A road traffic signal control parameter adaptive optimization method and system

Abstract

The application provides a road traffic signal control parameter adaptive optimization method and system, and belongs to the technical field of road traffic signal control, and comprises the following steps: acquiring road intersection point cloud data and forming vehicle trajectories; extracting approaching area speed and deceleration samples and conflict area entering and leaving time based on the vehicle trajectories; constructing a parking feasibility time sample set based on the approaching area speed and deceleration samples; constructing a conflict area occupancy time sample set based on the conflict area entering and leaving time; obtaining a yellow light risk boundary lower limit based on the parking feasibility time sample set, and obtaining a full red risk boundary lower limit based on the conflict area occupancy time sample set; adding a risk change rate constraint and a minimum residence time constraint based on the yellow light risk boundary lower limit and the full red risk boundary lower limit; online solving yellow light time and full red time, calculating emptying time, and integrating the solving results to obtain optimized road traffic signal control parameters.

Description

Technical Field

[0001] This invention belongs to the field of road traffic signal control technology, and in particular relates to an adaptive optimization method and system for road traffic signal control parameters. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As pilot programs for intelligent connected vehicles on public roads strengthen engineering requirements for safety, controllability, auditability, and rollback capability, the role of roadside infrastructure in safety governance and operation control becomes more prominent. The standard system sets requirements for signal display consistency, controller interface, and basic performance. However, there is a lack of a unified and interpretable risk constraint framework for methods to adjust yellow light, all-red light, and clearing time online using external roadside sensor data, resulting in engineering implementation relying heavily on fixed or empirical parameters.

[0004] In recent years, roadside lidar has been used to acquire traffic flow characteristics such as vehicle trajectories, speed distribution, and headway due to its advantages such as stable ranging and minimal impact from lighting conditions. This enables phase extension, allocation optimization, and traffic efficiency assessment. In research and engineering practice related to yellow light and all-red light control, common approaches either use fixed or time-segmented parameters, or rely on single-vehicle predictions of running yellow lights or entering conflict zones as triggering criteria.

[0005] The above methods face challenges in cross-scenario generalization, interpretability, and engineering acceptance. Their main shortcomings are:

[0006] The yellow light, all-red light, and clearing times are generally configured with fixed or roughly divided time periods, which makes it difficult to adapt to changes in vehicle speed distribution, vehicle composition, and driving behavior, resulting in a trade-off between safety and efficiency.

[0007] Some solutions base control decisions on predictions of bicycles running through yellow lines or trajectory extrapolation, which are highly dependent on models, data, and scenario transfer, making it difficult to form an engineering commitment to controllable risk upper bounds.

[0008] In existing technologies, even with roadside sensing input, there is a lack of constraint optimization and smooth update mechanisms that are tightly coupled with the signal parameter system, which can easily lead to parameter update jumps, frequent fluctuations, or inability to be stably implemented. Summary of the Invention

[0009] To overcome the shortcomings of the prior art, the present invention provides an adaptive optimization method and system for road traffic signal control parameters, which is an adaptive clearing control scheme that takes statistical risk boundary as the core, can give an auditable risk upper bound and is closely coupled with the signal parameter system.

[0010] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0011] Firstly, an adaptive optimization method for road traffic signal control parameters is disclosed, including:

[0012] Acquire point cloud data of road intersections and generate vehicle trajectories;

[0013] Based on the vehicle trajectory, samples of approach zone velocity and deceleration, and entry and exit times of the conflict zone are extracted.

[0014] A parking feasibility time sample set is constructed based on the speed and deceleration samples in the approach zone; a conflict zone occupancy time sample set is constructed based on the entry and exit times of the conflict zone.

[0015] The lower bound of the yellow light risk boundary is obtained based on the parking feasibility time sample set, and the lower bound of the all-red risk boundary is obtained based on the conflict zone occupancy time sample set.

[0016] Based on the lower bounds of the yellow light risk boundary and the full red risk boundary, risk change rate constraints and minimum residence time constraints are added.

[0017] The system calculates the yellow light duration and all-red light duration online, and then integrates the results to obtain optimized road traffic signal control parameters.

[0018] As a further technical solution, acquiring point cloud data of road intersections and forming vehicle trajectories specifically includes: collecting point cloud data of road intersections at a fixed frequency, mapping the point cloud data to the road coordinate system through external parameter calibration, and performing noise reduction, dynamic point filtering, and temporal correlation on the point cloud data to form a set of vehicle trajectories.

