A maximum pressure traffic signal coordination control method fusing dynamic sub-area division under an intelligent network connection environment

Abstract

The present application relates to a kind of intelligent network connection environment under the maximum traffic signal coordination control method of fusion dynamic subzone division, belong to traffic signal coordination control technical field.The method includes: based on the predicted intersection traffic flow the comprehensive pressure of each intersection is obtained and index amplification coefficient is constructed, to obtain super edge weight and clustering is carried out to intersection, complete the dynamic division of control subzone;The predicted queuing length of each entrance is calculated in each control subzone, and the instantaneous comprehensive prediction pressure of each phase is calculated in combination with turning rate, introduce penalty function to obtain the effective green time of final execution, to build the ideal green start time of upstream;Best compromise start time is constructed, and the final executed coordination phase switching time is obtained in combination with the phase switching time of last round, realize traffic signal coordination control.The technical problem of solving queuing back overflow, local deadlock and different intersection coordination deficiency in the region signal control under oversaturated traffic state is solved.

Description

Technical Field

[0001] This invention relates to a maximum pressure traffic signal coordination control method that integrates dynamic sub-region division in an intelligent connected environment, belonging to the field of traffic signal coordination control technology. Background Technology

[0002] Urban traffic congestion often manifests as localized oversaturation spreading to surrounding areas, leading to queue overflow and localized deadlocks. Traditional signal coordination methods are effective under stable demand, but they are insufficient in responding to dynamic congestion boundaries under oversaturated conditions. While maximum pressure (MP) control can ensure queue stability, it mainly relies on local pressure feedback and lacks modeling of regional path associations and forward coordination among multiple intersections.

[0003] Intelligent connected vehicles (CAVs) provide refined trajectory data, but under conditions of mixed traffic flow and low penetration, existing methods either over-rely on CAV data or struggle to achieve closed-loop coordination between dynamic sub-zone division and signal control. Therefore, there is an urgent need to develop a regional signal coordination control method that adapts to mixed traffic flow and dynamic congestion boundaries. Summary of the Invention

[0004] The purpose of this invention is to provide a maximum pressure traffic signal coordination control method that integrates dynamic sub-region division in an intelligent connected environment. It aims to solve the technical problems of queue overflow, local deadlock and insufficient coordination between different intersections in regional signal control under oversaturated traffic conditions. It can reduce the average regional delay, suppress queue spread and improve the stability of road network operation under high demand and different intelligent connected vehicle penetration rates.

[0005] To achieve the above objectives, the technical solution of this invention is: a maximum pressure traffic signal coordination control method integrating dynamic sub-zone division in an intelligent connected environment. This method first integrates short-time CAV prediction and human-driven vehicle (HDV) observation estimation to construct a comprehensive regional pressure characterization; secondly, it introduces a hypergraph to describe the path associations of multiple intersections and realizes dynamic sub-zone division based on comprehensive pressure weights; then, within each sub-zone, it achieves temporal coordination of adjacent intersections through core path green window calculation; finally, for phase conflicts when multiple core paths share common nodes, it performs compromise coordination based on minimizing weighted absolute deviation, achieving closed-loop coordination between dynamic sub-zone division and regional signal control, including the following steps:

[0006] S1: Based on the location, speed, and path information of CAV and the number of queued vehicles, saturation flow rate, number of lanes, and green ratio of HDV, predict the departure traffic flow of CAV and HDV respectively, and sum them by weight to obtain the predicted traffic flow of the intersection. Calculate the predicted pressure based on the predicted traffic flow of the intersection, and obtain the real-time observed pressure of the intersection. Perform weighted fusion of the predicted pressure and the real-time observed pressure to obtain the comprehensive pressure of each intersection.

[0007] S2: Based on the geometric connectivity of the upstream and downstream intersections and the measured HDV flow, the basic weights are obtained. Based on the comprehensive pressure, an exponential amplification coefficient is constructed. The product of the basic weights and the exponential amplification coefficient is used as the hyperedge weights. Based on the hyperedge weights, the intersections are clustered to complete the dynamic division of the control sub-regions.

[0008] S3: Within each control sub-area, the predicted queue length of each approach lane is calculated using the real-time trajectory information of the CAV, and the instantaneous comprehensive predicted pressure of each phase is calculated in combination with the turning rate. When the space occupancy rate of the downstream intersection approach lane exceeds the preset threshold, the effective pressure is obtained by introducing a penalty function. The green light duration is allocated based on the effective pressure and weighted by history to obtain the final effective green light duration.

[0009] S4: Based on the effective green light duration of the final execution, obtain the downstream green window center, and combine it with the real-time driving time and vehicle queue emptying time to obtain the ideal green light start time upstream;

[0010] S5: Based on the ideal green light start time of each path upstream, construct the optimal compromise start time, and perform linear weighted fusion with the phase switching time of the previous round to obtain the final coordinated phase switching time, thereby realizing coordinated traffic signal control.

[0011] Optionally, S1 specifically includes:

[0012] Get the future Queuing situation at intersections within a given time period, including future... HDV traffic volume leaving intersection i within a given time period Defined as:

[0013]

[0014] In the formula, This represents the real-time number of vehicles queuing at time t during coordination phase i at intersection i. This represents the saturation flow rate at intersection i; This represents the number of lanes from intersection i to the downstream intersection i+1; This indicates the coordinated phase green light ratio at intersection i;

[0015] Future HDV traffic volume at intersection i within a given time period Defined as:

[0016]

[0017] In the formula, This represents the real-time number of vehicles queuing at time t, which is the coordination phase of the k-th approach at the upstream intersection i-1. This represents the saturation flow rate at the k-th approach of the upstream intersection i-1. This represents the number of lanes at the k-th approach to the upstream intersection i-1. This represents the coordinated phase green light ratio of the k-th approach at upstream intersection i-1;

[0018] Thus, we can obtain the future. Predict traffic flow at intersections within a specified time period for:

[0019]

[0020] In the formula, For the future The number of vehicles arriving at intersection i within a given time period. For the future The number of vehicles leaving intersection i within a given time period;

[0021] Intersection i at Predicted pressure over time Defined as:

[0022]

[0023] Real-time observed pressure at intersection i at time t Defined as:

[0024]

[0025] Thus, the combined pressure at intersection i at time t is obtained. for:

[0026]

[0027] In the formula, Let be the time decay coefficient, and .

