A signal control system based on traffic confidence

Abstract

This invention provides a traffic flow confidence-based signal control system, including a traffic flow confidence calculation module. This module calculates a basic confidence level based on the number of abnormal headway distances and overspeeding incidents in the current period's traffic flow data. It then corrects the basic confidence level based on historical traffic flow data, current date attributes, and traffic flow data from spatial adjacencies to obtain a final traffic flow confidence level. The traffic flow confidence calculation module inputs this final confidence level to a signal control module, which then adopts different signal control strategies based on the final confidence level and the current traffic flow data. By introducing the "traffic flow confidence level" index, the system intelligently decides on the activation, degradation, and fault tolerance strategies of the signal control scheme, achieving real-time optimization driven by high-confidence data, steady-state control in medium-confidence scenarios, and automatic degradation and safety fallback in low-confidence states.

Description

Technical Field

[0001] This invention belongs to the field of traffic signal control technology, and particularly relates to a signal control system based on traffic flow confidence. Background Technology

[0002] Traffic flow-based traffic signal control is an intelligent traffic management method that dynamically adjusts traffic signal timing schemes by real-time monitoring and analysis of the number and flow of traffic participants such as vehicles and pedestrians on the road. This aims to optimize traffic flow, reduce congestion, improve road efficiency, and ensure traffic safety. Various sensor devices, such as cameras, geomagnetic sensors, and radar, are used to collect real-time traffic flow data, including information on vehicle flow, speed, and queue length in different directions. Based on the data analysis results, and following specific optimization algorithms and strategies, parameters such as green light time, red light time, and phase transition sequence of traffic lights are dynamically adjusted to match the signal timing with actual traffic flow.

[0003] The prior art publication number CN120690017A discloses a traffic signal control system based on traffic flow prediction, belonging to the field of traffic signal control technology. It includes an edge computing unit deployed at each intersection to carry the following modules: a digital twin module, used to provide data and topology information to a graph transformation traffic prediction module; a graph transformation traffic prediction module, which jointly predicts vehicle and pedestrian traffic flow at several adjacent intersections for the next 1-2 minutes using a multi-head self-attention mechanism and location coding; and a local twin simulation module, which uses a simulation time step of ≤5s to simulate the short-term congestion, queue length, and average delay effects of various signal timing strategies at the second level. This invention, by designing an edge digital twin and graph transformation prediction closed loop, achieves short-term multi-intersection traffic joint prediction, overcoming the problems of high latency, low prediction accuracy, and isolated decision-making at a single intersection in centralized cloud computing, improving prediction accuracy and response speed, reducing communication latency, and enabling autonomous decision-making at the edge. Summary of the Invention

[0004] Existing traffic flow-based or predicted traffic flow-based signal control methods do not perform reliability assessments on the collected traffic flow data; they may only remove outliers during data processing. However, in practical applications, data acquisition equipment can be affected by other factors, or unexpected accidents may occur on the road, leading to discrepancies between the collected traffic flow data and the actual situation. In such cases, directly making signal control decisions based on the collected traffic flow data can have adverse effects, exacerbating traffic congestion and negatively impacting road safety.

[0005] To address the aforementioned technical problems, the present invention provides the following technical solution: a signal control system based on traffic flow confidence, comprising a traffic flow confidence calculation module. The traffic flow confidence calculation module calculates a basic confidence based on the number of abnormal headway times and the number of abnormal speeding times in the current period's traffic flow data. It then corrects the basic confidence based on historical traffic flow data, the current date attribute, and traffic flow data from spatial adjacencies to obtain a final traffic flow confidence. The traffic flow confidence calculation module inputs the final traffic flow confidence to a signal control module, which then adopts different signal control strategies based on the final traffic flow confidence and the current traffic flow data.

[0006] Specifically, when the traffic confidence calculation module calculates the basic confidence score, it uses the proportion of abnormal headway and abnormal speeding in the current traffic data as the calculation indicators for the basic confidence score, and uses 100% minus half of the sum of the proportion of abnormal headway and abnormal speeding as the basic confidence score.

