Multi-intersection intelligent network connected vehicle aeb braking control method and system

Abstract

The present application relates to the technical field of intelligent networked vehicles, and particularly relates to a multi-intersection intelligent networked vehicle AEB braking control method and system, the present application collects the sensing quality parameter information of the vehicle-road cooperative sensing network; calculates the sensing quality feature representation value based on the sensing quality parameter information; determines whether the vehicle-road cooperative sensing data at the target time meets the standard based on the sensing quality feature representation value; collects the misjudgment risk parameter information of the historical AEB early warning event triggered by the sensing data; calculates the misjudgment risk feature representation value based on the misjudgment risk parameter information; determines whether the AEB early warning at the target time is abnormal based on the misjudgment risk feature representation value; determines the cause of the AEB early warning abnormality at the target time and the corresponding processing strategy based on the difference between the misjudgment risk feature representation value and the predetermined misjudgment risk feature representation threshold value. The present application improves the accuracy of vehicle-road cooperative emergency braking decision.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle technology, and in particular to an AEB braking control method and system for intelligent connected vehicles based on multiple intersections. Background Technology

[0002] In autonomous driving scenarios at complex urban intersections, traditional single-vehicle AEB technology is limited by blind spots and perception range, making it unable to handle multi-directional traffic conflicts. While existing vehicle-to-infrastructure (V2I) AEB technology expands perception capabilities through multi-source data fusion, it faces two major challenges in practical deployment: first, spatiotemporal misalignment of multi-sensor data due to network latency and clock asynchrony leads to ghosting targets and false braking; second, sudden interruptions in target trajectories caused by communication instability render prediction models ineffective, resulting in missed braking. The perception uncertainty caused by the quality of underlying communication severely restricts the reliability and safety of V2I AEB in critical scenarios.

[0003] Chinese Patent Publication No. CN114495499A discloses a multi-objective intelligent connected vehicle cooperative optimization control method, including the following steps: Step 1: Data acquisition and sequence generation: Collect target vehicle traffic data, including location, speed, target lane, and code; virtually map multi-lane vehicles to a single lane to generate an initial car-following sequence; Step 2: Conflict detection and screening: Determine whether there are traffic conflicts between vehicles based on the car-following sequence; screen out conflicting vehicles using a minimum vehicle spacing model; Step 3: Construct a target prediction module: Construct a convolutional neural network, with the vehicle code sequence as input and predicted time and energy consumption as output; the objective function is to minimize travel time and energy consumption; Step 4: Multi-objective evolutionary training and optimization: Employ a multi-objective discrete evolutionary algorithm, combined with Pareto optimality, to select the best travel order; evolve the population through crossover, exchange, shifting, and symmetry operations; utilize the neural network prediction results to feed back the evolutionary process, improving optimization efficiency. The invention achieves a dual improvement in efficiency and energy saving by simultaneously optimizing travel time and energy consumption; it significantly shortens computation time by combining neural network prediction and evolutionary algorithms to meet real-time scheduling requirements; it effectively reduces vehicle delays and alleviates traffic congestion through optimal traffic sequence scheduling; and its conflict detection and virtual mapping mechanisms ensure vehicle traffic safety and prevent traffic accidents. It is applicable to various conflict area scenarios such as intersections and entrance / exit lanes. However, the core of the invention lies in optimizing the macroscopic traffic sequence under given reliable perception data, and it does not involve dynamic evaluation, diagnosis, and feedback adjustment mechanisms for the underlying vehicle-road cooperative perception network itself.

[0004] Therefore, it is evident that the existing technology has the following problems:

[0005] Existing technologies do not consider the problem that the collaborative AEB technology of intelligent connected vehicles in autonomous driving multi-intelligent connected vehicle scenarios may misjudge the actual vehicle's driving trajectory due to the lack of network communication quality monitoring, resulting in low accuracy of collaborative emergency braking in complex network environments. Summary of the Invention

[0006] To address this issue, the present invention provides an AEB braking control method and system for intelligent connected vehicles based on multiple intersections, in order to overcome the problem that the existing intelligent connected vehicle cooperative AEB technology in unmanned multi-intelligent connected vehicle scenarios suffers from misjudgment of the actual vehicle's driving trajectory due to the lack of network communication quality monitoring, resulting in low accuracy of cooperative emergency braking in complex network environments.

