An urban traffic signal intelligent regulation system based on big data and AI algorithm

Abstract

The application provides a kind of city traffic signal intelligent regulation and control system based on big data and AI algorithm, belongs to traffic signal regulation and control technical field, the system includes: through the vehicle passing details of each green light period in each signal cycle of target intersection, identify the saturated traffic segment of each green light period, and calculate the saturated headway of green light period in current signal cycle;According to the saturated headway, the inefficient traffic segment in the green light period is traversed, and the invalid green light duration of the green light period in the current signal cycle is determined;According to the invalid green light duration of each green light period in the current signal cycle, the distribution difference between each green light period is combined to identify the traffic dispersion type of the target intersection, when the traffic dispersion type is demand structure mismatch, the green light period with the highest invalid green light duration ratio is adaptively adjusted.The technical scheme provided by the application can reduce the hidden green loss caused by the flow dissipation and flow interruption in the release stage in traffic signal control.

Description

Technical Field

[0001] This application relates to the technical field of traffic signal control, and more specifically, to an intelligent urban traffic signal control system based on big data and AI algorithms. Background Technology

[0002] With the acceleration of urbanization and the continuous growth of motor vehicle ownership, traffic congestion and declining traffic efficiency have become increasingly prominent problems. Traditional fixed-time signal control is difficult to adapt to real-time changes in traffic flow, which can easily lead to idle green lights or queue overflow. To address this, modern traffic signal control has gradually introduced traffic flow detection, short-term flow prediction and dynamic timing optimization technologies to dynamically adjust the signal cycle, phase difference and green light ratio, which can significantly improve the traffic capacity of the road network.

[0003] In existing traffic signal control, the signal phase is first set based on the intersection's geometric layout and traffic demand. Then, key timing parameters, including cycle length, green light ratio, and phase difference, are calculated. The control objective is to balance the traffic demand of each flow direction while ensuring safety, typically minimizing average delay, queue length, or maximizing capacity to achieve traffic signal control. However, in intelligent urban traffic signal control based on big data and AI algorithms, the signal control logic based on cross-section detectors or conventional radar defaults to green light clearance flow equal to the dissipation flow leaving the intersection. In scenarios with high-density interconnected urban road networks, this control logic may be affected by downstream intersections. Queue overflow failure occurs when a queue backflows to the downstream exit of a target intersection. During the green light period, vehicles exiting the target intersection will be forced to stop because the downstream parking space is completely occupied. From the stop line section of the target intersection, the detector can record the vehicles passing through, but physically the vehicles have not left the intersection area. This leads to a hidden green light loss that cannot be identified by occupancy or flow parameters in classic signal control (i.e., the structural mismatch between green light idling and phase green light ratio caused by the dissipation and interruption of traffic flow during the release phase in traffic signal control). Therefore, how to reduce the hidden green light loss caused by the dissipation and interruption of traffic flow during the release phase in traffic signal control has become a difficult problem for the industry. Summary of the Invention

[0004] This application provides an intelligent urban traffic signal control system based on big data and AI algorithms, which can reduce the hidden green light loss caused by the dissipation and interruption of traffic flow during the release phase in traffic signal control.

[0005] This application provides an intelligent urban traffic signal control system based on big data and AI algorithms, the system comprising: The acquisition module is used to obtain vehicle passage details for each green light period within each signal cycle of the target intersection through the traffic big data platform; The processing module is used to identify saturated traffic flow segments that continuously pass through the stop line in each green light period based on the vehicle passage details of each green light period, and to calculate the saturated headway of the green light period in the current signal cycle based on the headway between each adjacent vehicle in the saturated traffic flow segment and the historical highest free-flow vehicle speed of the corresponding green light period. The processing module is also used to traverse inefficient passage segments within the green light period based on the saturated headway, and determine the invalid green light duration within the current signal cycle based on the inefficient passage segments. The execution module is used to identify the traffic flow dispersion type of the target intersection based on the invalid green light duration of each green light period in the current signal cycle and the distribution differences between each green light period. When the traffic flow dispersion type is demand structure mismatch, the green light period with the highest proportion of invalid green light duration is adaptively adjusted before the next signal cycle arrives.

[0006] In some embodiments, obtaining vehicle passage details for each green light period within each signal cycle at a target intersection through a traffic big data platform specifically includes: The traffic big data platform is used to obtain the vehicle passage data and the traffic light color timing of the target intersection. The vehicle passage data and the traffic light color timing are spatiotemporally aligned to obtain the vehicle passage time sequence and the corresponding traffic light color status of each lane. Based on the detection time of each vehicle passing the stop line in the vehicle passage time sequence, the headway between adjacent vehicles is calculated; Based on the signal light color status, the start and end times of each green light release period within each signal cycle are extracted, and the vehicle passage time sequence and its corresponding headway are combined into the vehicle passage details for the corresponding green light release period by combining the detection time falling within the start and end times of the same green light release period.

[0007] In some embodiments, identifying saturated traffic flow segments that continuously cross the stop line during each green light period, based on vehicle passage details for each green light period, specifically includes: For each green light period, the headway of each adjacent vehicle is extracted from the vehicle passing details corresponding to the green light period. The headway of each adjacent vehicle is traversed in chronological order. The starting vehicle where the headway is first less than the preset saturation threshold is taken as the starting point of the saturated traffic flow segment. The traversal continues until the headway is greater than or equal to the saturation threshold. The vehicle before the starting point of the saturated traffic flow segment is taken as the ending point of the saturated traffic flow segment. The vehicle sequence between the start and end points of the saturated traffic flow segment is determined as the saturated traffic flow segment within the corresponding green light period, thereby obtaining the saturated traffic flow segment that continuously passes through the stop line in each green light period.