[0007] Specifically, the traffic confidence calculation module includes a rule verification layer, a statistical comparison layer, and a spatial collaboration layer. The rule verification layer detects the integrity, legality, and continuity of the current traffic data based on historical traffic data, and corrects the basic confidence based on the degree of anomaly. The statistical comparison layer detects the historical deviation, volatility, and whether there are any sudden changes in the current traffic data based on the current date attribute, and corrects the basic confidence. The spatial collaboration layer verifies the upstream and downstream traffic differences and adjacent intersections of the current traffic data based on spatial adjacency point traffic data, and corrects the basic confidence.

[0008] Specifically, the traffic confidence calculation module also includes a fusion scoring layer, which applies different weights to the correction results of the rule verification layer, statistical comparison layer and spatial coordination layer according to the current traffic conditions, and merges them into the final traffic confidence.

[0009] Specifically, the rule verification layer performs data anomaly detection by comparing the rolling traffic statistics of the hour preceding the current traffic data collection with historical data from a sliding seven-day window. It also designs horizontal comparison rules for different types of detection devices to detect whether there are any anomalies in the data and adjusts the basic confidence level to different degrees based on the degree of anomaly.

[0010] Specifically, the statistical comparison layer uses time period characteristics and weekday / rest day / holiday attributes to construct a statistical model. Based on the time, date attributes, and business period characteristics of historical traffic data, it constructs a historical traffic baseline and updates it on a daily basis. Based on the historical traffic baseline with the same time period and date attributes as the current traffic data, it judges whether there is a significant deviation in the traffic flow at the current moment and adjusts the basic confidence level.

[0011] Specifically, the spatial coordination layer identifies anomalies such as missed detections, false detections, and incorrect direction identification by comparing the traffic flow consistency between the current intersection and its spatial neighbors, and corrects the basic confidence level. At the same time, it uses a rolling sliding window with a fixed time interval to match upstream and downstream traffic flow, thus offsetting short-term mismatch problems caused by traffic light control.

[0012] Specifically, when the flow confidence level is greater than or equal to 80%, the signal control module generates a dynamic adaptive signal optimization scheme based on real-time flow. The adjustment range of signal cycle duration, phase green ratio, and phase sequence optimization is widened as the flow changes.

[0013] Specifically, when the flow confidence level is greater than 60% but less than 80%, the signal control module uses historical flow data to generate an adaptive scheme, fine-tuning the signal cycle duration and phase green ratio, without adjusting the phase sequence.

[0014] Specifically, when the traffic confidence level is less than 60%, the signal control module adopts a preset safety plan, using a fixed time period plan or a holiday contingency plan to take over.

[0015] The beneficial effects of this invention are that, by introducing the "flow confidence level" index, it enables intelligent decision-making regarding the activation, degradation, and fault tolerance strategies of signal control schemes, achieving: real-time optimization driven by high-confidence data, steady-state control in medium-confidence scenarios, and automatic degradation and safety fallback in low-confidence states. While ensuring operational safety and efficiency, it allows signal control to "adjust only based on reliable data" and automatically takes protective measures against suspicious data. Attached Figure Description

[0016] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0018] Example 1: A signal control system based on flow confidence, such as Figure 1 As shown, the system includes a traffic flow confidence calculation module. This module calculates a basic confidence level based on the number of abnormal headway distances and speeding anomalies in the current period's traffic flow data. It then corrects the basic confidence level based on historical traffic flow data, the current date attribute, and traffic flow data from spatial adjacencies to obtain the final traffic flow confidence level. The traffic flow confidence calculation module inputs the final traffic flow confidence level to the signal control module. The signal control module then adopts different signal control strategies based on the final traffic flow confidence level and the current traffic flow data.

[0019] When calculating the base confidence level, half of the sum of the proportions of abnormal headway and abnormal speeding is subtracted from 100% to obtain the base confidence level. The traffic flow confidence calculation module includes a rule validation layer, a statistical comparison layer, and a spatial coordination layer. The rule validation layer, based on historical traffic flow data, checks the completeness, legality, and continuity of the current traffic flow data and corrects the base confidence level based on the degree of anomalies. The statistical comparison layer, based on the current date attribute, checks the historical deviation, volatility, and presence of abrupt changes in the current traffic flow data and corrects the base confidence level. The spatial coordination layer, based on spatially adjacent point traffic flow data, verifies the current traffic flow data by comparing upstream and downstream traffic differences and adjacent intersections, and corrects the base confidence level.

