An intersection signal safety control method based on behavior perception and dynamic scheduling

Abstract

This invention belongs to the field of intelligent transportation systems, and particularly relates to a method for intersection signal safety control based on behavior perception and dynamic scheduling. The method includes: S1, identifying the travel intention of the last vehicle in the queue of the current phase within a preset time window before the current phase of the green light ends; S2, determining whether the phase switching conditions are met; if so, proceeding to S3; S3, calculating the residual value of each red light phase using a green light residual value model; selecting the phase with the highest residual value as the next permitted phase; S4, monitoring whether any vehicles have mistakenly entered the intersection within a monitoring time window after the red light of the current phase turns on; if so, postponing the green light activation time of the next permitted phase. This method can accurately and dynamically adjust the timing of phase switching while ensuring the traffic efficiency of the intersection, and actively monitor and prevent mistaken entry after the red light turns on.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation systems, and in particular relates to a method for safe control of intersection signals based on behavior perception and dynamic scheduling. Background Technology

[0002] In urban road traffic systems, at-grade intersections, as road network nodes, have signal control strategies that directly affect traffic efficiency and operational safety.

[0003] Currently, many intersections still use fixed-cycle or traffic-statistic-based adaptive control methods, uniformly switching phases at the end of the green light, without considering the differences in actual traffic behavior of vehicles at different positions within the same phase. Especially during peak hours, vehicles at the rear of a convoy in the same direction often exhibit heterogeneous driving behavior: some vehicles accelerate through due to being far from the stop line and traveling at high speeds; others brake in advance to prepare to stop due to the deceleration of the vehicle in front or their own decision. If the traffic signal controller forcibly switches to red based solely on the preset green light duration, it can easily lead to accelerating vehicles being forced to brake suddenly or even run a red light, or cause braked vehicles to remain at the intersection, creating secondary delays. More seriously, if a vehicle mistakenly enters the conflict zone after the red light comes on, the existing control system lacks real-time monitoring and dynamic delay mechanisms, failing to allow sufficient time to clear the conflict zone, which can easily induce cross-collision accidents.

[0004] Therefore, how to ensure the efficiency of traffic flow at intersections while accurately and dynamically adjusting the timing of phase switching, and actively monitoring and preventing unauthorized incursions after the red light is turned on, thus realizing the closed-loop evolution of signal control from passive response to behavior perception, risk prediction, and dynamic intervention, has become an urgent problem to be solved. Summary of the Invention

[0005] To address the aforementioned shortcomings of existing technologies, the present invention aims to provide a traffic signal safety control method based on behavior perception and dynamic scheduling, which can accurately and dynamically adjust the timing of phase switching while ensuring the traffic efficiency of intersections, and actively monitor and prevent unauthorized entry after the red light is turned on.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] A traffic signal safety control method based on behavior perception and dynamic scheduling for traffic signal control systems at urban road intersections includes the following steps:

[0008] S1. Within a preset time window before the current phase of the green light ends, identify the passage intention of the last vehicle in the queue in real time; the passage intention includes accelerating through and preparing to stop.

[0009] S2. Determine if the phase switching conditions are met: the current green light duration is not less than the minimum green light duration, and there are no new vehicles added to the current phase or the maximum green light limit has been reached; if met, proceed to S3.

[0010] S3. Using the Green Light Residual Value Model Calculate the residual value of each red light phase; where, The residual value of the i-th red light phase. This represents the current number of vehicles queuing for this phase. The expected traffic efficiency per unit time for this phase;

[0011] If the vehicle at the rear of the queue in the current phase intends to accelerate through, then a temporary maneuver suppression factor is applied to all red light phases j∈C that conflict with the current phase. <1, correct residual value ;

[0012] Select the phase with the highest residual value as the next release phase;

[0013] The green light of the current phase ends, and after a preset yellow light time, it switches to red; the moment when the green light of the next traffic phase turns on is marked as the reference switching point T. base And switch to S4;

[0014] S4. Within the monitoring time window after the red light of the current phase turns on, monitor whether any vehicles have mistakenly entered the intersection in that phase; if so, based on the real-time position and speed of the mistakenly entered vehicle, predict the clearing time T required for it to completely pass through the conflict area from its current position. clear ; and postpone the green light activation time for the next clearance phase by at least T. clear ;

[0015] The monitoring time window starts at the moment the red light of the current phase turns on, and its duration does not exceed the sum of the preset yellow light time and the all-red light time, so that in T... base The previous step was to complete the error detection.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] 1. Real-time recognition of the tail vehicle's intention to proceed. By classifying the behavior of the tail vehicle in the current phase (accelerating through vs. preparing to stop) within a preset time window before the green light ends, this mechanism breaks through the static judgment mode of traditional signal control that relies solely on traffic flow or queue length. Compared to the one-size-fits-all green light cutoff strategy in existing technologies, this mechanism avoids emergency braking or invalid green light extensions caused by misjudging the tail vehicle's behavior, significantly reducing the risk of conflict and the probability of traffic interruption during the yellow light phase.