[0019] As a further technical solution, an approach zone and a conflict zone are divided in the road coordinate system: the approach zone is located within a certain distance before the stop line and is used to collect the approach speed and deceleration behavior of vehicles; the conflict zone is located inside the intersection and is used to count the time process of vehicles entering and completely leaving the conflict zone.

[0020] As a further technical solution, it also includes: taking a specified quantile from the sample set of parking feasibility times to form a lower bound of the risk boundary for yellow light time, which provides a constraint with clear risk semantics for yellow light time.

[0021] As a further technical solution, under the premise of satisfying the lower bound constraints of the risk boundaries of yellow light and all-red light respectively, the system uses the efficiency cost function as the objective to jointly optimize the yellow light time and all-red light time; a parameter change rate limit is introduced in the optimization process; and a minimum dwell time constraint is set at the same time.

[0022] As a further technical solution, it also includes: continuously monitoring parking feasibility exceeding the boundary event and conflict zone occupation exceeding the boundary event in actual operation, and estimating the corresponding exceeding the boundary probability;

[0023] Further calculate the upper confidence bound of the out-of-bounds probability to verify whether the actual operational risk meets the preset risk target;

[0024] When the upper confidence bound exceeds the risk threshold, the quantile parameters are automatically adjusted to make the lower risk boundary more conservative; when the risk is significantly lower than the threshold and the minimum dwell time constraint is met during long-term operation, the parameters are slowly recovered to improve traffic efficiency.

[0025] Secondly, an adaptive optimization system for road traffic signal control parameters is disclosed, comprising:

[0026] LiDAR, edge computing units, and signal control units installed on the roadside;

[0027] The lidar acquires point cloud data of the road intersection and transmits it to the edge computing unit;

[0028] The edge computing unit is configured to include;

[0029] The preprocessing module is configured to preprocess the acquired point cloud data of road intersections.

[0030] The trajectory and sample module is configured to: generate vehicle trajectories based on preprocessed data, and extract approach zone velocity and deceleration samples, and collision zone entry and exit times based on the vehicle trajectories.

[0031] A parking feasibility time sample set is constructed based on the speed and deceleration samples in the approach zone; a conflict zone occupancy time sample set is constructed based on the entry and exit times of the conflict zone.

[0032] The risk boundary and optimization module is configured to: obtain the lower bound of the yellow light risk boundary based on the parking feasibility time sample set, and obtain the lower bound of the all-red risk boundary based on the conflict zone occupancy time sample set.

[0033] Based on the lower bounds of the yellow light risk boundary and the full red risk boundary, risk change rate constraints and minimum residence time constraints are added.

[0034] The system calculates the yellow light and red light times online and the clearing time, and integrates the results to obtain optimized road traffic signal control parameters.

[0035] The signal controller is configured to include: a parameter receiving and execution module, the parameter receiving and execution module being configured to receive optimized road traffic signal control parameters and control traffic signals.

[0036] As a further technical solution, it also includes: a log and interface module, which is configured to record the sample distribution, quantiles, confidence upper bound, out-of-bounds events and final parameters in each parameter calculation process to form a complete risk evidence chain for project acceptance, operation and maintenance audit and regulatory verification.

[0037] The above one or more technical solutions have the following beneficial effects:

[0038] The technical solution of this invention first collects point clouds and forms vehicle trajectories using roadside lidar in the approach and conflict zones of intersections. Then, it constructs parking feasibility time samples and conflict zone occupancy time samples within a sliding window, and constructs the lower bounds of the risk boundaries for yellow lights and all-red lights in a statistical sense using quantiles and confidence upper bounds. Then, under the combined effect of risk boundary constraints, rate of change constraints, and minimum dwell time constraints, it solves online for the optimal parameters of yellow light time and all-red time and calculates the clearing time. Finally, the parameters are integrated and sent to the signal controller for execution, and closed-loop verification is performed on out-of-bounds events after execution to ensure long-term stable alignment with risk targets, thereby maximizing traffic efficiency while ensuring that the risk upper bound is controlled.