[0020] The rule validation layer corrects the basic confidence level by deducting points for exceeding the traffic limit, deducting points for traffic deviation comparisons from multiple data sources, and judging abnormal traffic flow at night. The correction for exceeding the traffic limit mainly involves comparing the lane traffic flow in the traffic data with the lane saturation flow and the lane's historical traffic limit. When the lane traffic flow exceeds the smaller of the lane saturation flow and the lane's historical traffic limit, the basic confidence level is corrected once; in this embodiment, the first correction is a reduction of 5% from the basic confidence level. When the lane traffic flow exceeds both the lane saturation flow and the lane's historical traffic flow, the basic confidence level is corrected a second time; in this embodiment, the second correction is a reduction of 20% from the basic confidence level.

[0021] If multiple data sources are used when collecting traffic data, the traffic deviation ratio of the multiple data sources is judged. When the current traffic is >0, the traffic data collected by the current data source is compared with the traffic data collected by the next priority data source. If the traffic deviation ratio is greater than or equal to 50% and the traffic difference is greater than 5 units, the basic confidence level is corrected three times. In this embodiment, the three corrections are specifically the basic confidence level -20%.

[0022] Meanwhile, the rule validation layer also determines whether there is abnormal traffic late at night. If the current time is not late at night (6:00-22:00) but the current traffic is 0, it is determined that the traffic data is abnormal and the basic confidence level is set to 0; if the current time is late at night (22:00-6:00) and the current traffic is 0, it is determined that there may be abnormal data and the basic confidence level is set to 50%.

[0023] When comparing current traffic data collected by the rule verification layer, "rolling traffic statistics from the previous hour" is used instead of the average over a fixed time period. This effectively suppresses short-term fluctuations at the 5-minute level and more accurately identifies persistent anomalies (such as prolonged periods of zero traffic or extremely high traffic). Simultaneously, by differentiating anomalies of varying severity, a hierarchical confidence level correction dynamically reflects the degree of anomaly impact, avoiding over-penalization or over-leniency. Historical traffic data is not compared using a fixed threshold but is automatically adjusted based on the highest sliding traffic flow over the past 7 days. This allows for real-time convergence with changes in traffic conditions, preventing false alarms caused by pattern shifts (such as holidays or construction). Through data source traffic deviation ratios, horizontal comparison rules are designed for various types of detection equipment (inductive loops, radar, video detection, etc.) present on actual urban roads. Consistency assessments are performed across devices according to data source priority, resulting in quantifiable "cross-source confidence," enhancing the robustness and reliability of anomaly identification. Overall, the rule verification layer's data anomaly detection covers typical scenarios such as zero traffic outside of late night, lane saturation traffic limits, historical limit breaches, significant differences between multiple sources, insufficient headway, and speeding anomalies. Each type of anomaly has defined triggering conditions and processing logic, demonstrating good scalability and engineering feasibility. In specific implementation, this embodiment's rule verification layer also designs a unified anomaly detection rule table, which hierarchically corrects the basic confidence level based on the current traffic data according to the anomaly detection rule table.

[0024] The statistical comparison layer utilizes "time period characteristics + weekday / rest day / holiday attributes" to construct a statistical model to determine whether there is a significant deviation (i.e., "not meeting expectations") in the current traffic flow and outputs a confidence score. The statistical model comprises three main modules: an input data module, a historical data modeling module, and a real-time scoring module. The input data module splits the current traffic information into specific fields, such as intersection ID / direction, current timestamp, current traffic value, date attribute, day of the week, time slice, and business period (high / off-peak / evening / night). The historical data modeling module constructs a historical traffic baseline based on historical traffic data. Under each "intersection + direction + time slice" dimension, it statistically analyzes the characteristics of the historical traffic data, including time slice, day of the week, date attribute, traffic statistics, and business period category. The historical traffic baseline is updated daily (using a 15-minute sliding window before and after the same moment in the last 30 / 60 days), while excluding the influence of outlier data (confidence below 50%). The traffic statistics of the historical traffic baseline include the mean μ, standard deviation σ, P90 / P10, minimum, and maximum values.

