AI traffic signal control method and system based on multi-source perception and edge computing

Abstract

The application provides an AI traffic signal control method and system based on multi-source perception and edge computing, the method comprising collecting multi-source traffic flow data according to a preset period to generate a multi-class basic data table; aggregating phase-level traffic flow indicators; determining a current scene; matching a corresponding control strategy according to the determined current scene, and calculating the saturation of each phase based on edge computing according to the total phase flow, executing the corresponding control strategy, and generating a signal timing instruction; executing phase sequence switching and green light duration adjustment, and uploading the phase-level traffic flow indicators and the signal timing instruction to the cloud for global optimization based on multi-intersection data. The application adapts to traffic flow characteristics in different periods; by constructing a closed-loop control architecture from perception to processing, decision-making, execution and global optimization, the data islands between systems are broken.

Description

Technical Field

[0001] This invention belongs to the field of intelligent traffic signal control technology, specifically relating to an AI traffic signal control method and system based on multi-source sensing and edge computing. Background Technology

[0002] Traffic signal control refers to the core means by which traffic management departments regulate traffic flow and ensure traffic order at urban intersections through signal control equipment. In the transportation sector, traditional signal control schemes include fixed control schemes based on historical traffic flow data with pre-set multi-period time slots; simple inductive control schemes that rely on loop detectors and ordinary video vehicle detectors to detect the arrival of vehicles or pedestrians on secondary roads and trigger signal switching; and adaptive control schemes that incorporate basic algorithms such as Webster to adjust signal cycles and green light durations. Currently, urban intersection signal control still primarily relies on fixed timing and manual on-site adjustments. Data transmission between roadside sensing devices, computing units, and traffic signal controllers is mostly unidirectional, lacking deep technological collaboration. Especially in complex scenarios such as sudden changes in traffic flow and oversaturated queuing, continuous optimization and adjustments are needed after deployment to enhance control effectiveness.

[0003] Among existing technologies, traffic signal control solutions suffer from several shortcomings. First, their perception capabilities are weak. Traditional detectors are highly susceptible to environmental interference, and their accuracy is insufficient under conditions of rain, fog, and strong light. They cannot accurately acquire key traffic flow parameters such as lane-specific flow, queue length, and headway, and they lack integrated monitoring capabilities for pedestrians and non-motorized vehicles. Second, data processing and control are lagging. The reliance on centralized cloud computing for signal light switching leads to excessively long data transmission and processing times, and the control response time cannot promptly address sudden changes in traffic flow and occasional congestion. Third, the control strategies are simplistic. Fixed timing for multiple time periods cannot adapt to dynamic changes in traffic flow. Existing adaptive algorithms only cover a single, unsaturated condition and do not design differentiated control strategies for different scenarios such as off-peak surges, peak oversaturation, and low nighttime flow, resulting in poor queue reduction during peak hours. Furthermore, the signal timing optimization lacks scientific rigor, the signal phase sequence design relies on human experience and does not incorporate dynamic adjustments based on lane turning traffic flow, and the allocation of cycle length and green light duration is not integrated with AI intelligent algorithms, which can easily lead to wasted or insufficient green light time. Finally, there is insufficient system coordination, with data silos existing between sensing devices, computing units, signal controllers, and the cloud platform, making it impossible to achieve closed-loop control with real-time perception, dynamic decision-making, precise execution, and effect feedback. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an AI traffic signal control method and system based on multi-source perception and edge computing, thereby overcoming the shortcomings of the prior art.

[0005] In a first aspect, the present invention provides an AI traffic signal control method based on multi-source sensing and edge computing, the method comprising: Collect multi-source traffic flow data at a preset cycle, and generate multiple types of basic data tables based on the multi-source traffic flow data; Calculate the PCU flow rate and maximum queue length for each lane based on the aforementioned multiple basic data tables; Based on the lane-specific PCU flow rate, the maximum queue length of the lane, the intersection phase scheme, the number of lanes at the approach, and the turning relationship of each lane, the single-lane data is aggregated into a phase-level traffic flow index. Based on edge computing and the aforementioned phase-level traffic flow index, and using the signal cycle as the determination cycle, the current scenario is determined by combining time period division and working condition characteristics. The corresponding control strategy is matched according to the current scene as determined, and the saturation of each phase is calculated based on the total phase flow, combined with the phase saturation flow, the current cycle, and the current green light duration, based on the edge computing. The corresponding control strategy is executed according to the phase saturation, the current scene, and the maximum rank length of the phase, and a signal timing instruction is generated. Based on the signal timing instructions, the traffic lights are driven to perform phase sequence switching and green light duration adjustment, and the phase-level traffic flow indicators and the signal timing instructions are uploaded to the cloud for global optimization based on multi-intersection data.

[0006] Compared with existing technologies, the beneficial effects of this invention are as follows: By collecting multi-source traffic flow data at all times with high precision, it effectively overcomes the shortcomings of traditional detectors that are affected by environmental interference such as rain, fog, and strong light; by aggregating single-lane traffic flow indicators into phase-level traffic flow indicators, a data transmission chain is established between the raw sensing data and the input parameters of the control algorithm, so that the calculation of phase saturation and scene determination are based on a unified phase-level indicator, ensuring the data correlation and logical consistency between each step; by realizing real-time data processing and AI algorithm decision-making through edge computing units, the data transmission and processing latency is significantly reduced; by designing differentiated control strategies, it adapts to the traffic flow characteristics of different time periods; and by constructing a closed-loop control architecture from perception to processing, decision-making, execution, and global optimization, it breaks down the data silos between various systems.

[0007] Furthermore, the step of collecting multi-source traffic flow data at a preset period and generating multiple types of basic data tables based on the multi-source traffic flow data includes: Parameter settings for the Rayvision all-in-one machine are performed via the cloud. The system acquires multi-source traffic flow data collected by the integrated radar-visual system at a preset cycle, generates multiple types of basic data tables based on the multi-source traffic flow data, and constructs two types of statistical data tables based on control requirements.

[0008] Furthermore, the steps of calculating the lane-specific PCU traffic flow and the maximum queue length for each lane based on the multiple types of basic data tables include: Obtain the start and end times of the latest execution plan from the signal timing scheme information table, accumulate the number of vehicles within the start and end times, and calculate the PCU flow rate and maximum queue length of each lane respectively. The maximum allowable flow rate for a given period is calculated based on the lane saturation flow rate and the green light duration. It is then determined whether the flow rate of the PCU in each lane is greater than the maximum allowable flow rate for that period. If so, the data is considered abnormal, and the calculated flow rate of the PCU in each lane is verified.