[0039] This embodiment of the technical solution utilizes vehicle trajectories and kinematic statistics generated from roadside lidar point clouds at urban signalized intersections to construct statistical risk boundaries for yellow light, all-red, and clearing times. Under controlled constraints of the upper bound of safety risk, it provides a method and system for online solving and issuing adaptive control parameters for yellow light, all-red, and clearing times. This achieves an interpretable trade-off between safety and efficiency without relying on single-vehicle red light prediction models, and ensures project feasibility and auditability through parameter change rate limits and risk confidence upper bound verification.

[0040] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0041] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0042] Figure 1 This is an overall flowchart of the method in an embodiment of the present invention;

[0043] Figure 2 This is a system structure block diagram of an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram illustrating the spatial definition of the proximity zone and the conflict zone in an embodiment of the present invention. Detailed Implementation

[0045] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0047] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0048] Related terms:

[0049] Approach zone: A spatial area within a certain distance in front of the stop line used to collect data on the approach status of vehicles, and to statistically analyze the distribution characteristics of vehicle speed, deceleration, arrival behavior, and following distance.

[0050] Conflict zone: The spatial area within an intersection where the travel paths of vehicles in different phases conflict. The distribution of conflict zone occupancy time and clearance time is obtained by statistically analyzing the time when vehicles enter and leave the conflict zone.

[0051] Yellow light time At discrete update time Calculate and issue the duration of the yellow light display in the next execution cycle.

[0052] All Red Time The duration of the all-red zone is set during phase switching to ensure that the collision zone is completely cleared.

[0053] Clear time The combined safe interval between yellow and all-red lights meets the requirements. .

[0054] Sliding window : The time length used for statistical distribution.

[0055] Update cycle : The time interval for online updates of control parameters.

[0056] Quantile operator : Take from the sample set Quantile operators.

[0057] Out-of-boundary events: Events observed in actual operation where the parking feasibility time or conflict zone occupancy time exceeds the issued parameters. The out-of-boundary probability is estimated by the out-of-boundary ratio, and the upper confidence bound is used to evaluate whether the risk objective has been achieved.

[0058] Example 1

[0059] See appendix Figure 1-3 As shown, this embodiment discloses an adaptive optimization method for road traffic signal control parameters. Taking an adaptive clearing control of the risk boundary of a crossroads with a standard urban signalized plane as an example, the method operates continuously on the roadside of the intersection. It is used to adaptively adjust the yellow light time, the red light time, and the clearing time online under different vehicle speed distributions, traffic flow states, and driving behavior conditions, thereby improving traffic efficiency while ensuring that the upper limit of risk is controlled.

[0060] The overall process of this embodiment is as follows: Figure 1 As shown, an adaptive optimization method for road traffic signal control parameters, executed by an intersection roadside edge computing unit and linked with the road traffic signal controller, includes:

[0061] Step 1: Roadside lidar collects point clouds and forms vehicle trajectories;

[0062] Step 2: Construct parking feasibility time samples and conflict zone occupancy time samples within the sliding window;

[0063] The lower bound of the risk boundary between yellow and all-red lights is determined based on statistical quantiles;

[0064] Step 3: Solve the yellow light time and all-red light time online under risk constraints, rate of change constraints and minimum residence time constraints, and calculate the clearing time;

[0065] Step 4: Send the parameters to the signal controller for execution; at the same time, perform statistics and confidence upper bound verification on out-of-bounds events during operation to achieve long-term stable alignment of risk targets.

[0066] In this implementation example, regarding step one, the roadside lidar collects point cloud data of the intersection at a fixed frequency and maps the point cloud to the road coordinate system through extrinsic parameter calibration. The system performs noise reduction, dynamic point filtering, and temporal correlation on the point cloud to form a set of vehicle trajectories.

[0067] like Figure 3 As shown, this embodiment clearly delineates the approach zone and the conflict zone in the road coordinate system: the approach zone is located within a certain distance before the stop line and is used to collect the approach speed and deceleration behavior of vehicles; the conflict zone is located inside the intersection and is used to statistically analyze the time process of vehicles entering and completely leaving the conflict zone. Key events such as vehicles entering the approach zone, reaching the stop line, entering the conflict zone, and leaving the conflict zone are identified at the trajectory level, providing a foundation for subsequent time sample construction.

[0068] Specifically, to ensure symbol consistency, the discrete update time is defined as... The point cloud frame of the roadside lidar is ,in For the first The point in the radar coordinate system The three-dimensional coordinate vector below, The number of points; define the road coordinate system as... The radar coordinate system is Radar extrinsic parameters are ,in For rotation matrix, It is a translation vector.

