An intersection roadside laser radar credible gate signal control method and system

Abstract

The application provides a crossroad roadside laser radar credible gate signal control method and system, and belongs to the technical field of traffic signal control. The method comprises the following steps: obtaining a purified dynamic point cloud set; performing time sequence correlation on the dynamic point cloud in the set under a road coordinate system based on the purified dynamic point cloud set, and generating a vehicle target trajectory set; after completing the external parameter update determination, constructing a health diagnosis index based on the point cloud and the trajectory data under the unified external parameter; mapping the obtained health diagnosis index to a traffic parameter credibility vector, wherein each dimension of the traffic parameter credibility vector corresponds to a traffic control input parameter; under the credibility vector constraint, calculating a lane-level traffic parameter based on the vehicle trajectory data, weighting and collecting each lane-level traffic parameter to form a crossroad-level traffic state input quantity; and taking the traffic parameter credibility vector and the external parameter observability criterion as inputs to construct a hierarchical degradation signal control.

Description

Technical Field

[0001] This invention belongs to the field of traffic signal control technology, and in particular relates to a reliable gating signal control method and system for roadside lidar at intersections. 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] With the deep integration of new-generation information technology and the transportation industry, the development of intelligent transportation and the improvement of traffic safety governance capabilities have been further advanced, making signal control, operation monitoring and safety governance supported by roadside perception an important application direction.

[0004] In the field of road traffic signal control, my country has established a relatively complete standard system, which emphasizes the consistency of signal display, control equipment interfaces, and basic performance. However, when the control input comes from an external roadside sensing system, the reliability of the input data is not guaranteed by the signal controller itself, but depends on the long-term stable operation of the roadside sensors, the stability of external parameters, point cloud quality, and data link quality. If the roadside sensing gradually degrades under non-hard fault conditions, the signal control system may be driven by erroneous data without its knowledge, resulting in efficiency fluctuations or even safety risks.

[0005] In recent years, roadside lidar has been used for lane-level vehicle presence detection, speed estimation, arrival rate estimation, and queue length estimation, providing input for adaptive timing, phase extension, and coordinated control. However, in engineering deployment, roadside lidar is prone to extrinsic parameter drift due to factors such as pole tilting, loose mounting components, temperature drift, and vibration; point cloud quality degrades due to rain, fog, dust, echo attenuation, and device aging; and timestamp jitter can lead to trajectory breaks and speed estimation errors due to network and clock interference.

[0006] Existing solutions typically separate sensing output, equipment diagnostics, and signal control, or use diagnostics only for maintenance alarms rather than as part of the control loop. This leads to the following defects under long-term operating conditions: ① When extrinsic parameter drift or point cloud degradation causes deviations in key inputs such as queue length and arrival rate, the control system lacks interpretable input reliability constraints, making it prone to erroneous adaptation and efficiency fluctuations; ② When data degrades only in some lanes or distance segments, the system struggles to selectively use data at the granularity of the indicators, resulting in sudden shifts between using all data and not using it at all; ③ When sensing reliability fluctuates in the short term, the lack of hysteresis and degradation mechanisms that match signal control easily leads to frequent switching and control instability. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, the present invention provides a reliable gating signal control method and system for roadside lidar at intersections, which integrates online self-calibration, health diagnosis and signal control strategy gating of roadside lidar into a unified closed-loop system solution.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0009] In the first aspect, a reliable gating signal control method for roadside lidar at intersections is disclosed, including:

[0010] Point cloud data of intersections is acquired based on roadside lidar and the point cloud data is uniformly converted to the road coordinate system. Then, the point cloud data in the road coordinate system is purified to obtain a purified dynamic point cloud set.

[0011] Based on the purified dynamic point cloud set, temporal correlation is performed on the dynamic point cloud in the set under the road coordinate system to generate a set of vehicle target trajectories;

[0012] Based on the set of vehicle target trajectories, online self-calibration is performed using traffic flow trajectory structure constraints, and external parameter update determination is completed.

[0013] After completing the external parameter update determination, a health diagnosis index is constructed based on the point cloud and trajectory data under the unified external parameters. The health diagnosis index includes point cloud layer health index and trajectory layer health index.

[0014] The obtained health diagnostic indicators are mapped to a traffic parameter confidence vector, and each dimension of the traffic parameter confidence vector corresponds to a traffic control input parameter.

[0015] Under the credibility vector constraint, lane-level traffic parameters are calculated based on vehicle trajectory data, and the traffic parameters of each lane are weighted and summarized to form the intersection-level traffic state input.

[0016] Using traffic parameter confidence vectors and extrinsic observability criteria as inputs, a hierarchical degraded signal control state machine is constructed, including normal adaptive mode, conservative adaptive mode and backoff control mode. Based on the above modes, signal control is realized.

[0017] As a further technical solution, for each trajectory in the set of vehicle target trajectories, the average speed and driving direction angle of the vehicle are estimated in the approach zone before the stop line, and the mapping between the trajectory and the lane number is completed by combining the lane geometry prior of the intersection.

[0018] During continuous operation, multiple trajectories within the same lane are statistically analyzed for subsequent calculation of lane-level traffic parameters.

[0019] As a further technical solution, online self-calibration is performed using traffic flow trajectory structure constraints: the consistency of vehicle trajectory direction, the consistency of stop line position, and the consistency of ground plane are used as structural constraints to construct residual functions, and the perturbation of radar extrinsic parameters is solved. At the same time, an observability criterion is introduced to gate the extrinsic parameter update process: when the number of available trajectories in the approach zone is insufficient or the trajectory direction distribution is too concentrated, the extrinsic parameter update is paused to avoid erroneous updates under conditions of insufficient or degraded traffic samples.

[0020] As a further technical solution, the point cloud layer health index is used to characterize the effective point density, echo intensity stability, and outlier ratio within the region of interest.

[0021] The trajectory layer health index is used to characterize the trajectory breakage rate, speed jitter, and lane mapping consistency.

[0022] The above health indicators are all statistically calculated within a fixed time window to reflect the reliability of the current roadside perception system in depicting traffic conditions.

