Big data based dynamic adaptive control system of urban arterial green wave band

Abstract

The application relates to the field of intelligent traffic control and traffic signal optimization, in particular to a city trunk road green wave dynamic self-adaptive control system based on big data, which comprises the following modules: a multi-source data sensing module, which is used for receiving trunk road traffic flow data, intersecting road traffic flow data and multi-intersection traffic data, and performing space-time alignment processing to obtain target fusion traffic flow data; a vehicle fleet feature solving module, which is used for calculating vehicle fleet real-time dispersion and spatial distribution density; an adaptive control module, which is used for determining phase elastic expansion and contraction tolerance, and solving node state activity based on a preset topological adjustment model containing a preset control non-should period parameter, determining instruction issuing weight and outputting a green wave dynamic control instruction; and an instruction issuing execution module, which is used for driving a traffic signal control device to execute green light phase adjustment; the application can realize active capture of trunk road dispersed vehicle fleets and dynamic recovery of green wave breaking states.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control and traffic signal optimization, specifically to a dynamic adaptive control system for green wave zones on urban arterial roads based on big data. Background Technology

[0002] In urban arterial road traffic organization, signal coordination control at consecutive intersections usually improves the continuous traffic capacity of vehicles through green wave belts, so as to reduce the number of stops and delays and maintain the efficiency of mainline traffic flow. The effectiveness of green wave belt control directly depends on the ability to timely perceive and coordinate the arrival status of mainline convoys, traffic disturbances on intersecting roads, and traffic rhythm at multiple intersections.

[0003] In traditional methods, green wave control on main roads often uses preset timing schemes or adjustments based on single-point detection data. When the traffic flow becomes dispersed or fragmented due to factors such as bus stops, temporary lane occupancy, lane merging, and queuing on side roads, the existing control methods have weak spatiotemporal unification capabilities for multi-source heterogeneous traffic data. It is difficult to identify the degree of dispersion and spatial distribution characteristics of the traffic flow in a timely manner. As a result, the targeted and real-time adjustment of the green light phase cannot meet the preset control accuracy and delay requirements. At the same time, it is also easy to squeeze the release space of intersecting roads when optimizing traffic flow on main roads, affecting the overall coordinated operation stability of continuous intersection groups. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a dynamic adaptive control system for green wave corridors on urban arterial roads based on big data. Specifically, the technical solution of this invention includes:

[0005] The multi-source data perception module is used to receive multi-source heterogeneous traffic flow data through a preset data interface. The multi-source heterogeneous traffic flow data includes main road traffic flow data, intersecting road traffic flow data, and multi-intersection traffic data. Based on a preset road network coordinate system and a unified time axis, the multi-source heterogeneous traffic flow data is spatiotemporally aligned to obtain the target fused traffic flow data.

[0006] The fleet feature calculation module is connected to the multi-source data perception module. It is used to receive target fused traffic flow data and calculate the real-time dispersion and spatial distribution density of the fleet based on the target fused traffic flow data.

[0007] The adaptive control module is used to determine the phase elastic scaling tolerance based on the real-time dispersion and spatial distribution density of the fleet, and uses the phase elastic scaling tolerance as the input of the preset topology adjustment model. The preset topology adjustment model includes a preset control refractory period parameter to characterize the control intervention time interval of continuous intersections. It is used to calculate the node state activity based on the real-time dispersion of the fleet, and determine the command issuance weight based on the node state activity. Combined with the phase elastic scaling tolerance, it outputs a green wave dynamic control command containing the command issuance weight.

[0008] The instruction issuance and execution module is used to send green wave dynamic control instructions to the traffic signal control equipment at the corresponding intersection, so as to drive the traffic signal control equipment to perform green light phase adjustment based on the instruction issuance weight.

[0009] Preferably, the process of spatiotemporal alignment of multi-source heterogeneous traffic flow data includes: performing timestamp synchronization processing on the multi-source heterogeneous traffic flow data to convert the multi-source heterogeneous traffic flow data from an asynchronous state to a synchronous state; and performing spatial mapping processing on the synchronous multi-source heterogeneous traffic flow data based on a preset road network coordinate system to generate target fused traffic flow data.

[0010] Preferably, the fleet feature calculation module includes:

[0011] The shock wave feature extraction unit is used to extract speed change waves and deceleration transmission behavior of vehicles in car-following state as features of the incoming shock wave based on the target fused traffic flow data.

[0012] The dispersion calculation unit is used to determine the real-time dispersion of the vehicle fleet based on the velocity variance and average velocity corresponding to the characteristics of the incoming shock wave.

[0013] The density calculation unit is used to calculate the spatial distribution density based on the vehicle coordinate information in the target fused traffic flow data.

[0014] Preferably, the process of determining the phase elastic scaling tolerance includes: under the constraint of a preset global signal period, subtracting the sum of a preset minimum green light time and a preset clearing time from the preset global signal period, performing dimensionless processing on the spatial distribution density and the real-time dispersion of the vehicle fleet based on a reference benchmark, and using the product of the resulting difference and the spatial distribution density as the benchmark value of the phase elastic scaling tolerance, and using the product of the benchmark value and the reciprocal of the real-time dispersion of the vehicle fleet as the phase elastic scaling tolerance; performing elastic scaling processing on the preset initial green light phase according to the phase elastic scaling tolerance to generate a green wave dynamic control command, wherein the elastic scaling processing includes green light early start operation and green light delayed end operation.

[0015] Preferably, when the real-time dispersion of the fleet is greater than a preset dispersion threshold, an early green light start operation is triggered, and a pre-extension time of the preset initial green light phase is generated; when the real-time dispersion of the fleet is less than or equal to the preset dispersion threshold, a delayed green light stop operation is triggered, and a post-extension time of the preset initial green light phase is generated.

[0016] Preferably, the queue lengths of intersecting roads are extracted from the target fused traffic flow data, and the queue lengths of intersecting roads are compared with a preset queue length threshold. If the queue length of intersecting roads is greater than the preset queue length threshold, a suppression instruction is generated, and the elastic scaling process is canceled based on the suppression instruction. If the queue length of intersecting roads is less than or equal to the preset queue length threshold, a release instruction is generated, and the elastic scaling process is executed based on the release instruction.

[0017] Preferably, the system also includes a macro-evaluation module, which includes:

[0018] The reconstruction rate calculation unit is used to determine the discrete reconstruction rate of the fleet based on the actual number of vehicles passing through the green wave and the preset expected number of vehicles passing through.

[0019] The coherence index calculation unit is used to perform a weighted summation of the vehicle discrete reconstruction rate and the reciprocal of the travel time variance of adjacent intersections in the multi-intersection traffic data of the target fused traffic flow data, to obtain the spatiotemporal coherence index of the green wave band.

[0020] Preferably, the spatiotemporal coherence index of the green wave band is compared with a preset coherence benchmark value; if the spatiotemporal coherence index of the green wave band is lower than the preset coherence benchmark value, the preset control refractory period parameter in the preset topology adjustment model is increased; if the spatiotemporal coherence index of the green wave band is higher than or equal to the preset coherence benchmark value, the preset control refractory period parameter in the preset topology adjustment model is kept unchanged.

[0021] Preferably, the adjustment process of the preset topology adjustment model includes: multiplying the difference between the preset maximum dispersion and the real-time dispersion of the fleet with the preset control refractory period parameter as the node state activity; determining the weight adjustment coefficient based on the ratio of the node state activity to the preset basic activity; and using the weight adjustment coefficient to perform a multiplication operation on the preset initial value of the command issuance weight to realize the dynamic increase or decrease of the command issuance weight.

[0022] Preferably, the target fused traffic flow data is preprocessed before transmission. The preprocessing includes: compressing the target fused traffic flow data based on a preset encoding algorithm, and transmitting the compressed target fused traffic flow data to the fleet feature calculation module based on a block transmission strategy.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. This invention uses the asynchronous elimination and spatial mapping unit of the multi-source data sensing module to synchronize heterogeneous traffic flow data with timestamps and transform it into a unified coordinate system. This effectively solves the problem of spatiotemporal inconsistency of data in traditional methods, eliminates the observation bias of a single device, provides a unified and accurate physical object and time reference for feature calculation, and reduces the prediction error of downstream vehicle arrival time.

[0025] 2. This invention utilizes a shock wave feature extraction unit and a discreteness and density calculation unit, enabling the system to accurately separate normal following traffic flow from structurally broken traffic flow from fused data; by calculating speed change waves, the ratio of speed variance to average speed, and spatial distribution, the original data is transformed into feature indicators reflecting the breakup trend of the vehicle fleet, overcoming the technical deficiency of traditional methods in being unable to identify the degree of dispersion of the vehicle fleet in a timely manner.