[0018] 2. Dynamic Phase Switching Conditions. A dual-condition approach is introduced: "minimum green light time constraint + no new vehicles added / maximum green light limit reached." This decouples the green light duration from its fixed duration, allowing it to adaptively adjust based on real-time traffic conditions. Compared to traditional timed control or methods triggered solely by traffic flow thresholds, this strategy better aligns with actual traffic needs, promptly releasing right-of-way when traffic is sparse and ensuring basic traffic capacity during periods of heavy traffic, thus improving the rationality and flexibility of phase resource allocation.

[0019] 3. Phase optimization mechanism based on residual value model. Constructing the residual value function. And apply a temporary inhibition factor to vehicles that accelerate through the rear. This method modifies the value of conflicting phases, thereby prioritizing the next phase with higher overall benefits in multi-phase competition. It integrates traffic efficiency and safety constraints into the decision-making objective, overcoming the limitations of traditional methods that only pursue maximum traffic volume while ignoring conflict risks, thus making phase switching more system-optimal and safer.

[0020] 4. Proactive defense mechanism against intrusion risk after the red light is on. A monitoring time window is set after the red light turns on to detect whether any vehicles have inadvertently entered the red light, and the time T required to clear the red light is predicted based on the vehicle's location and speed. clear Dynamically postpone the green light activation time of the next phase to T base +T clear This closed-loop feedback mechanism is missing from existing signal systems—traditional control rigidly executes phase switching once the red light is on, which cannot cope with sudden intrusion events; this solution realizes the transformation from post-event remediation to in-event intervention, effectively preventing potential conflicts and significantly improving the operational safety of intersections.

[0021] In summary, this method can accurately identify the intention of the following vehicle and dynamically adjust the timing of phase switching while ensuring the traffic efficiency of the intersection. It can also actively monitor and prevent unauthorized entry after the red light is turned on, thus realizing the closed-loop evolution of signal control from passive response to behavior perception, risk prediction and dynamic intervention.

[0022] Preferably, the traffic signal control system includes at least two longitudinally arranged vehicle detectors, traffic light groups, and local controllers located in front of the stop lines of each approach lane;

[0023] In S1, the passage intention of the vehicle at the tail of the current phase queue is identified based on the continuous time-series data output by the vehicle detector.

[0024] This setup, through a dual-detector longitudinal layout and temporal data analysis, reliably determines the behavioral intent of the following vehicle. Traditional single-point detectors can only acquire instantaneous presence / passage information and cannot distinguish whether a vehicle is passing at a continuous constant speed or decelerating; however, this solution utilizes two longitudinally arranged detectors, combined with the time difference between the vehicle's passage through the two detectors and the speed change trend, to effectively infer its acceleration state and subsequent behavioral tendencies. This design overcomes the inherent ambiguity of single-point detection in behavior recognition, providing a high-confidence input basis for subsequent dynamic phase switching and conflict suppression, significantly improving the accuracy and robustness of intent recognition, and avoiding safety and efficiency losses caused by unnecessary green light extensions or premature red light cuts due to misjudgments.

[0025] Preferably, the vehicle detector includes a first detector and a second detector; the continuous time-series data includes timestamps of the vehicle triggering the first and second detectors; the process of identifying the passage intention of the vehicle at the tail of the current phase queue includes:

[0026] Calculate the instantaneous speed of the vehicle at the rear of the platoon. Where d is the distance between the two detectors; t1 and t2 are the times before and after the triggering of the detectors, respectively;

[0027] Obtain the average startup speed of this phase in the same historical period. ;

[0028] like And t2 is less than the preset threshold T when the current green light ends. pass If the intention to pass is to accelerate, then it is determined that the intention is to stop; otherwise, it is determined that the intention is to stop. Where Δv th T is the threshold for velocity increment. pass The latest trigger time to allow safe passage.

[0029] This setup, integrating kinematic features and spatiotemporal safety boundaries, achieves dual discrimination of passage intention, significantly improving recognition reliability. Existing methods mostly rely on a single speed threshold or simple time margin judgment, which is easily affected by short-term fluctuations (such as deceleration after a brief acceleration) and misjudgment. This solution innovatively incorporates instantaneous acceleration trends (through...) The determination is coupled with the safety margin of the remaining green light time, which not only eliminates interference from non-continuous acceleration, but also ensures that the decision to accelerate through is physically feasible, and that the vehicle has sufficient time to safely pass through the conflict zone. This dual constraint mechanism effectively avoids the high false alarm / false alarm risk of traditional single-dimensional criteria, making the intent recognition results more consistent with actual driving logic and traffic safety requirements.

[0030] This scheme achieves high-precision, high-confidence recognition of the following vehicle's passage intention with a low-cost detector, through the collaborative criteria of kinematic parameters and safe time windows, without the need for video or high-precision positioning, thus providing a solid behavioral perception foundation for subsequent dynamic signal scheduling.

[0031] Preferably, the velocity increment threshold Δv th Adjust dynamically based on weather conditions:

[0032] ;

[0033] in, =1.0m / s is the baseline threshold. The variable is a rainfall indicator; it is set to 1 if rainfall occurs and 0 otherwise. μ = 0.5.