[0023] As a further technical solution, when mapping the obtained health diagnosis indicators to traffic parameter confidence vectors, a parameterized monotonic mapping function is used to compress and weight the health indicators, so that the confidence of various traffic parameters is normalized to the [0,1] interval.

[0024] As a further technical solution, in normal adaptive mode, adaptive timing and phase control are performed using a complete set of traffic parameter inputs;

[0025] In conservative adaptive mode, only high-confidence parameters are enabled, and constraints are imposed on the phase extension and period adjustment amplitudes.

[0026] In rollback control mode, sensor input is stopped, and preset time-segmented timing or historical statistical timing schemes are output.

[0027] Secondly, a reliable gating signal control system for roadside lidar at intersections is disclosed, comprising:

[0028] The dynamic point cloud set construction module is configured to: acquire intersection point cloud data based on roadside lidar and convert the point cloud data to the road coordinate system, and then perform purification processing on the point cloud data in the road coordinate system to obtain the purified dynamic point cloud set.

[0029] The vehicle target trajectory set generation module is configured to: based on the purified dynamic point cloud set, perform temporal correlation on the dynamic point cloud in the set in the road coordinate system to generate the vehicle target trajectory set;

[0030] The online self-calibration module is configured to: perform online self-calibration based on the set of vehicle target trajectories and utilize traffic flow trajectory structure constraints, and complete the external parameter update determination;

[0031] The health diagnostic indicator construction module is configured to: after completing the external parameter update determination, construct health diagnostic indicators based on point cloud and trajectory data under unified external parameters, wherein the health diagnostic indicators include point cloud layer health indicators and trajectory layer health indicators.

[0032] The mapping module is configured to map the obtained health diagnosis indicators into a traffic parameter confidence vector, wherein each dimension of the traffic parameter confidence vector corresponds to a traffic control input parameter.

[0033] Under the credibility vector constraint, lane-level traffic parameters are calculated based on vehicle trajectory data, and the traffic parameters of each lane are weighted and summarized to form the intersection-level traffic state input.

[0034] The control module is configured to construct a hierarchical degraded signal control state machine, including a normal adaptive mode, a conservative adaptive mode, and a backoff control mode, based on the traffic parameter confidence vector and the external parameter observability criterion, and to control the signal based on the above modes.

[0035] The above one or more technical solutions have the following beneficial effects:

[0036] In the context of urban signalized intersections, addressing the engineering reliability and safety controllability requirements of signal control input data, this invention utilizes roadside lidar point clouds to achieve online self-calibration and health diagnosis. The diagnostic results are then transformed into a traffic parameter credibility vector, which is used as a gating variable to embed adaptive traffic signal control and hierarchical degradation control. This enables a smooth switching method and system between normal adaptive, conservative adaptive, and backoff control.

[0037] 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

[0038] 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.

[0039] Figure 1 This is an overall flowchart of the method in an embodiment of the present invention;

[0040] Figure 2 This is a system structure block diagram according to an embodiment of the present invention;

[0041] Figure 3This is a schematic diagram of traffic flow structure constraints and self-calibration residuals in an embodiment of the present invention. Detailed Implementation

[0042] 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.

[0043] 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.

[0044] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0045] Example 1

[0046] See appendix Figure 1 As shown, this embodiment discloses a reliable gating signal control method for roadside lidar at intersections. The flowchart includes a complete link encompassing point cloud acquisition and preprocessing, trajectory generation, self-calibration constrained by observability criterion, construction of health indicators, reliability mapping, calculation of lane-level traffic parameters, calculation of gating quantity, hierarchical degradation control state machine, and parameter distribution and execution. Specifically, it includes:

[0047] Point cloud data of intersections is acquired based on roadside lidar and the point cloud data is uniformly converted to the road coordinate system. Then, the point cloud data in the road coordinate system is purified to obtain a purified dynamic point cloud set.

[0048] Based on the purified dynamic point cloud set, temporal correlation is performed on the dynamic point cloud in the set under the road coordinate system to generate a set of vehicle target trajectories;

[0049] Based on the set of vehicle target trajectories, online self-calibration is performed using traffic flow trajectory structure constraints, and external parameter update determination is completed.

[0050] After completing the external parameter update determination, a health diagnosis index is constructed based on the point cloud and trajectory data under the unified external parameters. The health diagnosis index includes point cloud layer health index and trajectory layer health index.

[0051] The obtained health diagnostic indicators are mapped to a traffic parameter confidence vector, and each dimension of the traffic parameter confidence vector corresponds to a traffic control input parameter.

[0052] Under the credibility vector constraint, lane-level traffic parameters are calculated based on vehicle trajectory data, and the traffic parameters of each lane are weighted and summarized to form the intersection-level traffic state input.

[0053] Using traffic parameter confidence vectors and extrinsic observability criteria as inputs, a hierarchical degraded signal control state machine is constructed, including normal adaptive mode, conservative adaptive mode and backoff control mode. Based on the above modes, signal control is realized.

[0054] This embodiment's sub-solution collects LiDAR point clouds at the roadside of intersections and forms vehicle trajectories. During continuous operation, it utilizes traffic flow structure constraints to achieve online extrinsic parameter perturbation estimation. Simultaneously, it constructs health indicators for the point cloud and trajectory layers and maps them into a traffic parameter credibility vector for control. The credibility is used as a gating variable to drive control mode selection and control input set pruning, smoothly switching between normal adaptive, conservative adaptive, and backoff control. It also maintains consistent archiving of extrinsic parameter update status, health indicators, credibility, and control parameters to form an auditable closed loop, thereby maintaining stable, interpretable, and safe control of signal control even under sensor degradation or drift conditions.

[0055] A more detailed embodiment of a trusted gating signal control method for roadside lidar at intersections, implemented by a roadside edge computing unit and linked with a road traffic signal controller, includes the following steps:

[0056] Step 1: Deployment of the roadside sensing system and point cloud acquisition.

[0057] A roadside lidar sensing system is deployed on the approach lane poles of a typical signalized intersection on an urban main road. The roadside lidar sensing system includes lidar and a roadside edge computing unit.