[0025] The real-time scoring module extracts historical traffic baselines with the same intersection, direction, time slice, and date type as the current traffic from the historical data modeling module based on the current traffic information input from the input data module. It compares the two traffic statistics and calculates the historical deviation confidence score based on the current traffic value, historical mean, historical standard deviation, and standardized deviation. The module then corrects the basic confidence score based on the historical deviation confidence score.

[0026] The statistical comparison layer is constructed using a multi-dimensional historical statistical model. Compared to traditional baseline models that only count by the hour or the whole day, it introduces multi-dimensional fine-grained modeling, capturing minute fluctuations at the time slice level (5 minutes). Further subdivision based on date type (weekday / restday) + day of the week ensures accurate matching of historical comparisons and significantly improves the contextual rationality of traffic anomaly identification. Simultaneously, through a sliding window and daily update mechanism, historical statistics use a sliding window of nearly 60 days, automatically updated daily, removing low-confidence data. This dynamically evolving historical baseline allows the model to continuously adapt to seasonality, construction adjustments, or changes in travel patterns, possessing self-learning characteristics.

[0027] The spatial coordination layer identifies the following anomalies by comparing the traffic flow consistency between the current intersection and its spatial neighbors (upstream and downstream, same road segment, adjacent directions): High upstream flow → Current flow is close to 0 (suspected missed detection); High current traffic → No downstream traffic (suspected false alarm); The current direction is much higher than the adjacent direction (suspected direction identification error).

[0028] When the upstream flow rate of the current road segment is detected to be >200 (indicating significant inflow), but the current flow rate of the current road segment is less than 20 vehicles (almost 0), it is considered a missed detection scenario (high upstream flow rate, low current flow rate). This indicates that the current device may not have identified the expected traffic flow, and the basic confidence level is corrected by multiplying it by 0.4.

[0029] When the current traffic of the current road segment is detected to be more than twice the upstream traffic (current / upstream > 3.0) and (upstream traffic > 20), it indicates that the current value is abnormally large compared to the upstream, which may be due to reverse false alarm or collection misalignment. The basic confidence score is then corrected by multiplying it by 0.6.

[0030] When the current traffic flow is detected to be >200, but the downstream traffic flow is less than 50, it indicates a false alarm scenario (high current traffic flow, low downstream traffic flow), suggesting that vehicles have "nowhere to go" and that the traffic flow is an unreasonable surge. The baseline confidence level is then adjusted by multiplying it by 0.5. Alternatively, if the current traffic flow is more than twice the downstream traffic flow (current / downstream > 3.0) and the downstream traffic flow is >20, it still indicates a false alarm scenario, suggesting the data may be inflated. The baseline confidence level is then adjusted by multiplying it by 0.7.

[0031] The spatial coordination layer introduces spatial continuity constraints, enabling the identification of abnormal flow directions and missed detections through upstream and downstream linkages, making confidence assessments more consistent with the logical structure of urban road networks. Simultaneously, when comparing the flow rates of spatially adjacent electrical systems, the concept of "travel delay compensation" is introduced, employing a rolling 15-minute sliding window to match upstream and downstream flow rates. This effectively offsets short-term mismatches caused by traffic light control, achieving true flow continuity identification across phases.

[0032] The traffic confidence calculation module also includes a fusion scoring layer, which applies different weights to the correction results of the rule verification layer, statistical comparison layer, and spatial coordination layer based on the current traffic conditions, and merges them into the final traffic confidence score. The fusion scoring layer uses dynamic weight fusion based on rules, switching weights according to business rules. During peak periods, the statistical comparison layer is emphasized, while during periods of downtime, the rule verification layer is emphasized.