[0034] This setup, by adaptively adjusting the sensitivity of intent discrimination based on the environment, improves the robustness and safety of behavior recognition under complex weather conditions. In low-traction road conditions such as rain, vehicle braking distance increases significantly, and drivers generally tend to drive conservatively. Even if the vehicle's instantaneous speed is slightly higher than the average starting speed, drivers may actually choose to slow down and stop due to caution. If a fixed Δv is still used... th This can easily lead to vehicles that should be stopping being misjudged as accelerating through, resulting in improper extensions or delays in the green light transition to red, increasing the risk of running a red light or causing a conflict. This solution improves Δv... th This mechanism raises the threshold for determining whether a vehicle is accelerating through a conservatist road, making the system more inclined to make conservative decisions in adverse weather conditions; that is, it assumes the last vehicle is about to stop. This aligns with human driving behavior trends and avoids safety hazards caused by mechanical criteria. Compared to existing fixed threshold methods, this dynamic adjustment mechanism enables the perception layer to proactively adapt to environmental disturbances, enhancing the reliability of the system's decisions under varying operating conditions.

[0035] Preferably, the latest trigger time The dynamic correlation between intersection geometric parameters and vehicle dynamic limits is calculated using the following formula:

[0036] ;

[0037] in, The distance from the first detector to the parking line. The speed limit for this approach lane is [value to be inserted here]. This represents the driver's average reaction time.

[0038] This setup, based on road geometry and ergonomics parameters, constructs a safe time threshold, ensuring the interpretability and physical consistency of the decision to allow safe passage. Traditional methods often use empirically fixed times as the latest trigger time, failing to differentiate between different approach lane lengths, speed limits, or driving behavior, easily leading to overly conservative approaches for short approach lanes and overlooking risks for long approach lanes; while this solution... Explicit modeling, which is the sum of travel time and reaction time, not only conforms to the basic principle of clearing time in traffic flow theory but also incorporates the physiological response characteristics of human drivers. This ensures that, under any intersection configuration, there is sufficient time to completely pass through the conflict zone before the green light ends. Compared to the empirical threshold method, this mechanism significantly improves the scientific rigor and scenario adaptability of traffic intention determination and phase switching decisions.

[0039] Preferably, in step S3, if the vehicle at the tail of the current phase queue intends to accelerate through and a phase switch has been performed, then in step S4, the stop time of the safety monitoring window is extended to:

[0040] , .

[0041] This setup, by introducing a controllable behavior verification window, effectively suppresses the risk of conflict caused by misjudgment of intent or sudden interference (such as sudden braking or skidding), thus improving operational safety after phase switching. Although S1–S3 has comprehensively considered multiple factors such as speed trend, weather, and time margin to determine the following vehicle's acceleration, uncertainties still exist in actual driving, such as driver hesitation at the critical point, sudden drop in acceleration due to slippery road surface, and sudden obstacles ahead. If the phase switches immediately and monitoring terminates, subsequent vehicles may enter the green light phase before the conflict zone is completely cleared, leading to serious conflicts. This solution adds a monitoring delay of 0.5–1.0 s to ensure that the system continues to track the vehicle's actual trajectory after the phase switch—if it decelerates below the safety threshold during the delay, emergency intervention can be triggered, such as early red light switching or conflict warning activation; if it continues to accelerate, it will naturally complete the passage. This mechanism adds a "behavioral verification layer" to high-confidence intent recognition without sacrificing traffic efficiency, significantly enhancing the system's fault tolerance to real-world driving uncertainties.

[0042] Preferably, the average startup speed during the same historical period Weighted sliding window estimation is used:

[0043] , ;

[0044] in, Let the instantaneous velocity of the k-th historical sample be denoted as . instantaneous velocity The moment of occurrence, Let λ be the current time, λ be the time decay coefficient, and N be the number of historical samples.

[0045] This setup, by dynamically focusing on recent driving behavior characteristics through time decay weights, makes the average start speed estimation more closely match the current traffic conditions and driver habits, significantly improving the adaptability and timeliness of intent recognition. Traditional simple arithmetic averages are easily affected by long-term historical outliers (such as low traffic during holidays), leading to… This approach deviates from real-time driving trends; however, this scheme uses an exponential decay of long-term sample weights. Follow - The increased rapid descent allows recent, contemporaneous acceleration behaviors to dominate the estimation results, thus more accurately reflecting the average acceleration tendency of drivers at the intersection. For example, faster acceleration during the morning rush hour congestion relief period and more cautious acceleration at the end of the evening rush hour. This mechanism does not require manual time segmentation or external detection; it adaptively captures the short-term evolution of driving behavior solely based on timestamps, ensuring that v> +Δv th The criteria are always based on the current context to avoid systematic misjudgments caused by static benchmark mismatch, such as misjudging normal acceleration as abnormal or misjudging conservative start as preparation to stop.

[0046] Preferably, when S1 determines that the intention to pass is to accelerate passage, the local controller performs the following cooperative operation:

[0047] Temporarily suspend the submission of the switchover condition signal that there are no new vehicles in the current phase;

[0048] Simultaneously, the estimated entry time of the vehicle into the intersection is preloaded. ;

[0049] in, This is the distance from the second detector to the parking line.