[0058] The lidar is installed at a height of approximately 6–8 m, covering an approach zone of about 60 m in front of the stop line when viewed from above. The lidar is connected to a roadside edge computing unit via Ethernet, which receives and analyzes the 3D point cloud data output by the lidar at fixed intervals.

[0059] The edge computing unit performs ground separation, static background suppression, and outlier filtering on the acquired raw point cloud to obtain a purified dynamic point cloud set, providing basic data for subsequent vehicle trajectory generation. See the attached system architecture diagram. Figure 2 As shown, this illustrates the integration relationship between roadside lidar, edge computing units, communication networks, and signal control units.

[0060] Specifically, at any given moment For point clouds The purified point cloud is obtained by performing ground separation, static background suppression, and outlier filtering. .

[0061] In this embodiment, lidar point cloud data is collected at fixed time intervals, and the point clouds are uniformly converted to road coordinate system for expression.

[0062] To ensure symbol consistency, the discrete update time is defined as... Roadside radar point cloud data is ,in For the first The point in the radar coordinate system The three-dimensional coordinate vector below, This represents the number of points in the frame; the road coordinate system is defined as follows: Radar extrinsic parameters are rigid body transformations ,in For rotation matrix, Let be the translation vector; define the road coordinates after the extrinsic parameter transformation as . And satisfy the following formula.

[0063] (1)

[0064] in, For point In the road coordinate system The coordinate vector below, For rotation matrix, It is a translation vector.

[0065] In the specific implementation process, rotation matrix With translation vector The acquisition of data typically employs a combination of pre-calibration and feature adaptation: First, by utilizing lane lines or pre-set targets with known geometric features within the monitoring area, the RANSAC (Random Sample Consensus) algorithm is used to fit the road plane equation to calculate the installation height, pitch angle, and roll angle of the lidar. Then, the axial orientation of the road coordinate system is determined by combining this with the statistically significant principal direction of the vehicle's trajectory, thereby calculating the extrinsic parameter matrix reflecting the sensor's initial pose offline. In implementation schemes involving the spatiotemporal correlation of multi-frame point clouds... The rotation transformation matrix can be constructed from the change in the target's heading angle between adjacent sampling times. It can then be dynamically calculated based on the product of the target's real-time velocity vector and the sampling time interval, thereby achieving precise unification of point cloud data in both spatial and temporal dimensions.

[0066] The purpose of step 1 above is to establish a unified spatiotemporal reference and a high-purity dynamic point set for all subsequent calculations: on the one hand, it transforms the point cloud data into the road coordinate system, providing unified coordinates for lane mapping, stop line / lane geometric constraints, and self-calibration residual construction; on the other hand, it obtains a purified dynamic point cloud set through ground separation, background suppression, and outlier filtering, reducing the interference of static facilities and noise on trajectory correlation, improving trajectory continuity and statistical stability of health indicators, thereby improving the reliability of subsequent credibility gating and control decisions.

[0067] Step 2: Vehicle trajectory generation and lane mapping.

[0068] Using the cleaned point cloud obtained in step 1 as input, the system performs temporal correlation on the dynamic point cloud in the road coordinate system to generate a set of vehicle target trajectories. For each trajectory, the average speed and driving direction angle of the vehicle are estimated in the approach zone before the stop line, and the mapping between the trajectory and the lane number is completed by combining the lane geometry prior of the intersection.

[0069] The mapping from trajectory to lane can be achieved based on lane geometric priors, such as lane centerlines / lane polygon regions: a set of trajectory points is selected in the approach zone before the stop line, the trajectory points are projected onto the centerlines of each lane and the average lateral distance / direction angle is calculated, and the lane with the smallest distance and the highest direction consistency is selected as the lane number of the trajectory; when there are cross-lane / lane change situations, a majority frame attribution and hysteresis strategy can be adopted to keep the lane number stable under short-term noise.

[0070] During continuous operation, the system statistically analyzes multiple trajectories within the same lane for subsequent calculation of lane-level traffic parameters. This trajectory generation and lane mapping process provides a unified spatiotemporal basis for subsequent online self-calibration, health diagnosis, and signal control.

[0071] More specifically, for the purified point cloud, it is unified to the road coordinate system to form a set using formula (1). .based on The dynamic point set is used to perform time-series correlation to obtain the target trajectory set. ,in Indicates the first The trajectory sequence of a target within a time window This represents the number of tracks within the window. For each track... Estimate its average speed in the approach area before the stop line. With direction angle The trajectory lane number is obtained based on lane geometry priors. To form a subset of lanes The lane assignment result will be used in step 5 for lane-level traffic parameter calculation.

[0072] Specifically, temporal correlation can be achieved through single-frame clustering and cross-frame data correlation: clustering is performed on the dynamic point cloud of each frame to obtain target clusters, specifically features such as centroids or bounding boxes. Then, cross-frame matching is performed based on motion model predictions such as uniform velocity / Kalman prediction and spatial thresholds such as nearest neighbor or maximum IoU matching. Consistent IDs are assigned to the same target and they are concatenated to form a trajectory sequence. When occlusion or frame loss occurs, trajectory breakage can be reduced by short-term trajectory maintenance and reconnection thresholds.

[0073] Step 2 outputs a set of vehicle target trajectories, including lane number, average speed / direction angle in the approach zone, and other attributes. Its advantage is that it improves the instability of single-frame point cloud into continuous and statistically significant trajectory data, thereby providing structural constraints such as trajectory direction consistency / stop line consistency for Step 3 to perform online self-calibration and observability criterion calculation; and providing trajectory layer health indicators such as trajectory breakage rate, speed jitter, and lane mapping consistency for Step 4; and providing a sample basis for calculating traffic parameters such as lane arrival rate, speed, occupancy rate, and queuing for Step 6. Finally, it enters Step 7 for state machine gating and parameter distribution.

[0074] Step 3: Online self-calibrated extrinsic parameter perturbation estimation and observability-gated update.

[0075] During long-term operation, the external parameters of the roadside lidar may experience slight drift due to factors such as slight tilting of the poles, loosening of installation components, and environmental vibration. To ensure the long-term consistency of the point cloud and trajectory in the road coordinate system, the system performs online self-calibration using traffic flow trajectory structure constraints without deploying artificial targets or affecting traffic operation.