[0028] Optionally, the step of using the product of the base weight and the exponential amplification factor as the hyperedge weight specifically involves:

[0029] The weights of basic connectivity and measured traffic flow are defined as basic weights. The expression is:

[0030]

[0031] In the formula, This represents the geometric connectivity coefficient of the upstream and downstream intersections. This represents the HDV traffic weighting coefficient. Indicates the intersection of upstream and downstream. HDV measured flow rate within the time window;

[0032] Constructing the exponential amplification factor The expression is:

[0033]

[0034] In the formula, Indicates the pressure sensitivity coefficient; This represents the combined pressure at intersection i at time t. This represents the combined pressure at intersection i at time t+1;

[0035] Multiplying the base weight by the exponential amplification factor yields the hyperedge weight, expressed as:

[0036]

[0037] In the formula, This represents the weight of the superedge.

[0038] Optionally, the penalty function is specifically:

[0039]

[0040] In the formula, For the penalty function, The threshold for the occupancy rate of the downstream intersection approach lane. This represents the real-time lane space occupancy rate.

[0041] Optionally, S3 specifically includes:

[0042] Arrival time of a single intersection approach lane l Discretized into time The vehicle arrival curve is obtained, and its expression is:

[0043]

[0044] In the formula, In time At that time, the number of CAVs arriving at the approach lane l of phase j at intersection i. Indicates the approach lane l of phase j at intersection i; This indicates the estimated time when vehicle v will arrive at intersection entrance lane l.

[0045] This allows us to obtain the predicted queue length of lane l at the intersection entrance. The expression is:

[0046]

[0047] In the formula, This represents the saturation flow rate of lane l at the intersection entrance. In time At that time, the number of CAVs arriving at the import lane l ;

[0048] Thus, instantaneous comprehensive predicted pressure is obtained. The expression is:

[0049]

[0050] In the formula, For a single point in time used to construct an instantaneous comprehensive forecast of pressure, This represents the maximum queuing capacity of the import lane. Let j be the set of downstream exit channels. Let k be the maximum queuing capacity of the exit lane. Indicates the spatial coupling influence factor. This indicates the turning rate when a vehicle turns from entrance lane l to exit lane k. For the future At that time, the predicted queue length of exit lane k, For the future At that time, the predicted queue length of the exit lane k, which incorporates the turning rate, is combined;

[0051] By constructing a penalty function, the effective pressure at phase j of intersection i is obtained. for:

[0052]

[0053] In the formula, Indicates the pressure attenuation coefficient;

[0054] Before each phase begins, the phase for which the green light duration is to be calculated is considered as the fourth phase of a cycle. The fourth phase and the preceding three signal phases are combined to form a complete signal cycle, creating a new cycle, which is defined as the virtual cycle, for the purpose of allocating green light durations. Specifically:

[0055]

[0056] In the formula, This represents the green light ratio of phase j. This represents the sum of the pressures at intersection i in phase j compared to the three phases preceding it; p ij This represents the pressure value at phase j in intersection i;

[0057] The green light duration allocated to the current phase requiring passage is obtained based on the imaginary period and the green light ratio. for:

[0058]

[0059] In the formula, This indicates the duration of the imaginary period formed by phase j and its three preceding phases;

[0060] Green light duration By performing historical weighting, the final effective green light duration is obtained. for:

[0061]

[0062] In the formula, Represents the smoothing coefficient. Indicates the number of cycles.

[0063] Optionally, S4 specifically includes:

[0064] The traffic signal controller at each intersection along the defined path broadcasts predicted values ​​of pressure and green light intensity once within a preset time before phase release, thereby obtaining the downstream green window center. for:

[0065]

[0066] In the formula, The green window center time for phase j at the downstream intersection. The predicted green light duration for phase j at the downstream intersection;

[0067] For vehicles traveling from upstream to downstream, the travel time at 85% of the locations uploaded by CAV in the first 30-60 seconds is taken as the upstream to downstream travel time. ;

[0068] Calculate the downstream vehicle queue emptying time The expression is:

[0069]

[0070] In the formula, The queue length based on CAV at the downstream intersection. The saturation flow rate at the downstream intersection;

[0071] Ultimately, the ideal green light start time for upstream is obtained. for:

[0072]

[0073] In the formula, To ensure a safety margin, vehicles must pass through the green wave at downstream intersections without getting stuck, and ,in, The variance of travel time, As weight.

[0074] Optionally, S5 specifically includes:

[0075] Along each path The set of ideal green light start times upstream is obtained by reverse recursion. To find the optimal compromise moment for unifying the ideal green light start time across all phases. The expression is:

[0076]

[0077] in, >0 indicates the weight of the given path; This represents the error value between the actual time of switching the green light phase and the ideal start time of path R;

[0078] Introduce a threshold This is used to maintain the continuity of vehicle queuing, thereby obtaining the final coordinated phase switching time. for:

[0079]

[0080]

[0081] In the formula, As weight, This refers to the phase switching moment of the previous round.

[0082] The beneficial effects of this invention are:

[0083] 1. This invention introduces a hypergraph to describe the path association of multiple intersections, integrates CAV prediction and HDV measurement to construct a comprehensive pressure weight, realizes the dynamic division of control sub-regions, and avoids congestion boundary cutting.

[0084] 2. This invention establishes variable period maximum pressure control within the sub-region, and combines core path green window calculation with L1 compromise strategy to achieve timing coordination between adjacent intersections and resolution of phase conflicts at common nodes.

[0085] 3. This invention integrates CAV trajectory information and HDV estimation to construct a comprehensive stress characterization, which can operate stably under low penetration rates and cope with high-density rigid vehicle fleets by trading space for time under high penetration rates, thus avoiding performance collapse. Attached Figure Description

[0086] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0087] Figure 2 This is a schematic diagram illustrating the calculation of the maximum pressure green time in each phase of the present invention;

[0088] Figure 3 This is the regional average delay-time curve of the present invention;

[0089] Figure 4 This is the regional average queuing time curve of the present invention;

[0090] Figure 5 This is the average parking time curve for the region in this invention;

[0091] Figure 6 This is a graph showing the regional throughput or cumulative number of vehicles completed in this invention.

[0092] Figure 7 This is a graph showing the changes in the average delay index of each method under different penetration rates according to the present invention;

[0093] Figure 8 This is a graph showing the variation of the average queue length index for each method under different penetration rates according to the present invention. Detailed Implementation

[0094] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0095] Example 1: As Figure 1 As shown, a maximum pressure traffic signal coordination control method integrating dynamic sub-region division in an intelligent connected environment includes the following steps:

[0096] S1: Based on the location, speed, and path information of CAV and the number of queued vehicles, saturation flow rate, number of lanes, and green ratio of HDV, predict the departure traffic flow of CAV and HDV respectively, and sum them by weight to obtain the predicted traffic flow of the intersection. Calculate the predicted pressure based on the predicted traffic flow of the intersection, and obtain the real-time observed pressure of the intersection. Perform weighted fusion of the predicted pressure and the real-time observed pressure to obtain the comprehensive pressure of each intersection.