[0008] In some embodiments, calculating the saturated headway of the green light period within the current signal cycle based on the headway between adjacent vehicles in the saturated traffic flow segment and the historical highest free-flow vehicle speed during the corresponding green light period specifically includes: Obtain the historical highest free-flow vehicle speed corresponding to the green light period, and determine the theoretical minimum headway for the green light period based on the historical highest free-flow vehicle speed; After removing abnormal headway distances that are less than the theoretical minimum headway distance from the headway distances between adjacent vehicles in the saturated traffic flow segment, the saturated headway distance of the green light period in the current signal cycle is determined based on the remaining headway distances.

[0009] In some embodiments, traversing inefficient traffic segments within the green light period based on the saturated headway specifically includes: A nonlinear tolerance interval function is constructed using the saturated headway as a reference value. The nonlinear tolerance interval function is the dynamic threshold curve output by the trained lightweight time series prediction model. The headway of adjacent vehicles detected during the green light period is input into the nonlinear tolerance interval function. If the actual headway exceeds the upper limit boundary corresponding to the dynamic threshold curve, the moment is marked as the starting point of inefficient passage. Then, the passage time segment corresponding to the sequence of vehicles that continuously exceed the upper limit boundary is taken as the inefficient passage segment during the green light period.

[0010] In some embodiments, determining the invalid green light duration within the current signal cycle based on the inefficient traffic segment specifically includes: If the average headway in the inefficient traffic segment does not exceed the constraint of the saturated headway, the time when the last vehicle in the inefficient traffic segment passes the stop line and the time when the green light ends are identified, and the duration corresponding to the time when the last vehicle passes the stop line and the time when the green light ends is determined as the invalid green light duration. If the average headway in the inefficient traffic segment exceeds the constraint of the saturated headway, the implicit efficiency loss duration in the inefficient traffic segment is calculated, and the invalid green light duration is determined based on the explicit invalid green light duration and the implicit efficiency loss duration.

[0011] In some embodiments, the vehicle transit details include the transit time and headway of each vehicle.

[0012] In some embodiments, the traffic flow dispersion type includes demand structure mismatch type and demand structure equilibrium type.

[0013] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The intelligent traffic signal control system for urban traffic provided in this application, based on big data and AI algorithms, firstly obtains vehicle passage details for each green light period within each signal cycle at the target intersection through a traffic big data platform; secondly, based on the vehicle passage details for each green light period, it identifies saturated traffic flow segments that continuously pass through the stop line during each green light period, and calculates the saturated headway of the green light period within the current signal cycle based on the headway between adjacent vehicles in the saturated traffic flow segment and the historical highest free-flow speed of the corresponding green light period; then, it iterates through inefficient passage segments within the green light period based on the saturated headway, and determines the invalid green light duration within the current signal cycle based on the inefficient passage segments; finally, it identifies the traffic flow dispersion type of the target intersection based on the invalid green light duration of each green light period within the current signal cycle and the distribution differences between each green light period. When the traffic flow dispersion type is demand structure mismatch, the green light period with the highest proportion of invalid green light duration is adaptively adjusted before the next signal cycle arrives.

[0014] Therefore, this application can reduce the hidden green light losses caused by traffic flow dissipation and interruption during the release phase in traffic signal control. First, by obtaining vehicle passage details for each green light period within each signal cycle at the target intersection through a traffic big data platform, the data foundation for signal control can be upgraded to single-vehicle-level time-series behavior records, avoiding the obscuring of traffic efficiency details by aggregated data. Second, based on the vehicle passage details for each green light period, the saturated traffic flow segments that continuously cross the stop line in each green light period are identified, and the saturated headway is calculated accordingly. This ensures that the calibration of the traffic capacity reference benchmark is always based on the actual saturated following state in the current cycle, eliminating the time-varying deviation introduced by using fixed historical experience values ​​or offline calibration parameters. This ensures the dynamic adaptability of the saturated headway to the current traffic environment and driving behavior. Then, based on the saturated headway, inefficient passage segments within the green light period are identified, and invalid green light durations are determined based on these inefficient passage segments. This transforms the evaluation of green light time utilization efficiency from a vague flow ratio to a precise measurement of inefficient time loss. This invention achieves complete capture and quantification of ineffective green light consumption, avoiding the hidden green light losses that cannot be identified through occupancy or flow parameters in classic signal control. Then, based on the ineffective green light duration of each green light period within the current signal cycle and the distribution differences between these periods, the traffic flow dispersion type of the target intersection is identified. When the traffic flow dispersion type indicates a demand structure mismatch, the green light period with the highest proportion of ineffective green light duration is adaptively adjusted. This ensures that the triggering timing of signal timing optimization strictly converges to the moment when there is a significant mismatch in the supply and demand structure of each phase. The adjustment target is precisely located at the phase with the most severe inefficient consumption and the phase most in need of compensation. The adjustment magnitude is directly based on the wasted ineffective green light duration as the trimming benchmark. Thus, under the constraint of maintaining a constant total cycle length, the green light duration is directionally transferred from redundant phases to shortage phases, achieving the steady-state optimization goal of maximizing traffic efficiency with minimal intervention. In summary, the technical solution provided in this application can reduce the hidden green light losses caused by traffic flow dissipation and interruption during the release phase in traffic signal control. Attached Figure Description

[0015] Figure 1 This is a modular structure diagram of an intelligent urban traffic signal control system based on big data and AI algorithms, as shown in some embodiments of this application. Figure 2 This is an exemplary flowchart illustrating the determination of vehicle passage details according to some embodiments of this application. Detailed Implementation

[0016] In the embodiments of this application, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner to facilitate understanding.