[0050] This setup, by linking the delayed submission of the phase switching signal with the preloaded entry time, achieves refined and forward-looking phase extension decisions. It avoids premature phase switching that could force the last vehicle to brake suddenly or run a red light, while ensuring the continuity of subsequent traffic flow. Traditional control logic often triggers phase switching immediately upon detecting no new vehicles, but it doesn't consider whether a vehicle already in the detection zone and accelerating can safely clear the conflict zone. This solution, after determining that the vehicle is accelerating, actively suppresses the reporting of the phase switching signal, thus retaining the green light. Simultaneously, it accurately predicts the vehicle's arrival time at the stop line based on real-time speed and geometric distance; this prediction can be used to dynamically calibrate the phase end time. This upgrades the phase switching decision from a discrete event judgment to continuous time window matching, preventing erroneous interception of legally passing vehicles and avoiding the risk of conflict caused by the last vehicle actually decelerating. Especially at high-density intersections, this mechanism can effectively reduce queue accumulation and secondary start delays caused by the last vehicle being stopped.

[0051] Preferably, in S3, the temporary maneuvering inhibition factor The formula for calculation is:

[0052] ;

[0053] In the formula, v is the instantaneous speed of the vehicle at the tail of the platoon identified in S1. For reference speed.

[0054] This setup, through a speed-dependent exponential suppression mechanism, allows for flexible release of high-confidence, accelerating vehicles and cautious constraint of hesitant, low-speed vehicles, resulting in a system response that is both sensitive and robust. When v≪ For example, vehicles moving slowly or appearing to be preparing to stop. ≈1, strong inhibitory effect, the system tends to adopt conservative strategies, such as prematurely terminating the green light or strengthening monitoring; when v→ Or higher (clearly indicating an intention to accelerate). →e −1 With a threshold of ≈0.37 or lower, suppression is significantly reduced, allowing for more sufficient passage time and phase extension. This design avoids the jitter risk associated with hard switching in traditional threshold methods, instead reflecting the gradual change in driving intention with a smooth transition. For example, in critical acceleration scenarios, the vehicle transitions from 0.7... Accelerate to 0.9 When the inhibition factor decreases from 0.49 to 0.41, the control strategy is adjusted gradually rather than abruptly. This improves the system's accuracy in fitting real driving behavior and prevents frequent phase oscillations caused by small speed fluctuations, significantly enhancing control stability and user experience consistency.

[0055] Preferably, the temporary maneuvering suppression factor The duration of the effect is:

[0056] ;

[0057] in, This is the maximum clearing distance at the intersection. For safety redundancy time.

[0058] This setup, with its dynamic duration design, achieves strict coupling between the suppression strategy and the vehicle's physical motion, serving as a key time anchor point to ensure the reliability of the closed loop of intent recognition → suppression execution → phase switching. Attached Figure Description

[0059] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0060] Figure 1 This is a flowchart of the method. Detailed Implementation

[0061] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to represent selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0063] It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the figures, or the orientation or positional relationship commonly used when the product is in use. They are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance. In addition, the terms "horizontal," "vertical," etc., do not indicate that the component is required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0064] Example:

[0065] like Figure 1 As shown, this invention provides a traffic signal safety control method based on behavior perception and dynamic scheduling for use in traffic signal control systems at urban road intersections.

[0066] The traffic signal control system includes at least two longitudinally arranged vehicle detectors, traffic lights, and local controllers located in front of the stop lines at each approach lane.

[0067] In practical implementation, a millimeter-wave radar + video fusion detector is preferred: the radar provides continuous speed and distance sequences (sampling rate ≥ 10Hz), and the video provides license plate and vehicle type information for queue identification. The longitudinal distance d ∈ [8, 20]m between the two detectors satisfies: d ≥ 2 × average vehicle length to ensure the ability to distinguish adjacent vehicles; d ≤ 0.5 × minimum safe following distance to avoid missed detections. A single detector cannot reliably distinguish between vehicles at the end of the queue and those in the middle, and may easily misidentify stationary vehicles as being at the end of the queue; therefore, at least two longitudinal detectors are necessary.

[0068] A dual-detector longitudinal deployment scheme is adopted: the first detector D1 is set at d1=5m before the stop line of the entrance lane, and the second detector D2 is set at d2=15m.

[0069] This includes the following steps:

[0070] S1. Within a preset time window before the current phase of the green light ends, identify the passage intention of the last vehicle in the queue in real time; the passage intention includes accelerating through and preparing to stop.

[0071] Definition: The last vehicle in the queue refers to the last motor vehicle to trigger the upstream detector within a preset time window before the current green light phase ends, and there are no other vehicles in the queue ahead of it. Its driving status is not subject to the constraints of the vehicle in front following it.

[0072] In practice, the vehicle at the tail of the platoon is defined as: within the continuous time window Δt w =Within 2 seconds, there is a vehicle where only D1 is triggered while D2 is not, and this vehicle is in the subsequent t gap It was not covered by a new vehicle within 1.5 seconds.