[0097] It's important to understand that obtaining all CAV (Carrier Amplitude Volume) data in urban traffic is impractical. Due to its limited penetration, CAV cannot directly reflect the actual traffic flow volume and state at intersections. Therefore, in calculating predicted intersection pressure, CAV traffic volume cannot be simply used as the statistical standard; instead, it must be weighted and calculated by incorporating general traffic volume. Since the traffic flow structure consists of both CAV and HDV (High-Voltage Traffic Volume), in this embodiment, to predict future pressure levels, it is necessary to obtain future... The queuing situation at the intersection within a given time period. Specifically, S1 is:

[0098] Traffic flow passes through various intersections during its operation; to achieve time... Predicting internal traffic flow only requires forecasting the volume of traffic leaving and entering the intersection during a given time period. Due to the availability of CAV data (path, speed, and location), predicting the pressure on this segment of vehicles is relatively easy. However, obtaining HDV data requires adherence to the following definitions:

[0099] Future HDV traffic volume leaving intersection i within a given time period Defined as:

[0100]

[0101] In the formula, This represents the real-time number of queued vehicles (including CAV and HDV) at time t of coordination phase i at intersection i. This represents the saturation flow rate at intersection i; This represents the number of lanes from intersection i to the downstream intersection i+1; Indicates the coordinated phase green light ratio at intersection i;

[0102] Future HDV traffic volume at intersection i within a given time period Defined as:

[0103]

[0104] In the formula, This represents the real-time number of vehicles queuing at time t, which is the coordination phase of the k-th approach at the upstream intersection i-1. This represents the saturation flow rate at the k-th approach of the upstream intersection i-1. This represents the number of lanes at the k-th approach to the upstream intersection i-1. This represents the coordinated phase green light ratio of the k-th approach at upstream intersection i-1;

[0105] Alternatively, intelligent connected vehicles can accurately obtain future trajectory information through real-time CAV (Continuous Ambient Vehicle) trajectory broadcasting. The number of vehicles arriving at and leaving the intersection within the time window is calculated, so unlike HDV, it does not require prediction based on traffic flow, thus providing future data. Predict traffic flow at intersections within a specified time period for:

[0106]

[0107] In the formula, For the future The number of vehicles arriving at intersection i within a given time period. For the future The number of vehicles leaving intersection i within a given time period;

[0108] Intersection i at Predicted pressure over time Defined as:

[0109]

[0110] Optionally, predicted pressure alone is insufficient to reflect actual traffic conditions. Therefore, this embodiment introduces real-time observed pressure at the intersection as a comprehensive weighting term for prediction, making the calculation of predicted pressure more reasonable, thereby incorporating the real-time observed pressure at intersection i at time t. Defined as:

[0111]

[0112] Thus, the combined pressure at intersection i at time t is obtained. for:

[0113]

[0114] In the formula, This is the time decay coefficient, used to balance the confidence levels of predicted pressure and real-time observed pressure. .

[0115] S2: Based on the geometric connectivity of the upstream and downstream intersections and the measured HDV flow, the basic weights are obtained. Based on the comprehensive pressure, an exponential amplification coefficient is constructed. The product of the basic weights and the exponential amplification coefficient is used as the hyperedge weights. Based on the hyperedge weights, the intersections are clustered to complete the dynamic division of the control sub-regions.

[0116] It's important to understand that traditional dynamic sub-region partitioning methods are mostly based on ordinary graph theory models, calculating the correlation between pairs of intersections and then using classical clustering algorithms to partition the signal control area. However, traditional correlation models often focus on measuring the relationship between two independent upstream and downstream intersections, neglecting the discrete state of convoys among multiple intersections and the various complex paths existing at multiple intersections within a region. Furthermore, in an intelligent connected vehicle environment, to perform dynamic sub-region partitioning based on connected vehicle path and location information, it is first necessary to correlate multiple intersections along the intelligent connected vehicle path and calculate the path pressure at upstream and downstream intersections. Therefore, in this embodiment, hypergraph theory is introduced to replace traditional ordinary graph theory for dynamic sub-region partitioning of regional signal control, to better describe the group correlation relationships between multiple intersections in the regional road network, thereby obtaining a control sub-region dynamic partitioning method that is more suitable for the intelligent connected vehicle environment.

[0117] Furthermore, traditional sub-zone division methods use the geometric features of intersections, traffic flow volume, and signal cycle duration as the division criteria. Since the signal control method used in this embodiment is a variable-cycle maximum pressure signal control method based on intelligent connected vehicle data, the cycle duration of each intersection changes constantly with the traffic flow volume. Therefore, the cycle lengths of different intersections are not the same, and signal cycle duration cannot be used as the sub-zone division criterion. However, since the signal cycle duration is determined by the traffic flow volume, the traffic flow volume can be used as the standard. Therefore, in constructing the hypergraph and hyperedge weights, removing the signal cycle duration parameter satisfies the requirements of this embodiment without loss of generality.

[0118] Furthermore, in a mixed traffic environment, the dynamic sub-region division of a region must not only consider the spatial topological relationships of intersections within the regional system, but also require precise control over the traffic conditions faced by the system. While retaining parameters such as the geometric features of intersections and traffic flow magnitude, this embodiment uses the predicted pressure of intersections in an intelligent connected environment as a major criterion for sub-region division, constructing a dynamic hyperedge weight that integrates traffic flow dynamics and intersection phase pressure, specifically:

[0119] Optionally, the step of using the product of the base weight and the exponential amplification factor as the hyperedge weight specifically involves:

[0120] The weights of basic connectivity and measured traffic flow are defined as basic weights. The expression is:

[0121]

[0122] In the formula, This represents the geometric connectivity coefficient of the upstream and downstream intersections. This represents the HDV traffic weighting coefficient. Indicates the intersection of upstream and downstream. Measured HDV flow rate within the time window;

[0123] Constructing the exponential amplification factor The expression is:

[0124]

[0125] In the formula, Indicates the pressure sensitivity coefficient; This represents the combined pressure at intersection i at time t. This represents the combined pressure at intersection i at time t+1;

[0126] Multiplying the base weight by the exponential amplification factor yields the hyperedge weight, expressed as:

[0127]

[0128] In the formula, This represents the weight of the superedge.

[0129] Understandably, the purpose of introducing the exponential amplification factor in this embodiment is to prevent network deadlock when the road network system is under oversaturated traffic conditions. It forces adjacent intersections with excessive traffic pressure to be assigned to the same sub-zone for signal coordination and control.