[0017] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] refer to Figure 1 As shown in the figure, this is a modular structure diagram of an intelligent urban traffic signal control system based on big data and AI algorithms according to this embodiment of the application. The system includes: an acquisition module 100, a processing module 200, and an execution module 300, which are described below: The acquisition module 100 is used to obtain vehicle passage details for each green light period within each signal cycle of the target intersection through the traffic big data platform.

[0019] refer to Figure 2 As shown, this diagram is an exemplary flowchart illustrating the determination of vehicle passage details according to some embodiments of this application. In this embodiment, obtaining vehicle passage details for each green light period within each signal cycle of the target intersection through a traffic big data platform can be achieved using the following steps: In step 1001, the vehicle passage data and the traffic light color timing of the target intersection are obtained through the traffic big data platform. The vehicle passage data and the traffic light color timing are spatiotemporally aligned to obtain the vehicle passage time sequence and the corresponding traffic light color status of each lane. In step 1002, the headway between adjacent vehicles is calculated based on the detection time of each vehicle passing the stop line in the vehicle passing time sequence; In step 1003, the start and end times of each green light release period within each signal cycle are extracted based on the signal light color status, and the vehicle passage time sequence and its corresponding headway are combined into the vehicle passage details for the corresponding green light release period by combining the detection time falling within the start and end times of the same green light release period.

[0020] In practical implementation, firstly, the sensing devices and traffic signal controllers at the target intersection are connected to a traffic big data platform. Real-time vehicle data is collected using sensing devices deployed at the target intersection (such as geomagnetic coils, radar-visual fusion detectors, or electronic police checkpoints). This vehicle data consists of vehicle identification marks in the lanes and the detection time recorded when a vehicle crosses the stop line. Simultaneously, the traffic light color sequence is obtained from the traffic signal controller. This traffic light color sequence refers to the start and end time sequence of the green light period for each phase. In this application, the sensing device uses a radar-visual fusion detector, and subsequently, Apache... The Flink streaming computing framework performs spatiotemporal alignment processing on vehicle passage data and traffic light color timing. This alignment process involves synchronizing the timestamps of both types of data using a unified global time base (e.g., NTP network time protocol) to eliminate time deviations caused by device clock drift or transmission delays. Based on the detection times of each vehicle passing the stop line in the aligned vehicle passage data, a vehicle passage time sequence for each lane is constructed in chronological order. This time sequence is a set of vehicle passage events arranged chronologically. The vehicle identifiers of vehicles in each lane in the passage data are associated and matched with the phases that are simultaneously in the green light state, so that each vehicle passage record is assigned a corresponding traffic light color status marker to obtain the traffic light color status for each lane. The "state" refers to the light color attribute of the lane where the vehicle passes the stop line when the traffic light is green. Then, based on the time difference between two adjacent vehicles passing through the same detection section in the vehicle passage time sequence, the headway between adjacent vehicles is calculated. The headway is the time interval between the rear bumper of the preceding vehicle passing through the section and the front bumper of the following vehicle passing through the same section. The headway represents the closeness of vehicle following. Finally, based on the start and end times of each green light period indicated by the traffic light color state as the time window boundary, and using the vehicle detection time as the index key, the vehicle passage time sequences whose detection times fall within the same green light period start and end time interval and their corresponding headway are combined to form the vehicle passage details for that green light period. The vehicle passage details include the passage time and headway of each vehicle.

[0021] It should be noted that, in this application, the vehicle passage details refer to the sequence of vehicles arranged in ascending order of passage time during the green light period and the headway between each adjacent vehicle. Since traditional aggregated statistical indicators based on cross-sectional flow or occupancy cannot reveal the temporal distribution characteristics and the degree of car-following closeness of vehicles passing through the stop line, it is difficult to distinguish the time boundary between effective passage and ineffective dissipation. As a structured time-series data record with single-vehicle passage time and headway as the basic unit, the vehicle passage details can effectively provide a unique data input source and feature reference benchmark for subsequent steps.

[0022] Processing module 200 is used to identify saturated traffic flow segments that continuously pass through the stop line in each green light period based on the vehicle passage details of each green light period, and to calculate the saturated headway of the green light period in the current signal cycle based on the headway between adjacent vehicles in the saturated traffic flow segment and the historical highest free-flow vehicle speed of the corresponding green light period.

[0023] Based on the vehicle passage details for each green light period, the following steps are used to identify the saturated traffic flow segment that continuously crosses the stop line during each green light period: For each green light period, the headway of each adjacent vehicle is extracted from the vehicle passing details corresponding to the green light period. The headway of each adjacent vehicle is traversed in chronological order. The starting vehicle where the headway is first less than the preset saturation threshold is taken as the starting point of the saturated traffic flow segment. The traversal continues until the headway is greater than or equal to the saturation threshold. The vehicle before the starting point of the saturated traffic flow segment is taken as the ending point of the saturated traffic flow segment. The vehicle sequence between the start and end points of the saturated traffic flow segment is determined as the saturated traffic flow segment within the corresponding green light period, thereby obtaining the saturated traffic flow segment that continuously passes through the stop line in each green light period.