[0009] Furthermore, the calculation expression for the lane-specific PCU traffic flow is as follows: ; In the formula, This indicates the PCU flow rate for each lane. , , , These represent the number of small vehicles, medium-sized vehicles, heavy vehicles, and electric vehicles counted within the period, respectively. The formula for calculating the maximum queue length for each lane is: ; In the formula, This indicates the maximum queue length for each lane. Indicates the detection length of the Ravec integrated machine. This indicates the distance between the location of the integrated radar monitor and the parking line. Indicates the length of the blind zone illuminated by the integrated radar vision system; The expression for calculating the maximum allowable flow rate per period is: ; In the formula, Indicates the maximum allowed flow rate during the period. Indicates lane saturation flow. Indicates the duration of the green light. Indicates the supersaturation coefficient. Indicates the duration of the period.

[0010] Furthermore, the expression for the aggregation into phase-level traffic flow indicators is: ; In the formula, Indicates the total phase flow. This indicates the number of lanes included in that phase. Indicates the first The flow rate of each lane's PCU. Indicates the maximum queue length of the phase. , ... These represent the maximum queue length of lane 1, the maximum queue length of lane 2, and the maximum queue length of lane 3, respectively. The maximum queue length for each lane of a single lane.

[0011] Furthermore, the expression for calculating the saturation of each phase is as follows: ; In the formula, Indicates phase saturation. Indicates the duration of the current period. Indicates phase saturation flow rate, Indicates the first Each phase saturation flow, Indicates the duration of the green light in the current phase. This represents the mean value of phase saturation. Indicates the number of phases. Indicates the first Phase saturation.

[0012] Furthermore, the step of performing global optimization based on multi-intersection data includes: The Webster algorithm is used to calculate the basic value of the cycle length, and the phase switching order is adjusted according to the current scene to obtain the phase sequence optimization result; The signal timing command is globally optimized based on edge computing, the phase sequence optimization results, multi-intersection data, lane configuration, and traffic distribution characteristics.

[0013] Secondly, the present invention also provides an AI traffic signal control system based on multi-source sensing and edge computing, the system comprising: The data acquisition module is used to collect multi-source traffic flow data at a preset period and generate multiple types of basic data tables based on the multi-source traffic flow data. The calculation module is used to calculate the PCU flow rate and the maximum queue length of each lane based on the aforementioned multiple basic data tables. The aggregation module is used to aggregate single-lane data into phase-level traffic flow indicators based on the PCU flow of each lane, the maximum queue length of each lane, the intersection phase scheme, the number of lanes at the approach, and the turning relationship of each lane. The determination module is used to determine the current scenario based on edge computing and the phase-level traffic flow index, with the signal cycle as the determination cycle, combined with time period division and working condition characteristics. The execution module is used to match the corresponding control strategy according to the determined current scene, and calculate the saturation of each phase based on the edge computing according to the total phase flow, combined with the phase saturation flow, the current cycle, and the current green light duration. It also executes the corresponding control strategy according to the phase saturation, the current scene, and the maximum rank length of the phase, and generates signal timing instructions. The optimization module is used to drive the traffic lights to perform phase sequence switching and green light duration adjustment based on the signal timing instructions, and upload the phase-level traffic flow indicators and the signal timing instructions to the cloud for global optimization based on multi-intersection data.

[0014] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described AI traffic signal control method based on multi-source perception and edge computing.

[0015] Fourthly, the present invention also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described AI traffic signal control method based on multi-source perception and edge computing. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the AI ​​traffic signal control method based on multi-source perception and edge computing in the first embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the control logic in the first embodiment of the present invention; Figure 3 This is a structural block diagram of the AI ​​traffic signal control system based on multi-source perception and edge computing in the second embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of the electronic device in the third embodiment of the present invention.

[0018] Explanation of key component symbols: 10. Data Acquisition Module; 20. Calculation Module; 30. Aggregation Module; 40. Decision Module; 50. Execution Module; 60. Optimization Module; 70. Bus; 71. Processor; 72. Memory; 73. Communication interface.

[0019] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0020] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0021] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0023] Example 1 Please see Figure 1 The figure shows an AI traffic signal control method based on multi-source perception and edge computing in the first embodiment of the present invention, the method including steps S1 to S6: S1. Collect multi-source traffic flow data according to a preset cycle, and generate multiple types of basic data tables based on the multi-source traffic flow data; Specifically, step S1 includes steps S11 to S12: S11 allows for parameter settings of the Rayvision all-in-one machine via the cloud; S12, acquire the multi-source traffic flow data collected by the radar-visual integrated machine according to a preset cycle, generate multiple types of basic data tables based on the multi-source traffic flow data, and construct two types of statistical data tables based on control requirements; It should be noted that the radar-visual integrated unit is deployed on the existing traffic light poles at the intersection. After deployment, parameters are first set via the cloud, including the number of detection lanes, virtual detection line layout, data upload frequency, and statistical cycle. After configuration, the detection area coverage of the entire lane is verified through the preview function. Subsequently, traffic flow data is collected according to the preset cycle, generating multiple basic data tables such as the intersection information table and the approach lane information table. At the same time, two types of statistical data tables are built based on the platform control requirements. Both types of tables are associated with the intersection ID, radar-visual integrated unit number, lane number, and statistical timestamp to ensure data traceability. Among them, the five-minute statistical data table is for traffic flow prediction service. After the raw data is entered into the database, the traffic flow and queue length indicators of a single lane are calculated according to the five-minute time interval, and then the five-minute phase traffic flow indicators are calculated. Finally, the data is used for traffic flow prediction training. The data covers lane-specific PCU converted traffic flow, lane-specific average speed, time occupancy rate, number of pedestrians waiting at the stop line, and number of non-motorized vehicle categories. The signal cycle statistics table provides real-time traffic flow data. It requires calculating the single-lane traffic flow and queue length indices for the duration of each signal cycle in the last second of that cycle, then calculating the phase traffic flow indices for that cycle, and inputting these into a prediction algorithm to forecast the phase traffic flow indices two cycles later. Finally, this data is fed into the signal control algorithm to generate the signal timing scheme two cycles later. The data includes lane-by-lane cycle-based calculated traffic flow, maximum queue length within the cycle, lane saturation, and minimum vehicle headway.