[0130] S3: Within each control sub-area, the predicted queue length of each approach lane is calculated using the real-time trajectory information of the CAV, and the instantaneous comprehensive predicted pressure of each phase is calculated in combination with the turning rate. When the space occupancy rate of the downstream intersection approach lane exceeds the preset threshold, the effective pressure is obtained by introducing a penalty function. The green light duration is allocated based on the effective pressure and weighted by history to obtain the final effective green light duration.

[0131] Optionally, due to the availability of CAV information, the road network control system continuously broadcasts the location, speed, and route of each CAV in real time, including the arrival time of each intersection approach lane l. Discretized into time (Unit: seconds), the vehicle arrival curve is obtained, and the expression is:

[0132]

[0133] In the formula, In time At that time, the number of CAVs arriving at the approach lane l of phase j at intersection i. Indicates the approach lane l of phase j at intersection i; This indicates the estimated time when vehicle v will arrive at intersection entrance lane l.

[0134] This allows us to obtain the predicted queue length of lane l at the intersection entrance. The expression is:

[0135]

[0136] In the formula, This represents the saturation flow rate of lane l at the intersection entrance. In time At that time, the number of CAVs arriving at the import lane l ;

[0137] It is important to understand that the maximum pressure control in this embodiment determines the direction of maximum pressure by real-time monitoring of traffic flow at intersection entrances and exits and calculating the pressure difference at the intersection. At each intersection, the green light duration for each phase depends on the pressure difference of the corresponding lane. To meet the requirements of dynamic coordination, the selection of pressure parameters is not limited to simple queue length, but rather the traffic flow density of the road segment is selected as the pressure parameter for maximum pressure, thereby calculating the pressure value of the phase.

[0138] Furthermore, in the supermap sub-region division stage, since dynamic sub-region division needs to consider multiple events such as traffic flow, queue configuration, and path selection in the next few seconds of CAV prediction, the selection of comprehensive pressure cannot only consider a single moment. Instead, it should incorporate the accumulated demand of each traffic flow within the overall prediction window and the available downstream supply traffic flow into a single formula, resulting in a relatively large computational workload. In the variable cycle maximum pressure control stage, in the intelligent connected vehicle-maximum pressure control environment, the required comprehensive prediction pressure must be completed within a short period (during the yellow light period) for each decision. Therefore, this embodiment adopts a single time point. Construct snapshot pressure (instantaneous pressure) to reduce the computational volume.

[0139] Optionally, the instantaneous comprehensive predicted pressure is obtained. The expression is:

[0140]

[0141] In the formula, For a single point in time used to construct an instantaneous comprehensive forecast of pressure, This is the maximum queuing capacity of the import lane. Let j be the set of downstream exit channels. Let k be the maximum queuing capacity of the exit lane. Indicates the spatial coupling influence factor. This indicates the turning rate when a vehicle turns from entrance lane l to exit lane k. For the future At that time, the predicted queue length of exit lane k, For the future At that time, the predicted queue length of the exit lane k, which incorporates the turning rate, is combined;

[0142] Optionally, since traditional MP signal control methods are prone to continuously releasing vehicles to downstream intersections that are already congested when the road network system is oversaturated, causing downstream intersections to transition from congestion to vehicle overflow, and gradually leading to network deadlock in the block road network, this embodiment introduces a network deadlock overflow penalty mechanism at the single-intersection control level: defining a threshold for the space occupancy rate of the downstream intersection approach lanes. When the real-time lane space occupancy rate Greater than the threshold At that time, physical truncation is implemented to ensure that the traffic conditions of the road network system are within a safe range, and the penalty function is applied. Defined as:

[0143]

[0144] Based on the penalty function, the effective pressure of phase j at intersection i is obtained. for:

[0145]

[0146] In the formula, This represents the pressure attenuation coefficient, used to balance the green light duration distribution across different phases.

[0147] Furthermore, compared to the traditional method of using the sum of pressure differences at the inlet corresponding to each phase of the next cycle for maximum pressure, this embodiment improves the pressure summation object: before the start of each phase, the phase for which the green light duration is to be calculated is regarded as the fourth phase of a cycle. The fourth phase and the preceding three signal phases are combined to form a complete signal cycle, forming a new cycle, which is defined as the virtual cycle, thereby allocating the green light duration. In this embodiment, to avoid green light ratio jitter caused by variable cycles, the green light duration of each phase is allocated based on the effective pressure, specifically:

[0148]

[0149] In the formula, This represents the green light ratio of phase j. This represents the sum of the pressures at intersection i in phase j compared to the three phases preceding it; p ij This represents the pressure value at phase j in intersection i;

[0150] The green light duration allocated to the current phase requiring passage is obtained based on the imaginary period and the green light ratio. for:

[0151]

[0152] In the formula, This indicates the duration of the imaginary period formed by phase j and its three preceding phases;

[0153] Green light duration By performing historical weighting, the final effective green light duration is obtained. for:

[0154]

[0155] In the formula, Represents the smoothing coefficient. Indicates the number of cycles.

[0156] Optionally, such as Figure 2 As shown, taking a standard four-phase signalized intersection as an example, This represents the green light duration of the j-th phase in the i-th cycle of the intersection. To obtain... The size of, then let and The first three signal phases (i.e. , , The green ratio can be calculated by forming a complete signal period (i.e., imaginary period 1). Similarly, the green ratio can be calculated. , , And the magnitude of the subsequent phase.

[0157] S4: Based on the effective green light duration of the final execution, obtain the downstream green window center, and combine it with the real-time driving time and vehicle queue emptying time to obtain the ideal green light start time upstream;

[0158] Optionally, traditional signal coordination control requires setting a uniform cycle during signal coordination control. Then, under the same cycle, a large-scale linear or nonlinear mixed-integer programming model is used to determine the phase difference between intersections on the coordination path. This fixed-cycle, fixed-phase-difference control mode has good control performance when traffic volume remains constant, but it struggles to meet the real-time traffic flow demands. Furthermore, within a signal control area, the traffic status of each route and node changes in real time. Therefore, to achieve dynamic regional signal coordination control, this embodiment introduces a core coordination path-level time estimation method to adapt to the variable-cycle maximum pressure signal control strategy while also meeting the needs of real-time changing coordination path schemes. Moreover, since the selection of the core path within each sub-area is not singular, these paths may have overlapping signalized intersections. This embodiment also proposes a core path common intersection node-level compromise strategy to resolve conflicts between different coordination phases at common intersections.

