict weight (0.85), time window offset (0s), and environmental impact factor (0.12), the system calculates the conflict confidence level to be 0.84. Using a higher vehicle quantity estimate of 15 vehicles and a lower vehicle quantity estimate of 10 vehicles, the freeze requirement is calculated as follows: 0.84 × (15 - 10) = 4.2 vehicles.

[0147] After rounding, the system obtained a freeze requirement of 4 vehicles and generated a conflict freeze certificate CF1. The lane group number of CF1 is G1, the conflict time window is from 07:32:05 to 07:32:10, the associated event fragments are E1 and E2, the conflict type is occlusion conflict, the conflict confidence level is 0.84, and the freeze status is pending.

[0148] The system then establishes a demand reserve account for G1. The unserved demand from the previous control period is 2 vehicles; the observable queuing demand for this period is 14 vehicles, based on the most reliable radar queuing events; the potential arrival demand is calculated as "upstream estimated arrivals 9 vehicles - locally confirmed vehicles 4 vehicles - stop line passage 3 vehicles", yielding 2 vehicles; frozen demand is 4 vehicles; there are no bus priority requests this period, so the equivalent priority traffic demand is 0 vehicles. Therefore, the book demand for G1 participating in quota calculation this control period is: 2 + 14 + 2 + 4 + 0 = 22 vehicles.

[0149] The system calculates the downstream acceptance space. The maximum available storage capacity for D1 is 32 vehicles, with 21 currently occupied. There are 4 vehicles already booked but not yet arrived. Therefore, the downstream acceptance space is: 32 - 21 - 4 = 7 vehicles.

[0150] The saturation release rate for G1 is 0.50 vehicles / s. The available green light time for this cycle is 18 seconds. Therefore, the number of vehicles that can be released based on the saturation release capacity is: 0.50 × 18 = 9 vehicles.

[0151] The system generates an initial passage quota based on the book demand of 22 vehicles, downstream acceptance space of 7 vehicles, and saturated release capacity of 9 vehicles, resulting in: min(22, 7, 9) = 7 vehicles.

[0152] Since the vehicles released from G1 will enter the downstream intersection I2 within the expected arrival time window, the system triggers cross-intersection acceptance. The upstream corridor agent sends an acceptance request AR1 to the downstream corridor agent, proposing to release 7 vehicles, with an expected arrival time window from 07:32:38 to 07:32:52. The downstream corridor agent reads the downstream acceptance space of D1 for 7 vehicles, and considers the potential arrival demand of 2 vehicles and the frozen demand of 1 vehicle in the downstream associated lane group, calculating the risk occupancy to be 3 vehicles. Therefore, the downstream acceptable capacity is: 7 - 3 = 4 vehicles.

[0153] Because the downstream available capacity of 4 vehicles is less than the planned release of 7 vehicles, the system returns a partial acceptance status, with 4 vehicles accepted. Based on this, the upstream corridor agent corrects the final passage quota of G1 to 4 vehicles, and the 3 unaccepted requests are written into the unserved requests of G1.

[0154] The phase agent reads G1's final passage quota of 4 vehicles. G1 belongs to phase P1, with a phase saturation release rate of 0.50 vehicles / s. The minimum green light time is 12s, the maximum green light time is 40s, the yellow light time is 3s, and the all-red light time is 2s. Based on the final passage quota, the requested green light time is: 4 ÷ 0.50 = 8s.

[0155] Since the requested green light time is less than the minimum green light time, the executable signal control action generated by the system is: maintain or switch to phase P1, execute green light for 12s, yellow light for 3s, and all red for 2s. The control takes effect at 07:32:03, and the associated acceptance request number is AR1.

[0156] After the control action was executed, the system read the stop line crossing event from 07:32:03 to 07:32:15. After deduplication of the coil detector L1, video line crossing event, and radar trajectory, it was confirmed that G1 actually released 4 vehicles. From 07:32:41 to 07:32:56, the system read the downstream arrival event, confirming that 4 vehicles arrived at the downstream receiving object D1. Therefore, the relative acceptance deviation is: (4-4) ÷ 4 = 0.

[0157] The relative deviation of release is: (4-4) ÷ 4 = 0.

[0158] The system releases the 4 reserved vehicles corresponding to AR1, records the serviced demand of G1 as 4 vehicles, and writes the remaining unserviced demand into the demand temporary storage account for the next control cycle.