[0073] In practice, the intention of the vehicle at the end of the current phase queue is identified based on the continuous time-series data output by the vehicle detector.

[0074] The vehicle detector includes a first detector and a second detector; the continuous time-series data includes the timestamps of the vehicle triggering the first and second detectors; the process of identifying the passage intention of the vehicle at the tail of the current phase queue includes:

[0075] Calculate the instantaneous speed of the vehicle at the rear of the platoon. Where d is the distance between the two detectors; t1 and t2 are the times before and after the triggering of the detectors, respectively;

[0076] Obtain the average startup speed of this phase in the same historical period. ;

[0077] like And t2 is less than the preset threshold T when the current green light ends. pass If the intention to pass is to accelerate, then it is determined that the intention is to stop; otherwise, it is determined that the intention is to stop. Where Δv th T is the threshold for velocity increment. pass The latest trigger time to allow safe passage.

[0078] This setup, integrating kinematic features and spatiotemporal safety boundaries, achieves dual discrimination of passage intention, significantly improving recognition reliability. Existing methods often rely solely on a single velocity threshold or simple time margin judgment, making them susceptible to short-term fluctuations (such as deceleration after a brief acceleration) and prone to misjudgment. This solution innovatively incorporates instantaneous acceleration trends (through...) The determination of intent is coupled with the safety margin of the remaining green light time, which eliminates interference from non-continuous acceleration while ensuring that the decision to accelerate through is physically feasible, giving the vehicle sufficient time to safely pass through the conflict zone. This dual-constraint mechanism effectively avoids the high false alarm / false alarm risk of traditional single-dimensional criteria, making the intent recognition results more consistent with actual driving logic and traffic safety requirements. Through the collaborative criteria of kinematic parameters and a safe time window, this scheme achieves high-precision, high-confidence recognition of the following vehicle's intent to pass using a low-cost detector, without requiring video or high-precision positioning, providing a solid behavioral perception foundation for subsequent dynamic signal scheduling.

[0079] In specific implementation, the velocity increment threshold Δv th Adjust dynamically based on weather conditions:

[0080] ;

[0081] in, =1.0m / s is the baseline threshold; This is a rainfall indicator variable; the value is 1 for rainfall and 0 otherwise. It is obtained from local weather sensors or cloud API; μ=0.5.

[0082] In low-traction road conditions such as rain, vehicle braking distance increases significantly, and drivers generally tend to drive conservatively. Even if the vehicle's instantaneous speed is slightly higher than the average starting speed, drivers may still choose to slow down and stop due to caution. If a fixed Δv is still used... th This can easily lead to vehicles that should be stopping being misjudged as accelerating through, resulting in improper extensions or delays in the green light transition to red, increasing the risk of running a red light or causing a conflict. This solution improves Δv... th This mechanism raises the threshold for determining whether a vehicle is accelerating through a conservatist road, making the system more inclined to make conservative decisions in adverse weather conditions; that is, it assumes the last vehicle is about to stop. This aligns with human driving behavior trends and avoids safety hazards caused by mechanical criteria. Compared to existing fixed threshold methods, this dynamic adjustment mechanism enables the perception layer to proactively adapt to environmental disturbances, enhancing the reliability of the system's decisions under varying operating conditions.

[0083] In practice, the latest trigger time The dynamic correlation between intersection geometric parameters and vehicle dynamic limits is calculated using the following formula:

[0084] ;

[0085] in, The distance from the first detector to the parking line. The speed limit for this approach lane is [value to be inserted here]. This refers to the driver's average reaction time. In practice, =1.0 s.

[0086] Traditional methods often use empirically fixed times as the latest trigger time, failing to differentiate between different approach lane lengths, speed limits, or driving behavior. This can easily lead to overly conservative approaches for short approach lanes and overlooked risks for long approach lanes. This solution, however, will... Explicit modeling, which is the sum of travel time and reaction time, not only conforms to the basic principle of clearing time in traffic flow theory but also incorporates the physiological response characteristics of human drivers. This ensures that, under any intersection configuration, there is sufficient time to completely pass through the conflict zone before the green light ends. Compared to the empirical threshold method, this mechanism significantly improves the scientific rigor and scenario adaptability of traffic intention determination and phase switching decisions.

[0087] Traffic intention is a binary label: if the time difference Δt between vehicles passing through D1 and D2 is ≤ T pass If the value is 0, it is determined to accelerate through; otherwise, it is determined to prepare to stop. Where T... pass The physical meaning is "the latest triggering time that allows safe passage through the conflict zone".

[0088] In practice, the average startup speed of the historical period is mentioned. Weighted sliding window estimation is used:

[0089] , ;

[0090] in, Let the instantaneous velocity of the k-th historical sample be denoted as . instantaneous velocity The moment of occurrence, Let λ be the time decay coefficient at the current moment, and N be the number of historical samples. In practice, λ ∈ [0.05, 0.2] h. −1 .

[0091] To better understand the average startup speed during the same period in history The calculation mechanism is explained below.