[0026] 3. This invention calculates the phase elastic scaling tolerance based on discreteness and spatial density, and intelligently triggers early green light start or delayed green light stop operation according to the discreteness threshold; this enables the system to accurately match the control direction for different fracture modes, transforming abstract disturbances into executable time quantities, and achieving efficient and dynamic reconstruction capture of discrete convoys on the main road without disrupting the global cycle.

[0027] 4. This invention dynamically corrects the refractory period parameter by calculating the spatiotemporal coherence index, while the topology adjustment model dynamically allocates weights based on node activity. This mechanism not only avoids timing conflicts caused by multiple intersections competing for control resources, but also effectively suppresses frequent and ineffective short-term interventions through closed-loop feedback of results, achieving coordination between the dynamic response capability of the entire corridor and long-term operational stability.

[0028] 5. This invention prioritizes the transmission of core data blocks directly related to discreteness calculation through encoding algorithms and block transmission strategies. This significantly reduces the network transmission burden of high-frequency sensing data without sacrificing the timeliness of key control information, ensuring the second-level real-time calculation and efficient implementation of green wave adaptive control commands under big data conditions. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the module of the dynamic adaptive control system for green wave belts on urban main roads based on big data provided in the embodiments of this application. Detailed Implementation

[0030] 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.

[0031] A big data-based dynamic adaptive control system for green wave belts on urban arterial roads, comprising:

[0032] The multi-source data perception module is used to receive multi-source heterogeneous traffic flow data through a preset data interface. The multi-source heterogeneous traffic flow data includes main road traffic flow data, intersecting road traffic flow data, and multi-intersection traffic data. Based on a preset road network coordinate system and a unified time axis, the multi-source heterogeneous traffic flow data is spatiotemporally aligned to obtain the target fused traffic flow data.

[0033] The fleet feature calculation module is connected to the multi-source data perception module. It is used to receive target fused traffic flow data and calculate the real-time dispersion and spatial distribution density of the fleet based on the target fused traffic flow data.

[0034] The adaptive control module is used to determine the phase elastic scaling tolerance based on the real-time dispersion and spatial distribution density of the fleet, and uses the phase elastic scaling tolerance as the input of the preset topology adjustment model;

[0035] The preset topology adjustment model includes a preset control refractory period parameter to characterize the control intervention time interval of continuous intersections. It is used to calculate the node status activity based on the real-time dispersion of the vehicle fleet, and to determine the command issuance weight based on the node status activity. Combined with the phase elastic scaling tolerance, it outputs a green wave dynamic control command containing the command issuance weight.

[0036] The instruction issuance and execution module is used to send green wave dynamic control instructions to the traffic signal control equipment at the corresponding intersection, so as to drive the traffic signal control equipment to perform green light phase adjustment based on the instruction issuance weight.

[0037] This embodiment provides a dynamic adaptive control mechanism for green wave zones at continuous intersections on urban arterial roads, such as... Figure 1 As shown; specifically, taking the central axis of a north-south main road in a certain city as the main scenario, the first intersection, second intersection, third intersection, fourth intersection and fifth intersection are set up in sequence on the central axis. The second intersection to the fourth intersection constitute the key coordination section; the system is deployed in the traffic control center and establishes data connection with roadside radar, video detectors, floating car platform and traffic signals at each intersection;

[0038] During the morning rush hour, a convoy that should have traveled continuously along the green wave is affected by roadside parking, bus stops, and lane merging between the first and second intersections, resulting in a discrete phenomenon where the front of the traffic is spread out and the rear of the traffic is piled up. The system continuously senses, calculates, adjusts, and issues control commands around this discrete process.

[0039] In one specific implementation, the multi-source data sensing module first receives three types of data: one type is main road traffic flow data, such as vehicle speed, flow rate, headway, and vehicle trajectory at each entrance lane from the first to the fourth intersection; another type is intersecting road traffic flow data, such as queue length, number of stops, and release rate of east-west side roads at the second and third intersections; and the third type is multi-intersection traffic data, such as arrival time, departure time, and travel time of the same license plate or anonymous trajectory identifier between adjacent intersections.

[0040] Taking a specific operation as an example: During a certain control cycle, 6 vehicles are detected upstream of the second intersection, with arrival times of 2s, 4s, 5s, 11s, 12s, and 13s respectively. If the original fixed green wave scheme is followed, these 6 vehicles will be divided into two sub-queues: the first 3 vehicles and the last 3 vehicles. After unifying the data from different sources to the same clock reference and the same road network coordinates, the system forms target fused traffic flow data, which includes both the position of each vehicle in the corridor and the queuing boundary information of the branch road.

[0041] After receiving the above fusion results, the fleet feature calculation module does not directly use a single flow value as the control basis. Instead, it calculates whether the fleet is discrete, the discreteness value, and whether the spatial distribution characteristics meet the downstream green light conditions. Here, the real-time discreteness can be understood as a comprehensive reflection of the speed and time difference within the fleet, and the spatial distribution density can be understood as the degree of vehicle aggregation within a unit road segment.

[0042] Taking the aforementioned 6 vehicles as an example, if the average speed of the first 3 vehicles is 42 km / h and the average speed of the last 3 vehicles is 28 km / h, and there is a time gap between the two groups that is greater than the preset time interval threshold, then the system determines that the convoy has changed from a compact formation to a discrete formation; if the 6 vehicles are still concentrated within a 200-meter range, then the spatial distribution density is high, which means that by appropriately stretching the downstream green light window, it is still possible to reintegrate them into the same green wave channel.

[0043] The adaptive control module takes the discreteness and spatial distribution density as inputs, first calculates the phase elastic scaling tolerance, and then sends the tolerance to the topology adjustment model. The preset topology adjustment model is used to characterize the dynamic coupling degree of the green wave control relationship between consecutive intersections. For ease of explanation, the current common cycle of the third intersection can be set to 120s. After system calculation, it is considered that this cycle can be advanced by 2s and extended by 5s. Then the phase elastic scaling tolerance can be expressed as an allowable interval [-2, +5].

[0044] Meanwhile, a preset control refractory period parameter is used to limit the time interval between two consecutive control intervention actions to be greater than a preset time threshold, so as to avoid the signal controller being repeatedly stretched in adjacent cycles, causing timing oscillations; the system calculates the node status activity based on real-time dispersion; when the activity is greater than or equal to the preset activity threshold, it indicates that there is a higher priority reconstruction demand at that node, and the command issuance weight increases accordingly; when the activity is less than the preset activity threshold, it indicates that maintaining the original configuration is more stable, and the command issuance weight decreases accordingly; the final output green wave dynamic control command includes both the time amount for early start or delayed end of the green light on the main road at the third intersection, as well as the execution weight of the command relative to other candidate adjustment actions;

[0045] The instruction issuance and execution module is responsible for sending the green wave dynamic control instructions generated by the control center to the traffic signal control equipment at the corresponding intersection. The traffic signal control equipment is equipped with a programmable logic controller. The instruction issuance and execution module establishes a communication connection with the programmable logic controller, and through the programmable logic controller, it parses the green wave dynamic control instructions and directly drives the load unit of the corresponding intersection signal to perform green light phase adjustment.

[0046] To prevent multiple nodes from simultaneously performing adjustments exceeding the preset phase stretching limit, the system prioritizes actions with higher weight. For example, if the second and third intersections simultaneously request to extend the green light on the main road, and the third intersection is closer to the break in the traffic flow and has higher activity, the system will first have the third intersection perform a 4-second delay, while the second intersection will only perform a 1-second fine-tuning. After receiving the instruction, the traffic signal controller will complete the green light start and end time correction without violating local safety constraints.

[0047] In abnormal conditions, if a data source is temporarily interrupted, such as if the floating car platform does not send a response within 3 seconds, the system can still use the roadside detector and video trajectory for degraded control; if the positioning deviation of different data sources for the same vehicle exceeds the preset tolerance value, the data source with higher reliability will be temporarily used as the master data, and the low reliability data will be marked as auxiliary reference.

[0048] If the node's activity level is high, but the local traffic signal is already in the yellow light transition or pedestrian clearance phase, the phase will not be changed directly in this cycle. Instead, the instruction will be postponed to the next executable window. If the quality of all data is below the minimum threshold, the system will revert to the preset coordination scheme to avoid abnormal data causing miscontrol.