[0069] Define the lower point of the road coordinate system as And it satisfies formula (1).

[0070] (1)

[0071] in, For point In the road coordinate system The coordinate vector below, For rotation matrix, It is a translation vector.

[0072] In this implementation example, point cloud frames are acquired. And perform preprocessing to obtain purified point cloud. A point set is formed according to formula (1). A trajectory set is formed by temporal correlation of dynamic points. And extract the velocity samples of the approach zone from the trajectory. With deceleration samples Simultaneously extract the entry time from the conflict zone. With departure time . Indicates the first The vehicle trajectory, or the first A target trajectory object, Indicates the number of available tracks within the window.

[0073] The spatial extent of the conflict zone is defined in the road coordinate system, which can be represented by a polygonal region or a grid region. For each trajectory, it is determined whether its centroid or vehicle bounding box intersects with the conflict zone. The moment when the trajectory first enters the conflict zone boundary is recorded as the entry moment; the moment when the trajectory last leaves the conflict zone boundary and does not re-enter it is recorded as the departure moment. If the trajectory exhibits jitter near the conflict zone boundary, a minimum duration frame count or a hysteresis threshold can be introduced to ensure stable entry / departure determination. The difference between the entry moment and the departure moment yields the conflict zone occupancy time sample.

[0074] In the road coordinate system, an approach zone is predefined, specifically a certain distance range before the stop line. For each trajectory, a corresponding trajectory segment is extracted within the approach zone. Based on the trajectory's position information and timestamp within this segment, a velocity sequence is calculated. The velocity sample can be the average velocity or the final velocity within the segment. The deceleration sample can be obtained from the velocity difference between adjacent time points, and can be further extracted as the minimum deceleration, average deceleration, or achievable deceleration under comfort braking constraints within the segment to construct a parking feasibility time sample. All of the above extractions can be calculated independently for each trajectory within a sliding window and then aggregated to form a sample set.

[0075] The formation process of the trajectory set is an implementable workflow of single-frame clustering and cross-frame association: In the road coordinate system, each frame of the cleaned point cloud is first dynamically filtered, and then spatially clustered to obtain target point cloud clusters, such as based on distance thresholds or voxel adjacency clustering; subsequently, data association is performed between adjacent frames based on the centroid position, velocity prediction, and spatial thresholds of the target clusters, assigning consistent target identifiers to the same vehicle and concatenating them in chronological order to obtain the vehicle trajectory; for short-term occlusion or frame loss, short-term trajectory preservation and reconnection thresholds can be used to reduce trajectory breaks. The resulting trajectory set is used for subsequent sample extraction and risk statistics.

[0076] In this implementation example, regarding step two, a parking feasibility time sample and a conflict zone occupancy time sample are constructed within the sliding window.

[0077] First, within a sliding time window, speed and deceleration samples are extracted based on vehicle trajectories within the proximity zone, and a conservative set of parking feasibility time samples is constructed. This sample set statistically describes the time scale required for a vehicle to safely stop under the current traffic conditions.

[0078] A specified quantile is taken from the sample set to form the lower bound of the risk boundary for yellow light time. This lower bound does not rely on the prediction of running a yellow light at the individual vehicle level, but directly expresses the risk level of the infeasibility of stopping in the form of a statistical distribution, thus providing a constraint with a clear risk semantic for yellow light time.

[0079] After obtaining the sample set within the sliding window, the samples are sorted in ascending order of value, and the sample value located at the specified quantile position is selected as the quantile result. For example, when the quantile parameter is p, the quantile can be the sample located at the position p times the number of samples after sorting. This quantile serves as the lower bound of the risk boundary, ensuring that, statistically, only a predetermined proportion of samples will exceed this boundary, thus providing an interpretable risk constraint for yellow or all-red lights.

[0080] The boundary value obtained by taking a specified quantile on the sliding window sample set serves as the minimum safe lower limit for the yellow light time or all-red time, constraining the risk of events such as infeasibility of stopping or incomplete clearance of the conflict zone. In other words, the lower limit of the risk boundary is not a single-vehicle prediction result, but an executable parameter lower limit directly given by statistical distribution. Its risk semantics correspond to the engineering commitment that the proportion of samples exceeding the boundary does not exceed the set target, facilitating auditing and acceptance.