[0024] S2, calculate the PCU flow rate and maximum queue length of each lane based on the aforementioned multiple basic data tables; Specifically, step S2 includes steps S21 to S22: S21, obtain the start and end times of the latest execution plan from the signal timing scheme information table, accumulate the number of vehicles within the start and end times, and calculate the PCU flow rate and the maximum queue length of each lane respectively. It should be noted that the lane-specific PCU traffic flow calculation based on the forward-facing radar-visual integrated machine requires first locating the specific illuminated lane from the intersection information table - approach lane information table - forward-facing radar-visual integrated machine and actual collected lane association configuration table. Then, the start and end times of the latest execution plan are obtained from the signal timing scheme information table, and the number of small cars, medium-sized cars, heavy cars and electric vehicles in that section are accumulated. The calculation expression for the lane-specific PCU traffic flow is (rounded down): ; In the formula, This indicates the PCU flow rate for each lane. , , , These represent the number of small vehicles, medium-sized vehicles, heavy vehicles, and electric vehicles counted within the period, respectively. The formula for calculating the maximum queue length for each lane is: ; In the formula, This indicates the maximum queue length for each lane. Indicates the detection length of the Ravec integrated machine. This indicates the distance between the location of the integrated radar monitor and the parking line. Indicates the length of the blind zone illuminated by the integrated radar vision system; S22, calculate the maximum allowable flow rate for the cycle based on the lane saturation flow rate and the green light duration, and determine whether the PCU flow rate of the lane is greater than the maximum allowable flow rate for the cycle. If so, it is determined to be a data anomaly, so as to verify the calculated PCU flow rate of the lane. It should be noted that the maximum allowable flow rate per cycle is calculated based on lane saturation flow rate and green light duration. The expression for calculating the maximum allowable flow rate per cycle is as follows: ; In the formula, Indicates the maximum allowed flow rate during the period. Indicates lane saturation flow. Indicates the duration of the green light. Indicates the supersaturation coefficient. Indicates the duration of the period; It should be explained that if the calculated lane-specific PCU flow is greater than the maximum allowable flow for a period, it is directly judged as data anomaly, the data for that period is removed, and the subsequent signal timing update task is not executed.

[0025] It should be noted that the calculation of the maximum queue length per lane based on the reverse radar-visual integrated machine requires first locating the lane illuminated by the reverse radar-visual integrated machine using the intersection information table, the approach lane information table, and the configuration table of the association between the reverse radar-visual integrated machine and the actual collected lanes. Then, the maximum value of the queue length detected by the reverse radar-visual integrated machine within the current period is taken. Subsequently, based on the design length of the approach lane, the maximum allowable queue length is set, and the calculation expression is: ; In the formula, This indicates the maximum allowed queue length, in meters (m). Indicates the design length of the intersection approach lane. This represents a reasonable threshold, which is set to 1.5 in this implementation. If the calculated maximum queue length for each lane is greater than the maximum allowable queue length, it is directly judged as data anomaly, and the subsequent signal timing update task is not executed.

[0026] S3, based on the PCU flow rate of the lane, the maximum queue length of the lane, the intersection phase scheme, the number of lanes at the approach, and the turning relationship of each lane, the single-lane data is aggregated into a phase-level traffic flow index; It needs to be explained that, based on single-lane PCU flow rate, maximum queue length per lane, intersection phasing scheme, number of lanes at the approach, and turning relationships of each lane, single-lane data is aggregated into phase-level traffic flow indicators. The expression for the aggregated phase-level traffic flow indicators is as follows: ; In the formula, Indicates the total phase flow. This indicates the number of lanes included in that phase. Indicates the first The flow rate of each lane's PCU. Indicates the maximum queue length of the phase. , ... These represent the maximum queue length of lane 1, the maximum queue length of lane 2, and the maximum queue length of lane 3, respectively. Maximum queue length for each lane in a single lane; It should be noted that, in order to avoid invalid data from interfering with the decision-making accuracy of the control algorithm, this proposal sets the data collection and update frequency of the radar-vision all-in-one machine to once every 2 seconds, and designs differentiated thresholds and verification rules to identify and handle abnormal situations. To address the issue of missing traffic flow data for a specific period, traffic flow index interpolation requires obtaining the phase information corresponding to each lane. This necessitates clearly defining the start and end times of the green and yellow lights, as well as the start and end times of the red light for each lane. Two different algorithms are used, executed as needed. Traffic flow during the red light period is uniformly counted as 0. Specifically: If the number of missing data uploaded by the radar-visual all-in-one machine within a unit time (1 minute) does not exceed 10: If there are less than 2 valid data during the current green light period, but the data during the green light periods of the preceding and following periods are complete, then an assignment is made, and the expression is: ; In the formula, This represents the traffic interpolation value every 2 seconds during the current green light cycle. This represents the traffic data for every 2 seconds during the previous green light cycle. This represents all traffic data every 2 seconds during the green light period of the next cycle; during the green light period of the current cycle, each 2-second data point is assigned the value Q for the current green light, and during the red light period, each 2-second data point is assigned the value 0; If there are at least two valid traffic data points every 2 seconds during the current green light period, then the arithmetic mean imputation is performed on the missing data before, after, and in the middle of the existing valid data. The specific expression is as follows: ; In the formula, This indicates that data points are missing in the current period. This indicates the valid data points from the previous period. This indicates the valid data points for the next cycle; the radar-visual integrated machine traverses all data, determines the start and end times of the green light for the current cycle based on phase information, locates the missing data points during the green light period, and then processes the missing data points according to... If Q is missing, it will be assigned a value of 0 for each 2-second data point during the red light period.

[0027] If more than 10 data entries are missing from the data uploaded by the Rayvision all-in-one machine within a unit of time (1 minute), it will be directly judged as a data anomaly, and the subsequent signal timing update task will not be executed.

[0028] Regarding the missing data for queue length in a specific period: Queue length interpolation requires obtaining cycle information for each lane, clarifying the start and end times of the signal cycle for each lane, and focusing only on the interpolation of the maximum queue length within the cycle. The specific calculation expression is as follows: ; In the formula, This indicates the maximum queue length value that is missing in the current cycle. This indicates the maximum queue length in the previous valid period. This indicates the maximum queue length for the next valid period. The Rayvision all-in-one machine traverses all data to determine the time range of the current period. If the maximum queue length data for this period is missing, it traces back to the maximum queue length of the most recent valid period (such as the previous period) and then searches for the maximum queue length of the most recent valid period (such as the next period). Finally, it assigns a value to the missing data points by pressing L for missing.