[0007] To achieve the above objectives, the present invention provides an AEB braking control method for intelligent connected vehicles based on multiple intersections, comprising:

[0008] Collect perception quality parameter information of the vehicle-road cooperative perception network;

[0009] Calculate the perceived quality feature representation value based on the perceived quality parameter information;

[0010] Based on the perceived quality characteristic values, determine whether the vehicle-road cooperative perception data at the target time meets the standard;

[0011] When the vehicle-road cooperative perception data meets the standards, collect the misjudgment risk parameter information of historical AEB warning events triggered by the perception data;

[0012] Calculate the characteristic value of misjudgment risk based on the aforementioned misjudgment risk parameter information;

[0013] Based on the aforementioned misjudgment risk characteristic value, determine whether the AEB warning at the target time is abnormal;

[0014] When an AEB warning anomaly is detected at a target time, the difference between the misjudged risk feature characterization value and the predetermined misjudged risk feature characterization threshold is calculated.

[0015] Based on the difference, determine the cause of the AEB warning anomaly at the target time and its corresponding handling strategy; determine the increase in the fault tolerance threshold of time alignment in the fusion algorithm, and determine whether to trigger the link quality detection of the communication module to switch to the backup communication frequency band;

[0016] The perception quality parameters include the sensor's clock deviation and the variance of the communication message reception interval; the misjudgment risk parameters include the position fusion error of multi-source targets and the number of trajectory prediction interruptions.

[0017] Furthermore, the process of calculating the perceived quality feature representation value based on the perceived quality parameter information includes:

[0018] Extract the clock deviation of the sensors and the variance of the receiving interval of communication messages of the target vehicle-road cooperative perception network within the historical period;

[0019] The first perceived quality factor is determined by calculating the ratio of a predetermined clock skew threshold for a single cycle to the clock skew.

[0020] The second sensing quality factor is determined by calculating the ratio of a predetermined receive interval variance threshold for a single period to the receive interval variance.

[0021] The sum of the first perceived quality factor and the second perceived quality factor is determined as the perceived quality characteristic value.

[0022] Furthermore, the process of determining whether the vehicle-road cooperative sensing data at the target time meets the standard based on the perceived quality feature characterization value includes:

[0023] If the perceived quality feature value is less than or equal to the predetermined perceived quality feature threshold, then the vehicle-road cooperative perception data at the target time is determined to be non-compliant with the standard.

[0024] If the perceived quality feature value is greater than the predetermined perceived quality feature threshold, then the vehicle-road cooperative perception data at the target time is determined to meet the standard.

[0025] Furthermore, the process of calculating the misjudgment risk characteristic value based on the misjudgment risk parameter information includes:

[0026] Extract the location fusion error of multi-source targets and the number of interruptions in trajectory prediction for historical AEB early warning events triggered by sensing data within the historical period;

[0027] The ratio of a predetermined location fusion error threshold for a single period to the location fusion error is used to determine the first misjudgment risk factor.

[0028] The second misjudgment risk factor is determined by calculating the ratio of a predetermined threshold number of interruptions for a single cycle to the number of interruptions.

[0029] The sum of the first misjudgment risk factor and the second misjudgment risk factor is determined as the misjudgment risk characteristic value.

[0030] Furthermore, the process of determining whether the AEB warning at the target time is abnormal based on the aforementioned misjudgment risk characteristic representation value includes:

[0031] If the value of the misjudged risk feature is less than or equal to the predetermined misjudged risk feature threshold, then the AEB warning at the target time is determined to be abnormal.

[0032] If the value of the misjudged risk feature is greater than the predetermined threshold for misjudged risk feature, then the AEB warning at the target time is determined to be normal.