[0081] Secondly, within a sliding window, a sample of conflict zone occupancy time is extracted based on vehicle trajectories; that is, the time elapsed from when a vehicle enters the conflict zone to when it completely leaves the conflict zone. This sample set is used to reflect the time scale required for the conflict zone to be completely cleared under current traffic conditions.

[0082] Similarly, the lower bound of the risk boundary for all-red time is formed by statistically analyzing the conflict zone occupancy time samples using quantile operators. The system collects conflict zone occupancy time samples within a sliding window, sorts the samples, and then calculates the quantiles according to specified quantile parameters to obtain the lower bound of the risk boundary for all-red time. Since the conflict zone occupancy time directly corresponds to the time span between a vehicle entering and completely leaving the conflict zone, this quantile boundary can statistically ensure that the proportion of incompletely cleared conflict zones does not exceed the set target, thus providing a clear, executable, and auditable lower bound constraint for all-red time. Therefore, in this embodiment, yellow lights and all-red lights correspond to risk events with different physical meanings and are modeled using independent statistical boundaries.

[0083] Specifically, when constructing parking feasibility time samples and determining the lower bound of the yellow light risk boundary, the system constructs a sample set within a sliding window. The sample is defined as in formula (2).

[0084] (2)

[0085] in, For the first Conservative parking feasibility time for a sample The reaction time is always on. For speed samples, For deceleration samples, This is a divide-by-zero protection constant. This represents the number of yellow light samples within the window. The system uses quantiles to obtain the lower bound for the yellow lights. and apply safety constraints. .

[0086] Regarding the construction of conflict zone occupancy time samples and the determination of the lower bound of the all-red risk boundary.

[0087] The system in the sliding window Internal construction sample set The sample is defined as shown in formula (3).

[0088]

[0089] in, For the first The duration of conflict zone occupancy for each sample For the time of entering the conflict zone, For the time to leave the conflict zone, This represents the number of all-red samples within the window. The system uses quantiles to obtain the lower bound for all-red samples. and apply safety constraints. .

[0090] In one implementation example, regarding step three, the yellow light time and the all-red light time are solved online under risk constraints, rate of change constraints, and minimum dwell time constraints, and the clearing time is calculated.

[0091] In this embodiment, the system uniformly describes the yellow light time and the all-red time through the clearing time relationship, where the clearing time is the sum of the two. Under the premise of satisfying the lower bound constraints of the risk boundaries for both yellow light and all-red time, the system jointly optimizes the yellow light time and all-red time with an efficiency cost function as the objective.

[0092] To ensure the feasibility of the project, the system introduces a parameter change rate limit during the optimization process to prevent abrupt changes in control parameters within adjacent update cycles; at the same time, a minimum dwell time constraint is set to avoid frequent parameter adjustments in a short period of time, thereby ensuring the stability and consistency of the signal controller operation.

[0093] Specifically, the two-stage clearing relationship is established and the clearing time is calculated.

[0094] The system specifies that the clearing time must meet the following conditions:

[0095]

[0096] in, To clear the time, Yellow light time The time is when it is all red.

[0097] Online optimization and smooth updates under risk constraints. Specifically, in each update cycle, the system aims to minimize the traffic loss caused by the sum of yellow and all-red lights, while satisfying the lower bounds of the yellow and all-red risk boundaries. It also incorporates parameter change rate constraints, ensuring that the parameter change relative to the previous cycle does not exceed the maximum allowable change, and a minimum dwell time constraint, meaning that parameter adjustments for efficiency purposes are generally not allowed during the dwell time. Under these constraints, the system solves for the optimal yellow and all-red parameters for the current cycle and calculates the clearing time, which serves as the parameter set to be issued.

[0098] The system defines the efficiency cost function as follows:

[0099]

[0100] in, For the sake of efficiency, These are non-negative weighting coefficients. The system is subject to risk constraints. , Minimize To suppress frequent jumps, the system introduces a rate-of-change constraint as shown in formula (4).