[0024] In practice, firstly, the headway sequence of adjacent vehicles passing through the stop line section during the green light period is extracted from the vehicle passage details to obtain the headway of each adjacent vehicle. A preset saturation threshold is used as the state discrimination benchmark. The saturation threshold refers to the maximum acceptable headway empirical value observed under the historical saturated flow state of the target intersection. Specifically, it can be calibrated based on the headway distribution characteristics measured when the queued vehicles in the lane are continuously released during peak hours. The headway sequence during the green light period is traversed in ascending order of time. The vehicle whose headway value is first detected to be less than the saturation threshold during the traversal is marked as the starting point of the saturated traffic flow segment. The starting point of the saturated traffic flow segment is the first time when the queued vehicles begin to pass through the stop line section in a close following manner. The process begins by identifying the vehicle. Then, iterating through the headway sequence of adjacent vehicle pairs, whenever a headway value is greater than or equal to the saturation threshold, determines that the traffic flow has transitioned from a saturated following state to a discrete free state. The iteration ends at this point, and the preceding vehicle in the adjacent vehicle pair at that moment is marked as the end point of the saturated traffic flow segment. The end point of the saturated traffic flow segment refers to the last vehicle in the queue that continuously passes the stop line section in a close following manner. Finally, the ordered set of all vehicles between the start and end points of the saturated traffic flow segment is determined as the saturated traffic flow segment within the green light period. By applying the above iteration and determination process sequentially to each green light period, the saturated traffic flow segment continuously passing the stop line within each green light period can be obtained.

[0025] It should be noted that, in this application, a saturated traffic flow segment refers to a continuous queue of vehicles in a saturated flow state during the green light period. Identifying saturated traffic flow segments can provide a set of vehicle following samples that conforms to the definition of traffic flow theory for the accurate calibration of saturated headway. Since saturated headway must be measured in a stable following state where the traffic flow is in a continuous queue and is releasing maximum capacity, it has physical significance in characterizing traffic capacity. Discrete headway in an unsaturated state is affected by individual driver behavior differences and random arrival intervals and cannot accurately reflect the minimum headway level of the lane under extreme traffic conditions. By identifying saturated traffic flow segments, it is ensured that the estimated saturated headway truly reflects the actual saturated flow release characteristics of the approach lane in the current signal cycle, providing an accurate reference benchmark for the identification of invalid green light durations.

[0026] The saturated headway during the current signal cycle is calculated based on the headway between adjacent vehicles in a saturated traffic flow segment and the historical highest free-flow vehicle speed during the corresponding green light period. This is achieved through the following steps: Obtain the historical highest free-flow vehicle speed corresponding to the green light period, and determine the theoretical minimum headway for the green light period based on the historical highest free-flow vehicle speed; After removing abnormal headway distances that are less than the theoretical minimum headway distance from the headway distances between adjacent vehicles in the saturated traffic flow segment, the saturated headway distance of the green light period in the current signal cycle is determined based on the remaining headway distances.

[0027] Specifically, the historical maximum free-flow speed of the approach lane corresponding to the green light period is obtained from the historical operation database of the signal control system. The historical maximum free-flow speed is the highest driving speed value of vehicles passing through the stop line section obtained from the observation records of the approach lane of the target road segment or the highest design speed of the target road segment. The historical maximum free-flow speed represents the maximum capacity boundary that vehicles can reach in the approach lane under the condition of not being constrained by downstream queuing and signal obstruction. Subsequently, the theoretical minimum headway of the green light period is determined based on the historical maximum free-flow speed. The theoretical minimum headway refers to the minimum allowable time interval between two vehicles under ideal safe following conditions. For example, its determination method can be based on the basic relationship between headway, vehicle speed, and vehicle length in traffic engineering, that is, the theoretical minimum headway is equal to the sum of the vehicle length and the minimum safe distance divided by the historical maximum free-flow speed. The vehicle length can be taken as the converted length of a standard passenger car, and the minimum safe distance can be taken as the sum of the driving distance corresponding to the driver's reaction time and the braking safety margin at the corresponding vehicle speed. This calculation process is completed in advance in the parameter configuration module of the signal control system through AI algorithms.

[0028] Specifically, the headway sequence between adjacent vehicles in the saturated traffic flow segment is iterated one by one. Each headway observation value is compared with the theoretical minimum headway, and all headway observation values ​​smaller than the theoretical minimum headway are eliminated. This elimination operation is to constrain abnormally small headway observation values ​​caused by abnormal traffic events such as detector false triggering, vehicles crossing lanes, and motorcycles or non-motorized vehicles crossing. Based on the headway set of the remaining saturated traffic flow segments after elimination, the saturated headway of the green light period in the current signal cycle is calculated. The current signal cycle refers to the current signal cycle of the target intersection, which includes a complete red light duration, green light duration, and yellow light duration. The saturated headway refers to the average time interval between two adjacent vehicles passing through the stop line section in a stable following manner under saturated traffic flow conditions. Specifically, the saturated headway can be obtained by averaging the remaining headway observation values ​​using the arithmetic mean method, which will not be elaborated here.

[0029] It should be noted that the saturated headway in this application is used to characterize the degree of close following of vehicles when they are continuously released at maximum capacity under the current traffic environment and road geometry conditions at the target intersection entrance. It is a scalar parameter that quantifies the limit level of traffic efficiency during the green light period. Since the essential sign of the traffic flow transitioning from a saturated state to a discrete state during the green light period is that the headway gradually diverges and increases from being stably maintained near the saturated headway, determining the saturated headway can provide a physically meaningful dynamic reference benchmark for the traversal identification of inefficient traffic segments.