[0159] For the conflict freeze certificate CF1, the system confirmed that 4 vehicles had passed the stop line and arrived at the downstream receiving object within the current cycle and subsequent arrival time windows. Therefore, the freeze status of CF1 was updated from pending to cancelled. The reliability of video detection device V1 was corrected from 0.82 to 0.80 due to underestimating the number of vehicles in the obstructed scene. The reliability of millimeter-wave radar R1 was corrected from 0.90 to 0.91, consistent with the downstream arrival results. The saturation release rate of G1 was rewritten based on the actual release of 4 vehicles and the actual effective green light time of 12 seconds. Under the condition that no downstream congestion or signal protection interruption occurred, the updated saturation release rate remained around 0.50 vehicles / s.

[0160] Within this control cycle, the system sequentially generates video event clips, radar event clips, loop detector passes, conflict freeze certificates, and acceptance requests. Based on these events, it calculates the initial passage quota, the final passage quota, and the corresponding signal control actions. After the control actions are executed, the system settles and writes back the final passage quota, acceptance requests, conflict freeze certificates, and demand temporary ledger for this cycle based on stop line passes and downstream arrival events. Traffic demands not accepted by downstream entities are retained in the unserved demands of the next control cycle, while frozen demands confirmed by actual passes and downstream arrival events are updated to a cancelled status.

[0161] The data in the above embodiments are examples of operational data collected or recorded by the system within a control cycle after configuring road objects, data source parameters, and signal control parameters. The formulas illustrate the system's processing steps for calculating event fragments, conflict freeze certificates, demand temporary accounts, traffic quotas, acceptance requests, and settlement write-backs within that control cycle. Parameters involving different dimensions are normalized according to preset ranges before participating in comprehensive calculations. Relevant thresholds, weights, and upper and lower limits can be determined based on intersection channelization conditions, equipment deployment methods, historical operation records, and on-site calibration results.

[0162] The parameters in the above formula can be set according to the intersection channelization conditions, detection equipment type, signal control constraints, historical traffic operation data, and on-site calibration results. For weighting coefficients, thresholds, validity periods, associated time windows, associated distance windows, and upper and lower confidence limits, those skilled in the art can configure them according to the actual road grade, number of approach lanes, detection equipment deployment method, and traffic management requirements. Differences in dimensions involved in the formula calculation process can be eliminated through normalization. The normalization method is a conventional technique in traffic state assessment and multi-source data processing, and will not be elaborated upon here.

[0163] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0164] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0165] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for constructing a multimodal fusion urban transportation intelligent agent system, characterized in that, Includes the following steps: Acquire multimodal traffic data from the target road network and map it to intersections, lane groups, phases, and directed road segments; Delay correction is performed on each data source to generate traffic event segments; Consistency checks are performed on traffic event segments within the same lane group and the same control cycle. When conflicts exist, conflict freeze certificates are generated, and conflicting events are not directly merged into a single traffic state. Establish a demand temporary ledger based on observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, served demand, and unserved demand. The initial passage quota is generated from the demand temporary storage account, the acceptance request is generated for the initial passage quota released across the intersection, and the final passage quota is obtained according to the acceptance status. The final passage quota generates a signal to control the action, and settlement is performed after execution.

2. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The process of acquiring multimodal traffic data in the target road network and mapping it to intersections, lane groups, phases, and directed road segments includes: establishing a road object configuration table, a data source configuration table, and an intersection association configuration table. The road object configuration table records intersections, lane groups, phases, directed road segments, upstream source relationships, turning relationships, and downstream receiving objects; The data source configuration table records the data source number, data source type, installation location, detection range, corresponding lane group, delay parameters, and initial reliability value. Based on the equipment detection area, lane group spatial range, phase configuration, and steering relationship, data output from at least two of the following data sources—video detection equipment, millimeter-wave radar, geomagnetic detector, coil detector, V2X roadside unit, bus positioning access equipment, and floating car trajectory access equipment—are mapped to the corresponding lane group.

3. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The traffic event segment includes event number, data source number, modality type, lane group number, phase number, downstream receiving object, receiving time window, event time window, spatial location, vehicle quantity range, speed range, queue length, data source reliability, delay parameter, environmental impact factor, and event type; The event time window is obtained by correcting the reception time window according to the delay parameter, and the event types include arrival events, queuing events, stop line passing events, signal phase events, priority traffic events, and abnormal events.

4. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The consistency verification of traffic event segments within the same lane group and the same control cycle includes vehicle quantity range verification, queue length verification, time window verification, spatial attribution verification, and traffic conservation verification. When there are non-overlapping vehicle quantity ranges, queue length conflicts exceeding the preset tolerance, event time window offsets exceeding the preset time tolerance, spatial attribution verification results inconsistent with the original mapped lane group, stop line throughput and queue changes not satisfying the traffic conservation relationship, or upstream expected arrivals inconsistent with local confirmed arrivals, a conflict freeze certificate is determined to exist in the traffic event segment.

5. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The conflict freeze certificate includes certificate number, lane group number, conflict time window, associated event segment number, number of vehicle conflict, queue length conflict, time window offset, conflict type, conflict confidence level, freeze status, freeze requirement, and effective time window. The conflict types include obstruction conflicts, delay conflicts, spatial mismatch conflicts, inconsistency in stop line throughput conflicts, and inconsistency in upstream release throughput conflicts; The freezing demand corresponding to conflict-freezing vouchers that are in the pending status is recorded in the demand temporary storage account. Conflict-freezing vouchers that are in the cancelled status or have been transferred to the abnormal status will no longer participate in the passage quota calculation.

6. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The demand temporary ledger includes lane group number, observable queuing demand, potential arrival demand, frozen demand, priority traffic equivalent demand, served demand, abnormal transfer demand, unserved demand, and ledger status. The potential arrival demand is determined by the upstream estimated arrival volume, the locally confirmed number of vehicles, and the stop line throughput, and is used to represent the traffic demand that has been released upstream and should enter the current lane group according to the travel time but has not yet been confirmed by the local traffic event segment; The freeze request is determined by conflict freeze credentials that are in a pending state; The unserved demand is used to reserve traffic demand that has not been released by the final passage quota or has not been actually served through event confirmation.

7. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, After establishing the demand temporary ledger, the process also includes constructing a traffic intelligent agent; The traffic intelligent agent includes lane group intelligent agent, phase intelligent agent, corridor intelligent agent and data anomaly intelligent agent; The lane group intelligent agent takes a single lane group as the control responsibility boundary and maintains traffic event fragments, conflict freeze certificates, and demand temporary ledgers. The phase agent takes the set of lane groups controlled by the same phase as the control responsibility boundary and summarizes the demand status of the corresponding lane groups; The corridor intelligent agent takes the upstream intersection, downstream intersection and directional road segment as the control responsibility boundary to handle the acceptance of passage quota and arrival settlement; The data anomaly agent takes the data source or detection area as the control responsibility boundary, outputs reliability correction and anomaly markers, and does not directly output signal control actions to traffic signal controllers.

8. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The process of generating an initial traffic quota from the demand reserve account and generating an acceptance request for the initial traffic quota released across intersections includes: determining the number of vehicles to be released for a lane group based on the book demand in the demand reserve account, downstream acceptance space, saturation release rate, and allocable green light time. When the vehicles corresponding to the quota of vehicles to be released are expected to enter the downstream receiving objects within the expected arrival time window, the upstream corridor intelligent agent generates an acceptance request. The acceptance request includes the acceptance request number, upstream intersection number, downstream intersection number, upstream release lane group number, quota of vehicles to be released, estimated arrival time window, acceptance request credibility level, acceptance status, and acceptance constraint reason.

9. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The process of obtaining the final passage quota based on the acceptance status includes: the downstream corridor agent determining the downstream acceptable amount based on the downstream acceptance space, reserved occupancy capacity, potential arrival demand and frozen demand, and returning the acceptance status to the upstream corridor agent based on the downstream acceptable amount. The acceptance status includes full acceptance, partial acceptance, delayed acceptance, and rejection of acceptance; When the acceptance status is full acceptance, the final passage quota corresponds to the initial passage quota. When the acceptance status is partial acceptance, the final passage quota corresponds to the downstream acceptable volume, and the unaccepted demand is written into the unserved demand of the upstream lane group. When the acceptance status is "acceptance suspended" or "acceptance refused", the release of the corresponding lane group across the intersection is restricted, and the reason for the acceptance constraint is written into the corridor agent log.

10. The method for constructing a multimodal fusion urban transportation intelligent agent system according to claim 1, characterized in that, The action of generating a signal control based on the final passage quota and performing settlement and write-back after execution includes: the phase agent summarizing the final passage quota of each lane group under the same phase control to obtain the phase final quota; The green light request time is determined based on the final phase passage quota and the phase saturation release rate. Candidate control actions are determined from maintaining the current phase, extending the current phase, shortening the current phase, switching to the next phase, skipping phases with no service demand, and performing conservative control actions; Candidate control actions that do not meet the minimum green light time, maximum green light time, yellow light time, all-red time, pedestrian clearance time, phase conflict relationship, downstream acceptance constraint and acceptance constraint are eliminated, and the retained candidate control actions are converted into signal control actions. Within the travel time window after the signal control action is executed, stop line passing events and downstream arrival events are collected. The actual number of passing vehicles and the actual number of arriving vehicles are matched with the final passage quota, acceptance request and demand temporary account, respectively. Based on the matching results, the freeze status of conflict freeze vouchers, demand temporary account, data source reliability, travel time parameters and saturation release rate are updated.

Images