[0101] (4)

[0102] in, and These represent the yellow light time and the all-red light time that were issued and executed in the previous update cycle, respectively. and These represent the maximum allowable change in a single update. The system solves for... And calculate At the same time, a minimum dwell time is introduced. When it has been less than a year since the last update At this time, parameters can only be increased when risk constraints will be violated, and parameters cannot be decreased simply for efficiency. This represents the optimal solution for the yellow light time and the all-red light time obtained in this cycle under the constraints of risk boundary, rate of change, and minimum residence time. This represents the optimal clearance time calculated for this cycle, defined as the total safe interval between the yellow light time and the all-red time. The clearance time is a derived parameter used to align with the signal parameter system and as part of the issued parameter set; its risk semantics are shared by the infeasibility risk of stopping corresponding to the yellow light and the clearance risk of the conflict zone corresponding to the all-red light, avoiding the mistaken writing of the clearance time itself as an independent risk object.

[0103] Step 4: Send the parameters to the signal controller for execution; at the same time, perform statistics and confidence upper bound verification on out-of-bounds events during operation to achieve long-term stable alignment of risk targets with operational verification and self-consistent parameter tuning mechanisms.

[0104] Specifically, after the parameters are issued and executed, the system continuously monitors parking feasibility exceeding the boundary and conflict zone occupancy exceeding the boundary events during actual operation, and estimates the corresponding exceeding probability in a statistical sense. The system further calculates the upper confidence bound of the exceeding probability to verify whether the actual operational risk meets the preset risk target.

[0105] When the upper confidence bound exceeds the risk threshold, the system automatically adjusts the quantile parameters to make the lower bound of the risk boundary more conservative. When the risk is significantly lower than the threshold and the minimum residence time constraint is met during long-term operation, the system slowly retracts parameters to improve traffic efficiency. Through this closed-loop mechanism, the statistical risk boundary remains consistent with the actual operational performance in the long term.

[0106] Confidence upper bound verification and self-consistent parameter tuning for risk targets.

[0107]

[0108] in, The out-of-bounds event indicator variable representing the risk of a yellow light, when the parking feasibility time sample corresponding to the k-th sample... Greater than the optimal yellow light duration issued in this cycle hour, Select 1 if the value is 1, otherwise select 0. The overbounded event indicator variable representing the risk of all-red zones is the time occupied by the conflict zone corresponding to the k-th sample. Greater than the optimal value of the all-red time issued in this cycle hour, Set the value to 1, otherwise set it to 0. The above indicator variable is used to count the proportion of out-of-bounds operations within the sliding window, thereby estimating the actual probability of out-of-bounds operation.

[0109] And calculate the probability estimate of exceeding the limit as shown in formula (5).

[0110]

[0111] in, and These are the probability estimates for yellow and all-red lights exceeding the risk threshold. To provide an auditable upper bound on the risk, this invention uses a normal approximation to calculate the confidence upper bound:

[0112]

[0113] in, and These are the upper bounds of the probability confidence for yellow lights and all-red lights exceeding the limit, respectively. Confidence level The corresponding standard normal quantile. The system uses... and As a criterion for determining whether a risk level is met, among them and A preset upper limit threshold for risk is set; if the threshold is not met, the quantile parameter is adjusted upwards first. or To make the lower bound more conservative, it is slowly lowered when there is long-term surplus and the minimum residence time constraint is met in order to recover efficiency.

[0114] In this embodiment, the purpose of using the normal approximation to calculate the confidence upper bound is to transform the observed out-of-bounds proportion into a risk upper bound with confidence guarantees, thereby meeting the auditability requirements of the project: even with a limited number of window samples, a conservative risk upper bound estimate can be provided to determine whether the actual operational risk meets the preset risk target. When the confidence upper bound exceeds the risk threshold, the system can trigger a self-consistent parameter tuning mechanism, such as increasing the quantile parameter to make the boundary more conservative and reduce subsequent out-of-bounds risks; when long-term operation is significantly below the threshold and meets the minimum residence time constraint, the system can slowly recover parameters to improve traffic efficiency, achieving long-term closed-loop alignment of risk and efficiency.

[0115] In this embodiment, parameter distribution and evidence chain archiving are performed as follows: The system distributes the optimized yellow light time, all-red time, and clearing time, forming a parameter set, to the road traffic signal controller for execution. Simultaneously, the system records the sample distribution, quantiles, upper confidence bounds, out-of-bounds events, and final parameters for each parameter calculation process, forming a complete risk evidence chain for use in project acceptance, operation and maintenance audits, and regulatory verification.