[0030] The processing module 200 is further configured to traverse inefficient traffic segments within the green light period based on the saturated headway, and determine the invalid green light duration within the current signal cycle based on the inefficient traffic segments.

[0031] The following steps are used to traverse the inefficient traffic segments within the green light period based on the saturated headway: A nonlinear tolerance interval function is constructed using the saturated headway as a reference value. The nonlinear tolerance interval function is the dynamic threshold curve output by the trained lightweight time series prediction model. The headway of adjacent vehicles detected during the green light period is input into the nonlinear tolerance interval function. If the actual headway exceeds the upper limit boundary corresponding to the dynamic threshold curve, the moment is marked as the starting point of inefficient passage. Then, the passage time segment corresponding to the sequence of vehicles that continuously exceed the upper limit boundary is taken as the inefficient passage segment during the green light period.

[0032] Specifically, a nonlinear tolerance interval function is constructed using the saturated headway as a baseline value. This nonlinear tolerance interval function refers to a mapping relationship where the saturated headway is the input variable, and the output is a threshold upper limit curve that changes with the green light duration. This nonlinear tolerance interval function does not use a static threshold with a fixed multiplier; instead, it incorporates the cumulative effect of driver reaction delay into a dynamic threshold adjustment mechanism. Its construction involves deploying a trained lightweight time-series prediction model in the data processing module of the roadside edge computing unit and processing it using AI algorithms. Specifically, the model input features are the time step number after the green light turns on or the percentage of green light duration corresponding to that moment, and the output is the dynamic threshold corresponding to that moment, thus forming a dynamic threshold curve that evolves over time. The lightweight... During the training phase, the lightweight time series prediction model can construct a training sample set using observational data of historical saturated traffic flow segments. The headway sequence arranged chronologically during the green light period is used as the supervision signal. The input layer of the model receives the green light time step number and its normalized value. The encoding module, which includes a temporal convolutional network layer, extracts the temporal cumulative features of driver reaction lag. Then, the fully connected output layer generates the upper limit prediction value of the threshold corresponding to the input time step number. During training, the mean square error between the predicted threshold and the statistical upper bound of the actual headway is minimized as the optimization objective. An adaptive moment estimation optimizer is used for gradient backpropagation until the loss function converges. The model parameters after training are then solidified into the corresponding parameters of the lightweight time series prediction model to complete the training of the lightweight time series prediction model.

[0033] Preferably, the nonlinear tolerance interval function is defined by the following expression:

[0034] in, During the green light period The upper limit of the dynamic headway determination threshold at any given time is the boundary value for determining whether the headway at that time belongs to inefficient passage. The saturated headway is the reference headway value calibrated based on the saturated traffic flow segment within the current signal cycle; It is the cumulative value of the passage time calculated from the moment the green light turns on, in seconds, and its value range is a closed interval from zero to the actual duration of the green light period; The amplitude coefficient of the cumulative effect of driver reaction delay is determined based on the statistical increase rate of the headway distance as the release time changes in the historical saturated traffic flow segment of the approach lane, and the value is usually in the range of 0.2 to 0.8. This is the decay rate coefficient for the cumulative effect of driver reaction lag. Its value is determined based on the convergence speed at which the increase in headway stabilizes in the historical saturated traffic flow segment of the approach lane. The value is typically in the range of 0.05 to 0.25. Specifically, this value reflects the convergence speed at which driver reaction lag propagates along the queue and tends towards saturation in the queuing vehicles of the approach lane. In this expression, the baseline term... An initial discrimination boundary with reference to the saturation headway is provided, and an exponential correction term is given. This describes the nonlinear growth process of the cumulative effect of driver reaction lag as it gradually approaches the upper limit of saturation over the release time, with coefficients... This controls the final amplitude level of the cumulative effect. Together, the three constitute a monotonically increasing nonlinear discrimination threshold function with an upper bound, using the green light duration as the independent variable.

[0035] It should be noted that the dynamic threshold curve in this embodiment represents the traffic efficiency decay law caused by the accumulation of driver reaction delay during the green light release process. Driver reaction delay refers to the time delay in the perception and operation response of the driver behind in the queue to the start of the vehicle in front. This delay is accumulated and transmitted from the front to the back of the queue, causing the headway of the following vehicles to show a systematic increasing trend as the queue position moves backward. The traffic efficiency decay law refers to the phenomenon that the upper limit of the reasonable headway allowed during the green light release period should be gradually widened as the release time progresses due to the cumulative effect.

[0036] Specifically, the headway distances of adjacent vehicles detected during the green light period are input in real time into the nonlinear tolerance interval function. The nonlinear tolerance interval function outputs the upper limit boundary value of the corresponding dynamic threshold curve based on the current green light release time. The actual headway distance observation value is compared with the upper limit boundary value item by item. If the actual headway distance exceeds the upper limit boundary value corresponding to the dynamic threshold curve, it is determined that the traffic flow has changed from a saturated following state to an inefficient discrete state at that moment, and this moment is marked as the inefficient passage start point. The inefficient passage start point is the moment when the headway distance of vehicles first deviates from the dynamic discrimination threshold boundary during the green light release period. Subsequently, starting from the inefficient passage start point, the headway distance sequence of subsequent vehicles is continuously traversed, and the passage time segment corresponding to the vehicle sequence that continuously exceeds the upper limit boundary is defined as the inefficient passage segment within the green light period.