[0092] λ = 0.02 s⁻¹, corresponding to a time constant τ = 1 / λ = 50 s, and cross-validation shows an error of <5% in 95% of scenarios.

[0093] In practice, the system maintains a circular buffer to store historical samples of the most recent N=50 vehicles; when a new sample is added, the oldest sample is automatically discarded. The computation delay is ≤ 10 ms.

[0094] By dynamically focusing on recent driving behavior characteristics through time decay weighting, the average start speed estimation is made more closely aligned with current traffic conditions and driver habits, significantly improving the adaptability and timeliness of intent recognition. Traditional simple arithmetic averages are easily affected by long-term historical outliers (such as low traffic during holidays), leading to… This approach deviates from real-time driving trends; however, this scheme uses an exponential decay of long-term sample weights. Follow - The increased rapid descent allows recent, contemporaneous acceleration behaviors to dominate the estimation results, thus more accurately reflecting the average acceleration tendency of drivers at the intersection. For example, faster acceleration during the morning rush hour congestion relief period and more cautious acceleration at the end of the evening rush hour. This mechanism does not require manual time segmentation or external detection; it adaptively captures the short-term evolution of driving behavior solely based on timestamps, ensuring that v> +Δv th The criteria are always based on the current context to avoid systematic misjudgments caused by static benchmark mismatch, such as misjudging normal acceleration as abnormal or misjudging conservative start as preparation to stop.

[0095] In practice, when S1 determines that the intention to pass is to expedite passage, the local controller performs the following cooperative operations:

[0096] Temporarily suspend the submission of the switchover condition signal that there are no new vehicles in the current phase;

[0097] Simultaneously, the estimated entry time of the vehicle into the intersection is preloaded. ;

[0098] in, This is the distance from the second detector to the parking line.

[0099] By linking the delayed submission of the phase switching signal with the preloaded entry time, the phase extension decision can be made more refined and forward-looking. This avoids the tail vehicle being forced to brake suddenly or run a red light due to premature phase switching, while ensuring the continuity of subsequent traffic flow. Traditional control logic often triggers phase switching immediately upon detecting no new vehicles, but it does not consider whether the tail vehicle that has entered the detection zone and is accelerating through can actually safely clear the conflict zone. This solution actively suppresses the reporting of the phase switching signal after determining that it is accelerating through, that is, it retains the green light; at the same time, it accurately predicts the time when the tail vehicle will arrive at the stop line based on real-time speed and geometric distance; this predicted value can be used to dynamically calibrate the phase end time. This move upgrades the decision to switch from a discrete event judgment to continuous time window matching, which not only prevents the wrong interception of legally passing vehicles, but also avoids the conflict risk caused by the tail vehicle actually decelerating. Especially at high-density intersections, this mechanism can effectively reduce the queue accumulation and secondary start delay caused by the last vehicle being stopped.

[0100] S2. Determine if the phase switching conditions are met: the current green light duration is not less than the minimum green light duration, and there are no new vehicles added to the current phase or the maximum green light limit has been reached; if met, proceed to S3.

[0101] S3. Using the Green Light Residual Value Model Calculate the residual value of each red light phase; where, The residual value of the i-th red light phase; This represents the current number of vehicles queuing for this phase. The expected traffic efficiency per unit time for this phase is expressed in vehicles per second and is obtained by a moving average of historical data from the same period.

[0102] If the vehicle at the rear of the queue in the current phase intends to accelerate through, then a temporary maneuver suppression factor is applied to all red light phases j∈C that conflict with the current phase. <1, correct residual value ;

[0103] Select the phase with the highest residual value as the next release phase;

[0104] The green light of the current phase ends, and after a preset yellow light time, it switches to red; the moment when the green light of the next traffic phase turns on is marked as the reference switching point T. base Then switch to S4.

[0105] To facilitate a better understanding by those skilled in the art, the green light residual value model and its parameters are explained as follows.

[0106] ;

[0107] This represents the current number of vehicles queuing for that phase. In practice, it is defined as the total number of vehicles queuing within 30m behind the stop line at the moment the red light turns on in the current phase, including vehicles that are starting but have not yet crossed the stop line; this is obtained through the statistics of detector D.

[0108] The expected traffic efficiency per unit time for this phase, which is the average traffic efficiency for the same historical period of this phase, is calculated using the following formula:

[0109] ;

[0110] in, This represents the actual number of vehicles passing through in the previous period. The time for the green light to turn on. This refers to the time it takes for the first vehicle to cross the stop line after the green light turns on.

[0111] In practice, temporary maneuvering inhibition factor The formula for calculation is:

[0112] ;

[0113] In the formula, v is the instantaneous speed of the vehicle at the tail of the platoon identified in S1. For reference speed. In actual implementation, =8m / s.

[0114] The temporary maneuvering suppression factor The duration of the effect is:

[0115] ;

[0116] in, The maximum clearance distance at the intersection is the distance from the current position of the vehicle in the current phase to the farthest boundary of the conflict area. For safety redundancy time. In specific implementation, =0.8s.