[0116] Specifically, the system will Composition parameter set The data is then sent to the signal control unit for execution, and at the same time, the quantiles, confidence upper limits, out-of-bounds events, and final parameters are recorded to form a risk evidence chain to support project acceptance and long-term operation and maintenance audits.

[0117] Through the above implementation method, this embodiment achieves risk-controllable adaptive adjustment of intersection clearance time without relying on a single-vehicle red light prediction model. While ensuring that the upper limit of safety risk is controlled, it improves the responsiveness of signal control parameters to changes in traffic conditions.

[0118] This embodiment replaces the single-vehicle prediction triggering logic with an engineering closed loop of risk boundary, constraint optimization, and confidence upper bound verification. This embodiment proposes constructing the risk boundary using a sliding window statistical distribution, directly using the upper bound of the out-of-bounds probability as a safety constraint parameter instead of using a single-vehicle yellow light prediction model, thus making the risk objective interpretable, auditable, and engineering-acceptable. This embodiment proposes a two-stage clearing model for yellow and all-red lights, modeling and constraining the parking feasibility risk and the risk of completely clearing the conflict zone separately, then unifying them into a single optimization framework for collaborative solution, avoiding the overly conservative or insufficient approach caused by traditional single clearing time. This embodiment couples the risk boundary with the traffic signal's executable parameter system into a clearly defined constraint optimization problem, introducing parameter change rate limits and minimum dwell time to avoid frequent jumps and control instability, enhancing engineering feasibility. This embodiment introduces confidence upper bound verification and a self-consistent parameter tuning mechanism for out-of-bounds events, aligning the statistical boundary with the actual operational performance closed loop, achieving risk self-correction and sustainable compliance in long-term operation. This embodiment's sub-technical solution records the output parameter set and risk evidence chain together, forming a traceable log for management and operation and maintenance departments, so that security and controllability are not only described by algorithms but can be quantified and verified.

[0119] Example 2

[0120] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0121] Example 3

[0122] The purpose of this embodiment is to provide a computer-readable storage medium.

[0123] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0124] Example 4

[0125] The purpose of this embodiment is to provide an adaptive optimization system for road traffic signal control parameters, including:

[0126] LiDAR, edge computing units, and signal control units installed on the roadside;

[0127] The lidar acquires point cloud data of the road intersection and transmits it to the edge computing unit;

[0128] The edge computing unit is configured to include;

[0129] The preprocessing module is configured to preprocess the acquired point cloud data of road intersections.

[0130] The trajectory and sample module is configured to: generate vehicle trajectories based on preprocessed data, and extract approach zone velocity and deceleration samples, and collision zone entry and exit times based on the vehicle trajectories.

[0131] A parking feasibility time sample set is constructed based on the speed and deceleration samples in the approach zone; a conflict zone occupancy time sample set is constructed based on the entry and exit times of the conflict zone.

[0132] The risk boundary and optimization module is configured to: obtain the lower bound of the yellow light risk boundary based on the parking feasibility time sample set, and obtain the lower bound of the all-red risk boundary based on the conflict zone occupancy time sample set.

[0133] Based on the lower bounds of the yellow light risk boundary and the full red risk boundary, risk change rate constraints and minimum residence time constraints are added.

[0134] The system calculates the yellow light and red light times online and the clearing time, and integrates the results to obtain optimized road traffic signal control parameters.

[0135] The signal controller is configured to include: a parameter receiving and execution module, the parameter receiving and execution module being configured to receive optimized road traffic signal control parameters and control traffic signals.

[0136] In one implementation example, it also includes a log and interface module, which is configured to record the sample distribution, quantiles, confidence upper bound, out-of-bounds events and final parameters in each parameter calculation process to form a complete risk evidence chain for project acceptance, operation and maintenance audit and regulatory verification.

[0137] like Figure 2 As shown, in this embodiment, at least one set of lidar is installed on the roadside of the intersection, covering the approach zone before the stop line and the conflict zone inside the intersection. The lidar is connected to the roadside edge computing unit, which interacts with the road traffic signal controller via a standard communication interface to exchange control parameters. The system operates with a fixed update cycle ΔT, forming a closed-loop control structure of perception, calculation, distribution, and verification.

[0138] Example 5

[0139] The purpose of this embodiment is to provide a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods and functions involved in any of the above embodiments.