[0159] Specifically, the core idea of ​​the core coordinated path-level time estimation method is to treat the signals of two adjacent intersections on each core path as a "starting green window - ending green window." This means calculating when the coordinated phase at the upstream intersection should be activated so that the convoy can reach the center of the green window at the downstream intersection at the prescribed ideal speed after the queue has just emptied, thus smoothly passing through the downstream intersection. Therefore, the core coordinated path-level time estimation method does not require all intersections on the path to have a uniform signal cycle length, and it does not depend on any preset cycle length. Specifically:

[0160] The traffic signal controller at each intersection along the defined path broadcasts predicted values ​​of pressure and green light intensity once within a preset time (e.g., 2-3 seconds) before phase release, thereby obtaining the downstream green window center. for:

[0161]

[0162] In the formula, The green window center time for phase j at the downstream intersection. The predicted green light duration for phase j at the downstream intersection;

[0163] Optionally, due to real-time changes in traffic conditions, the travel time from upstream intersection i to downstream intersection j varies. Therefore, different travel times need to be adopted according to different traffic environments. In this embodiment, the travel time obtained through real-time analysis is used to calculate the travel time of 85% of the locations uploaded by CAV in the first 30-60 seconds from upstream to downstream as the travel time from upstream to downstream. ;

[0164] Calculate the downstream vehicle queue emptying time The expression is:

[0165]

[0166] In the formula, The queue length based on CAV at the downstream intersection. The saturation flow rate at the downstream intersection;

[0167] Ultimately, the ideal green light start time for upstream is obtained. for:

[0168]

[0169] In the formula, To ensure a safety margin, vehicles must pass through the green wave at downstream intersections without getting stuck, and ,in, The variance of travel time, As weight.

[0170] S5: Based on the ideal green light start time of each path upstream, construct the optimal compromise start time, and perform linear weighted fusion with the phase switching time of the previous round to obtain the final coordinated phase switching time, thereby realizing coordinated traffic signal control.

[0171] Optionally, in the dynamic sub-area core path selection proposed in this embodiment, the real-time pressure magnitude serves as a crucial condition for core path selection within each control sub-area. Through pressure magnitude selection, interweaving and overlap occur between the core coordinated paths, resulting in multiple signal phases at the same intersection being defined as coordinated phases. However, a coordinated control method considering only the sequential extension of a single coordinated phase is insufficient to satisfy the coordination of multiple signal phases, leading to corresponding shifts in the coordinated signal phases of upstream and downstream intersections centered on this intersection. If each core coordinated path is forced to switch its green light phase according to its ideal green light start time, it cannot guarantee that vehicle queues on any coordinated path are operating within the green wave band, and it also leads to complex phase switching within the same intersection, resulting in frequent staggered green light switching phenomena. Therefore, to simultaneously consider the collaborative work of multiple coordinated phases in a single intersection, this embodiment proposes a compromise strategy at the core path common intersection node level, specifically:

[0172] When multiple core coordination paths share the same intersection j, along each path The set of ideal green light start times upstream is obtained by reverse recursion. ;

[0173] At the node level, a weighted absolute value (L1) minimization linear programming problem is introduced to find an optimal compromise time that unifies the ideal green light start time for each phase, while ensuring the ideal start time of each core coordination path. The expression is:

[0174]

[0175] in, >0 indicates the weight of the given path; This represents the error value between the actual time of switching the green light phase and the ideal start time of path R;

[0176] It is understandable that this embodiment, by setting the weighted absolute deviation value to be minimized, allows the value to operate at the lowest cost of balancing all coordination paths without being interfered with by any extreme coordination path.

[0177] Furthermore, in actual phase cycling, if the optimal compromise time obtained from the compromise calculation is directly used... Sending the signal to the signal controller will pose the following two potential risks:

[0178] First, if the calculated start time of a certain cycle differs significantly from the phase switching time of the previous cycle, and the traffic signal switches the phase at that instant, it will cause a large number of vehicles to queue before the coordinated phase switches, resulting in traffic congestion. Second, a longer switching time amplifies the cycle length, leading to overlapping queue peaks and making it difficult for vehicles across the entire road network to converge to a stable green wave, resulting in saturation flow rate fluctuations.

[0179] Secondly, since this embodiment uses the intelligent connected variable period maximum pressure control method, this method simultaneously assigns the signal switching phase time when calculating the phase duration. This moment differs from the optimal moment calculated through compromise above. Once Less than The signal phase will switch ahead of time, which will affect the synergistic effect of measures between upstream and downstream of the path.

[0180] Therefore, this embodiment introduces a threshold. This is used to maintain the continuity of vehicle queuing, thereby obtaining the final coordinated phase switching time. for:

[0181]

[0182]

[0183] In the formula, As weight, This refers to the phase switching moment of the previous round.

[0184] It is understandable that this embodiment obtains the final coordinated phase switching time through linear weighted fusion, keeps the coordinated phase green light window moving slowly, and ensures that the green wave will not be lost due to excessive switching time. This allows the signal phase to be switched according to the variable period maximum pressure control method without disrupting the phase switching rhythm between multiple coordinated phases.

[0185] Based on the specific implementation details, the effectiveness of the technical solution of the present invention will be demonstrated through experiments.

[0186] Specifically, the experiment uses the road network of a city's first ring road area as an example. By constructing a simulated road network, multiple levels of traffic demand coverage were set, including undersaturation, medium saturation, critical saturation, and oversaturation. The differences in the performance of the method at different congestion stages were observed. Compared with low demand, medium-to-high demand scenarios are more prone to phenomena such as queue propagation, overflow congestion induced by bottleneck intersections, and repeated dispersal of corridor convoys. Therefore, it is more effective in verifying the substantial effect of regional coordination. In the qualitative analysis of control performance, the simulated traffic volume was set to 40,000 veh / h.

[0187] Furthermore, regarding the comparative methods, to avoid weakening the persuasiveness by only comparing with traditional fixed timing or simplified versions, this experiment selected four representative published cutting-edge coordination methods as baselines: First, the traditional maximum pressure method based on micro-position weighting (PWBP), representing classic pressure feedback control lacking explicit coordination; second, the smoothing enhancement method under maximum pressure (Smoothing-MP), emphasizing reducing start-stop and phase jitter while maintaining throughput characteristics; third, the coordinated maximum pressure method (C-MP), achieving process coordination under decentralized conditions through adjacent coupling or corridor tendency; and fourth, the connected vehicle priority control method based on rolling optimization (RH-CAV), representing an advanced baseline in the intelligent connected vehicle environment. These methods represent four mainstream research directions: traditional micro-pressure, homogeneous smoothing, homogeneous coordination, and local optimization of connected vehicles, making the comparison more comprehensive.

[0188] Furthermore, this invention employs three categories of indicators—regional-scale operational efficiency, congestion level, and operational smoothness—for comprehensive evaluation. Specifically, average regional delay is used to measure overall traffic efficiency loss; average queue length and queue peak are used to measure congestion level and spillover risk; number of stops (or start-stop frequency) is used to measure operational smoothness; and throughput / number of completed vehicles is used to measure network service capacity. Finally, the control effect of the control method is analyzed under different penetration rates.