[0076] Regarding online self-calibration, the residual function is constructed by constraining the natural structure of traffic flow trajectories: the clustering of trajectories in the same lane direction is used as the directional consistency constraint, the spatial position stability of deceleration / stop strips near the stop line is used as the stop line consistency constraint, and the plane stability of ground point fitting is used as the ground consistency constraint; the external parameter perturbation is solved under the condition of no target placement and no traffic interference, so that the point cloud has a long-term unified expression that is not distorted by slow drift such as slight tilting of poles.

[0077] In this embodiment, the consistency of vehicle trajectory direction, the consistency of stop line position, and the consistency of ground plane are used as structural constraints to construct the residual function, and the perturbation of radar extrinsic parameters is solved. At the same time, an observability criterion is introduced to gate the extrinsic parameter update process: when the number of available trajectories in the approach zone is insufficient or the trajectory direction distribution is too concentrated, the system suspends the extrinsic parameter update to avoid erroneous updates under conditions of insufficient or degraded traffic samples.

[0078] like Figure 3 As shown, by introducing an observability gating mechanism, the system can smoothly correct extrinsic parameters when traffic flow structure conditions are met, and keep extrinsic parameters frozen when conditions are not met, thereby preventing online self-calibration from deviating too much from the target.

[0079] Specifically, observability gating means that when the number of available trajectories in the approach zone is insufficient or the trajectories are too concentrated, the structural constraints degenerate and the self-calibration solution becomes unstable. Therefore, the external parameters are paused and frozen. The external parameters are only updated when the number of trajectories is sufficient and the directional distribution meets the threshold and the residual decreases, so as to avoid the deviation from calibration when there is low traffic at night or occlusion degradation.

[0080] Considering that engineering drift is usually small, the extrinsic parameter updates are defined in perturbation form. And construct the structural constraint residual function. The residual is composed of lane direction consistency residuals. Stop line consistency residual Residual with ground plane The weighted composition is as follows:

[0081]

[0082] in, , , These are non-negative weighting coefficients. , , These represent the direction consistency residual, stop line consistency residual, and ground plane residual, respectively. The optimal perturbation solution for the system is given by the following formula:

[0083]

[0084] in, For rotational perturbation, For translational perturbation, The optimal perturbation solution is used to update the current extrinsic parameters, obtaining the extrinsic parameters for unified transformation of the point cloud and trajectory generation in the next update cycle, thereby suppressing long-term extrinsic parameter drift; it is also used to form the extrinsic parameter update status / update amount and enter the recording and consistency archiving in step 8; the perturbation size and residual reduction rate can also be used as auxiliary quantities for health diagnosis or control gating, for example, if the perturbation is too large or the residual does not decrease, freeze update / degradation control will be triggered.

[0085] To avoid erroneous updates due to insufficient observability, this invention defines an observability index. It is used for gating extrinsic parameter updates, and an implementable form is given below:

[0086] .

[0087] in, This represents the number of available tracks within the window. It is a positive constant used for smoothing at low flow rates. This is the sensitivity coefficient. The set of trajectory direction angles within the window. Let be the sample variance of the set of orientation angles.

[0088] Number of available tracks within the window The sample variance of the orientation angle set is used to reflect whether the sample is sufficient. It is used to reflect whether the trajectory direction distribution is sufficiently dispersed and whether it can provide observable structural constraints such as direction / stop line. When Very small or When the direction is too concentrated, the structural constraints will degenerate, and the self-calibration results are prone to instability. Therefore, the criteria should be reduced and a freeze update should be triggered. Used for smoothing under low flow conditions to avoid numerical instability or division by zero issues; Used to adjust the sensitivity of the criterion to changes in directional distribution. The threshold can be calibrated using historical data or set based on engineering experience.

[0089] when Update the external parameters to:

[0090]

[0091] when External parameters are not updated to avoid becoming increasingly biased or skewed.

[0092] Step 4: Construction of point cloud and trajectory health diagnostic indicators.

[0093] After completing the extrinsic parameter update determination in step 3, the system constructs health diagnostic indicators based on point cloud and trajectory data under unified extrinsic parameters. The health indicators include two levels: the point cloud layer and the trajectory layer.

[0094] Among them, the point cloud layer health index is used to characterize the effective point density, echo intensity stability, and outlier ratio within the region of interest; the trajectory layer health index is used to characterize the trajectory breakage rate, speed jitter, and lane mapping consistency. All of these health indices are statistically calculated within a fixed time window to reflect the reliability of the current roadside perception system's characterization of traffic conditions.

[0095] In this embodiment, the system uses the residual statistics generated during the online self-calibration process as supplementary information or weight adjustment factors for health indicators, so that the health diagnosis results can simultaneously reflect the impact of external parameter stability on perceived quality.

[0096] In this embodiment, the health indicator vector is defined as:

[0097]

[0098] in, Represents the effective point density index and is defined as follows: ,in The effective points after purification. Area of ​​the region of interest; This indicates the stability index of echo intensity; Indicator representing outlier rate; The trajectory breakage rate index reflects the discontinuity of the trajectory caused by frame drops and timestamp jitter. The calculation of each of the above indices is based on the trajectory set generated after the external parameters are confirmed in step 2, to ensure that the timing of the health indicators is consistent with the control input.

[0099] Health indicators are statistically obtained from the purified point cloud in step 1 and the trajectory set in step 2 within a fixed window: echo intensity stability can be characterized by the variance / dispersion of the intensity distribution within the window; outlier rate is the proportion of outlier points; trajectory breakage rate can be calculated by the number of frame breaks within the window for the same target ID / total number of frames; speed jitter can be calculated by the variance of the trajectory speed sequence; lane mapping consistency can be calculated by the rate of change of lane number for the same trajectory. All of the above statistics can be calculated online in real time.

[0100] In one implementation, based on the trajectory set generated in step 2, the system calculates the direction consistency residual, stop line consistency residual, and ground plane residual in step 3, and calculates the mean, variance, or quantile of these residuals within a window; the above residual statistics can be used as supplementary items for health diagnostic indicators, or as auxiliary quantities for updating the weights / extrinsic parameters of health indicators.