[0140] The steps and methods involved in the apparatus of the above embodiments correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0141] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0142] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. An adaptive optimization method for road traffic signal control parameters, characterized in that, include: Acquire point cloud data of road intersections and generate vehicle trajectories; Based on the vehicle trajectory, samples of approach zone velocity and deceleration, and entry and exit times of the conflict zone are extracted. Construct a parking feasibility time sample set based on approach zone velocity and deceleration samples; Construct a sample set of conflict zone occupancy time based on the entry and exit times of the conflict zone; The lower bound of the yellow light risk boundary is obtained based on the parking feasibility time sample set, and the lower bound of the all-red risk boundary is obtained based on the conflict zone occupancy time sample set. Based on the lower bounds of the yellow light risk boundary and the full red risk boundary, risk change rate constraints and minimum residence time constraints are added. The system calculates the yellow light duration and all-red light duration online, and then integrates the results to obtain optimized road traffic signal control parameters.

2. The adaptive optimization method for road traffic signal control parameters as described in claim 1, characterized in that, Acquiring point cloud data of road intersections and forming vehicle trajectories specifically includes: collecting point cloud data of road intersections at a fixed frequency, mapping the point cloud data to the road coordinate system through external parameter calibration, denoising the point cloud data, dynamically filtering points and temporally associating them, and forming a set of vehicle trajectories.

3. The adaptive optimization method for road traffic signal control parameters as described in claim 1, characterized in that, In the road coordinate system, the approach zone and the conflict zone are divided: the approach zone is located within a certain distance before the stop line and is used to collect the approach speed and deceleration behavior of vehicles; The conflict zone is located inside the intersection and is used to track the time it takes for vehicles to enter and completely leave the conflict zone.

4. The adaptive optimization method for road traffic signal control parameters as described in claim 1, characterized in that, it further... include: By taking a specified quantile from the sample set of parking feasibility times, a lower bound of the risk boundary for yellow light time is formed. This lower bound of the risk boundary provides a constraint with a clear risk semantic for yellow light time.

5. The adaptive optimization method for road traffic signal control parameters as described in claim 1, characterized in that, Under the premise of satisfying the lower bound constraints of the risk boundaries of yellow light and all-red light respectively, the system uses the efficiency cost function as the objective to jointly optimize the yellow light time and all-red light time; a parameter change rate limit is introduced in the optimization process; and a minimum dwell time constraint is set at the same time.

6. The adaptive optimization method for road traffic signal control parameters as described in claim 1, characterized in that, it further... include: Continuously monitor parking feasibility exceeding the boundary and conflict zone occupation exceeding the boundary in actual operation, and estimate the corresponding exceeding the boundary probability; Calculate the upper confidence bound of the out-of-bounds probability to verify whether the actual operational risk meets the preset risk target; When the upper confidence bound exceeds the risk threshold, the quantile parameters are automatically adjusted to make the lower bound of the risk boundary more conservative. When the risk is significantly lower than the threshold during long-term operation and the minimum dwell time constraint is met, parameters are slowly recovered to improve traffic efficiency.

7. An adaptive optimization system for road traffic signal control parameters, characterized in that, include: LiDAR, edge computing units, and signal control units installed on the roadside; The lidar acquires point cloud data of the road intersection and transmits it to the edge computing unit; The edge computing unit is configured to include; The preprocessing module is configured to preprocess the acquired point cloud data of road intersections. The trajectory and sample module is configured to: generate vehicle trajectories based on preprocessed data, and extract approach zone velocity and deceleration samples, and collision zone entry and exit times based on the vehicle trajectories. Construct a parking feasibility time sample set based on approach zone velocity and deceleration samples; Construct a sample set of conflict zone occupancy time based on the entry and exit times of the conflict zone; The risk boundary and optimization module is configured to: obtain the lower bound of the yellow light risk boundary based on the parking feasibility time sample set, and obtain the lower bound of the all-red risk boundary based on the conflict zone occupancy time sample set. Based on the lower bounds of the yellow light risk boundary and the full red risk boundary, risk change rate constraints and minimum residence time constraints are added. The system calculates the yellow light and red light times online and the clearing time, and integrates the results to obtain optimized road traffic signal control parameters. The signal controller is configured to include: a parameter receiving and execution module, the parameter receiving and execution module being configured to receive optimized road traffic signal control parameters and control traffic signals.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method described in any one of claims 1-6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the method described in any one of claims 1-6 above.

Images