[0049] At 7:45 AM during the morning rush hour on the central avenue, a segment of discrete vehicles affected by bus stops appeared between the second and third intersections. The system completed fusion and calculation within 2 seconds, recognizing that although the vehicle group had split, the rear vehicles were still concentrated within a 250-meter range. Therefore, it issued a 4-second delay to the main road green light at the third intersection, executing a control command with a weight of 0.82, while simultaneously issuing a command to the second intersection to maintain the original plan, executing a weak adjustment command with a weight of 0.31. After the third intersection executed the command, the four rear vehicles that would have encountered a red light were reintegrated into the green light window, and the green wave breakpoint was partially repaired.

[0050] The purpose of this step is to construct a closed-loop control link from perception to solution to tolerance calculation, weight decision-making, and intersection execution, so as to actively capture discrete vehicle convoys on the main road and dynamically restore the green wave interruption state, while maintaining the basic traffic order of intersecting roads.

[0051] Furthermore, the multi-source data sensing module includes:

[0052] The asynchronous elimination unit is used to perform timestamp synchronization processing on multi-source heterogeneous traffic flow data to convert the multi-source heterogeneous traffic flow data from an asynchronous state to a synchronous state.

[0053] The spatial mapping unit is used to perform spatial mapping processing on multi-source heterogeneous traffic flow data in a synchronous state based on a preset road network coordinate system to generate target fused traffic flow data.

[0054] This embodiment provides a spatiotemporal alignment mechanism to address the problem of asynchronous and locational inconsistencies in multi-source traffic data. Specifically, in the aforementioned central axis avenue scenario, if only the original reports from each detection device are relied upon, it is easy for situations to arise where the video identifies that a vehicle has crossed the stop line, but the radar still shows that it is in the queuing section of the entrance lane, leading to incorrect predictions of the arrival time of the convoy at downstream intersections. Therefore, before the data enters the convoy feature calculation process, it is first processed uniformly through an asynchronous elimination unit and a spatial mapping unit.

[0055] In one specific implementation, the asynchronous elimination unit normalizes the timestamps of different data sources; a simplified example can be used: the video device uploads at 08:00:10.400, the radar uploads at 08:00:10.120, and the floating car platform actually arrives at the platform at 08:00:11.000 due to network backhaul delay, but its sampling generation time is 08:00:10.300;

[0056] The system does not use the reception time as the reference, but uses the sampling generation time as the main time axis, and constructs synchronization frames in 100-millisecond steps; therefore, the aforementioned three data will be mapped to the same control frame from 08:00:10.300 to 08:00:10.400; if a certain type of data is missing in a frame, it will be filled in using the short-time hold value of the previous frame or the interpolation value of the adjacent frame, but the effective filling time shall not exceed the preset upper limit, such as 1 second, otherwise it will be marked as a missing frame;

[0057] The spatial mapping unit further transforms data from different device coordinate systems to a unified road network coordinate system; for example, video equipment provides image pixel coordinates, radar provides polar coordinates relative to roadside equipment, and floating cars provide latitude and longitude; the system pre-establishes a reference coordinate system for the central axis road network, sets the stop line of the second intersection approach lane as the zero point of the longitudinal coordinate, and takes the northward direction along the main road as the positive direction;

[0058] For a first vehicle in the video frame, if its pixel position, after calibration and conversion, corresponds to 85 meters upstream of the stop line; for a second vehicle recorded by radar, if its polar coordinates, after conversion, correspond to 92 meters upstream of the stop line; for the first trajectory point of the floating car, if, after map matching, it falls on the midline position between the second and third intersections and corresponds to 310 meters upstream, then all three can be compared, clustered, and tracked on the same coordinate axis.

[0059] If we further perform a micro-level simulation, we can assume that there are three vehicles at the same moment after synchronization: vehicle A, vehicle B, and vehicle C. Vehicle A is derived from video and is located at -30 meters; vehicle B is derived from radar and is located at -32 meters; and vehicle C is derived from a floating car and is located at -145 meters.

[0060] Based on proximity and speed consistency, the system determines that vehicle A and vehicle B may be duplicate observations of the same target, and thus merges them into a unified vehicle object; vehicle C is retained separately. After this processing, the target fused traffic flow data is no longer a loose data patchwork, but a unified traffic flow state description for subsequent control.

[0061] In abnormal conditions, if the time difference between two types of data exceeds the preset synchronization tolerance, such as more than 2 seconds, they will not be forcibly merged, but will be assigned to different control frames. If the position difference of the same vehicle in different sources after spatial mapping exceeds the preset distance threshold, such as more than 25 meters, the system will not merge them directly, but will retain the dual-track records and hand them over to subsequent reliability screening. If the road network calibration parameters fail due to construction and rerouting, the system will automatically switch to the temporary coordinate layer, retain only the intersection-level aggregation statistics, and suspend fine control based on lane-level trajectory.

[0062] At the south entrance of the third intersection of Zhongzhou Avenue, a video detector misidentified a large bus as being too far forward due to angular obstruction, while the radar gave a stable position that was too far back. The system first corrected the timestamp, then compared the difference between the two in a unified coordinate system, and found that it fell within the allowed merging range. Finally, the weighted middle position was used as the fusion result for the bus. As a result, the fourth intersection downstream will not open the green light earlier or later due to the deviation of a single device when estimating the arrival time of the convoy.

[0063] The purpose of this mechanism is to first eliminate the inconsistency between the temporal and spatial representation of data, so that subsequent calculations of indicators such as dispersion, density, and travel time can be based on the same physical object and the same time reference, thereby reducing the probability of misjudgment.

[0064] Furthermore, the fleet feature calculation module includes:

[0065] The shock wave feature extraction unit is used to extract speed change waves and deceleration transmission behavior of vehicles in car-following state as features of the incoming shock wave based on the target fused traffic flow data.

[0066] The dispersion calculation unit is used to determine the real-time dispersion of the vehicle fleet based on the velocity variance and average velocity corresponding to the characteristics of the incoming shock wave.

[0067] The density calculation unit is used to calculate the spatial distribution density based on the vehicle coordinate information in the target fused traffic flow data.

[0068] This embodiment provides a feature calculation mechanism for fleet morphology recognition; specifically, the fused vehicle positions and speeds alone are insufficient to determine whether green wave state reconstruction needs to be initiated, because some vehicle speed reductions are normal following, while others indicate that the fleet has experienced structural breakage.

[0069] Therefore, in this embodiment, the characteristics of the incoming shock wave are further extracted, and the real-time dispersion and spatial distribution density are calculated accordingly. The formula for calculating the spatial distribution density is as follows:

[0070]

[0071] in, For absolute spatial distribution density, This represents the total number of vehicles in a fleet that are continuously distributed in the vehicle coordinate information. and These are the absolute position coordinates of the last and first vehicles in the convoy, respectively, within the preset road network coordinate system.

[0072] In one specific implementation, the shock wave feature extraction unit monitors the propagation of speed abrupt changes in vehicles within the same convoy. This can be illustrated with a very simple example: Suppose there are 5 consecutive vehicles between the second and third intersections, numbered M1 to M5; at a certain moment, the speeds of M1 and M2 are 40 km / h, M3 decreases to 22 km / h due to a bus exiting the station ahead, and M4 and M5 decrease to 24 km / h and 20 km / h respectively.

[0073] Specifically, the condition for determining a speed change wave in the shock wave feature extraction unit is: real-time monitoring of the speed difference and deceleration between two adjacent vehicles. When the absolute value of the deceleration of the rear vehicle relative to the front vehicle exceeds the preset shock wave trigger deceleration threshold within a continuous time period, and the number of vehicles that the deceleration feature is transmitted to the following vehicles in the preset time window reaches the preset following chain threshold, then it is extracted as a feature to be incorporated into the shock wave.

[0074] The system observed that the sudden drop in speed was not an isolated event of a single vehicle, but rather a propagation from front to back along the following chain. This propagation process was then identified as a merging shock wave feature. Here, merging emphasizes that it is often coupled with factors such as merging of side road vehicles, bus stops, and interference from non-motorized vehicles, resulting in uneven vehicle density or following distances exceeding the preset threshold.