[0033] The signal control module adopts different signal control strategies based on the final traffic confidence level and current traffic data. This embodiment is applied to adaptive signal control and coordinated control scenarios for urban arterial roads and intersections. By introducing the "traffic confidence level" index, it makes intelligent decisions on the activation, degradation, and fault tolerance strategies of signal control schemes, achieving real-time optimization driven by high-confidence data, steady-state control in medium-confidence scenarios, and automatic degradation and safety fallback in low-confidence states. The system design goal is to ensure operational safety and efficiency while allowing signal control to "adjust only based on reliable data" and automatically take protective measures for suspicious data. When the traffic confidence level is greater than or equal to 80%, the signal control module generates a dynamic adaptive signal optimization scheme based on real-time traffic. The adjustment range of signal cycle duration, phase green ratio, and phase sequence optimization is widened to follow the adjustable range of traffic changes. When the traffic confidence level is greater than 60% but less than 80%, the signal control module uses historical traffic data to generate an adaptive scheme, fine-tuning the signal cycle duration and phase green ratio, without adjusting the phase sequence. When the traffic confidence level is less than 60%, the signal control module adopts a preset safety plan, using a fixed time period plan or a holiday contingency plan to take over.

[0034] Example 2: A signal control system based on traffic flow confidence, including a traffic flow confidence calculation module. The traffic flow confidence calculation module calculates a basic confidence based on the number of abnormal headway times and the number of abnormal speeding times in the current period traffic flow data, and corrects the basic confidence based on historical traffic flow data, current date attributes, and traffic flow data of spatial adjacencies to obtain a final traffic flow confidence. The traffic flow confidence calculation module inputs the final traffic flow confidence to the signal control module, and the signal control module adopts different signal control strategies based on the final traffic flow confidence and the current traffic flow data.

[0035] When calculating the base confidence level, half of the sum of the proportions of abnormal headway and abnormal speeding is subtracted from 100% to obtain the base confidence level. The traffic flow confidence calculation module includes a rule validation layer, a statistical comparison layer, and a spatial coordination layer. The rule validation layer, based on historical traffic flow data, checks the completeness, legality, and continuity of the current traffic flow data and corrects the base confidence level based on the degree of anomalies. The statistical comparison layer, based on the current date attribute, checks the historical deviation, volatility, and presence of abrupt changes in the current traffic flow data and corrects the base confidence level. The spatial coordination layer, based on spatially adjacent point traffic flow data, verifies the current traffic flow data by comparing upstream and downstream traffic differences and adjacent intersections, and corrects the base confidence level.

[0036] The rule validation layer corrects the basic confidence level by deducting points for exceeding the traffic limit, deducting points for traffic deviation comparisons from multiple data sources, and judging abnormal traffic flow at night. The correction for exceeding the traffic limit mainly involves comparing the lane traffic flow in the traffic data with the lane saturation flow and the lane's historical traffic limit. When the lane traffic flow exceeds the smaller of the lane saturation flow and the lane's historical traffic limit, the basic confidence level is corrected once; in this embodiment, the first correction is a reduction of 5% from the basic confidence level. When the lane traffic flow exceeds both the lane saturation flow and the lane's historical traffic flow, the basic confidence level is corrected a second time; in this embodiment, the second correction is a reduction of 20% from the basic confidence level.

[0037] If multiple data sources are used when collecting traffic data, the traffic deviation ratio of the multiple data sources is judged. When the current traffic is >0, the traffic data collected by the current data source is compared with the traffic data collected by the next priority data source. If the traffic deviation ratio is greater than or equal to 50% and the traffic difference is greater than 5 units, the basic confidence level is corrected three times. In this embodiment, the three corrections are specifically the basic confidence level -20%.

[0038] Meanwhile, the rule validation layer also determines whether there is abnormal traffic late at night. If the current time is not late at night (6:00-22:00) but the current traffic is 0, it is determined that the traffic data is abnormal and the basic confidence level is set to 0; if the current time is late at night (22:00-6:00) and the current traffic is 0, it is determined that there may be abnormal data and the basic confidence level is set to 50%.