[0029] For the abnormal situation where traffic is consistently zero: When the radar-visual integrated machine continuously collects data showing that the selected vehicle types (small, medium, and heavy) for a certain lane are all zero, it will automatically record the time when the current traffic flow is zero. Simultaneously, it will trace back the historical data collected for that lane, searching for the most recent valid data point where the selected vehicle types (small, medium, and heavy) were not all zero, and calculate the time difference between the two. The specific expression is as follows: ; In the formula, This represents the time difference, i.e., the duration of zero traffic. This indicates the point in time when all traffic is zero. This indicates the most recent valid data point where the selected vehicle types were not all zero; subsequently, based on the current time period of the lane, a preset differential judgment threshold is matched, and the specific threshold standards are shown in Table 1: Table 1. Threshold Table for Time Period Division of the All-in-One Laser Viewer

[0030] When the calculated duration of zero traffic flow T is greater than or equal to the threshold T0 for the current time period of the lane, the detection data of the radar-visual integrated machine corresponding to the lane is determined to be abnormal (the abnormality may be due to lens obstruction, installation angle deviation, or the device being blown off course by the wind). At this time, the system automatically terminates the current and subsequent signal timing update tasks for the lane to avoid invalid data causing decision deviations in the control algorithm.

[0031] S4. Based on edge computing and the phase-level traffic flow index, and using the signal cycle as the determination cycle, the current scenario is determined by combining time period division and working condition characteristics. It should be noted that edge computing is used to analyze traffic flow parameters in real time to determine the current scenario and switch the corresponding control strategy accordingly. The timing optimization algorithm combines scenario targets and traffic flow data to dynamically adjust cycle length, phase sequence, and green light duration. Simultaneously, microscopic simulation and reinforcement learning techniques are incorporated to ensure the scientific validity and practicality of the control effect. The current scenarios include off-peak, peak, and nighttime scenarios.

[0032] Specifically, edge computing uses one signal cycle as the judgment period. Based on parameters such as lane-specific cycle-based traffic flow, maximum queue length within the cycle, and 5-minute statistical traffic flow collected in real time by the radar-view integrated machine, combined with time period division and operating condition characteristics, the scenario is determined according to the following rules. The judgment result is directly used for control strategy switching. Among them, the core parameter of scenario judgment, lane saturation, is calculated according to the following formula to ensure that the judgment standard is consistent across scenarios. The expression is: ; In the formula, This indicates the lane saturation level, with a value ranging from 0 to 1.5. A value greater than 1.0 indicates oversaturation. This indicates the critical number of vehicles that can pass through during the green light duration; Off-peak scene: The Ravis all-in-one machine is considered to be in a non-peak hours scenario when all of the following conditions are met.

[0033] A. The testing period is between 9:00-17:00 and 19:30-21:30.

[0034] B. All lane saturation S < 0.8, meaning there are no oversaturated lanes, and the green light duration basically matches the traffic flow demand.

[0035] C. The number of vehicles arriving at each entrance lane N is less than the critical number of vehicles N*.

[0036] D. There was no sustained surge in traffic lasting more than 5 minutes.

[0037] Peak Scenario: When any of the following core conditions are met, the Rayvision all-in-one machine is identified as a peak oversaturation scenario.

[0038] A. The testing period is between 7:00-9:00 and 17:00-19:00.

[0039] B. At least one approach lane has a saturation S ≥ 0.8 (or ≥ 0.9 for main roads), and the number of vehicles arriving at that approach lane N > the critical number of vehicles N*.

[0040] C. The maximum queue length Lmax of any entrance lane is greater than or equal to the set peak queue length threshold.

[0041] Nighttime scene: The Ravis all-in-one machine is identified as a low-traffic nighttime scenario when all of the following conditions are met.

[0042] A. The testing period is between 22:00 and 6:00 the next day.

[0043] B. There were no sustained traffic surges lasting more than 5 minutes.

[0044] C. The Qpcu of the lane entrance does not exceed the set threshold.

[0045] D. If a sudden short-term traffic surge occurs at night, monitor for multiple cycles. If the traffic volume drops back to within the set threshold range, maintain the nighttime scenario; otherwise, temporarily switch to the off-peak scenario.

[0046] S5, match the corresponding control strategy according to the current scene determined, and calculate the saturation of each phase based on the total phase flow, combined with the phase saturation flow, the current cycle, and the current green light duration based on the edge computing, and execute the corresponding control strategy according to the phase saturation, the current scene, and the maximum rank length of the phase, and generate signal timing instructions; It needs to be explained that the control approach is mainly based on the core characteristics of traffic flow in three scenarios: off-peak, peak, and nighttime. Differentiated control strategies are matched accordingly, and further subdivided within each scenario to optimize the control logic. For example... Figure 2 As shown, the off-peak scenario uses AI+adaptive control to adapt to situations where the number of vehicles arriving on the road network is less than the critical number of vehicles, focusing on dealing with unstable, sudden, and occasional traffic flow changes and congestion. The peak scenario also uses AI+adaptive control, but it is subdivided into two scenarios: the number of vehicles arriving on the road network is greater than the critical number of vehicles and the number is less than the critical number of vehicles. The former focuses on increasing throughput to reduce queue length, while the latter still aims to adapt to sudden changes in traffic flow. The nighttime scenario uses a sensor-based control method, which executes the control logic of keeping the main road green or switching the secondary road to traffic based on whether there are vehicles, pedestrians, and non-motorized vehicles arriving on the secondary road, thereby optimizing equipment energy consumption while ensuring traffic efficiency under low traffic conditions.

[0047] The integrated radar-visual system collects real-time traffic flow from the PCUs in each lane, aggregating it into total phase traffic. Edge computing combines phase saturation traffic, current cycle duration, and current green light duration to calculate phase saturation. The expression for calculating the saturation of each phase is as follows: ; In the formula, Indicates phase saturation. Indicates the duration of the current period. Indicates phase saturation flow rate, Indicates the first Each phase saturation flow, Indicates the duration of the green light in the current phase. This represents the mean value of phase saturation. Indicates the number of phases. Indicates the first Phase saturation.