[0075] The dispersion calculation unit rigorously calculates the velocity variance and average velocity of the vehicle groups corresponding to the shock wave, and uses the ratio of these two values ​​to characterize the real-time dispersion. Specifically, to ensure the rigor of the feature solution process, the set of vehicles corresponding to the shock wave features extracted is set as follows: ,Include There are 10 vehicles, and the instantaneous speed of each vehicle is 100 km / h. The system calculates the average speed. :

[0076]

[0077] The formula for calculating velocity variance is:

[0078]

[0079] The formula for calculating the real-time dispersion of a vehicle fleet is:

[0080]

[0081] in, To avoid a preset minimum speed protection benchmark value where the denominator is zero; To find the maximum value function; continuing with the previous example, if the speeds of M1 to M5 are 40, 39, 22, 24, and 20 respectively, then the average speed is approximately 29. The speed variance can accurately reflect the degree to which the speed of each vehicle deviates from the average value.

[0082] The larger the ratio, the more dispersed the vehicle speed distribution and the lower the average speed, indicating that the convoy is more unstable. If another group of vehicles has speeds of 36, 35, 34, 33, and 35 respectively, then although there are slight differences, the variance is small and the average speed is high, and its dispersion is significantly lower. In this way, the system decouples and distinguishes the speed reduction from the dispersion characteristics of the convoy, focusing on identifying the dispersion caused by unevenness.

[0083] The density calculation unit estimates the spatial distribution density based on vehicle coordinates. The critical control zone from the second intersection to the third intersection can be set at 300 meters. If eight vehicles from the same convoy are concentrated within 150 meters of this zone, the density is high. If the eight vehicles are dispersed to more than 400 meters or even across two adjacent intersections, the density is low.

[0084] For ease of control, the system can use a sliding window for statistics; for example, every 50 meters is a local segment, and the number of vehicles is counted separately to form a distribution profile of the convoy on the road segment; if the number of vehicles in a certain window suddenly drops from 4 to 0, and then 3 vehicles appear in the next window, the system can identify that there is a vehicleless section in the middle of the convoy that exceeds the preset distance threshold, which often corresponds to the characteristic of loss of convoy continuity.

[0085] In abnormal conditions, if the number of vehicles in the target fused traffic flow data is less than the minimum analysis threshold, such as less than 3 vehicles, then shock wave features will not be calculated for this cycle, and only basic flow features will be retained. If the average speed is close to 0, directly comparing the speed variance with the smallest positive value may lead to artificially high dispersion; therefore, the system uses a protective baseline value. Set a lower limit for the value to avoid the denominator being too small; if the vehicle coordinates change briefly due to occlusion, perform trajectory smoothing before density calculation to prevent the recognition error from being mistaken for a break in the vehicle convoy.

[0086] Between the north exit of the second intersection and the south entrance of the third intersection on the central axis avenue, the system observed in three consecutive sampling frames that the speed of a convoy rapidly differentiated from 38 km / h to 40 km / h in the front and 18 km / h in the rear, while the length of the convoy expanded from 120 meters to 280 meters. Based on this, the system determined that there was a significant shock wave propagation in this section of the road, with increased real-time dispersion, but the spatial distribution was still within the capture range. Therefore, the system passed the marker that needs to be prioritized for downstream repair to the subsequent control module.

[0087] The purpose of this step is to transform the raw traffic data into control features that can directly reflect the fragmentation trend and reconfigurability of the vehicle fleet, thereby enabling green wave regulation to shift from total volume assessment to structural analysis.

[0088] Furthermore, the adaptive control module includes:

[0089] The tolerance calculation unit is used to subtract the sum of the preset minimum green light time and the preset clearing time from the preset global signal period under the constraint of the preset global signal period. After performing dimensionless processing on the spatial distribution density and the real-time dispersion of the fleet based on the reference benchmark, the product of the difference and the spatial distribution density is used as the benchmark value of the phase elastic scaling tolerance. The product of the benchmark value and the reciprocal of the real-time dispersion of the fleet is used as the phase elastic scaling tolerance.

[0090] The phase elastic scaling unit is used to perform elastic scaling processing on the preset initial green light phase according to the phase elastic scaling tolerance in order to generate green wave dynamic control commands. The elastic scaling processing includes green light early start operation and green light delayed end operation.

[0091] This embodiment provides a calculation mechanism for mapping convoy status quantities to phase-adjustable time quantities. Specifically, in the aforementioned scenario, knowing only that the convoy is scattered is not enough to directly control the signal; it is also necessary to answer how many seconds earlier or how many seconds later the signal can be extended. If this step is missing, the control action will either be too conservative and unable to capture the broken convoy, or it will be overstretched and affect the release of side roads. Therefore, this embodiment sets up a tolerance calculation unit and a phase elastic stretching unit.

[0092] In one specific implementation, the tolerance calculation unit first calculates a baseline adjustable range under the premise that the global signal period remains unchanged; it can be derived using microscopic numerical analysis: assuming that the common period of the third intersection is 120s, the minimum green light time of the main road is 28s, and the clearing time of yellow light and all red light is 6s in total, then the remaining adjustable time base amount is 86s.

[0093] This baseline quantity does not mean that all of it can be allocated to the main road; rather, it must be combined with the spatial distribution density. If the density coefficient of the current vehicle fleet to be captured within the effective control range is 0.20, then the baseline value is 17.2s. This is then multiplied by the reciprocal of the real-time dispersion. If the dispersion is 2.0, then the phase elastic scaling tolerance calculated by the formula is 8.6s. If the dispersion increases to 4.0, it indicates that the vehicle fleet has become more fragmented. Although there is a need for capture, the effectiveness of overall stretching adjustment has decreased. The tolerance calculated by the formula is reduced to 4.3s.

[0094] Thus, the system provides greater adjustment space for situations with heavy traffic and a certain degree of clustering, while avoiding excessive phase adjustments for convoys with dispersion greater than the preset extreme value; the phase elastic expansion unit performs early start or late stop processing on the preset initial green light phase according to the obtained tolerance; the early start operation of the green light means that after ensuring the safe completion of the previous phase, the green light on the main road starts earlier than originally planned.

[0095] The green light delay operation refers to extending the green light on the main road for a short period after the original plan has ended. For example, if the original green light window for the main road at the third intersection was from the 30th to the 58th second, and the system has a tolerance of 5 seconds and determines that the following traffic will arrive between the 59th and 62nd second, then a dynamic control command to delay by 4 seconds is generated. If it is determined that the preceding traffic will arrive at the stop line earlier than expected and the traffic flow on the side road is low, then a dynamic control command to start early by 3 seconds is generated. Both operations are flexible adjustments to the original plan, rather than overturning the entire cycle structure.

[0096] In terms of fault tolerance protection, if the discreteness calculation result is abnormally close to 0, its reciprocal will be too large, and the system will cut off the tolerance value with a preset upper limit; if the spatial distribution density is too low, causing the calculated tolerance to be close to 0, then elastic scaling will not be performed, and the initial green light scheme will be maintained; if there are simultaneous early start and late stop requirements, the action that improves the passage efficiency of the convoy that is about to arrive will be given priority, or it will be split into two consecutive cycles to be executed separately to avoid excessive deformation in a single cycle.

[0097] At the third intersection of the central axis avenue, the system detected that a convoy of vehicles upstream of the second intersection was about to arrive in a pattern of converging at the front and rear and sparser in the middle. The calculation showed that the tolerance for this cycle was 6 seconds. Considering that the vehicles in the rear section were only 90 meters away from the stop line, while the vehicles in the front section had already entered the green light state, the system chose to delay the green light of the original main road by 5 seconds instead of starting it early. After the execution, the three vehicles in the rear section that would have been stuck at the red light were able to pass through one after another, and the convoy regrouped at the third intersection.

[0098] It should be noted that, in order to ensure that the final calculation result of the phase elastic scaling tolerance has a pure time dimension in seconds, the spatial distribution density and the real-time dispersion of the vehicle fleet will be dimensionless based on a reference benchmark before participating in the above multiplication operation.

[0099] Specifically, the system divides the calculated absolute spatial distribution density by the preset maximum saturation density of the road segment to obtain the dimensionless relative density coefficient, which is 0.20 in the aforementioned numerical deduction; similarly, the calculated real-time dispersion of the fleet is divided by the preset steady-state dispersion benchmark, which is consistent with the original physical dimension of the dispersion, thereby converting it into a dimensionless multiple value that characterizes the severity of dispersion, which is 2.0 in the aforementioned deduction.

[0100] This explicit structured processing not only strictly follows the logical mapping of using the reciprocal of density and dispersion for tolerance calculation, but also completely resolves the contradiction of inconsistent dimensions in cross-physical quantity calculation, ensuring that the final output tolerance value accurately corresponds to the executable time scale.

[0101] The purpose of this mechanism is to convert abstract traffic disturbance characteristics into directly executable time control quantities, thereby achieving dynamic time matching of the green light window for discrete convoys.