[0039] When comparing current traffic data collected by the rule verification layer, "rolling traffic statistics from the previous hour" is used instead of the average over a fixed time period. This effectively suppresses short-term fluctuations at the 5-minute level and more accurately identifies persistent anomalies (such as prolonged periods of zero traffic or extremely high traffic). Simultaneously, by differentiating anomalies of varying severity, a hierarchical confidence level correction dynamically reflects the degree of anomaly impact, avoiding over-penalization or over-leniency. Historical traffic data is not compared using a fixed threshold but is automatically adjusted based on the highest sliding traffic flow over the past 7 days. This allows for real-time convergence with changes in traffic conditions, preventing false alarms caused by pattern shifts (such as holidays or construction). Through data source traffic deviation ratios, horizontal comparison rules are designed for various types of detection equipment (inductive loops, radar, video detection, etc.) present on actual urban roads. Consistency assessments are performed across devices according to data source priority, resulting in quantifiable "cross-source confidence," enhancing the robustness and reliability of anomaly identification. Overall, the rule verification layer's data anomaly detection covers typical scenarios such as zero traffic outside of late night, lane saturation traffic limits, historical limit breaches, significant differences between multiple sources, insufficient headway, and speeding anomalies. Each type of anomaly has defined triggering conditions and processing logic, demonstrating good scalability and engineering feasibility. In specific implementation, this embodiment's rule verification layer also designs a unified anomaly detection rule table, which hierarchically corrects the basic confidence level based on the current traffic data according to the anomaly detection rule table.

[0040] The statistical comparison layer utilizes "time period characteristics + weekday / rest day / holiday attributes" to construct a statistical model to determine whether there is a significant deviation (i.e., "not meeting expectations") in the current traffic flow and outputs a confidence score. The statistical model comprises three main modules: an input data module, a historical data modeling module, and a real-time scoring module. The input data module splits the current traffic information into specific fields, such as intersection ID / direction, current timestamp, current traffic value, date attribute, day of the week, time slice, and business period (high / off-peak / evening / night). The historical data modeling module constructs a historical traffic baseline based on historical traffic data. Under each "intersection + direction + time slice" dimension, it statistically analyzes the characteristics of the historical traffic data, including time slice, day of the week, date attribute, traffic statistics, and business period category. The historical traffic baseline is updated daily (using a 15-minute sliding window before and after the same moment in the last 30 / 60 days), while excluding the influence of outlier data (confidence below 50%). The traffic statistics of the historical traffic baseline include the mean μ, standard deviation σ, P90 / P10, minimum, and maximum values.

[0041] The real-time scoring module extracts historical traffic baselines with the same intersection, direction, time slice, and date type as the current traffic from the historical data modeling module based on the current traffic information input from the input data module. It compares the two traffic statistics and calculates the historical deviation confidence score based on the current traffic value, historical mean, historical standard deviation, and standardized deviation. The module then corrects the basic confidence score based on the historical deviation confidence score.

[0042] The statistical comparison layer is constructed using a multi-dimensional historical statistical model. Compared to traditional baseline models that only count by the hour or the whole day, it introduces multi-dimensional fine-grained modeling, capturing minute fluctuations at the time slice level (5 minutes). Further subdivision based on date type (weekday / restday) + day of the week ensures accurate matching of historical comparisons and significantly improves the contextual rationality of traffic anomaly identification. Simultaneously, through a sliding window and daily update mechanism, historical statistics use a sliding window of nearly 60 days, automatically updated daily, removing low-confidence data. This dynamically evolving historical baseline allows the model to continuously adapt to seasonality, construction adjustments, or changes in travel patterns, possessing self-learning characteristics.

[0043] The spatial coordination layer identifies the following anomalies by comparing the traffic flow consistency between the current intersection and its spatial neighbors (upstream and downstream, same road segment, adjacent directions): High upstream flow → Current flow is close to 0 (suspected missed detection); High current traffic → No downstream traffic (suspected false alarm); The current direction is much higher than the adjacent direction (suspected direction identification error).

[0044] When the upstream flow rate of the current road segment is detected to be >200 (indicating significant inflow), but the current flow rate of the current road segment is less than 20 vehicles (almost 0), it is considered a missed detection scenario (high upstream flow rate, low current flow rate). This indicates that the current device may not have identified the expected traffic flow, and the basic confidence level is corrected by multiplying it by 0.4.