[0033] Furthermore, the process of determining the cause of the AEB warning anomaly at the target time based on the difference includes:

[0034] If the difference is less than or equal to a predetermined difference threshold, it is determined to be the first cause label;

[0035] If the difference is greater than a predetermined difference threshold, it is determined to be a second cause label.

[0036] Furthermore, the process of determining the corresponding handling strategy for the cause of the AEB warning anomaly at the target time based on the difference includes:

[0037] If it is the first cause label, then determine the increase in the fault tolerance threshold for time alignment in the fusion algorithm;

[0038] If the second cause label is selected, then the link quality test of the communication module is triggered to switch to the backup communication frequency band.

[0039] Furthermore, the method for collecting the position fusion error of multi-source targets is to detect the position of the same target vehicle at the same time through vehicle-mounted cameras and roadside radar, and calculate the Euclidean distance between the center points of the detected target vehicle positions after the two are time-aligned.

[0040] Furthermore, the number of interruptions in trajectory prediction is determined by the number of times the predicted trajectory of the target vehicle is lost per unit time due to data loss during the acquisition of the trajectory via the communication network.

[0041] Furthermore, the present invention also provides an AEB braking control system for intelligent connected vehicles based on multiple intersections, comprising:

[0042] The data acquisition module is used to collect perception quality parameters of the vehicle-road cooperative perception network and misjudgment risk parameters of historical AEB warning events triggered by perception data.

[0043] The analysis module is used to calculate the perceived quality feature representation value based on the perceived quality parameter information, and to calculate the misjudgment risk feature representation value based on the misjudgment risk parameter information;

[0044] The monitoring module is used to determine whether the vehicle-road cooperative perception data at the target time meets the standard based on the perception quality feature characterization value, and to determine whether the AEB warning at the target time is abnormal based on the misjudgment risk feature characterization value.

[0045] The diagnostic module is used to determine the cause of the AEB warning anomaly at the target time and its corresponding handling strategy based on the difference between the misjudged risk feature characterization value and the predetermined misjudged risk feature characterization threshold: determining the increase of the fault tolerance threshold for time alignment in the fusion algorithm, and triggering the link quality detection of the communication module to switch to the backup communication frequency band.

[0046] The perception quality parameters include the clock deviation of the sensor and the variance of the reception interval of the communication message, while the misjudgment risk parameters include the position fusion error of the multi-source target and the number of interruptions in trajectory prediction.

[0047] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides an AEB braking control method for intelligent connected vehicles based on multiple intersections. The beneficial effects of this method are that by collecting perception quality parameter information from the vehicle-road cooperative perception network, it can quickly identify the clock deviation of sensors and the variance of the reception interval of communication messages, so as to calculate the perception quality feature characterization value, thereby accurately determining whether the vehicle-road cooperative perception data at the target time meets the standard; if it meets the standard, by re-collecting the misjudgment risk parameter information of historical AEB warning events triggered by perception data, it can obtain the position fusion error of multi-source targets and the number of interruptions in trajectory prediction, so as to analyze the misjudgment risk feature characterization value, thereby accurately determining whether the AEB warning at the target time is abnormal. When an abnormal state occurs, the cause of the abnormality and the corresponding handling strategy are determined based on the difference between the misjudgment risk feature characterization value and the predetermined misjudgment risk feature characterization threshold. This overcomes the problem of low accuracy of cooperative emergency braking decisions in complex network environments caused by the lack of network communication quality and misjudgment of actual vehicle driving trajectory in multi-intelligent connected vehicle scenarios of autonomous driving.

[0048] In particular, this invention uses the clock deviation of sensors and the variance of the reception interval of communication messages as the basis for analyzing the stability of the sensing network link. It also uses the position fusion error of multi-source targets and the number of interruptions in trajectory prediction as the basic data for analyzing whether the AEB warning at the target time is abnormal. This invention can quantitatively reflect the stability of the sensing network link and the AEB warning status at the target time, thereby improving the accuracy of collaborative emergency braking decision-making in complex network environments of autonomous driving multi-intelligent connected vehicle scenarios.