[0102] Furthermore, the phase elastic scaling unit is also used to trigger an early green light start operation and generate a pre-extension time of the preset initial green light phase when the real-time dispersion of the fleet is greater than a preset dispersion threshold; and to trigger a delayed green light stop operation and generate a post-extension time of the preset initial green light phase when the real-time dispersion of the fleet is less than or equal to the preset dispersion threshold.

[0103] This embodiment provides a discrimination mechanism for selecting different extension directions for different discrete morphologies. Specifically, simply having the ability to start early or stop late is not enough to guarantee the control effect because the structure of discrete convoys is not the same. Some convoys are characterized by the front section arriving early while the rear section is still far away, in which case continuing to stop late is not very meaningful. Other convoys are characterized by the overall lag being slightly behind but still relatively concentrated, in which case starting too early would waste the green light. Therefore, this embodiment introduces a discreteness threshold to distinguish whether to prioritize starting early or stopping late.

[0104] In one specific implementation, the system presets a dispersion threshold, such as 1.5. When the real-time dispersion is greater than this threshold, it means that the speed differentiation within the convoy is obvious and the formation is stretched. Usually, it is necessary to start the green light early to capture the vehicles that have dispersed earlier, so as to prevent them from forming a new cut-off before the stop line.

[0105] Taking a specific operational process as an example: the first two vehicles in a convoy have reached the stop line at the third intersection, while the last four vehicles are still outside the preset distance threshold range of 150 meters upstream, with a dispersion of 2.3. If they still wait for the main road green light to start normally as originally planned, the first two vehicles may be stopped first and then become separated from the last four vehicles. Therefore, the system triggers an early start, generating an advance extension time, such as starting the main road green light 3 seconds in advance. Conversely, when the real-time dispersion is less than or equal to this threshold, it indicates that although the convoy is loose, it still has overall maintainability, and it is more suitable to cover the tail traffic flow by delaying the green light.

[0106] For example, if the spacing between 6 vehicles in the same convoy is slightly increased, but they are still within a continuous 200-meter interval, the dispersion is only 1.1. If an early start operation is performed at this time, it may only allow the first part to pass earlier, while the last part will still arrive after the original end time. Using a delayed interruption will allow the last part to fill in the same release window. Therefore, the system generates a post-extension time, such as extending the green light of the main road by 4 seconds.

[0107] To further illustrate, we can assume that the original green light interval at the fourth intersection was 40s to 68s. In the first scenario, the dispersion is 2.0, and the system output is extended forward by 2s, adjusting the green light to 38s to 68s. In the second scenario, the dispersion is 1.2, and the system output is extended backward by 5s, adjusting the green light to 40s to 73s. Although both scenarios involve phase elastic scaling, their directions of action and applicable objects are significantly different.

[0108] In abnormal conditions, if the dispersion is exactly equal to the threshold, the system will execute the delayed interruption branch to reduce the intrusion on the preceding phase; if the benefits of early start and delayed interruption are both low, such as predicting less than 1 new vehicle to pass, no adjustment will be made in this cycle; if the advance extension time will conflict with the pedestrian crossing protection period, the early start amount will be automatically compressed or the delayed interruption strategy will be selected.

[0109] At the fourth intersection of Zhongzhou Avenue, at 7:52 AM, a convoy affected by upstream lane merging was detected. The first two vehicles were rapidly approaching the stop line, while the latter were stuck in the middle of the road. The system calculated a dispersion of 1.9, which was higher than the threshold of 1.5, so it executed an early start of 2 seconds. Then at 7:55 AM, another relatively complete convoy released from the third intersection arrived at the fourth intersection, with a dispersion of only 1.0. The system then executed a delayed stop of 4 seconds. Through this continuous processing, the fourth intersection adopted different extension and retraction directions to control the convoys of different shapes.

[0110] The purpose of this step is to match the phase stretching direction with the discrete shape of the vehicle fleet, thereby enabling more targeted breakpoint repair.

[0111] Furthermore, the adaptive control module also includes:

[0112] The boundary review unit is used to extract the queue length of intersecting roads from the target fused traffic flow data and compare the queue length of intersecting roads with a preset queue length threshold.

[0113] If the queue length of intersecting roads exceeds a preset queue length threshold, a suppression command is generated, and the elastic scaling process is canceled based on the suppression command.

[0114] When the queue length of intersecting roads is less than or equal to a preset queue length threshold, a release instruction is generated, and elastic scaling is performed based on the release instruction.

[0115] This embodiment provides a boundary review mechanism to prevent the optimization of main roads from squeezing the traffic space of side roads. Specifically, extending or advancing the green light based solely on the discrete state of the main road traffic queues can easily cause queue overflow on intersecting roads in extreme cases, affecting the stability of the entire corridor. Therefore, before performing phase elastic scaling, the boundary review unit first performs constraint review on the queuing of side roads.

[0116] In one specific implementation, the boundary review unit extracts the queue length of intersecting roads from the fused data and compares it with a preset threshold. A simple example can be used: the available parking length of the east-west branch road at the third intersection is 180 meters, and the system sets the queue length threshold to 80% of that, i.e., 144 meters. If the queue at the east entrance reaches 152 meters in a certain period, it indicates that continuing to compress the green light in this direction may cause the queue to extend to the upstream community entrance. At this time, even if the main road has a strong demand for delayed interruption, a suppression instruction is generated to cancel this elastic expansion. If the queue at the east entrance is only 96 meters, a release instruction is generated, allowing the main road to execute the planned green light delay, such as 3 seconds.

[0117] This mechanism does not only consider a single instantaneous value of a branch, but also examines the growth trend. For example, if the current queue length is 138 meters, although it has not exceeded 144 meters, but shows a rapid upward trend for three consecutive sampling frames, the system can regard it as a critical state and reduce the scalable time instead of fully opening it. Conversely, if the branch queue once reached near the threshold, but has quickly fallen back with the release of the previous cycle, the main road can be allowed to perform a small-scale expansion and contraction again. To illustrate the control logic, two sets of comparisons can be constructed.

[0118] In Scenario 1, the convoy at the rear of the main road at the second intersection is expected to be able to release 5 more vehicles by adding a 4-second delay. However, the queue on the side road at the west entrance of the second intersection has reached 160 meters. The system issues a suppression command and cancels the delay. In Scenario 2, the main road at the fourth intersection needs to be opened 2 seconds earlier to accommodate the convoy that arrives earlier. However, the queues on its north-south side roads are all less than 60% of the threshold. The system issues a release command and allows the early opening. In this way, the main road adjustment is no longer a single-objective behavior, but a constrained optimization with rigid boundaries.

[0119] In abnormal situations, if branch queue length data is missing, the system prioritizes safety and limits the expansion range, for example, only allowing fine adjustments of no more than 1 second; if only one direction among multiple branches exceeds the limit, a local constraint method can be used to prohibit expansion that would intrude on the release time in that direction, while allowing fine adjustments that do not affect that direction; if a branch detector malfunction causes the queue length to be constantly 0, the system cross-checks video occupancy rate or historical trends to avoid misjudging that unlimited release is possible.

[0120] At the third intersection of the central axis avenue, the system originally planned to implement a 5-second delay to capture the discrete rear traffic flow from the second intersection. However, it simultaneously detected that the queue at the eastbound entrance was approaching the entrance / exit of the school on the side road, reaching 149 meters, exceeding the threshold of 144 meters. Therefore, the system immediately generated a suppression command, canceled the delay, and restored the current cycle to the initial plan. After two cycles, the queue on the side road dropped to 110 meters, and the main road was allowed to perform a 2-second fine-tuning again.

[0121] The purpose of this mechanism is to establish clear engineering boundaries for the green wave adaptive control of the main road, thereby achieving a balance between improving the efficiency of the main line and the stable operation of the branch roads.

[0122] Furthermore, the system also includes a macro-assessment module, which includes:

[0123] The reconstruction rate calculation unit is used to determine the discrete reconstruction rate of the fleet based on the actual number of vehicles passing through the green wave and the preset expected number of vehicles passing through.

[0124] The coherence index calculation unit is used to perform a weighted summation of the vehicle discrete reconstruction rate and the reciprocal of the travel time variance of adjacent intersections in the multi-intersection traffic data of the target fused traffic flow data, to obtain the spatiotemporal coherence index of the green wave band.

[0125] This embodiment provides a mechanism for corridor-level feedback evaluation of control results. Specifically, the aforementioned modules are already able to make real-time adjustments, but if we only look at whether more vehicles have passed through a single intersection, it is still impossible to determine whether the entire green wave has truly returned to continuity. Therefore, this embodiment introduces a macro-evaluation module to evaluate the control results across intersections.