[0045] When the current traffic of the current road segment is detected to be more than twice the upstream traffic (current / upstream > 3.0) and (upstream traffic > 20), it indicates that the current value is abnormally large compared to the upstream, which may be due to reverse false alarm or collection misalignment. The basic confidence score is then corrected by multiplying it by 0.6.

[0046] When the current traffic flow is detected to be >200, but the downstream traffic flow is less than 50, it indicates a false alarm scenario (high current traffic flow, low downstream traffic flow), suggesting that vehicles have "nowhere to go" and that the traffic flow is an unreasonable surge. The baseline confidence level is then adjusted by multiplying it by 0.5. Alternatively, if the current traffic flow is more than twice the downstream traffic flow (current / downstream > 3.0) and the downstream traffic flow is >20, it still indicates a false alarm scenario, suggesting the data may be inflated. The baseline confidence level is then adjusted by multiplying it by 0.7.

[0047] The spatial coordination layer introduces spatial continuity constraints, enabling the identification of abnormal flow directions and missed detections through upstream and downstream linkages, making confidence assessments more consistent with the logical structure of urban road networks. Simultaneously, when comparing the flow rates of spatially adjacent electrical systems, the concept of "travel delay compensation" is introduced, employing a rolling 15-minute sliding window to match upstream and downstream flow rates. This effectively offsets short-term mismatches caused by traffic light control, achieving true flow continuity identification across phases.

[0048] The traffic confidence calculation module also includes a fusion scoring layer, which applies different weights to the correction results of the rule verification layer, statistical comparison layer, and spatial coordination layer based on the current traffic conditions, and merges them into the final traffic confidence score. The fusion scoring layer uses dynamic weight fusion based on rules, switching weights according to business rules. During peak periods, the statistical comparison layer is emphasized, while during periods of downtime, the rule verification layer is emphasized.

[0049] The signal control module adopts different signal control strategies based on the final traffic confidence level and current traffic data. This embodiment is applied to adaptive signal control and coordinated control scenarios for urban arterial roads and intersections. By introducing the "traffic confidence level" index, it makes intelligent decisions on the activation, degradation, and fault tolerance strategies of signal control schemes, achieving real-time optimization driven by high-confidence data, steady-state control in medium-confidence scenarios, and automatic degradation and safety fallback in low-confidence states. The system design goal is to ensure operational safety and efficiency while allowing signal control to "adjust only based on reliable data" and automatically take protective measures for suspicious data. When the traffic confidence level is greater than or equal to 80%, the signal control module generates a dynamic adaptive signal optimization scheme based on real-time traffic. The adjustment range of signal cycle duration, phase green ratio, and phase sequence optimization is widened to follow the adjustable range of traffic changes. When the traffic confidence level is greater than 60% but less than 80%, the signal control module uses historical traffic data to generate an adaptive scheme, fine-tuning the signal cycle duration and phase green ratio, without adjusting the phase sequence. When the traffic confidence level is less than 60%, the signal control module adopts a preset safety plan, using a fixed time period plan or a holiday contingency plan to take over.

[0050] The specific signal control is mainly achieved through green wave road coordination control. By adjusting the start time of traffic lights, vehicles encounter green lights consecutively at the prescribed speed, thereby improving traffic efficiency, optimizing the driving experience, and enhancing travel punctuality. The confidence scores of each intersection on the green wave road are aggregated into a road-level confidence score. The specific control strategy is as follows: High confidence interval → Overall coordination: Enable green wave coordination and adaptive calculation, with uplink and downlink calculated separately; Medium confidence interval → Steady state priority + conditional segmentation: Adopt a high confidence coordinated control scheme based on the same historical period. If there is no high confidence scheme, it is downgraded to a low confidence interval. Identify low confidence breakpoints in the path and use local segmentation. Low confidence interval: The scheme is downgraded to a fixed green wave coordinated timing scheme, and green wave coordinated adaptive timing is turned off.

[0051] When the final confidence level of the fusion scoring layer is fused, if a significant negative deviation is found in the statistical comparison layer, it indicates that "the current traffic is significantly lower than the historical average"; and the rule verification layer and spatial coordination layer are normal (the rules are normal and the space is coherent). The signal control module changes the control strategy, without interrupting the adaptive mode, but adjusts to the holiday light load adaptive mode. The specific adjustments are as follows. Cycle adjustment: The upper and lower limits of the cycle duration are appropriately lowered to reduce the idle time on main roads. Green light ratio allocation: Reduce travel time in the main direction and increase travel time in the secondary direction to improve traffic efficiency.