[0189] Specifically, such as Figure 3As shown, traffic flow changes exhibit a clear two-stage phenomenon. In the first 1500 seconds of the simulation, the road network is in the process of loading and absorbing traffic. The present invention (CAV-DZMP) shows a large delay peak (about 10 seconds / veh) in the initial period of the simulation (around 250 seconds), which then gradually decreases. Other baseline methods show a more gradual increase in delay in this regard. From the perspective of control theory, this is due to the initial reconstruction and active flow control caused by the dynamic sub-region division concept in the CAV-DZMP algorithm. In the initial stage of road network loading, the CAV-DZMP algorithm is performing dynamic calculation and boundary division of CAV sub-region topology based on the traffic flow initialization of each intersection. In order to establish a robust cooperative structure in space and align the lower green windows in time, the algorithm actively cuts off a portion of the convoy at peripheral intersections that are not in the center, preventing vehicles from entering the more central intersections downstream. Although this proactive adjustment causes a short-term increase in delay in the road network, it lays the groundwork for the stable operation of the road network bottleneck later. When the simulation time exceeds 1500 seconds, a large number of vehicles gather in the road network, the saturation of each intersection increases significantly, the traffic condition enters the saturation period and the congestion continues to worsen. At this time, it is easy to see that the delay curves of the four baseline algorithms, PWBP, Smoothing-MP, C-MP, and RH-CAV, rise sharply in an almost linear manner. After the simulation, the results of these algorithms are as high as 70s / veh or even 90s / veh. Among them, PWBP, as a locally greedy algorithm, lacks the ability to predict the remaining degrees of freedom of the downstream segment of the next road segment and continues to allow traffic to the downstream congested segment, causing serious backflow of waiting and the delay immediately aggravated. Although C-MP and Smoothing-MP made smoothing attempts in the early stage, their cooperative weights, which could not be adjusted under such high pressure, could not match the trend of dynamically transferred traffic flow, and as a result, they eventually lost control stability. After some initial fluctuations, the delay curve of CAV-DZMP exhibits strong stability after 1500 seconds. The CAV-DZMP curve no longer follows the baseline method with large fluctuations, but instead smoothly stabilizes at a low level of 20-30 seconds per vehicle. Experimental results demonstrate that the dynamic sub-region division and core path coordination mechanism proposed in this invention reduces the local pressure on the road network to a uniform regional pressure across the entire network, blocking the spread of congestion and enabling the network to maintain an excellent service level even under high pressure.

[0190] Specifically, such as Figure 4As shown, observing the queuing length evolution trend of each method, it can be seen that the queuing length of all control methods changes very little within 1000s, initially showing an S-shaped upward trend. During the rapid growth phase from 1000s to 2000s, the queuing length of all control methods increases rapidly. However, after 2000s, the ability of various algorithms to suppress queuing growth diverges. The queuing curves of PWBP and C-MP basically maintain the highest upward trend and eventually approach and stabilize at around 37veh / edge. For the limited length of urban road segments, this level of average queuing means that many road segments have at least reached the physical limit and overflow obstruction has occurred, resulting in congestion. The reason why RH-CAV rises sharply and significantly is that RH-CAV has a strong queuing suppression effect in the range of 1500s to 2500s. However, due to the limited field of view of RH-CAV rolling optimization, RH-CAV experienced another rebound in queuing length at the end of the simulation experiment. In contrast, the CAV-DZMP queue length curve exhibits a near-horizontal wave-like oscillation after reaching 25veh / edge to 35veh / edge, rather than a monotonically increasing state. This means that under CAV-DZMP, the inflow rate and outflow rate at each intersection of the road network are basically equal. When the queue in the inflow sub-area is about to be full, the algorithm can automatically divide new boundaries to redistribute green lights, clear the queue in time, and suppress the average queue of the entire network within a safe range, so as to prevent local intersections from being completely paralyzed.

[0191] Specifically, such as Figure 5 As shown, the number of stops directly reflects the smoothness of vehicle movement on the road network and the rationality of phase switching in the relevant control methods. It can be seen that after 1000 seconds, the number of stops using the traditional baseline method begins to rise continuously and slowly. After 3500 seconds, the average number of stops for Smoothing-MP, C-MP, and PWBP all exceed an extremely high 0.8 times / vehicle. In actual traffic operation, this means that more than 90% of vehicles cannot pass through each intersection in one go and must experience multiple stops and waits. From 1500 seconds to 3600 seconds, the number of stops for CAV-DZMP increases slowly, as shown at the bottom of several curves. This is a result of the present invention retaining the green window alignment mechanism based on the CAV forward trajectory within the sub-region. The alignment mechanism enables downstream intersections to perceive the arrival time of the convoy as early as possible, and can use the L1 minimization trade-off algorithm to fine-tune the start time of the green light duration, providing the core convoy with a stop-free trajectory as much as possible, still providing vehicles with the condition of passing through without stopping under oversaturated traffic conditions.

[0192] Specifically, such as Figure 6As shown, system throughput is the baseline for measuring the network's ability to maintain traffic operation. It reveals the entire process of changes in road network vehicle capacity. From 0 to 1500 seconds, with the continuous input of traffic demand, the throughput of all methods increases rapidly. However, fundamental principles of traffic engineering indicate that once the road network reaches its saturation boundary, if not properly controlled, the system's capacity will decrease. Figure 6 This clearly demonstrates the fundamental principle: around 2000 seconds, the throughput of RH-CAV, PWBP, Smoothing-MP, and C-MP all experienced a severe precipitous drop after reaching a peak of 10000-12000 veh / h, falling to the 6000-8000 veh / h range by the end of the simulation. This reverse plunge in throughput under high demand reflects a large-scale grid deadlock within the road network—although the green light at the intersection is on, vehicles behind cannot enter because the road ahead is already filled with queuing vehicles, resulting in a waste of effective green light resources. In terms of throughput, CAV-DZMP demonstrated absolute system resilience. Although its red line exhibited some sawtooth fluctuations during the climbing phase, reflecting the micro-adjustments during real-time reconstruction of dynamic sub-regions, CAV-DZMP's throughput maintained relatively good stability during the congestion period from 2000 to 3600 seconds, oscillating densely above the 10000veh / h baseline without the catastrophic drop seen in the baseline method. This proves that the dynamic sub-region division combined with maximum pressure control in this invention can always ensure that traffic flow moves in the direction of remaining capacity, while the green window calculation mechanism for the core path maximizes the emptying efficiency within the intersection.