[0037] It should be noted that, in this application, inefficient traffic segments refer to continuous time intervals within the green light period where traffic flow is discrete and therefore in a state of lower than saturation flow efficiency. Identifying inefficient traffic segments can provide a clear spatiotemporal boundary for the precise quantification of ineffective green light duration. Since inefficient traffic segments directly correspond to continuous time intervals within the green light period where traffic flow has transitioned from a saturated following state to a discrete free state and the traffic efficiency is significantly lower than the lane's saturation capacity, the difference between the start and end times of this segment can be accumulated as the duration of the segment to obtain the ineffective green light duration that objectively reflects the loss of green light time utilization efficiency. This transforms the optimization basis of signal control from traditional traffic balance indicators to a precise measurement of inefficient green light consumption.

[0038] Determining the invalid green light duration within the current signal cycle based on the inefficient traffic segments is achieved through the following steps: If the average headway in the inefficient traffic segment does not exceed the constraint of the saturated headway, the time when the last vehicle in the inefficient traffic segment passes the stop line and the time when the green light ends are identified, and the duration corresponding to the time when the last vehicle passes the stop line and the time when the green light ends is determined as the invalid green light duration. If the average headway in the inefficient traffic segment exceeds the constraint of the saturated headway, the implicit efficiency loss duration in the inefficient traffic segment is calculated, and the invalid green light duration is determined based on the explicit invalid green light duration and the implicit efficiency loss duration.

[0039] In specific implementation, firstly, if the average headway within the inefficient traffic segment does not exceed the constraint of the saturated headway (i.e., the average headway within the inefficient traffic segment does not exceed the saturated headway), where the constraint refers to whether the ratio of the average headway to the saturated headway exceeds a preset upper limit, the time when the last vehicle in the inefficient traffic segment crosses the stop line and the green light ends are identified. The duration corresponding to the time when the last vehicle crosses the stop line and the green light ends is determined as the invalid green light duration. The invalid green light duration refers to the length of time the traffic is idle when there are no vehicles queuing in front of the stop line during the green light period, but the signal remains in a green light state. Then, for the state within the inefficient traffic segment that has not yet reached physical idleness but has deviated from the saturated traffic efficiency due to the dispersion of traffic flow, a mechanism for measuring implicit efficiency loss duration is introduced. Implicit efficiency loss duration refers to the difference between the actual number of vehicles passing through the inefficient traffic segment and the number of vehicles that can pass according to the saturated flow rate theory, caused by the headway being greater than the saturated headway. The equivalent time loss corresponding to the implicit efficiency loss duration is calculated based on the constraint that the average headway within an inefficient traffic segment exceeds the saturation headway. The triggering process involves extracting the headway of each adjacent vehicle within the inefficient traffic segment from the vehicle passage details and calculating its arithmetic mean as the average headway. This average headway is then compared to the saturation headway. If the average headway exceeds the saturation headway, the implicit efficiency loss duration calculation logic is initiated. The efficiency loss duration is calculated by multiplying the duration of the inefficient traffic segment by an efficiency loss factor derived from the ratio of average headway to saturation headway. This efficiency loss factor is equal to the difference between the ratio of saturation headway to average headway and represents the proportion of theoretical capacity lost due to headway divergence. The explicit ineffective green light duration and the implicit efficiency loss duration are summed to obtain the ineffective green light duration of the green light period within the current signal cycle.

[0040] It should be noted that, in this application, invalid green light duration refers to the cumulative time loss during the green light release period that cannot be used for the saturation release of vehicles. Since traditional signal control methods usually use the flow ratio or saturation of each phase as the basis for green light ratio allocation, they cannot distinguish the composition ratio of effective passage time and inefficient passage time during the green light release period. This leads to the erroneous allocation of green light time to phases that are in a discrete vacant state with no closely following traffic flow in the demand structure mismatch scenario. However, it is impossible to accurately quantify this wasted green light duration and transfer it to other phases with implicit queuing pressure. Therefore, by determining the invalid green light duration, it is possible to identify whether the current signal cycle is in a demand structure mismatch state, and thus transform the goal of signal timing optimization from fuzzy flow balance to precise inefficient time recovery and redistribution.

[0041] The execution module 300 is used to identify the traffic flow dispersion type of the target intersection based on the invalid green light duration of each green light period in the current signal cycle and the distribution differences between each green light period. When the traffic flow dispersion type is demand structure mismatch, the green light period with the highest proportion of invalid green light duration is adaptively adjusted before the next signal cycle arrives.

[0042] The traffic flow dispersion type of the target intersection is identified by combining the invalid green light duration of each green light period within the current signal cycle with the distribution differences between each green light period. This is achieved through the following steps: Calculate the average invalid green light duration for each green light period within the current signal cycle, and determine the distribution difference between each green light period based on the degree of deviation between the invalid green light duration for each green light period and the average invalid green light duration. The distribution differences between different green light periods are compared with a preset dispersion classification threshold, and the traffic flow dispersion type of the target intersection is identified based on the comparison results.