[0101] Step 5: Traffic parameter credibility vector mapping.

[0102] To directly apply the health diagnosis results to signal control, the system maps the health indicators obtained in step 4 into a traffic parameter confidence vector. Each dimension of the confidence vector corresponds to traffic control input parameters such as queue length, arrival rate, speed, and occupancy rate.

[0103] In this embodiment, the system uses a parameterized monotonic mapping function to compress and weight health indicators, normalizing the credibility of various traffic parameters to the [0,1] interval. Through this mapping method, the influence relationship between different health indicators and different traffic parameters can be explicitly modeled, thereby achieving selective credibility constraints at the parameter granularity, rather than a simple binary judgment of availability / unavailability.

[0104] Regarding the reliability vector mapping from health indicators to traffic parameters.

[0105] Define the traffic parameter confidence vector as follows:

[0106]

[0107] in , , , These correspond to the reliability of queue length, arrival rate, speed, and occupancy rate, respectively. To ensure interpretability, this invention employs a parameterizable monotonic mapping. An element-wise compression form is given below:

[0108]

[0109] in For the mapping matrix, For bias vectors, This is an element-wise compression function that ensures the confidence level falls within the [0,1] interval.

[0110] Element-wise compression function The purpose is to normalize the result after linear mapping to the range of 0 to 1, ensuring that the reliability output is monotonic and comparable, and facilitating direct connection with subsequent gating thresholds. An feasible but not limited example can be used, such as the sigmoid function (1 divided by 1 plus a negative exponent of e), or a piecewise linear function (mapped to 0 below the threshold, mapped to 1 above the threshold, with a linear transition in between). In practical engineering, one of these methods can be chosen based on the data distribution.

[0111] The mapping matrix has the same number of rows as the traffic parameter credibility vector and the same number of columns as the health indicator vector, which is used to describe the influence of different health indicators on the credibility of various traffic parameters.

[0112] The mapping process includes: first, obtaining a health indicator vector statistically within a fixed time window; second, multiplying it by a linear transformation. In addition First, aggregate the multidimensional health indicators according to the traffic parameter dimension, with each row of A corresponding to a traffic parameter confidence component; second, apply a compression function element by element to obtain the confidence vector. , Each dimension corresponds one-to-one with a control input parameter, such as queuing, arrival rate, speed, and occupancy rate, which are used for subsequent weighting, pruning, and control gating.

[0113] Step 6: Calculate lane-level traffic parameters under credibility constraints.

[0114] Under the credibility vector constraint, the system calculates lane-level traffic parameters based on vehicle trajectory data, including lane arrival rate, average speed, occupancy rate, and queue length. To ensure the stability of the intersection-level signal control input, the system weights and summarizes the parameters of each lane to form the intersection-level traffic state input.

[0115] It should be noted that each type of traffic parameter is constrained by its corresponding confidence component when participating in control decisions. There are three approaches: first, it is used as a weight in the aggregation, with lower confidence levels resulting in smaller weights; second, it is used as a gating condition, preventing the parameter from entering the control input set when the confidence level is below a threshold; third, it is used as a degradation condition, allowing the parameter to influence control in conservative mode only when the confidence level is in the middle range, and limiting the magnitude of control adjustments. These constraints prevent degraded data-driven control decisions from being made.

[0116] The system takes the target trajectory set as input and calculates features such as average speed, direction angle, and entry / exit timestamps for each trajectory within the approach zone before the stop line. Trajectories are then grouped and statistically analyzed by lane number. Lane-level traffic parameters such as lane arrival rate, average lane speed, lane occupancy rate, and lane queuing / tail end are calculated within a fixed window and then aggregated into intersection-level or phase-level inputs for subsequent phase demand calculations and hierarchical control state machines.

[0117] When the confidence level of a certain type of traffic parameter is low, the system automatically reduces the weight of that parameter in subsequent control calculations, or removes it from the control input set when necessary, thereby avoiding the direct driving of signal control strategies by degraded perception data.

[0118] For each lane Define the statistical window length as ,based on Calculate lane arrival rate:

[0119]

[0120] in For those entering the approach area within the window and whose lane number is The number of trajectories.

[0121] Calculate the average speed of the lane:

[0122]

[0123] in, It is the lane number; lane The effective duration within the window; Indicates the number of elements belonging to this set. The sampling time or the first sampling time Track index; This represents the observed velocity value at the corresponding sampling time or the average velocity of the corresponding trajectory within the proximity zone. Indicates the lane exit at update time t. The average window speed.

[0124] Calculate market share:

[0125]

[0126] in This represents the total time spent within the window.

[0127] Calculate queue length ,in This is the distance between the tail of the queue and the stop line.

[0128] To align with the confidence vector, the intersection-level input is defined as follows:

[0129]

[0130] in, For lane Queuing-related parameters (such as queue length or number of queues); For lane Arrival rate; For lane The average speed; For lane market share; Lane weights can be determined by lane function type, saturation flow rate, or phase assignment, and can be dynamically weighted by multiplying by the corresponding confidence level. , , , This is an intersection-level aggregated input, used for subsequent calculation of phase demand and execution of hierarchical control decisions.

[0131] Step 7: Layered degradation signal control for credibility gating.

[0132] The system takes traffic parameter confidence vector and external parameter observability criteria as inputs to construct a hierarchical degraded signal control state machine, including normal adaptive mode, conservative adaptive mode and backoff control mode.

[0133] In normal adaptive mode, the system uses the complete set of traffic parameter inputs to perform adaptive timing and phase control; in conservative adaptive mode, the system only enables high-confidence parameters and constrains the phase extension and cycle adjustment magnitude; in backoff control mode, the system stops using sensing inputs and outputs preset time-segmented timing or historical statistical timing schemes.

[0134] To avoid instability caused by frequent switching, the system sets entry and exit thresholds during mode switching and introduces a minimum dwell time mechanism to achieve a smooth transition between control modes.