[0193] Furthermore, to comprehensively evaluate the effectiveness of this invention, the experiment not only needs to examine its resistance to pressure under supersaturation conditions, but also needs to verify its performance evolution boundaries and backward compatibility under moderate normal flow conditions and with different CAV permeability rates. Therefore, this experiment also sets the traffic demand as a moderate normal flow condition of 20,000 veh / h, and sets four typical permeability gradients of 20%, 50%, 80%, and 100%. To intuitively reflect the changing law of control indicators with permeability jumps, Figure 7 and Figure 8 A stepped line graph was used to show the stage evolution characteristics of each method in terms of the two core macro indicators of average delay and average queue length.

[0194] Specifically, traditional physical detection-driven methods exhibit static curing characteristics, from Figure 7 and Figure 8 The evolution of the step curves for traditional baseline methods (PWBP, C-MP, Smoothing-MP) can be observed. It can be seen that the indicators of traditional baseline methods exhibit a completely horizontal, static baseline shape at different permeability steps. Figure 7 In the data analysis, the average delays for PWBP, C-MP, and Smoothing-MP were deadlocked at 32.27s, 29.58s, and 26.24s, respectively. Traditional maximum pressure signal control methods rely entirely on physical loop detectors (PLDs) buried upstream of intersections for data input. These PLDs can only count traffic flow and occupancy at the cross-section, but cannot analyze the underlying communication intentions of vehicles. Therefore, regardless of the proportion of CAVs (Carrier Availability Vehicles) in the road network, the information transparency of traffic flow remains unchanged for these traditional algorithms. This inherent limitation of the underlying perception architecture determines that its control effectiveness is insensitive to the penetration rate of connected vehicles, and it cannot fully benefit from the additional information dividends brought by the intelligent connected environment.

[0195] Furthermore, the RH-CAV method, which also belongs to the intelligent connected data-driven architecture, is analyzed. Figure 7 In the evolution trajectory of RH-CAV, under moderate traffic flow of 20,000 veh / h, its average delay fluctuates between 24 and 27 seconds within a penetration rate range of 20% to 80%. However, when the penetration rate reaches 100% coverage, its average delay not only fails to decrease but instead abruptly surges to 36.40 seconds / veh, and the queue length also deteriorates to 17.32 veh / edge. At 100% penetration, all CAVs have extremely small following distances. Under moderate traffic flow, this means that vehicles will quickly converge into high-density, rigid platoons with very small headway. RH-CAVs heavily rely on single-point or local micro-trajectory rolling optimization. When such high-density CAV platoons arrive simultaneously from multiple directions, the local controller falls into a multi-objective game constraint trap: to avoid disrupting one of the rigid platoons, RH-CAVs must allocate extremely long green lights, directly causing another high-density platoon in the conflicting direction to be forced to brake suddenly and experience a long red light wait. Due to the lack of a network-level macro-compromise mechanism, its local optimization computing power diverged when dealing with the control struggle of high-density connected vehicle fleets across the entire network, ultimately leading to an abnormal surge in system-level latency.

[0196] Furthermore, in Figure 7 and Figure 8 In this study, CAV-DZMP exhibited a unique spatiotemporal resource trade-off evolution characteristic. In terms of time efficiency (average delay), CAV-DZMP consistently maintained an absolute leading advantage across all penetration rate gradients. Even at 100% penetration, facing the game-theoretic conflict of high-density rigid vehicle convoys, its delay only slightly increased to 19.68 s / veh, far lower than PWBP (32.27 s) and RH-CAV (36.40 s), demonstrating that the core logic of this invention can significantly reduce vehicle time loss in the road network. Furthermore, in terms of space occupancy (queue length), from... Figure 8It can be observed that at a penetration rate of 20%, CAV-DZMP exhibits the lowest queue length (12.47 veh / edge), demonstrating good backward compatibility and initial queue suppression capabilities. However, as the penetration rate increases to 80% and 100%, its average queue length rises significantly, even surpassing traditional methods such as PWBP and Smoothing-MP in numerical terms. This seemingly contradictory phenomenon of low latency but long queues is actually a system-level control strategy adopted by the CAV-DZMP algorithm to trade space for time when dealing with high-density connected vehicle fleets. At a 100% penetration rate, CAV's smaller minimum following distance allows a denser fleet of vehicles than manually driven vehicles to be accommodated within a unit space of the road segment. To prevent fatal grid lock-ups when multiple densely packed convoys cross key intersections, CAV-DZMP's dynamic sub-zone mechanism proactively implements parking and flow control at peripheral intersections. This utilizes the physical space of road segments to safely store these densely packed connected convoys, statistically increasing the average queue length and allowing for accurate calculation and alignment with downstream green light windows. Once the coordinated green window opens, these tightly queued vehicles can be cleared in a single, high-volume burst, avoiding repeated starts and stops within the key intersection.

[0197] In summary, this invention can balance queuing suppression and delay reduction in the low-penetration stage where information is scarce. In the high-level autonomous driving stage with full coverage, it can break through the limitations of single-point optimization and actively utilize the spatial energy storage of the road network to achieve the ultimate time efficiency, demonstrating a robust and intelligent macro-control method. Therefore, this invention can effectively solve the deadlock of oversaturated grids, resolve multi-path coordination conflicts at minimal cost, and achieve forward-looking management of regional traffic.

[0198] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A maximum pressure traffic signal coordination control method integrating dynamic sub-region division in an intelligent connected environment, characterized in that, The method includes the following steps: S1: Based on the location, speed, and path information of CAV and the number of queued vehicles, saturation flow rate, number of lanes, and green ratio of HDV, predict the departure traffic flow of CAV and HDV respectively, and sum them by weight to obtain the predicted traffic flow of the intersection. Calculate the predicted pressure based on the predicted traffic flow of the intersection, and obtain the real-time observed pressure of the intersection. Perform weighted fusion of the predicted pressure and the real-time observed pressure to obtain the comprehensive pressure of each intersection. S2: Based on the geometric connectivity of the upstream and downstream intersections and the measured HDV flow, the basic weights are obtained. Based on the comprehensive pressure, an exponential amplification coefficient is constructed. The product of the basic weights and the exponential amplification coefficient is used as the hyperedge weights. Based on the hyperedge weights, the intersections are clustered to complete the dynamic division of the control sub-regions. S3: Within each control sub-area, the predicted queue length of each approach lane is calculated using the real-time trajectory information of the CAV, and the instantaneous comprehensive predicted pressure of each phase is calculated in combination with the turning rate. When the space occupancy rate of the downstream intersection approach lane exceeds the preset threshold, the effective pressure is obtained by introducing a penalty function. The green light duration is allocated based on the effective pressure and weighted by history to obtain the final effective green light duration. S4: Based on the effective green light duration of the final execution, obtain the downstream green window center, and combine it with the real-time driving time and vehicle queue emptying time to obtain the ideal green light start time upstream; S5: Based on the ideal green light start time of each path upstream, construct the optimal compromise start time, and perform linear weighted fusion with the phase switching time of the previous round to obtain the final coordinated phase switching time, thereby realizing coordinated traffic signal control.