[0048] Edge computing calculates the saturation of each phase in real time, and determines the small-step increase / decrease mode based on the set upper and lower limits of saturation. The thresholds for the upper and lower limits of saturation are shown in Table 2. Table 2 Thresholds for Setting Upper and Lower Limits of Saturation in Small-Step Increment / Decrease Mode

[0049] Off-peak traffic scenarios often exhibit characteristics such as unstable flow, occasional short-term surges, and no continuous queues. Typically, their control strategies focus on adapting to sudden changes in traffic flow, reducing average vehicle delays, and avoiding wasted green light time. Their incremental adjustment model is as follows: A. When the calculated Sat of all phases is greater than the upper limit of the corresponding saturated flow scenario (Qs), the increment is increased according to the set increment duration_3 (3 seconds).

[0050] B. When the calculated Sat of all phases is less than the lower limit of the corresponding saturated flow scenario (Qs), then the reduction is performed according to the set reduction duration of duration_6 (6 seconds) or duration_4 (4 seconds). (The remaining text appears to be incomplete and possibly contains errors.) The phase reduction of Sat is a longer green light time (such as duration_6), less than or equal to the average. The phase of Sat is reduced by less green light time (duration_4).

[0051] C. When the calculated Sat falls within the range of [lower saturation limit, upper saturation limit], then for Sat≤ The phase of Sat is appropriately reduced, for Sat > The phase of Sat maintains the current green light duration, and saturation is calculated once every 2 cycles to avoid missing traffic fluctuations.

[0052] D. When the calculated Sat of some phases is greater than the upper limit of the corresponding saturated flow scenario (Qs) while Sat of some phases is less than or equal to the upper limit of the corresponding saturated flow scenario (Qs), i.e., there is a phenomenon that requires simultaneous mixed increment and decrement, then for the phases where Sat is greater than the upper limit, increment is performed according to rule A. For the phases where Sat is less than or equal to the upper limit of the corresponding saturated flow scenario (Qs), there are two cases: first, if it is less than the lower limit of the corresponding saturated flow scenario (Qs), then decrement is performed according to rule B; second, if it is within the range of [lower limit of saturation, upper limit of saturation], then remain unchanged according to rule C.

[0053] The final constraint should ensure that the overall increase or decrease in green light duration is consistent with the increase or decrease in total reduction duration, so as to avoid changes in cycle duration.

[0054] E. When the calculated phase Sat is within the range of [lower saturation limit, upper saturation limit] and the phase Sat is less than the lower limit of the corresponding saturated flow scenario (Qs), that is, when there is a phenomenon that requires simultaneous mixed maintenance and reduction, the phases with Sat less than the lower limit of the corresponding saturated flow scenario (Qs) are reduced according to rule B, and the phases within the range of [lower saturation limit, upper saturation limit] are kept unchanged according to rule C.

[0055] F. If abnormal data is detected from the integrated radar screen, the small step increase / decrease algorithm is paused, and the fixed signal timing scheme is switched to the fixed signal timing scheme.

[0056] Peak traffic conditions typically exhibit characteristics such as concentrated and continuous flow, occasional short-term surges, and no sustained queues. Control strategies generally focus on reducing queue length, increasing capacity on main roads, and controlling average vehicle delay. When all phases have high saturation, queue length reduction is necessary, and the designated green light duration for each phase needs to be extended. In this case, if: A. If no reverse radar-monitoring unit is currently installed at the intersection, meaning the queue length indicator cannot be obtained, then: a. When the calculated Sat is greater than the upper limit of the corresponding saturated flow scenario (Qs), then according to the set combination of morning and evening peak hours, duration_7 (7 seconds) and duration_5 (5 seconds), the phases with a green light time greater than the average value are increased by a longer green light time (such as duration_7), and the phases with a green light time less than or equal to the average value are increased by a smaller green light time (such as duration_5).

[0057] b. When the calculated Sat is less than the lower limit of the corresponding saturated flow scenario (Qs), the flow will be reduced according to the set reduction duration_3 (3 seconds).

[0058] c. The remaining situations are consistent with the small-step increase / decrease pattern in off-peak scenarios, the difference being the amount of increase or decrease, which will not be elaborated on further.

[0059] B. Since a reverse radar-monitoring unit is currently installed at the intersection, and queue length indicators are obtained, the incremental calculation of queue length is considered to strive for a one-step increase in green light time. The specific calculation formula is as follows: ; In the formula, Indicates the number of vehicles in the queue. This represents the average headway of the j-th lane in the i-th cycle. This indicates the increase in queue size, i.e., the additional number of vehicles to be evacuated. This indicates the green light increment duration parameter. The saturation flow rate of the lane is represented; based on the calculated value of Δg, the lanes are categorized to ensure that queuing is dispersed in one step and meets the green light duration constraint, as shown in Table 3. Table 3 Green Light Increase Table

[0060] Nighttime scenarios often exhibit traffic flow characteristics such as sparse traffic, fewer pedestrians and non-motorized vehicles, and drivers' sensitivity to waiting time. The control strategy focuses on ensuring traffic efficiency under low traffic conditions and reducing equipment energy consumption, and mainly adopts inductive control mode.

[0061] A. Main road is always green Based on nearly one month of nighttime 5-minute statistical data, the radar-guided all-in-one machine identifies roads with a traffic flow rate (Yi) of ≥60% as main roads. The straight-ahead phase of the main road remains constantly green. If the radar-guided all-in-one machine does not detect any vehicles on the main road for several consecutive cycles, the green light on the main road switches to "flashing mode," and returns to constant green within 1 second after a vehicle is detected.

[0062] B. Secondary induction triggering Vehicle detection employs dual virtual detection lines, with one line at 50m and another at 100m from the secondary road entrance. A valid arrival is determined when the integrated radar-view camera detects at least one vehicle. Pedestrian detection uses a 2m x 5m rectangular area at the secondary road stop line. A valid request is determined when both the spherical camera and the integrated radar-view camera detect at least two pedestrians. The green light duration for the secondary road is then calculated using the following formula: ; In the formula, Indicates the duration of the green light for the next road. This indicates the number of vehicles detected on each road. During this process, the delay from the detection of a valid demand by the radar-visual integrated machine to the green light turning on must meet a threshold of ≤2 seconds to avoid vehicles waiting for a long time.