[0135] Define control mode ,in This is the normal adaptive mode. It is a conservative adaptive mode. This is for fallback mode. Define the comprehensive gating quantity. As shown below:

[0136]

[0137] in The observability index is the extrinsic observability criterion defined in step 2, which is used to reflect the degree to which the current traffic flow structure supports online self-calibration and perception reliability. and These are arrival rate and queue reliability, respectively. , , The weighting coefficients are satisfied. System settings entry threshold With exit threshold and satisfy And set a minimum stay time Used for hysteresis gating.

[0138] The observability index is the extrinsic observability criterion defined in step 3 of this invention. It is obtained by statistically analyzing the trajectory set generated in step 2 within the window, used for gating extrinsic parameter updates, and as one of the inputs to the control state machine.

[0139] In this implementation example, the health index vector is first calculated in step 4, and then the confidence components corresponding to each traffic parameter are obtained through mapping relationships. The component corresponding to the arrival rate is denoted as... The corresponding queuing weight is denoted as In other words, and These are not additional, manually set quantities, but rather two confidence components output by the health indicators through a mapping matrix according to the parameter dimensions.

[0140] when And continue for no less than Time to enter and allows the use of control input sets ;when lie in or Time to enter And based on the credibility, the input set is trimmed to obtain ;when or multiple consecutive cycles Time to enter It can output preset time-segmented timing or historical statistical timing.

[0141] This represents the complete set of control inputs, including intersection-level queuing input, intersection-level arrival rate input, intersection-level speed input, and intersection-level occupancy rate input. These inputs are aggregated in step 6, and each item has a corresponding confidence component in step 5, used for subsequent gating, weighting, or pruning to support switching between normal adaptive mode and conservative / backoff mode.

[0142] This represents the subset of input that has been pruned based on credibility, satisfying... yes A subset of. The pruning rule can be explicitly stated as: if the confidence level of an input is below a threshold, then from... Remove the input from the list; if the confidence level is in the middle range, retain it but reduce its weight and limit its influence on the control parameters. (Pruned) Decisions are made for conservative adaptive or backoff modes, thereby enabling the selective use of inputs based on reliability.

[0143] exist and In this mode, the system uses phase sets Generation cycle and green light parameters, the cycle is defined as follows Define phase Effective green light is and satisfy ,in To compensate for time loss, the system employs a reliability-weighted phase demand. Based on this, green lights are allocated, and one feasible implementation is given as follows:

[0144]

[0145] Where Dt,p represents the confidence-weighted demand for phase p at update time t; qt,p represents the queue aggregation quantity of the lane group corresponding to phase p (obtained by summarizing lane queues according to phase affiliation); ft,p represents the arrival rate aggregation quantity of the lane group corresponding to phase p; Ctq and Ctf are the queue and arrival rate confidence quantities, respectively. This is a balancing factor used to adjust the contribution of arrival rate to demand; it can be set based on historical calibration or engineering experience. If a certain confidence level is too low, the corresponding item can be removed or its weight significantly reduced to avoid unreliable inputs affecting green light allocation.

[0146] In addition to valid green lights, a cycle also includes time periods that cannot be used for passage, such as yellow lights, all red lights, and vehicle start-up losses. The total of these time periods is defined as lost time. Yellow and all-red lights can be obtained directly from signal display parameters or determined from predetermined timing parameters; starting losses can be set based on traffic engineering experience or historical statistics. Therefore... It can be determined by signal display parameters and empirical / statistical items, and used to calculate the total number of available effective green lights.

[0147] in and Phases Queuing and arrival rates for corresponding entrance lanes or lane groups This is the arrival rate weighting coefficient.

[0148]

[0149] It's time to update. Lower phase The allocated green light; Phase The minimum guaranteed green light value; It is a cycle; It is a set of phases; It is the lost time; the minimum green light time for all phases is... ; the text in parentheses Subtract the minimum green light total and then subtract This indicates the remaining valid green lights that can be allocated proportionally to demand. It is the phase requirement. It is the sum of the demand for all phases. If a phase is lower than the minimum green light requirement after allocation, it will be truncated according to the minimum green light constraint, and the remaining phases will be reallocated.

[0150] The algorithm described above uses the confidence vector and the external parameter observability criterion as a unified gating input, enabling the system to automatically reduce the aggressiveness of control and downgrade to a conservative or fallback mode according to rules when the perception degrades or the sample is insufficient, thus avoiding erroneous control driven by false perception. At the same time, by forming a hysteresis through the entry / exit threshold and the minimum dwell time, it suppresses frequent switching caused by short-term fluctuations and improves control stability. Furthermore, it is linked with the record archiving in step 8 to achieve an interpretable and auditable engineering closed loop.

[0151] Step 8: Closed-loop consistency verification and auditable output.

[0152] Within each control update cycle, the system consistently records and archives external parameter update status, health indicators, traffic parameter reliability, control mode, and issued signal control parameters. For example... Figure 2 As shown, this recording mechanism covers the entire link between roadside lidar, edge computing unit and signal control unit, which facilitates subsequent project acceptance, operation analysis and maintenance audit.

[0153] Specifically, the system records the external parameter update status in each update cycle. , , , , and the distribution of parameter sets And maintain consistent records to meet the traceability requirements for project operation, maintenance and acceptance.

[0154] in, It is the set of parameters issued at update time t; Lt is the period; P is the set of valid green lights for each phase; P is the set of phases. It specifies the control mode (normal, conservative, rollback). It also indicates that the system records the external parameter update status in each update cycle. Health indicator vector Ht, credibility vector Ct, gating quantity Control Mode and the distribution of parameter sets It is used for operation and maintenance diagnosis, acceptance audit and post-event traceability.

[0155] Through the above steps, this embodiment achieves reliable gating and hierarchical degradation control of signal control input under the condition of roadside lidar perception degradation without relying on manual calibration and intervention, so that the intersection signal control system has stability, interpretability and safe rollback capability in long-term operation.