[0117] By employing a speed-dependent exponential suppression mechanism, flexible release of high-confidence, accelerating vehicles and cautious constraint of low-speed, hesitant vehicles can be achieved, enabling the system response to be both sensitive and robust. When v≪ For example, vehicles moving slowly or appearing to be preparing to stop. ≈1, strong inhibitory effect, the system tends to adopt conservative strategies, such as prematurely terminating the green light or strengthening monitoring; when v→ Or higher (clearly indicating an intention to accelerate). →e −1 With a threshold of ≈0.37 or lower, suppression is significantly reduced, allowing for more sufficient passage time and phase extension. This design avoids the jitter risk associated with hard switching in traditional threshold methods, instead reflecting the gradual change in driving intention with a smooth transition. For example, in critical acceleration scenarios, the vehicle transitions from 0.7... Accelerate to 0.9 When the inhibition factor decreases from 0.49 to 0.41, the control strategy is adjusted gradually rather than abruptly. This improves the system's accuracy in fitting real driving behavior and prevents frequent phase oscillations caused by small speed fluctuations, significantly enhancing control stability and user experience consistency.

[0118] S4. Within the monitoring time window after the red light of the current phase turns on, monitor whether any vehicles have mistakenly entered the intersection in that phase; if so, based on the real-time position and speed of the mistakenly entered vehicle, predict the clearing time T required for it to completely pass through the conflict area from its current position. clear ; and postpone the green light activation time for the next clearance phase by at least T. clear ;

[0119] The monitoring time window starts at the moment the red light of the current phase turns on, and its duration does not exceed the sum of the preset yellow light time and the all-red light time, so that in T... base The previous step was to complete the error detection.

[0120] In specific implementation, in S3, if the vehicle at the tail of the current phase queue intends to accelerate through and performs a phase switch, then in S4, the stop time of the safety monitoring window is extended to:

[0121] , .

[0122] To facilitate understanding of the mechanism for extending the stop time of the safety monitoring window, the following explanation is provided.

[0123] Designed based on the following safety margins:

[0124] 0.5 s: Covers the average driver reaction time (0.3~0.7 s) + signal instruction delay (≤0.2 s);

[0125] 1.0 s: An additional 0.5 s is reserved to handle small vehicles, such as electric bicycles, passing through at low speeds.

[0126] In practice, those skilled in the art can adjust the implementation according to actual needs. Make adaptive adjustments.

[0127] By introducing a controllable behavior verification window, the risk of conflict caused by misjudgment of intent or sudden interference (such as sudden braking or skidding) can be effectively suppressed, improving operational safety after phase switching. Although S1–S3 has determined the rear vehicle's acceleration to pass by considering multiple factors such as speed trend, weather, and time margin, uncertainties still exist in actual driving, such as driver hesitation at the critical point, sudden drop in acceleration due to slippery road surface, and sudden obstacles ahead. If the phase switches immediately and monitoring stops, subsequent vehicles may enter the green light phase before the conflict zone is completely cleared, causing serious conflicts. This solution adds a monitoring delay of 0.5–1.0 s to ensure that the system continues to track the vehicle's actual trajectory after the phase switch—if it decelerates below the safety threshold during the delay, emergency intervention can be triggered, such as switching to red light in advance or activating a conflict warning; if it continues to accelerate, it will naturally complete the passage. This mechanism adds a "behavioral verification layer" to high-confidence intent recognition without sacrificing traffic efficiency, significantly enhancing the system's fault tolerance to uncertainties in real-world driving.

[0128] Compared to existing technologies, this invention breaks through the static judgment mode of traditional signal control that relies solely on traffic flow or queue length by classifying the behavior of vehicles at the tail of the current phase queue (accelerating through vs. preparing to stop) within a preset time window before the green light ends. Compared to the one-size-fits-all green light cutoff strategy in existing technologies, this mechanism avoids emergency braking or invalid green light extensions caused by misjudging the behavior of the tail vehicle, significantly reducing the risk of conflict and traffic interruption during the yellow light phase. In addition, this method introduces a dual-condition judgment of whether to switch phases based on "minimum green light time constraint + no new vehicles / reaching the maximum green light limit," so that the duration of the green light is no longer fixed, but adaptively adjusted according to the real-time traffic conditions. Compared to traditional timed control or switching triggered only by traffic flow thresholds, this strategy is more in line with actual traffic needs, releasing right-of-way in a timely manner when traffic flow is sparse, and ensuring basic traffic capacity when traffic flow is dense, thus improving the rationality and flexibility of phase resource allocation.

[0129] In addition, by constructing a residual value function And apply a temporary inhibition factor to vehicles that accelerate through the rear. This method modifies the value of conflicting phases, thus prioritizing the next phase with higher overall efficiency in multi-phase competition. It integrates traffic efficiency and safety constraints into the decision-making objective, overcoming the limitations of traditional methods that only pursue maximum traffic volume while ignoring conflict risks, making phase switching more system-optimal and safer. Furthermore, a monitoring time window is set after the red light turns on to detect whether any vehicles have mistakenly entered the red light, and the time T required to clear the red light is predicted based on their location and speed. clear Dynamically postpone the green light activation time of the next phase to T base +T clear This closed-loop feedback mechanism is missing from existing signal systems—traditional control rigidly executes phase switching once the red light is on, which cannot cope with sudden intrusion events; this solution realizes the transformation from post-event remediation to in-event intervention, effectively preventing potential conflicts and significantly improving the operational safety of intersections.