[0043] In specific implementation, firstly, the average invalid green light duration for each green light period within the current signal cycle is calculated. Using this average invalid green light duration as a reference center, the deviation between the invalid green light duration for each green light period and the average invalid green light duration is calculated one by one. The deviation refers to the dispersion of the observed invalid green light duration for a single green light period relative to the center of the mean. For example, the absolute value of the difference between each invalid green light duration and the mean can be used to determine the deviation between the invalid green light duration for each green light period and the average invalid green light duration. Then, the distribution difference between each green light period is determined by the deviation of each green light period. The distribution difference refers to the statistical measure of the uneven distribution of invalid green light durations among different phases within the current signal cycle. It can be determined by calculating the standard deviation of each deviation as a distribution difference index. Then, the distribution difference is compared with a preset dispersion classification threshold. The dispersion classification threshold is determined based on the historical operation of the intersection. The threshold values ​​obtained from data statistics are used to distinguish the types of traffic flow dispersion. These values ​​can be set according to actual needs or expert knowledge. Based on the comparison results, the traffic flow dispersion type of the target intersection is identified. Traffic flow dispersion type refers to a qualitative classification label of the degree of matching between traffic flow demand and signal timing in each phase within the current signal cycle. When the distribution difference is less than or equal to the dispersion classification threshold, the traffic flow dispersion type is determined to be demand structure balanced. A balanced demand structure indicates that the distribution of invalid green light durations among phases is relatively uniform, and the current cycle's timing scheme basically matches the demand structure of traffic arriving from each direction. When the distribution difference is greater than the dispersion classification threshold, the traffic flow dispersion type is determined to be demand structure mismatched. A demand structure mismatch indicates that the distribution of invalid green light durations among phases is significantly uneven, with a structural mismatch where one phase has significantly redundant green light time while another phase has significantly insufficient green light time. Targeted adjustments to the green light ratio allocation scheme are required in the next signal cycle.

[0044] It should be noted that, in this application, the traffic flow dispersion type refers to a qualitative classification label of the degree of balance in the distribution of invalid green light duration consumed by each phase during the green light period in the current signal cycle. The traffic flow dispersion type includes demand structure mismatch type and demand structure balance type. The determination of the traffic flow dispersion type can provide a clear decision trigger criterion for whether to initiate the green light ratio reduction and adjustment. Since forcibly adjusting the timing in the demand structure balance type will cause the originally matched phases to generate a new supply and demand imbalance, while in the demand structure mismatch type, if it is not adjusted in time, the invalid green light duration of redundant phases will continue to accumulate, and the queuing pressure of the shortage phases will be further aggravated. By determining the traffic flow dispersion type, the green light ratio adjustment process can be reliably triggered when the demand structure mismatch is identified in multiple consecutive signal cycles, so as to ensure that the intervention action of the signal control only acts on the most needed phase pair with the minimum amplitude at the necessary time, and avoids system oscillation and traffic efficiency loss caused by frequent or blind adjustment of timing parameters.

[0045] When the traffic flow dispersion type is demand structure mismatch, the following steps are used to adaptively adjust the green light period with the highest proportion of invalid green light duration before the next signal cycle arrives: Obtain the invalid green light duration and the actual green light duration for each green light period within the current signal cycle, and calculate the ratio of the invalid green light duration to the actual green light duration for each green light period. The green light period with the largest ratio is identified as the green light period to be cut, and the green light period with the smallest ratio is identified as the green light period to be compensated. The amount of reduction is determined based on the invalid green light duration of the green light period to be reduced. The amount of reduction is then subtracted from the green light duration of the green light period to be reduced in the next signal cycle to obtain the reduced green light duration. The amount of reduction is then added to the green light duration of the green light period to be compensated in the next signal cycle to obtain the compensated green light duration. Based on the trimmed green light duration and the compensated green light duration, a phase-green ratio allocation scheme for the next signal cycle is generated, and the phase-green ratio allocation scheme is sent to the signal control mechanism for execution.

[0046] In practice, firstly, the invalid green light duration and the actual green light duration of each green light period within the current signal cycle are obtained. The ratio of the invalid green light duration to the actual green light duration of each green light period is calculated. The value of this ratio represents the proportion of inefficiently consumed or idle portions of the green light time for that phase. The actual green light duration refers to the duration of the green light period assigned to the signal timing scheme of the current signal cycle, from the green light activation time to the green light end time. Secondly, the green light period with the largest ratio is identified as the green light period to be cut, and the green light period with the smallest ratio is identified as the green light period to be added. The green light period is divided into two categories: the green light period to be cut refers to the phase whose green light duration will be reduced in the green light ratio adjustment of the next signal cycle, and the green light period to be compensated refers to the phase whose green light duration will be increased in the green light ratio adjustment of the next signal cycle. Then, the cutting amount is determined based on the invalid green light duration of the green light period to be cut. The cutting amount is the time length deducted from the green light duration of the next cycle of the green light period to be cut. Specifically, it can be obtained by multiplying the invalid green light duration by a preset cutting ratio coefficient. The preset cutting ratio coefficient is a convergence control coefficient between zero and one, calibrated based on the historical adjustment effect feedback of the intersection. The parameters are set to prevent timing oscillations caused by excessively large single adjustments. The green light duration to be clipped is obtained by subtracting the clipping amount from the green light duration to be allocated in the next signal cycle. Simultaneously, the green light duration to be compensated is added to the green light duration to be allocated in the next signal cycle to obtain the compensated green light duration. The clipped and compensated green light durations are the actual new green light durations for the clipped and compensated phases in the next signal cycle, respectively. Finally, the clipped green light duration is assigned to the green light duration parameter of the phase corresponding to the green light duration to be clipped, and the compensated green light duration is assigned to... The green light duration parameter corresponding to the green light period to be compensated is assigned, while the green light duration of other phases remains unchanged in the current cycle. Based on this, a phase green light ratio allocation scheme for the next signal cycle is generated. The phase green light ratio allocation scheme refers to a structured timing parameter set of the green light duration of each phase and its arrangement order in the next signal cycle. This phase green light ratio allocation scheme is sent to the signal control agency for execution through the communication protocol between the signal control system and the intersection signal control agency. At the beginning of the next signal cycle, the signal control agency controls the switching of the light color status of each light group according to the updated phase green light ratio allocation scheme.For example, in a certain four-phase signal cycle, if the calculated proportion of invalid duration for the north-south straight-ahead phase is the maximum among all phases, and the proportion of invalid duration for the east-west left-turn phase is the minimum among all phases, then the north-south straight-ahead phase is determined as the green light period to be cut, and the east-west left-turn phase is determined as the green light period to be compensated. If the invalid green light duration for the north-south straight-ahead phase is a certain value, the cutting amount is determined according to the preset cutting ratio coefficient. The green light duration for the north-south straight-ahead phase in the next cycle is reduced by this cutting amount, and the green light duration for the east-west left-turn phase in the next cycle is increased by the same cutting amount. This generates the adjusted timing scheme for the next cycle and sends it to the intersection signal control agency for execution.