2. The maximum pressure traffic signal coordination control method for intelligent connected environments integrating dynamic sub-region division as described in claim 1, characterized in that, Specifically, S1 is: Get the future Queuing situation at intersections within a given time period, including future... HDV traffic volume leaving intersection i within a given time period Defined as: ； In the formula, This represents the real-time number of vehicles queuing at time t during coordination phase i at intersection i. This represents the saturation flow rate at intersection i; This represents the number of lanes from intersection i to the downstream intersection i+1; This indicates the coordinated phase green light ratio at intersection i; Future HDV traffic volume at intersection i within a given time period Defined as: ； In the formula, This represents the real-time number of vehicles queuing at time t, which is the coordination phase of the k-th approach at the upstream intersection i-1. This represents the saturation flow rate at the k-th approach of the upstream intersection i-1. This represents the number of lanes at the k-th approach to the upstream intersection i-1. This represents the coordinated phase green light ratio of the k-th approach at upstream intersection i-1; Thus, we can obtain the future. Predict traffic flow at intersections within a specified time period for: ； In the formula, For the future The number of vehicles arriving at intersection i within a given time period. For the future The number of vehicles leaving intersection i within a given time period; Intersection i at Predicted pressure over time Defined as: ； Real-time observed pressure at intersection i at time t Defined as: ； Thus, the combined pressure at intersection i at time t is obtained. for: ； In the formula, Let be the time decay coefficient, and .

3. The maximum pressure traffic signal coordination control method for intelligent connected environments integrating dynamic sub-region division as described in claim 1, characterized in that, Specifically, the step of using the product of the base weight and the exponential amplification factor as the hyperedge weight is as follows: The weights of basic connectivity and measured traffic flow are defined as basic weights. The expression is: ； In the formula, This represents the geometric connectivity coefficient of the upstream and downstream intersections. This represents the HDV traffic weighting coefficient. Indicates the intersection of upstream and downstream. HDV measured flow rate within the time window; Constructing the exponential amplification factor The expression is: ； In the formula, Indicates the pressure sensitivity coefficient; This represents the combined pressure at intersection i at time t. This represents the combined pressure at intersection i at time t+1; Multiplying the base weight by the exponential amplification factor yields the hyperedge weight, expressed as: ； In the formula, This represents the weight of the superedge.

4. The maximum pressure traffic signal coordination control method for intelligent connected environments integrating dynamic sub-region division as described in claim 1, characterized in that, The penalty function is specifically: ； In the formula, For the penalty function, The threshold for the occupancy rate of the downstream intersection approach lane. This represents the real-time lane space occupancy rate.

5. The maximum pressure traffic signal coordination control method for intelligent connected environments integrating dynamic sub-region division according to claim 4, characterized in that, Specifically, S3 is: Arrival time of a single intersection approach lane l Discretized into time The vehicle arrival curve is obtained, and its expression is: ； In the formula, In time At that time, the number of CAVs arriving at the approach lane l of phase j at intersection i. Indicates the approach lane l of phase j at intersection i; This indicates the estimated time when vehicle v will arrive at intersection entrance lane l. This allows us to obtain the predicted queue length of lane l at the intersection entrance. The expression is: ； In the formula, This represents the saturation flow rate of lane l at the intersection entrance. In time At that time, the number of CAVs arriving at the import lane l ; Thus, instantaneous comprehensive predicted pressure is obtained. The expression is: ； In the formula, For a single point in time used to construct an instantaneous comprehensive forecast of pressure, This is the maximum queuing capacity of the import lane. Let j be the set of downstream exit channels. Let k be the maximum queuing capacity of the exit lane. Indicates the spatial coupling influence factor. This indicates the turning rate when a vehicle turns from entrance lane l to exit lane k. For the future At that time, the predicted queue length of exit lane k, For the future At that time, the predicted queue length of the exit lane k, which incorporates the turning rate, is combined; By constructing a penalty function, the effective pressure at phase j of intersection i is obtained. for: ； In the formula, Indicates the pressure attenuation coefficient; Before each phase begins, the phase for which the green light duration is to be calculated is considered as the fourth phase of a cycle. The fourth phase and the preceding three signal phases are combined to form a complete signal cycle, creating a new cycle, which is defined as the virtual cycle, for the purpose of allocating green light durations. Specifically: ； In the formula, This represents the green light ratio of phase j. This represents the sum of the pressures at intersection i in phase j compared to the three phases preceding it; p ij This represents the pressure value at phase j in intersection i; The green light duration allocated to the current phase requiring passage is obtained based on the imaginary period and the green light ratio. for: ； In the formula, This indicates the duration of the imaginary period formed by phase j and its three preceding phases; Green light duration By performing historical weighting, the final effective green light duration is obtained. for: ； In the formula, Represents the smoothing coefficient. Indicates the number of cycles.

6. The maximum pressure traffic signal coordination control method for intelligent connected environments integrating dynamic sub-region division as described in claim 1, characterized in that, Specifically, S4 is: The traffic signal controller at each intersection along the defined path broadcasts predicted values ​​of pressure and green light intensity once within a preset time before phase release, thereby obtaining the downstream green window center. for: ； In the formula, The green window center time for phase j at the downstream intersection. The predicted green light duration for phase j at the downstream intersection; For vehicles traveling from upstream to downstream, the travel time at 85% of the locations uploaded by CAV in the first 30-60 seconds is taken as the upstream to downstream travel time. ; Calculate the downstream vehicle queue emptying time The expression is: ； In the formula, The queue length based on CAV at the downstream intersection. The saturation flow rate at the downstream intersection; Ultimately, the ideal green light start time for upstream is obtained. for: ； In the formula, To ensure a safety margin, vehicles must pass through the green wave at downstream intersections without getting stuck, and ,in, The variance of travel time, As weight.

7. The maximum pressure traffic signal coordination control method for intelligent connected environments integrating dynamic sub-region division as described in claim 1, characterized in that, Specifically, S5 is: Along each path The set of ideal green light start times upstream is obtained by reverse recursion. To find the optimal compromise moment for unifying the ideal green light start time across all phases. The expression is: ； in, >0 indicates the weight of the given path; This represents the error value between the actual time of switching the green light phase and the ideal start time of path R; Introduce a threshold This is used to maintain the continuity of vehicle queuing, thereby obtaining the final coordinated phase switching time. for: ； ； In the formula, As weight, This refers to the phase switching moment of the previous round.

Images