[0126] In one specific implementation, the reconstruction rate calculation unit first calculates the discrete reconstruction rate of the convoy. For example, before a certain control action, the system predicts that 10 vehicles in a discrete convoy have the potential to be reintegrated into the green wave. After executing an early green light start or delayed green light stop, if 8 vehicles actually pass through the third intersection continuously and remain uninterrupted at the fourth intersection, then the reconstruction rate for this action can be recorded as 0.8. If only 3 vehicles actually enter the effective green wave window, the reconstruction rate is 0.3, indicating that the repair effect of this local control is weak.

[0127] The preset expectation here can be obtained by predicting the number of vehicles based on the trajectory before control, rather than by retrospective calculation; the coherence index calculation unit forms a spatiotemporal coherence index based on the reconstruction rate and the stability of the travel time between adjacent intersections.

[0128] To facilitate understanding, a simplified deduction can be used: If the travel times for 5 vehicles between the second and third intersections are 24s, 25s, 24s, 26s, and 25s respectively, the variance is small and its reciprocal is large, indicating a stable traffic rhythm; if the travel times become 18s, 31s, 27s, 35s, and 20s, the variance increases and the reciprocal decreases, indicating that although some vehicles have passed, the overall traffic rhythm is disordered; the system sums the reconstruction rate and this reciprocal according to preset weights to obtain the spatiotemporal coherence index of the green wave band; in this way, the evaluation no longer only looks at how many vehicles have passed, but also whether the traffic flow has passed through multiple intersections at a stable rhythm;

[0129] To illustrate with a more concrete numerical example: if the reconstruction rate after a certain control is 0.75, and the inverse of the travel time variance of adjacent intersections is converted to 0.60, with weights of 0.6 and 0.4 respectively, then the coherence index is 0.75 × 0.6 plus 0.60 × 0.4, resulting in 0.69. If, in another round of control, the reconstruction rate reaches 0.85, but the inverse of the travel time variance is only 0.20, indicating unstable results, then the index is only 0.59, reflecting that although the strategy increases the traffic volume at local intersections, it fails to create a sustainable green wave.

[0130] Regarding the fault tolerance mechanism, if the number of sample vehicles is insufficient in a certain evaluation period, such as less than 5 vehicles, the travel time variance is not representative. The system can extend the evaluation window or use the results of the previous effective period as a smooth substitute. If the expected number of vehicles passing is 0, the reconstruction rate calculation will lose its meaning. In this case, the period is directly marked as having no reconstruction task and is not included in the index update. If individual vehicles have abnormally long travel times, such as outliers caused by temporary stops, the system first performs outlier removal and then calculates the variance.

[0131] In the three consecutive intersections from the second to the fourth intersection of the central axis avenue, the system evaluated the two-round control actions between 7:45 and 7:50. In the first round, after a 4-second delay at the third intersection, 7 out of the originally predicted 9 discrete vehicles continued to pass through continuously, with a reconstruction rate of 0.78. Moreover, the travel time fluctuation from the second to the fourth intersection was small, resulting in a high continuity index.

[0132] In the second round, although starting the fourth intersection 3 seconds earlier allowed more vehicles to rush through the current intersection, they subsequently dispersed again before J5, resulting in significant fluctuations in travel time and a decrease in the continuity index; based on this, the system identified that the second round strategy was not suitable for long-term maintenance.

[0133] It is worth noting that since the reconstruction rate is a dimensionless ratio between 0 and 1, while the reciprocal of the travel time variance has a specific physical dimension and fluctuates greatly, direct weighted summation can easily lead to excessively large differences in the range of values ​​of the coherence index and loss of evaluation benchmark. Therefore, before including the reciprocal of the travel time variance in the weighted summation, the system has a pre-set extreme value normalization mapping mechanism to convert it into a standard interval value.

[0134] Specifically, to clarify the calculation process of the normalization mapping, the system presets a baseline value for the upper limit of the variance corresponding to the ideal stationary traffic state. And a baseline value for the lower bound of the inverse variance corresponding to a severely disordered state. Let the travel time at the adjacent intersection be Its variance is Its reciprocal is:

[0135]

[0136] The system uses an extremum normalization formula to calculate the mapped standard reciprocal metric:

[0137]

[0138] This allows for a structured projection onto a standard dimensionless scale of 0 to 1, which is the process of converting to 0.60 and 0.20 in the aforementioned numerical deduction, based on a preset reconstruction rate weight. With coherence weight Using the formula:

[0139]

[0140] in, To determine the discrete reconstruction rate of the vehicle fleet, the spatiotemporal coherence index of the green wave band is accurately calculated. Through this explicit dimensional alignment action, the system completely eliminates the hidden danger of overwhelming the evaluation results of multiple objectives due to the excessive size of a single feature, and ensures that the weighted coherence index has real and effective spatiotemporal evaluation significance.

[0141] The purpose of this mechanism is to establish a quantifiable closed-loop result for real-time control, thereby enabling a joint evaluation of two objectives: optimizing the capacity of local intersections and ensuring the overall connectivity of the corridor.

[0142] Furthermore, the macro-assessment module also includes:

[0143] The parameter correction unit is used to compare the spatiotemporal coherence index of the green band with a preset coherence benchmark value.

[0144] When the spatiotemporal coherence index of the green wave band is lower than the preset coherence benchmark value, increase the preset control refractory period parameter in the preset topology adjustment model;

[0145] When the spatiotemporal coherence index of the green wave band is higher than or equal to the preset coherence benchmark value, the preset control refractory period parameter in the preset topology adjustment model remains unchanged.

[0146] This embodiment provides a mechanism for reverse correction of control rhythm using macroscopic evaluation results; specifically, without parameter correction, the system can only know whether it is good or bad, but cannot change the stability of subsequent control.

[0147] In continuous multi-cycle operation, if the system makes too frequent fine-tuning of the phases at each intersection, although it may seem to respond positively in the short term, it may cause the control rhythm of the entire corridor to become fragmented. Therefore, this embodiment uses the spatiotemporal coherence index to correct the control refractory period parameter.

[0148] In one specific implementation, the parameter correction unit compares the currently evaluated coherence index with a preset coherence benchmark value; the benchmark value can be set to 0.65; if the index is only 0.52 in a certain evaluation window, it indicates that although the previous rounds of dynamic control made adjustments locally, the corridor as a whole is still not coherent, and there may even be problems such as excessively dense adjustments between intersections and mutual interference; at this time, the system increases the control refractory period parameter, for example, from 2 cycles to 3 cycles, indicating that after a node completes a large adjustment, it should wait at least longer before allowing another strong intervention to be triggered; by lengthening the interval, the system suppresses the tendency to continuously stretch the phase in a short period of time;

[0149] Conversely, if the coherence index is higher than or equal to the benchmark value, for example, reaching 0.71, it indicates that the current control rhythm is well matched with the road conditions, and there is no need to further increase the suppression intensity. At this time, the control refractory period parameter is kept unchanged, and the system continues to operate with the current sensitivity. The one-way correction method of increasing when it is lower and keeping it when it is higher or equal is to make the control system more stable when the coherence is insufficient, rather than immediately returning to the oversensitive state when the coherence is good.

[0150] In one specific embodiment: Suppose that the third intersection node is adjusted every cycle between 08:00 and 08:05. Although this increases the throughput per cycle, the travel time fluctuation between the second and fourth intersections continues to increase, causing the consistency index to drop to 0.58. The system then increases the control refractory period of the third intersection from 1 cycle corresponding to 120s to 2 cycles corresponding to 240s. The third intersection no longer immediately scales up or down for every slight discrete event, but only responds to significant discrete events, thereby reducing the coupling disturbance with the second and fourth intersections.

[0151] In abnormal conditions, if the consistency index is extremely low for several consecutive windows, such as below 0.4, the system can trigger an alarm in addition to increasing the control refractory period, indicating abnormal construction, equipment failure, or signal scheme mismatch. If the index drops intermittently but recovers quickly in the next window, a smoothing rule can be adopted to correct the index only after it has fallen below the benchmark value twice in a row to avoid parameter jitter. If the parameter has reached the preset upper limit, it will no longer be increased, and the system will switch to manual review or high-level strategy switching.