[0130] This method can accurately identify the intention of the following vehicle while ensuring the efficiency of traffic flow at intersections, dynamically adjust the timing of phase switching, and actively monitor and prevent unauthorized incursions after the red light is turned on, thus realizing the closed-loop evolution of signal control from passive response to behavior perception, risk prediction, and dynamic intervention.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for intersection signal safety control based on behavior perception and dynamic scheduling, characterized in that, A traffic signal control system for urban road intersections includes the following steps: S1. Within a preset time window before the current phase of the green light ends, identify the passage intention of the last vehicle in the queue in real time; the passage intention includes accelerating through and preparing to stop. S2. Determine if the phase switching conditions are met: the current green light duration is not less than the minimum green light duration, and there are no new vehicles added to the current phase or the maximum green light limit has been reached; if met, proceed to S3. S3. Using the Green Light Residual Value Model Calculate the residual value of each red light phase; where, The residual value of the i-th red light phase. This represents the current number of vehicles queuing for this phase. The expected traffic efficiency per unit time for this phase; If the vehicle at the rear of the queue in the current phase intends to accelerate through, then a temporary maneuver suppression factor is applied to all red light phases j∈C that conflict with the current phase. <1, correct residual value ; Select the phase with the highest residual value as the next release phase; The green light of the current phase is ended, and the red light is switched after the preset yellow light time; the green light of the next release phase is marked as the reference switching point T base and go to S4; S4. Within the monitoring time window after the red light of the current phase turns on, monitor whether any vehicles have mistakenly entered the intersection in that phase; if so, based on the real-time position and speed of the mistakenly entered vehicle, predict the clearing time T required for it to completely pass through the conflict area from its current position. clear ; and postpone the green light activation time for the next clearance phase by at least T. clear ; The monitoring time window starts at the moment the red light of the current phase turns on, and its duration does not exceed the sum of the preset yellow light time and the all-red light time, so that in T... base The previous step was to complete the error detection.

2. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 1, characterized in that, The traffic signal control system includes at least two longitudinally arranged vehicle detectors, traffic lights, and local controllers located in front of the stop lines at each approach lane. In S1, the passage intention of the vehicle at the tail of the current phase queue is identified based on the continuous time-series data output by the vehicle detector.

3. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 2, characterized in that, The vehicle detector includes a first detector and a second detector; The continuous time-series data includes the timestamps of the vehicle triggering the first and second detectors; The process of identifying the passage intention of the vehicle at the tail of the current phase queue includes: Calculate the instantaneous speed of the vehicle at the rear of the platoon. Where d is the distance between the two detectors; t1 and t2 are the times before and after the triggering of the detectors, respectively; Obtain the average startup speed of this phase in the same historical period. ; like And t2 is less than the preset threshold T when the current green light ends. pass If the intention to pass is to accelerate, then it is determined that the intention is to stop; otherwise, it is determined that the intention is to stop. Where Δv th T is the threshold for velocity increment. pass The latest trigger time to allow safe passage.

4. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 3, characterized in that, The velocity increment threshold Δv th Adjust dynamically based on weather conditions: ； in, =1.0m / s is the baseline threshold. The variable is a rainfall indicator; it is set to 1 if rainfall occurs and 0 otherwise. μ = 0.

5.

5. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 3, characterized in that, The latest trigger time The dynamic correlation between intersection geometric parameters and vehicle dynamic limits is calculated using the following formula: ； in, The distance from the first detector to the parking line. The speed limit for this approach lane is [value to be inserted here]. This represents the driver's average reaction time.

6. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 5, characterized in that, In step S3, if the vehicle at the tail of the current phase queue intends to accelerate through and a phase switch is performed, then in step S4, the stop time of the safety monitoring window is extended to: ， 。 7. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 3, characterized in that, The average startup speed during the same period in history Weighted sliding window estimation is used: ， ； in, Let the instantaneous velocity of the k-th historical sample be denoted as . instantaneous velocity The moment of occurrence, Let λ be the current time, λ be the time decay coefficient, and N be the number of historical samples.

8. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 3, characterized in that, When S1 determines that the intention to pass is to expedite passage, the local controller performs the following cooperative operation: Temporarily suspend the submission of the switchover condition signal that there are no new vehicles in the current phase; Simultaneously, the estimated entry time of the vehicle into the intersection is preloaded. ; in, This is the distance from the second detector to the parking line.

9. In the intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 1, in S3, the temporary maneuver suppression factor... The formula for calculation is: ； In the formula, v is the instantaneous speed of the vehicle at the tail of the platoon identified in S1. For reference speed.

10. The intersection signal safety control method based on behavior perception and dynamic scheduling according to claim 9, characterized in that, The temporary maneuvering suppression factor The duration of the effect is: ； in, This is the maximum clearing distance at the intersection. For safety redundancy time.

Images