[0156] In this embodiment, to address the issue of extrinsic parameter drift / point cloud degradation causing deviations in key inputs such as queue length and arrival rate, and the lack of interpretable credibility constraints, this embodiment introduces traffic flow structure constraints and observability gating updates in step 3 of online self-calibration. In step 4 of health diagnosis and step 5 of credibility mapping, it constructs point cloud layer / trajectory layer health indicators and maps them to traffic parameter credibility vectors, enabling the control side to impose explicit credibility constraints on the inputs, thereby suppressing error adaptation and efficiency fluctuations.

[0157] To address the problem in existing technologies where selective use at the index granularity is difficult when only some lanes / distance segments degrade, this embodiment designs the confidence vector in step 5 to output it according to the traffic parameter dimension. Here, queuing, arrival rate, speed, occupancy rate, etc., each correspond to a one-dimensional confidence. In step 6, the confidence is used for weighted summarization of lane-level parameters and pruning when necessary, so as to achieve selective activation at the index granularity and avoid sudden changes from using all to not using at all.

[0158] To address the issue of frequent switching caused by the lack of hysteresis and degradation mechanisms due to short-term reliability fluctuations, this example constructs a hierarchical degradation state machine with reliability gating in step 7, sets entry / exit thresholds and minimum dwell time, and in step 8, records and archives the external parameter update status, health indicators, reliability, control mode and issued parameters in a consistent manner, forming a control loop that is rollbackable and auditable.

[0159] Regarding roadside lidar: It refers to laser ranging and scanning equipment installed on poles, crossarms or special brackets near intersections. Its output is three-dimensional point cloud data sampled in time series, which is used to characterize the dynamic and static structure within the road space.

[0160] Online self-calibration refers to the process of estimating and updating the small drift of radar extrinsic parameters online, without deploying artificial targets or interrupting traffic flow, by utilizing the structural constraints formed by traffic flow trajectories and road geometry priors, so that the point cloud can be represented in road coordinates in a long-term stable manner.

[0161] Structural constraints include: trajectory direction consistency constraints, where trajectory direction angles within the same lane should converge; stop line consistency constraints, where the spatial position of vehicle deceleration / stopping strips near the stop line should be relatively stable in the road coordinate system, serving as a soft constraint on the stop line position; and ground plane consistency constraints, where the residuals of the plane fitted by ground points should be stable, used to constrain extrinsic parameter components such as pitch and altitude. These constraints are obtained from the trajectory set and point cloud geometric statistics, and together constitute an online self-calibrated residual function to solve for extrinsic parameter perturbations.

[0162] Regarding health indicators: These are a set of statistics used to quantify the quality of point clouds and trajectories. They include point cloud layer indicators such as point density, intensity distribution, noise outlier rate, and frame drop rate, as well as trajectory layer indicators such as trajectory breakage rate, speed jitter, and lane mapping consistency.

[0163] Regarding the traffic parameter confidence vector: it refers to the quantitative vector that maps health indicators to control availability, where each dimension corresponds to the confidence level or error bound of a traffic control input, such as queue length confidence, arrival rate confidence, speed confidence, and occupancy confidence, which are used to determine whether the gating control algorithm enables the input and the intensity of its activation.

[0164] Regarding the observability criterion: it refers to the judgment quantity used to determine whether the current traffic flow and point cloud quality can support reliable online self-calibration. It is used to decide whether to allow the update of external parameters, thereby avoiding erroneous updates under conditions of insufficient samples or degradation.

[0165] The observability criterion can be specifically defined as a criterion consisting of sample sufficiency and orientation distribution sufficiency: sample sufficiency is characterized by the number of available trajectories within the window, and orientation distribution sufficiency is characterized by the variance or distribution entropy of the set of trajectory orientation angles. When the number of available trajectories within the window is too small or the variance of the set of trajectory orientation angles is too small, it is judged as unobservable and the update of extrinsic parameters is frozen; updates of extrinsic parameters are only allowed when the criterion reaches the threshold and the residual convergence condition is met, thereby avoiding erroneous updates under conditions of insufficient samples or degradation.

[0166] Trustworthiness gating refers to a mechanism that uses a trustworthiness vector as input to select, prune, and assign weights to control modes and control input sets. Its goal is to limit the aggressiveness of control strategies and achieve smooth rollback when data degrades, thereby avoiding control instability and security risks.

[0167] The innovation of this invention lies in its closed-loop integration of diagnosis, reliability, gating, and control. First, it introduces an observability criterion to constrain online self-calibration updates, updating extrinsic parameters only when the traffic flow geometry is sufficient to support reliable calibration, avoiding the engineering risk of online calibration becoming increasingly biased under degenerate conditions. Second, it elevates health diagnosis from equipment status to control-oriented traffic parameter error boundaries and reliability vectors, enabling selective activation at the indicator granularity, rather than simple fault alarms or a single overall score. Third, it uses the reliability vector as a signal control gating variable to construct a hierarchical degradation control state machine with hysteresis and minimum dwell time, giving the control system rollback capability and preventing erroneous control driven by incorrect data. Fourth, it introduces structured pruning and weight adjustment of the control input set during mode switching, allowing the system to smoothly transition from fine adaptive to conservative adaptive and rollback control within the same control framework, and suppressing abrupt changes in control performance. Fifth, it forms a traceable log for gating decisions, extrinsic parameter updates, and control parameter distribution, facilitating engineering acceptance and long-term operation and maintenance audits.

[0168] Example 2

[0169] 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.

[0170] Example 3

[0171] The purpose of this embodiment is to provide a computer-readable storage medium.

[0172] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0173] Example 4

[0174] The purpose of this embodiment is to provide a reliable gating signal control system for roadside lidar at intersections, including:

[0175] The dynamic point cloud set construction module is configured to: acquire intersection point cloud data based on roadside lidar and convert the point cloud data to the road coordinate system, and then perform purification processing on the point cloud data in the road coordinate system to obtain the purified dynamic point cloud set.

[0176] The vehicle target trajectory set generation module is configured to: based on the purified dynamic point cloud set, perform temporal correlation on the dynamic point cloud in the set in the road coordinate system to generate the vehicle target trajectory set;

[0177] The online self-calibration module is configured to: perform online self-calibration based on the set of vehicle target trajectories and utilize traffic flow trajectory structure constraints, and complete the external parameter update determination;

[0178] The health diagnostic indicator construction module is configured to: after completing the external parameter update determination, construct health diagnostic indicators based on point cloud and trajectory data under unified external parameters, wherein the health diagnostic indicators include point cloud layer health indicators and trajectory layer health indicators.