[0063] C. Pedestrian crossing priority Regardless of whether there are vehicles on the main road, if both the integrated radar-sensing camera and the spherical camera detect at least one pedestrian waiting at the stop line, the pedestrian crossing phase will be immediately triggered. The green light duration is adapted to the crossing length, as shown in Table 4. Table 4 Green Light Combined with Pedestrian Crossing Length Adaptation Table

[0064] In addition, at intersections with a central safety island (width ≥ 2m), the pedestrian green light is divided into two segments, each lasting 10 seconds, with a 3-second red interval in between to ensure pedestrians can reach their destination safely.

[0065] S6, based on the signal timing instruction, drive the traffic lights to perform phase sequence switching and green light duration adjustment, and upload the phase-level traffic flow index and the signal timing instruction to the cloud for global optimization based on multi-intersection data; Specifically, step S6 includes steps S61 to S62: S61 uses the Webster algorithm to calculate the basic value of the cycle length and adjusts the phase switching order according to the current scene to obtain the phase sequence optimization result; S62, globally optimize the signal timing command based on edge computing, the phase sequence optimization results, multi-intersection data, lane configuration, and traffic distribution characteristics; It should be noted that the signal timing optimization algorithm needs to achieve coordinated optimization of cycle length, phase sequence, and green light duration to ensure that each parameter is well-suited to traffic flow characteristics and scene objectives. Initially, the Webster algorithm is used to calculate the baseline value of the cycle length and to dynamically adjust it based on real-time traffic flow, balancing theoretical optimality with practical operating conditions. The calculation formula is as follows: ; In the formula, Indicates the basic period. Indicates the total lost time. This indicates the ratio of flow to total. Indicates the startup time loss. Indicates the yellow light time. Indicates the time when all colors are red. Indicates the first Phase flow ratio, Indicates the first Lane-level phase-level flow, Indicates the first The saturation flow of each lane; If the calculated Y > 0.95, indicating a high degree of lane oversaturation within the cycle, then the Webster algorithm is considered to have a large calculation error. In this case, a correction is performed, and the calculation expression is as follows: ; In the formula, This indicates the corrected cycle length. As the arrival flow increases, the cycle length increases to reduce the number of cycles. Correspondingly, this reduces the start-up loss time and phase transition time of each cycle to improve green light utilization and throughput while reducing queue length. If radar data is missing, making it impossible to calculate Y, historical cycle values ​​from the same period are used as a substitute. If the missing duration exceeds a threshold, the system automatically switches to a preset fixed timing. Furthermore, if a phase saturation > 1.0 is detected after cycle adjustment, the cycle is not extended further; instead, the green light duration for that phase is extended until the saturation < 0.9.

[0066] Phase sequence optimization revolves around the logic of adapting to intersection scenarios, improving traffic efficiency, and ensuring traffic safety. First, the phase type is determined based on the intersection type and traffic flow characteristics. Then, the switching sequence is optimized based on time-of-day demand. Finally, control strategies are formulated for conflict points, forming a complete phase sequence optimization system. The selection of phase types must closely match the intersection form, lane configuration, and traffic flow distribution characteristics to ensure that phase settings are deeply aligned with actual traffic needs. The applicable scenarios and constraints for different phase types are shown in Table 5. Table 5 Phase Type Selection Table

[0067] After determining the phase type, the switching sequence needs to be optimized based on the core traffic flow demands at different times, ensuring that limited travel time is allocated to phases with high demand. For off-peak scenarios, traffic flow is mainly characterized by random fluctuations; therefore, it should be optimized by prioritizing high-volume phases to reduce wasted green light time. For peak scenarios, traffic flow is mainly characterized by concentrated queuing; therefore, it should be optimized by prioritizing main roads and queuing lengths to disperse long queues and prevent congestion from spreading. For nighttime scenarios, traffic flow is mainly characterized by sparse passage and rapid transit; therefore, it should be ordered by keeping the main road green light constantly and triggering secondary road phases on demand. The main road straight-through phase remains green, while secondary road phases are only triggered when vehicles / pedestrians are detected. After triggering, the secondary road straight-through phase is executed in the order of secondary road left turn, and after execution, the main road green light is restored to reduce vehicle waiting time.

[0068] Optimizing green light duration should follow the logic of basic allocation, dynamic balancing, and special adaptation. First, a basic duration is calculated based on the traffic flow ratio. Then, balancing is achieved through fine-tuning of saturation. Finally, adjustments are prioritized for special scenarios to ensure that the green light duration accurately matches traffic flow and safety requirements. The specific calculation expression is as follows: ; In the formula, Indicates the first The basic green light duration for each cycle This indicates the effective green light time within the cycle. The calculation result is rounded to the nearest integer. If the result is less than 9 seconds, it does not meet the minimum pedestrian crossing requirement, so it is taken as 9 seconds. If the result is greater than 40% of the cycle length, it is taken as 40% of the cycle length to avoid the green light of a single phase being too long and encroaching on the time of other phases. The basic green light duration can only meet the static flow ratio requirements. In actual operation, due to traffic flow fluctuations, the saturation of each phase may differ. Dynamic fine-tuning is needed to achieve saturation balance and avoid insufficient green light in some phases and wasted green light in others.

[0069] In summary, the AI ​​traffic signal control method based on multi-source perception and edge computing in the above embodiments of the present invention is noteworthy in that the cloud management platform downloads the control strategy, algorithm, and parameters to the edge computing end, and then the edge computing end runs the AI ​​traffic signal control; the actual effects of the AI ​​traffic signal control platform based on radar-visual integrated machine and edge computing in improving intersection traffic efficiency, reducing queue length, reducing average vehicle delay, and ensuring the safety of motorized and non-motorized traffic are accurately realized and quantitatively optimized. First, the system fully considers the factors affecting the effectiveness of traffic signal control. Addressing issues in existing technologies such as susceptibility of sensing devices to environmental interference, data processing lag, and disconnect between control strategies and scenarios, it utilizes a radar-visual integrated machine to collect multi-dimensional parameters including lane-specific PCU traffic flow, maximum queue length per cycle, and pedestrian and non-motorized vehicle crossing status. Combined with edge computing, it performs real-time anomaly data processing to ensure the accuracy and availability of core data. Second, it constructs a scientific mathematical model, establishes a scenario judgment system based on multiple formulas, and designs a complete implementation process for scenario-specific AI adaptive control algorithms and signal timing optimization algorithms. Finally, it quantitatively evaluates the effectiveness of traffic signal control, avoiding the problems of traditional control methods that rely on manual experience to set timing, are highly subjective, have weak ability to cope with sudden changes in traffic flow and oversaturated queues, and lack sufficient unidirectional transmission and coordination among systems.