[0152] During the morning rush hour on the central avenue, the second, third, and fourth intersections responded to multiple discrete traffic flows in succession. Although the local traffic volume increased, the overall traffic flow rhythm began to become unstable. The system calculated the coherence index to be 0.61, which was lower than the benchmark value of 0.65. Based on this, the parameter correction unit raised the control refractory period by one level. Within 20 minutes, the nodes only performed adjustments on high-priority discrete events, and frequent fine-tuning was significantly suppressed, and the corridor rhythm gradually recovered.

[0153] The purpose of this mechanism is to adjust the control rhythm through result feedback, thereby achieving a balance between dynamic response capability and long-term operational stability.

[0154] Furthermore, the preset topology adjustment model includes:

[0155] The node status evaluation unit is used to take the product of the difference between the preset maximum dispersion and the real-time dispersion of the fleet and the preset control refractory period parameter as the node status activity.

[0156] The weight dynamic adjustment unit is used to determine the weight adjustment coefficient based on the ratio of node status activity to preset base activity, and to use the weight adjustment coefficient to perform a multiplication operation on the preset initial value of command issuance weight, so as to realize the dynamic increase or decrease of command issuance weight.

[0157] This embodiment provides a topology adjustment mechanism for allocating control priorities among multiple intersections. Specifically, in a continuous corridor, different intersections may simultaneously request adjustment. If priorities are not distinguished, multiple nodes may compete for the same elastic time resource, leading to control conflicts. Therefore, this embodiment uses dynamic adjustment of node activity and weight to determine which intersection's instruction is more worthy of priority execution.

[0158] In one specific implementation, the node status assessment unit calculates the node status activity based on the preset maximum dispersion, real-time dispersion, and control refractory period parameter. A simplified numerical example illustrates this: assuming the system's preset maximum dispersion is 5.0, the current real-time dispersion of the third intersection is 2.0, and the control refractory period parameter is 2, then the node status activity of the third intersection can be obtained by multiplying the difference between the two by the refractory period parameter, which is 6.0. If the real-time dispersion of the fourth intersection is 4.2, and the refractory period parameter is also 2, then its activity is only 1.6.

[0159] This can be understood as follows: when the difference between the real-time dispersion and the preset maximum dispersion is greater than the first threshold, and the current control refractory period parameter meets the preset intervention conditions, it is determined that the node has priority to execute the reconfiguration control task; this definition helps to prioritize the selection of nodes that still have adjustment needs and are not in the frequent intervention period.

[0160] The dynamic weight adjustment unit then compares the node status activity with the preset base activity to obtain the weight adjustment coefficient. For example, if the base activity is set to 3.0 and the activity of the third intersection is 6.0, the coefficient is 2.0; if the activity of the fourth intersection is 1.6, the coefficient is approximately 0.53. If the initial weight values ​​of the original instructions issued by both nodes are 0.5, after multiplication, the updated weight of the third intersection is 1.0, and the updated weight of the fourth intersection is approximately 0.265. The system can normalize the weights, prioritize issuing instructions to the third intersection, and reduce the intervention intensity of the fourth intersection in this cycle. In this way, control resources will be concentrated on nodes with higher adjustment necessity and better intervention conditions.

[0161] In continuous corridors, this topology adjustment plays a crucial role. For example, the second intersection detects that the preceding convoy has arrived slightly ahead of schedule, the third intersection detects a significant break in the middle convoy, and the fourth intersection only observes a slight fluctuation at the tail end. If each of these three intersections were to adjust forcibly, the entire green wave channel could be segmented and cut off. Through the activity and weighting mechanism, the system prioritizes the reconstruction task for the third intersection, makes only minor adjustments to the second intersection, and maintains the original plan for the fourth intersection, thus forming a coordinated control with a primary and secondary focus.

[0162] In the boundary protection mechanism, if the real-time dispersion exceeds the preset maximum dispersion and the difference between the two is negative, it indicates that the node is close to the out-of-control boundary. At this time, the activity level is no longer used directly as a negative value, but is truncated to 0, indicating that it is not advisable to continue to apply strong intervention in this cycle. If the basic activity level is set too low, resulting in abnormal amplification of the coefficient, the system adopts the maximum weight upper limit protection. If the weights of multiple nodes are close to the same after calculation, the node located in the center of the vehicle breakdown or closest to the downstream critical bottleneck is selected for execution first.

[0163] At 7:48 AM on the central avenue, adjustment suggestions were simultaneously reported at the second, third, and fourth intersections. Calculations showed that the third intersection had the highest activity level, the second intersection was in the middle, and the fourth intersection had the lowest. The system then issued the delayed interruption instruction for the third intersection as the main instruction, compressed the early start suggestion for the second intersection to a 1-second fine adjustment, and the fourth intersection remained inactive for the current cycle. As a result, the traffic vehicles that were originally scattered between the second and third intersections were prioritized to complete their reconstruction at the third intersection, avoiding timing interference caused by multiple nodes working simultaneously.

[0164] Furthermore, to ensure that the increased preset control refractory period parameter can produce a substantial control and suppression effect and eliminate the potential logical ambiguity that the increase in parameter leads to a reverse increase in weight, the system will assign a dynamic state polarity to the refractory period parameter before performing the multiplication operation.

[0165] Specifically, to eliminate the ambiguity in the control and suppression option logic and to clarify the trigger boundary for polarity switching, the system introduces a dynamic polarity sign function. ,in, Let the input be the control refractory period parameter variable; let the base value of the control refractory period parameter be... The current actual refractory period parameter is When the node has not entered the inhibition period and normal regulation is allowed, i.e. At that time, the system settings At this point, the preset control refractory period parameter is used as a normal positive gain, namely the positive number 2 in the previous deduction, and participates in the product calculation;

[0166] However, once the node triggers the constraint conditions as described in the previous embodiment, causing the parameter to be forcibly increased, that is... At that time, the system settings This converts the parameter into a penalty factor with inherent negative polarity during multiplication; calculated using the formula:

[0167]

[0168] in, To calculate the node state activity by setting the maximum dispersion. Here, This refers to the real-time dispersion of the fleet calculated in the aforementioned embodiments; and the preset maximum dispersion involved in the calculation. Real-time dispersion of the fleet Maintaining dimensional consistency; due to the intervention of the penalty factor, node state activity... It will plummet to a deep negative minimum;

[0169] At this point, the system then corrects the operator. Node state activity Bottom truncation is performed, thereby forcibly clearing or minimizing the obtained weight adjustment coefficient and the final instruction issuance weight; based on this polarity conversion solution rule, the system successfully constructs a self-consistent closed loop in the underlying operation mechanism in which the forced increase of parameters is equivalent to strong suppression intervention, thus completely preventing frequent ineffective intervention of nodes in the refractory period.

[0170] The purpose of this mechanism is to establish a priority allocation logic for a continuous group of intersections that is oriented towards the entire corridor, thereby achieving orderly coordination of control actions.

[0171] Furthermore, the multi-source data perception module includes a data compression unit, which is used to compress the target fused traffic flow data based on a preset encoding algorithm, and transmit the compressed target fused traffic flow data to the fleet feature calculation module based on a block transmission strategy.

[0172] This embodiment provides a compression and segmentation mechanism suitable for high-frequency traffic sensing data transmission. Specifically, as multiple intersections on the aforementioned central axis avenue simultaneously upload trajectory, speed, queuing, and travel time data, directly transmitting the full fusion result could easily cause link congestion, leading to delays in feature calculation and control decisions. Especially in the green wave state reconstruction scenario, the algorithm's value depends on second-level response, so a data compression unit and segmentation transmission strategy need to be set between the sensing layer and the calculation layer.

[0173] In one specific implementation, the data compression unit performs structured encoding on the target fused traffic flow data; for ease of explanation, the fusion result within a control cycle can be viewed as three blocks: the first block stores the basic vehicle status, including vehicle identification, location, and speed; the second block stores lane and intersection association information; and the third block stores branch road queuing and adjacent intersection travel time statistics.

[0174] For data that does not change or changes very little in consecutive sampling frames, the system does not repeatedly transmit the complete content, but uses incremental encoding; for example, if the position of vehicle A in the previous frame was -120 meters and in this frame it is -113 meters, then only 7 meters forward is recorded; if the queue at the east entrance of the third intersection in the previous frame was 96 meters and in this frame it is 98 meters, then only 2 meters is recorded; thus, the complete field that originally needed to be transmitted is compressed into a small amount of change.

[0175] The segmented transmission strategy further ensures that key data arrives at the fleet feature calculation module first. Taking S1, S2, and S3 as examples, S1, which is directly related to the discreteness calculation, has the highest priority. S3 is required for branch boundary review, and S2 is the extended information used for longer-term evaluation. When the network bandwidth is sufficient, the three segments are sent completely.