[0179] The mapping module is configured to map the obtained health diagnosis indicators into a traffic parameter confidence vector, wherein each dimension of the traffic parameter confidence vector corresponds to a traffic control input parameter.

[0180] Under the credibility vector constraint, lane-level traffic parameters are calculated based on vehicle trajectory data, and the traffic parameters of each lane are weighted and summarized to form the intersection-level traffic state input.

[0181] The control module is configured to construct a hierarchical degraded signal control state machine, including a normal adaptive mode, a conservative adaptive mode, and a backoff control mode, based on the traffic parameter confidence vector and the external parameter observability criterion, and to control the signal based on the above modes.

[0182] Example 5

[0183] 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.

[0184] 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.

[0185] 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.

[0186] 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. A reliable gating signal control method for roadside lidar at intersections, characterized in that, include: Point cloud data of intersections is acquired based on roadside lidar and the point cloud data is uniformly converted to the road coordinate system. Then, the point cloud data in the road coordinate system is purified to obtain a purified dynamic point cloud set. Based on the purified dynamic point cloud set, temporal correlation is performed on the dynamic point cloud in the set under the road coordinate system to generate a set of vehicle target trajectories; Based on the set of vehicle target trajectories, online self-calibration is performed using traffic flow trajectory structure constraints, and external parameter update determination is completed. After completing the external parameter update determination, a health diagnosis index is constructed based on the point cloud and trajectory data under the unified external parameters. The health diagnosis index includes point cloud layer health index and trajectory layer health index. The obtained health diagnostic indicators are mapped to a traffic parameter confidence vector, and each dimension of the traffic parameter confidence vector corresponds to a traffic control input parameter. Under the credibility vector constraint, lane-level traffic parameters are calculated based on vehicle trajectory data, and the traffic parameters of each lane are weighted and summarized to form the intersection-level traffic state input. Using traffic parameter confidence vectors and extrinsic observability criteria as inputs, a hierarchical degraded signal control state machine is constructed, including normal adaptive mode, conservative adaptive mode and backoff control mode. Based on the above modes, signal control is realized.

2. The reliable gating signal control method for roadside lidar at intersections as described in claim 1, characterized in that, For each trajectory in the set of vehicle target trajectories, estimate the average speed and driving direction angle of the vehicle in the approach zone before the stop line, and combine the lane geometry prior of the intersection to complete the mapping between the trajectory and the lane number; During continuous operation, multiple trajectories within the same lane are statistically analyzed for subsequent calculation of lane-level traffic parameters.

3. The reliable gating signal control method for roadside lidar at intersections as described in claim 1, characterized in that, Online self-calibration is performed using traffic flow trajectory structure constraints: the residual function is constructed using the consistency of vehicle trajectory direction, the consistency of stop line position, and the consistency of ground plane as structural constraints, and the perturbation of radar extrinsic parameters is solved. At the same time, an observability criterion is introduced to gate the extrinsic parameter update process: when the number of available trajectories in the approach zone is insufficient or the trajectory direction distribution is too concentrated, the extrinsic parameter update is paused to avoid erroneous updates under conditions of insufficient or degraded traffic samples.

4. The reliable gating signal control method for roadside lidar at intersections as described in claim 1, characterized in that, The point cloud health index is used to characterize the effective point density, echo intensity stability, and outlier ratio within the region of interest. The trajectory layer health index is used to characterize the trajectory breakage rate, speed jitter, and lane mapping consistency. The above health indicators are all statistically calculated within a fixed time window to reflect the reliability of the current roadside perception system in depicting traffic conditions.

5. The reliable gating signal control method for roadside lidar at intersections as described in claim 1, characterized in that, When mapping the obtained health diagnostic indicators to traffic parameter confidence vectors, a parameterized monotonic mapping function is used to compress and weight the health indicators, so that the confidence of various traffic parameters is normalized to the [0,1] interval.

6. The reliable gating signal control method for roadside lidar at intersections as described in claim 1, characterized in that, in In normal adaptive mode, adaptive timing and phase control are performed using a complete set of traffic parameter inputs; In conservative adaptive mode, only high-confidence parameters are enabled, and constraints are imposed on the phase extension and period adjustment amplitudes. In rollback control mode, sensor input is stopped, and preset time-segmented timing or historical statistical timing schemes are output.

7. A reliable gating signal control system for roadside lidar at intersections, characterized in that, include: The dynamic point cloud set construction module is configured to: acquire intersection point cloud data based on roadside lidar and convert the point cloud data to the road coordinate system, and then perform purification processing on the point cloud data in the road coordinate system to obtain the purified dynamic point cloud set. The vehicle target trajectory set generation module is configured to: based on the purified dynamic point cloud set, perform temporal correlation on the dynamic point cloud in the set in the road coordinate system to generate the vehicle target trajectory set; The online self-calibration module is configured to: perform online self-calibration based on the set of vehicle target trajectories and utilize traffic flow trajectory structure constraints, and complete the external parameter update determination; The health diagnostic indicator construction module is configured to: after completing the external parameter update determination, construct health diagnostic indicators based on point cloud and trajectory data under unified external parameters, wherein the health diagnostic indicators include point cloud layer health indicators and trajectory layer health indicators. The mapping module is configured to map the obtained health diagnosis indicators into a traffic parameter confidence vector, wherein each dimension of the traffic parameter confidence vector corresponds to a traffic control input parameter. Under the credibility vector constraint, lane-level traffic parameters are calculated based on vehicle trajectory data, and the traffic parameters of each lane are weighted and summarized to form the intersection-level traffic state input. The control module is configured to construct a hierarchical degraded signal control state machine, including a normal adaptive mode, a conservative adaptive mode, and a backoff control mode, based on the traffic parameter confidence vector and the external parameter observability criterion, and to control the signal based on the above modes.

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.

Images