[0070] Example 2 Please see Figure 3 The figure shows an AI traffic signal control system based on multi-source perception and edge computing according to a second embodiment of the present invention. The system includes: The data acquisition module 10 is used to collect multi-source traffic flow data at a preset period and generate multiple types of basic data tables based on the multi-source traffic flow data. Calculation module 20 is used to calculate the lane-specific PCU flow rate and the maximum queue length of each lane based on the aforementioned multi-class basic data tables. Aggregation module 30 is used to aggregate single-lane data into phase-level traffic flow indicators based on the lane-specific PCU flow rate, the maximum queue length of the lane, the intersection phase scheme, the number of approach lanes, and the turning relationships of each lane. The expression for the aggregated phase-level traffic flow indicators is as follows: ; In the formula, Indicates the total phase flow. This indicates the number of lanes included in that phase. Indicates the first The flow rate of each lane's PCU. Indicates the maximum queue length of the phase. , ... These represent the maximum queue length of lane 1, the maximum queue length of lane 2, and the maximum queue length of lane 3, respectively. Maximum queue length for each lane in a single lane; The determination module 40 is used to determine the current scenario based on edge computing and the phase-level traffic flow index, with the signal cycle as the determination cycle, combined with time period division and working condition characteristics. The execution module 50 is used to match the corresponding control strategy according to the determined current scene, and calculate the saturation of each phase based on the edge computing, the total phase flow, the phase saturation flow, the current cycle, and the current green light duration. It then executes the corresponding control strategy based on the phase saturation, the current scene, and the maximum phase ranking length, and generates signal timing instructions. The expression for calculating the saturation of each phase is: ; In the formula, Indicates phase saturation. Indicates the duration of the current period. Indicates phase saturation flow rate, Indicates the first Each phase saturation flow, Indicates the duration of the green light in the current phase. This represents the mean value of phase saturation. Indicates the number of phases. Indicates the first Phase saturation; The optimization module 60 is used to drive the traffic lights to perform phase sequence switching and green light duration adjustment based on the signal timing instructions, and upload the phase-level traffic flow indicators and the signal timing instructions to the cloud for global optimization based on multi-intersection data.

[0071] In some alternative embodiments, the acquisition module 10 includes: The settings unit is used to set parameters for the Rayvision all-in-one machine via the cloud. The acquisition and construction unit is used to acquire multi-source traffic flow data collected by the radar-visual integrated machine at a preset period, generate multiple types of basic data tables based on the multi-source traffic flow data, and construct two types of statistical data tables based on control requirements.

[0072] In some alternative embodiments, the computing module 20 includes: The calculation unit is used to obtain the start and end times of the latest execution plan from the signal timing scheme information table, accumulate the number of vehicles within the start and end times of the execution, and calculate the PCU flow rate and the maximum queue length of each lane respectively. The calculation and judgment unit is used to calculate the maximum allowable flow rate of the cycle based on the lane saturation flow rate and the green light duration, and to determine whether the flow rate of the PCU in each lane is greater than the maximum allowable flow rate of the cycle. If so, it is determined to be a data anomaly, so as to verify the calculated flow rate of the PCU in each lane. The calculation expression for the lane-specific PCU traffic flow is as follows: ; In the formula, This indicates the PCU flow rate for each lane. , , , These represent the number of small vehicles, medium-sized vehicles, heavy vehicles, and electric vehicles counted within the period, respectively. The formula for calculating the maximum queue length for each lane is: ; In the formula, This indicates the maximum queue length for each lane. Indicates the detection length of the Ravec integrated machine. This indicates the distance between the location of the integrated radar monitor and the parking line. Indicates the length of the blind zone illuminated by the integrated radar vision system; The expression for calculating the maximum allowable flow rate per period is: ; In the formula, Indicates the maximum allowed flow rate during the period. Indicates lane saturation flow. Indicates the duration of the green light. Indicates the supersaturation coefficient. Indicates the duration of the period.

[0073] In some alternative embodiments, the optimization module 60 includes: The calculation and adjustment unit is used to calculate the basic value of the cycle length using the Webster algorithm and adjust the phase switching order according to the current scene to obtain the phase sequence optimization result; The optimization unit is used to perform global optimization of the signal timing command based on edge computing, the phase sequence optimization results, multi-intersection data, lane configuration, and traffic distribution characteristics.

[0074] The functions or operation steps implemented by the above modules and units are largely the same as those in the above method embodiments, and will not be repeated here.

[0075] The AI ​​traffic signal control system based on multi-source perception and edge computing provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0076] Example 3 The third embodiment of the present invention also proposes an electronic device, please refer to [link / reference]. Figure 4 The image shows an electronic device according to a third embodiment of the present invention.

[0077] The electronic device may include a processor 71 and a memory 72 storing computer program instructions.

[0078] Specifically, the processor 71 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement this application.

[0079] The memory 72 may include a mass storage device for data or instructions. For example, and not limitingly, the memory 72 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 72 may include removable or non-removable (or fixed) media. Where appropriate, the memory 72 may be internal or external to a data processing device. In a particular embodiment, the memory 72 is non-volatile memory. In a particular embodiment, the memory 72 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0080] The memory 72 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 71.

[0081] The processor 71 reads and executes the computer program instructions stored in the memory 72 to implement the AI ​​traffic signal control method based on multi-source perception and edge computing described in Embodiment 1 above.

[0082] In some embodiments, the electronic device may further include a communication interface 73 and a bus 70. For example, Figure 4 As shown, the processor 71, memory 72, and communication interface 73 are connected through bus 70 and complete communication with each other.

[0083] The communication interface 73 is used to enable communication between the various modules, devices, units, and / or equipment in this application. The communication interface 73 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0084] Bus 70 includes hardware, software, or both, that couples the components of a device together. Bus 70 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 70 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 70 may include one or more buses. Although this application describes and illustrates a specific bus, this application considers any suitable bus or interconnection.

[0085] The electronic device can acquire an AI traffic signal control system based on multi-source perception and edge computing, and execute the AI ​​traffic signal control method based on multi-source perception and edge computing in this embodiment.