[0176] When bandwidth decreases, the system sends S1 first, then S3, and delays S2 if necessary. A micro-deduction can be made: if the total amount of data after compression in a certain period is 90 units, and the link currently only allows the transmission of 60 units, then 30 units of S1 and 20 units of S3 are sent first, and the remaining 10 units are sent to make up the key digest of S2, instead of waiting for all the data to be collected and sent together.

[0177] This strategy can also be used in conjunction with the solution module for streaming processing; that is, the solution module can start calculating the speed variance and preliminary dispersion as soon as it receives S1; it can supplement the branch queuing constraints after receiving S3; if S2 arrives later, it can update a more complete traffic association relationship; thus forming a working mode of calculating while transmitting, rather than calculating after all of them arrive; this is especially important for time-sensitive scenarios such as traffic signal control.

[0178] In abnormal conditions, if the compressed data packet fails to be verified, the system will prioritize retransmitting the key segments rather than retransmitting the entire packet; if a segment does not arrive for a long time, the calculation module will use the corresponding digest value of the most recent valid period to temporarily fill the gap and add a conservative coefficient to the control result; if the network interruption exceeds the preset time, the system will stop the fine adjustment based on the high-frequency trajectory and switch to local intersection independent control or pre-stored coordination scheme.

[0179] During the morning rush hour on the central avenue, the fused vehicle trajectories were uploaded simultaneously from the second to the fourth intersection, causing congestion on the link between the control center and the edge calculation node. The system compressed the incremental information of vehicle position and speed and sent it first, so that the calculation module could still calculate the dispersion of the upstream platoon at the third intersection within 1 second. The branch road queue summary arrived and the boundary review was completed simultaneously. The lower priority extended traffic statistics were resent in the next time slot, without affecting the generation of control instructions for the current cycle.

[0180] The purpose of this mechanism is to reduce the transmission burden without sacrificing the timeliness of key control information, thereby enabling the real-time implementation of green wave adaptive control under big data conditions.

[0181] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A dynamic adaptive control system for green wave belts on urban arterial roads based on big data, characterized in that: The system includes: The multi-source data sensing module is used to receive multi-source heterogeneous traffic flow data through a preset data interface. The multi-source heterogeneous traffic flow data includes main road traffic flow data, intersecting road traffic flow data, and multi-intersection traffic data. The multi-source heterogeneous traffic flow data is spatiotemporally aligned based on a preset road network coordinate system and a unified time axis to obtain target fused traffic flow data. The fleet feature calculation module is connected to the multi-source data perception module and is used to receive the target fused traffic flow data and calculate the real-time dispersion and spatial distribution density of the fleet based on the target fused traffic flow data. An adaptive control module is used to determine a phase elastic scaling tolerance based on the real-time dispersion of the vehicle fleet and the spatial distribution density, and to use the phase elastic scaling tolerance as input to a preset topology adjustment model. The preset topology adjustment model includes a preset control refractory period parameter to characterize the control intervention time interval of continuous intersections. It is used to calculate the node state activity based on the real-time dispersion of the vehicle fleet, and to determine the command issuance weight based on the node state activity. Combined with the phase elastic scaling tolerance, it outputs a green wave dynamic control command containing the command issuance weight. The instruction issuance and execution module is used to send the green wave dynamic control instruction to the traffic signal control equipment at the corresponding intersection, so as to drive the traffic signal control equipment to perform green light phase adjustment based on the instruction issuance weight.

2. The dynamic adaptive control system for green wave strips on urban main roads based on big data as described in claim 1, characterized in that, The multi-source data sensing module includes: An asynchronous elimination unit is used to perform timestamp synchronization processing on the multi-source heterogeneous traffic flow data to convert the multi-source heterogeneous traffic flow data from an asynchronous state to a synchronous state. The spatial mapping unit is used to perform spatial mapping processing on the multi-source heterogeneous traffic flow data in the synchronous state based on a preset road network coordinate system to generate the target fused traffic flow data.

3. The dynamic adaptive control system for green wave strips on urban arterial roads based on big data as described in claim 1, characterized in that, The fleet feature calculation module includes: The shock wave feature extraction unit is used to extract speed change waves and deceleration transmission behavior of vehicles in the car-following state as features of the incoming shock wave based on the target fused traffic flow data. The discreteness calculation unit is used to determine the real-time discreteness of the convoy based on the velocity variance and average velocity corresponding to the characteristics of the incoming shock wave. The density calculation unit is used to calculate the spatial distribution density based on the vehicle coordinate information in the target fused traffic flow data.

4. The dynamic adaptive control system for green wave strips on urban main roads based on big data as described in claim 1, characterized in that, The adaptive control module includes: The tolerance calculation unit is used to, under the constraint of a preset global signal period, subtract the sum of a preset minimum green light time and a preset clearing time from the preset global signal period, perform dimensionless processing on the spatial distribution density and the real-time dispersion of the fleet based on a reference benchmark, and use the product of the difference and the spatial distribution density as the benchmark value of the phase elastic scaling tolerance, and then use the product of the benchmark value and the reciprocal of the real-time dispersion of the fleet as the phase elastic scaling tolerance. The phase elastic scaling unit is used to perform elastic scaling processing on the preset initial green light phase according to the phase elastic scaling tolerance, so as to generate the green wave dynamic control command, wherein the elastic scaling processing includes green light early start operation and green light delayed end operation.

5. The dynamic adaptive control system for green wave strips on urban main roads based on big data as described in claim 4, characterized in that, The phase elastic stretching unit is also used for, If the real-time dispersion of the fleet exceeds a preset dispersion threshold, the early green light start operation is triggered, and the pre-extension time of the preset initial green light phase is generated. If the real-time dispersion of the fleet is less than or equal to the preset dispersion threshold, the green light delay operation is triggered, and the post-extension time of the preset initial green light phase is generated.

6. The dynamic adaptive control system for green wave strips on urban main roads based on big data as described in claim 4, characterized in that, The adaptive control module further includes: A boundary review unit is used to extract the queue length of intersecting roads from the target fused traffic flow data and compare the queue length of intersecting roads with a preset queue length threshold. If the queue length of the intersecting roads is greater than the preset queue length threshold, a suppression command is generated, and the elastic scaling process is canceled based on the suppression command. If the queue length of the intersecting roads is less than or equal to the preset queue length threshold, a release instruction is generated, and the elastic scaling process is executed based on the release instruction.

7. The dynamic adaptive control system for green wave strips on urban arterial roads based on big data as described in claim 1, characterized in that, The system also includes a macro-assessment module, which includes: The reconstruction rate calculation unit is used to determine the discrete reconstruction rate of the fleet based on the actual number of vehicles passing through the green wave and the preset expected number of vehicles passing through. The coherence index calculation unit is used to perform a weighted summation of the vehicle discrete reconstruction rate and the reciprocal of the travel time variance of adjacent intersections in the multi-intersection traffic data of the target fused traffic flow data to obtain the spatiotemporal coherence index of the green wave band.

8. The dynamic adaptive control system for green wave strips on urban main roads based on big data as described in claim 7, characterized in that, The macro-assessment module also includes: The parameter correction unit is used to compare the green band spatiotemporal coherence index with a preset coherence benchmark value. If the spatiotemporal coherence index of the green wave band is lower than the preset coherence benchmark value, the preset control refractory period parameter in the preset topology adjustment model is increased. When the spatiotemporal coherence index of the green wave band is higher than or equal to the preset coherence benchmark value, the preset control refractory period parameter in the preset topology adjustment model remains unchanged.

9. The dynamic adaptive control system for green wave strips on urban main roads based on big data as described in claim 1, characterized in that, The preset topology adjustment model includes: The node status evaluation unit is used to take the product of the difference between the preset maximum dispersion and the real-time dispersion of the fleet and the preset control refractory period parameter as the node status activity. The weight dynamic adjustment unit is used to determine the weight adjustment coefficient based on the ratio of the node's state activity to a preset base activity, and to use the weight adjustment coefficient to perform a multiplication operation on a preset initial value of the instruction issuance weight, so as to realize the dynamic increase or decrease of the instruction issuance weight.

10. The dynamic adaptive control system for green wave strips on urban arterial roads based on big data as described in claim 1, characterized in that, The multi-source data sensing module includes: The data compression unit is used to compress the target fused traffic flow data based on a preset encoding algorithm, and to transmit the compressed target fused traffic flow data to the fleet feature calculation module based on a block transmission strategy.

Images