[0049] In particular, this invention implements a two-level judgment process. By evaluating whether the vehicle-road cooperative perception data at the target time meets the standard, and only when the vehicle-road cooperative perception data meets the standard, it further determines whether the AEB warning at the target time is abnormal. This can prioritize the elimination of interference caused by poor source data quality, and concentrate the diagnosis of the cause of AEB warning abnormality within the scope of data fusion and prediction algorithms, thereby improving the diagnostic efficiency of AEB warning abnormality.

[0050] In particular, this invention also provides an AEB braking control system for intelligent connected vehicles at multiple intersections, including a data acquisition module, an analysis module, a monitoring module, and a diagnostic module. The data acquisition and analysis modules collect and analyze sensor clock deviations, communication message reception interval variances, multi-source target position fusion errors, and trajectory prediction interruption counts to obtain perception quality characteristic values ​​and misjudgment risk characteristic values, thus providing standardized indicators for evaluation. The monitoring module determines whether the vehicle-road cooperative perception data at the target time meets the standards. Only when the vehicle-road cooperative perception data meets the standards is the AEB warning at the target time further determined to be abnormal, thus assessing the reliability of the AEB warning. After confirming the warning anomaly, the diagnostic module calculates the difference between the misjudgment risk characteristic value and a predetermined misjudgment risk characteristic threshold to distinguish whether the warning anomaly is due to data fusion failure or communication link instability, triggering corresponding processing strategies: determining the increase in the time alignment tolerance threshold in the fusion algorithm, and triggering the communication module's link quality detection to switch to a backup communication frequency band. This invention improves the accuracy of cooperative emergency braking decisions in complex network environments for autonomous multi-intelligent connected vehicles. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the steps of the AEB braking control method for intelligent connected vehicles at multiple intersections according to an embodiment of the present invention.

[0052] Figure 2 This invention provides a logical decision diagram for determining whether the vehicle-road cooperative sensing data at the target time conforms to the standard based on the perceived quality feature characterization value.

[0053] Figure 3 This invention provides a logical decision diagram for determining whether an AEB warning at a target time is abnormal based on the aforementioned false positive risk characteristic representation value.

[0054] Figure 4 This invention provides a logical decision diagram for determining the cause of AEB warning anomalies at the target time and the corresponding processing strategies based on the difference. Detailed Implementation

[0055] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0056] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0057] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0058] Please see Figure 1 The diagram shown is a flowchart illustrating the steps of an AEB braking control method for intelligent connected vehicles at multiple intersections according to an embodiment of the present invention. The present invention provides an AEB braking control method for intelligent connected vehicles at multiple intersections, comprising:

[0059] Step S1: Collect the perception quality parameter information of the vehicle-road cooperative perception network;

[0060] Step S2: Calculate the perceived quality feature representation value based on the perceived quality parameter information;

[0061] Step S3: Determine whether the vehicle-road cooperative sensing data at the target time meets the standard based on the perceived quality feature characterization value;

[0062] Step S4: In response to the vehicle-road cooperative perception data meeting the credible standard, collect the misjudgment risk parameter information of historical AEB warning events triggered by the perception data;

[0063] Step S5: Calculate the misjudgment risk characteristic representation value based on the misjudgment risk parameter information;

[0064] Step S6: Determine whether the AEB warning at the target time is abnormal based on the aforementioned misjudgment risk characteristic representation value;

[0065] Step S7: In response to an AEB warning anomaly at the target time, calculate the difference between the misjudged risk feature characterization value and the predetermined misjudged risk feature characterization threshold.

[0066] Step S8: Determine the cause of the AEB warning anomaly at the target time and its corresponding handling strategy based on the difference;

[0067] Determine the increase in the fault tolerance threshold for time alignment in the fusion algorithm, and determine whether to trigger the link quality detection of the communication module to switch to the backup communication frequency band;

[0068] The perception quality parameters include the sensor's clock deviation and the variance of the communication message reception interval; the misjudgment risk parameters include the position fusion error of multi-source targets and the number of trajectory prediction interruptions.