[0086] Furthermore, in conjunction with the AI ​​traffic signal control method based on multi-source perception and edge computing in Embodiment 1 above, this application can provide a storage medium for implementation. This storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement the AI ​​traffic signal control method based on multi-source perception and edge computing in Embodiment 1 above.

[0087] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. An AI-based traffic signal control method based on multi-source sensing and edge computing, characterized in that, The method includes: Collect multi-source traffic flow data at a preset cycle, and generate multiple types of basic data tables based on the multi-source traffic flow data; Calculate the PCU flow rate and maximum queue length for each lane based on the aforementioned multiple basic data tables; Based on the lane-specific PCU flow rate, the maximum queue length of the lane, the intersection phase scheme, the number of lanes at the approach, and the turning relationship of each lane, the single-lane data is aggregated into a phase-level traffic flow index. Based on edge computing and the aforementioned phase-level traffic flow index, and using the signal cycle as the determination cycle, the current scenario is determined by combining time period division and working condition characteristics. The corresponding control strategy is matched according to the current scene as determined, and the saturation of each phase is calculated based on the total phase flow, combined with the phase saturation flow, the current cycle, and the current green light duration, based on the edge computing. The corresponding control strategy is executed according to the phase saturation, the current scene, and the maximum rank length of the phase, and a signal timing instruction is generated. Based on the signal timing instructions, the traffic lights are driven to perform phase sequence switching and green light duration adjustment, and the phase-level traffic flow indicators and the signal timing instructions are uploaded to the cloud for global optimization based on multi-intersection data.

2. The AI ​​traffic signal control method based on multi-source perception and edge computing according to claim 1, characterized in that, The steps of collecting multi-source traffic flow data at a preset period and generating multiple types of basic data tables based on the multi-source traffic flow data include: Parameter settings for the Rayvision all-in-one machine are performed via the cloud. The system acquires multi-source traffic flow data collected by the integrated radar-visual system at a preset cycle, generates multiple types of basic data tables based on the multi-source traffic flow data, and constructs two types of statistical data tables based on control requirements.

3. The AI ​​traffic signal control method based on multi-source perception and edge computing according to claim 1, characterized in that, The steps for calculating the lane-specific PCU traffic flow and the maximum queue length for each lane based on the aforementioned multiple types of basic data tables include: Obtain the start and end times of the latest execution plan from the signal timing scheme information table, accumulate the number of vehicles within the start and end times, and calculate the PCU flow rate and maximum queue length of each lane respectively. The maximum allowable flow rate for a given period is calculated based on the lane saturation flow rate and the green light duration. It is then determined whether the flow rate of the PCU in each lane is greater than the maximum allowable flow rate for that period. If so, the data is considered abnormal, and the calculated flow rate of the PCU in each lane is verified.

4. The AI ​​traffic signal control method based on multi-source perception and edge computing according to claim 3, characterized in that, The formula for calculating the traffic flow of the PCU in each lane is: ； In the formula, This indicates the PCU flow rate for each lane. , , , These represent the number of small vehicles, medium-sized vehicles, heavy vehicles, and electric vehicles counted within the period, respectively. The formula for calculating the maximum queue length for each lane is: ； In the formula, This indicates the maximum queue length for each lane. Indicates the detection length of the Ravec integrated machine. This indicates the distance between the location of the integrated radar monitor and the parking line. Indicates the length of the blind zone illuminated by the integrated radar vision system; The expression for calculating the maximum allowable flow rate per period is: ； In the formula, Indicates the maximum allowed flow rate during the period. Indicates lane saturation flow. Indicates the duration of the green light. Indicates the supersaturation coefficient. Indicates the duration of the period.

5. The AI ​​traffic signal control method based on multi-source perception and edge computing according to claim 1, characterized in that, The expression for the aggregation into phase-level traffic flow indicators is: ； In the formula, Indicates the total phase flow. This indicates the number of lanes included in that phase. Indicates the first The flow rate of each lane's PCU. Indicates the maximum queue length of the phase. , ... These represent the maximum queue length of lane 1, the maximum queue length of lane 2, and the maximum queue length of lane 3, respectively. The maximum queue length for each lane of a single lane.

6. The AI ​​traffic signal control method based on multi-source perception and edge computing according to claim 1, characterized in that, The expression for calculating the saturation of each phase is as follows: ； In the formula, Indicates phase saturation. Indicates the duration of the current period. Indicates phase saturation flow rate, Indicates the first Each phase saturation flow, Indicates the duration of the green light in the current phase. This represents the mean value of phase saturation. Indicates the number of phases. Indicates the first Phase saturation.

7. The AI ​​traffic signal control method based on multi-source perception and edge computing according to claim 1, characterized in that, The steps for global optimization based on multi-intersection data include: The Webster algorithm is used to calculate the basic value of the cycle length, and the phase switching order is adjusted according to the current scene to obtain the phase sequence optimization result; The signal timing command is globally optimized based on edge computing, the phase sequence optimization results, multi-intersection data, lane configuration, and traffic distribution characteristics.

8. An AI traffic signal control system based on multi-source sensing and edge computing, characterized in that, The system includes: The data acquisition module is used to collect multi-source traffic flow data at a preset period and generate multiple types of basic data tables based on the multi-source traffic flow data. The calculation module is used to calculate the PCU flow rate and the maximum queue length of each lane based on the aforementioned multiple basic data tables. The aggregation module is used to aggregate single-lane data into phase-level traffic flow indicators based on the PCU flow of each lane, the maximum queue length of each lane, the intersection phase scheme, the number of lanes at the approach, and the turning relationship of each lane. The determination module is used to determine the current scenario based on edge computing and the phase-level traffic flow index, with the signal cycle as the determination cycle, combined with time period division and working condition characteristics. The execution module is used to match the corresponding control strategy according to the determined current scene, and calculate the saturation of each phase based on the edge computing according to the total phase flow, combined with the phase saturation flow, the current cycle, and the current green light duration. It also executes the corresponding control strategy according to the phase saturation, the current scene, and the maximum rank length of the phase, and generates signal timing instructions. The optimization module is used to drive the traffic lights to perform phase sequence switching and green light duration adjustment based on the signal timing instructions, and upload the phase-level traffic flow indicators and the signal timing instructions to the cloud for global optimization based on multi-intersection data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the AI ​​traffic signal control method based on multi-source perception and edge computing as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the AI ​​traffic signal control method based on multi-source perception and edge computing as described in any one of claims 1 to 7.

Images