[0047] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0048] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A smart traffic signal control system for cities based on big data and AI algorithms, characterized in that, The system includes: The acquisition module is used to obtain vehicle passage details for each green light period within each signal cycle of the target intersection through the traffic big data platform; The processing module is used to identify saturated traffic flow segments that continuously pass through the stop line in each green light period based on the vehicle passage details of each green light period, and to calculate the saturated headway of the green light period in the current signal cycle based on the headway between each adjacent vehicle in the saturated traffic flow segment and the historical highest free-flow vehicle speed of the corresponding green light period. The processing module is also used to traverse inefficient passage segments within the green light period based on the saturated headway, and determine the invalid green light duration within the current signal cycle based on the inefficient passage segments. The execution module is used to identify the traffic flow dispersion type of the target intersection based on the invalid green light duration of each green light period in the current signal cycle and the distribution differences between each green light period. When the traffic flow dispersion type is demand structure mismatch, the green light period with the highest proportion of invalid green light duration is adaptively adjusted before the next signal cycle arrives.

2. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, The traffic big data platform obtains detailed vehicle passage information for each green light period within each signal cycle at the target intersection, including: The traffic big data platform is used to obtain the vehicle passage data and the traffic light color timing of the target intersection. The vehicle passage data and the traffic light color timing are spatiotemporally aligned to obtain the vehicle passage time sequence and the corresponding traffic light color status of each lane. Based on the detection time of each vehicle passing the stop line in the vehicle passage time sequence, the headway between adjacent vehicles is calculated; Based on the signal light color status, the start and end times of each green light release period within each signal cycle are extracted, and the vehicle passage time sequence and its corresponding headway are combined into the vehicle passage details for the corresponding green light release period by combining the detection time falling within the start and end times of the same green light release period.

3. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, Based on the vehicle passage details for each green light period, the saturated traffic flow segment that continuously crosses the stop line during each green light period is identified, specifically including: For each green light period, the headway of each adjacent vehicle is extracted from the vehicle passing details corresponding to the green light period. The headway of each adjacent vehicle is traversed in chronological order. The starting vehicle where the headway is first less than the preset saturation threshold is taken as the starting point of the saturated traffic flow segment. The traversal continues until the headway is greater than or equal to the saturation threshold. The vehicle before the starting point of the saturated traffic flow segment is taken as the ending point of the saturated traffic flow segment. The vehicle sequence between the start and end points of the saturated traffic flow segment is determined as the saturated traffic flow segment within the corresponding green light period, thereby obtaining the saturated traffic flow segment that continuously passes through the stop line in each green light period.

4. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, The calculation of the saturated headway during the current signal cycle based on the headway between adjacent vehicles in the saturated traffic flow segment and the historical highest free-flow vehicle speed during the corresponding green light period specifically includes: Obtain the historical highest free-flow vehicle speed corresponding to the green light period, and determine the theoretical minimum headway for the green light period based on the historical highest free-flow vehicle speed; After removing abnormal headway distances that are less than the theoretical minimum headway distance from the headway distances between adjacent vehicles in the saturated traffic flow segment, the saturated headway distance of the green light period in the current signal cycle is determined based on the remaining headway distances.

5. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, The inefficient passage segments within the green light period, traversed based on the saturated headway, specifically include: A nonlinear tolerance interval function is constructed using the saturated headway as a reference value. The nonlinear tolerance interval function is the dynamic threshold curve output by the trained lightweight time series prediction model. The headway of adjacent vehicles detected during the green light period is input into the nonlinear tolerance interval function. If the actual headway exceeds the upper limit boundary corresponding to the dynamic threshold curve, the moment is marked as the starting point of inefficient passage. Then, the passage time segment corresponding to the sequence of vehicles that continuously exceed the upper limit boundary is taken as the inefficient passage segment during the green light period.

6. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, Determining the invalid green light duration within the current signal cycle based on the inefficient traffic segments specifically includes: If the average headway in the inefficient traffic segment does not exceed the constraint of the saturated headway, the time when the last vehicle in the inefficient traffic segment passes the stop line and the time when the green light ends are identified, and the duration corresponding to the time when the last vehicle passes the stop line and the time when the green light ends is determined as the invalid green light duration. If the average headway in the inefficient traffic segment exceeds the constraint of the saturated headway, the implicit efficiency loss duration in the inefficient traffic segment is calculated, and the invalid green light duration is determined based on the explicit invalid green light duration and the implicit efficiency loss duration.

7. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, The vehicle passage details include the passage time and headway of each vehicle.

8. The intelligent urban traffic signal control system based on big data and AI algorithms as described in claim 1, characterized in that, The traffic flow dispersion types include demand structure mismatch type and demand structure balanced type.

Images