Claims

1. A signal control system based on flow confidence, characterized in that, It includes a traffic confidence calculation module, which calculates the basic confidence based on the number of abnormal headway distances and the number of abnormal speeding in the current period traffic data, and corrects the basic confidence based on historical traffic data, current date attributes and spatial adjacency point traffic data to obtain the final traffic confidence. The flow confidence calculation module inputs the final flow confidence into the signal control module, which then adopts different signal control strategies based on the final flow confidence and the current flow data.

2. The signal control system based on flow confidence as described in claim 1, characterized in that, When calculating the basic confidence level, the traffic confidence calculation module uses the proportion of abnormal headway and abnormal speeding in the current traffic data as the calculation indicators for the basic confidence level, and takes 100% minus half of the sum of the proportion of abnormal headway and abnormal speeding as the basic confidence level.

3. The signal control system based on flow confidence as described in claim 1, characterized in that, The traffic confidence calculation module includes a rule verification layer, a statistical comparison layer, and a spatial coordination layer. The rule verification layer detects the integrity, legality, and continuity of the current traffic data based on historical traffic data, and corrects the basic confidence based on the degree of anomalies. The statistical comparison layer detects the historical deviation, volatility, and whether there are any sudden changes in the current traffic data based on the current date attribute, and corrects the basic confidence. The spatial coordination layer verifies the upstream and downstream traffic differences and adjacent intersections of the current traffic data based on the traffic data of spatial adjacent points, and corrects the basic confidence.

4. The signal control system based on flow confidence according to claim 1 or 3, characterized in that, The traffic confidence calculation module also includes a fusion scoring layer, which applies different weights to the correction results of the rule verification layer, statistical comparison layer and spatial coordination layer according to the current traffic conditions, and merges them into the final traffic confidence.

5. The signal control system based on flow confidence according to claim 3, characterized in that, The rule verification layer collects rolling traffic statistics from the hour preceding the current traffic data collection, compares them with historical data from a sliding seven-day window to detect data anomalies, and designs horizontal comparison rules for different types of detection devices to detect whether there are any anomalies in the data. Based on the degree of anomaly, the basic confidence level is corrected to different degrees.

6. The signal control system based on flow confidence according to claim 3, characterized in that, The statistical comparison layer uses time period characteristics and weekday / rest day / holiday attributes to build a statistical model. Based on the time, date attributes, and business time period characteristics of historical traffic data, it constructs a historical traffic baseline and updates it on a daily basis. Based on the historical traffic baseline with the same time period and date attributes as the current traffic data, determine whether there is a significant deviation in the traffic flow at the current moment, and adjust the basic confidence level accordingly.

7. The signal control system based on flow confidence according to claim 3, characterized in that, The spatial coordination layer identifies anomalies such as missed detections, false detections, and incorrect direction identification by comparing the traffic flow consistency between the current intersection and its spatial neighbors, and corrects the basic confidence level. At the same time, it uses a rolling sliding window with a fixed time interval to match upstream and downstream traffic flow, thus offsetting short-term mismatch problems caused by traffic light control.

8. The signal control system based on flow confidence as described in claim 1, characterized in that, When the flow confidence level is greater than or equal to 80%, the signal control module generates a dynamic adaptive signal optimization scheme based on real-time flow. The adjustment range of signal cycle duration, phase green ratio, and phase sequence optimization is widened as the flow changes.

9. The signal control system based on flow confidence as described in claim 1, characterized in that, When the flow confidence level is greater than 60% but less than 80%, the signal control module uses historical flow data to generate an adaptive scheme, fine-tuning the signal cycle duration and phase green ratio, without adjusting the phase sequence.

10. The signal control system based on flow confidence according to claim 1, characterized in that, When the traffic confidence level is less than 60%, the signal control module adopts a preset safety plan, using a fixed time period plan or a holiday contingency plan to take over.

Images