[0069] It is understood that whether the vehicle-road cooperative perception data conforms to the standard refers to whether the vehicle-road cooperative perception data meets the credible standard for use as AEB decision input. If the vehicle-road cooperative perception data conforms to the standard, it means that the vehicle-road cooperative perception data meets the credible standard for use as AEB decision input; conversely, if the vehicle-road cooperative perception data does not conform to the standard, it means that the vehicle-road cooperative perception data does not meet the credible standard for use as AEB decision input.

[0070] It is understood that whether the AEB warning at the target time is abnormal refers to whether the AEB warning at the target time meets the high-confidence warning criteria. If the AEB warning at the target time is not abnormal, it means that the AEB warning at the target time meets the high-confidence warning criteria; conversely, if the AEB warning at the target time is abnormal, it means that the AEB warning at the target time does not meet the high-confidence warning criteria.

[0071] It is understandable that the sensor clock deviation refers to the maximum clock difference between the roadside unit and the vehicle OBU, which is obtained directly from the logs of the time synchronization server through the NTP / PTP synchronization protocol. The calculation method is to calculate the absolute difference between the maximum client clock recorded in multiple rounds of time synchronization interaction and the server reference clock, and the unit is ms.

[0072] It is understandable that the variance of the received interval of communication messages refers to the variance of the time intervals between consecutively received BSM / RSM messages. It is obtained by statistically analyzing the received time intervals of N messages within a historical period and calculating their variance. The calculation formula is based on N-1 time interval samples within the target period, and the unit is milliseconds (ms). 2 The smaller the value, the more stable the communication and the less network jitter.

[0073] In this embodiment, the value of N messages is predetermined, and N is preferably 50.

[0074] Understandably, the position fusion error of multi-source targets represents the difference in positional distance when the same target vehicle is detected by both the vehicle-mounted camera and roadside radar. This is achieved by calculating the Euclidean distance between the center points of the time-aligned multi-sensor detection boxes. The formula is the Euclidean distance between two 3D spatial coordinate points of the same target vehicle ID from different sensors at the same timestamp, expressed in meters (m). A smaller value indicates higher fusion accuracy. When sensor clock deviation increases, detection boxes from different time points are incorrectly fused to the same moment. For example, if the vehicle actually moves 0.5 meters, but due to a 100ms time deviation, one sensor sees the position at time t, while another sees the position at t-100ms. The direct distance difference between these two positions is the position fusion error. The larger the clock deviation, the greater the corresponding physical distance difference in the misaligned fusion.

[0075] It is understandable that the number of trajectory prediction interruptions refers to the number of times the predicted trajectory of the target vehicle is interrupted due to data loss per unit time. This is obtained by statistically analyzing the frequency of trajectory IDs reappearing after being lost in consecutive frames. This data is collected through the status output interface of the roadside multi-target tracking algorithm. The calculation formula is the accumulation of the number of times the trajectory ID of the same vehicle identifier undergoes discontinuous changes within the statistical period, measured in times per second. A smaller value indicates more continuous trajectory tracking. The variance of the communication message reception interval directly reflects the stability of communication. When the variance increases, it means that the message arrival time is unstable, making message loss more likely. Once two or more critical messages are lost consecutively, the trajectory tracking algorithm will determine that the target is lost, assign a new trajectory ID, and thus cause a trajectory continuity interruption. Therefore, an increase in the variance of the reception interval directly leads to an increase in the number of interruptions per unit time. Excessive clock deviation can also cause errors in the time stamp of sensor data. Even if the data is not lost, the trajectory tracking algorithm may misjudge abnormal target movement due to incorrect timestamps, thus incorrectly terminating and restarting the trajectory, increasing the number of trajectory prediction interruptions.

[0076] In this embodiment, the statistical period is preset, and the preferred statistical period is 15 minutes.

[0077] This invention provides an AEB braking control method for intelligent connected vehicles at multiple intersections. By collecting perception quality parameters from the vehicle-road cooperative perception network, it can quickly identify sensor clock deviations and communication message reception interval variances to calculate perception quality feature values, thereby accurately determining whether the vehicle-road cooperative perception data at the target time meets the standards. If it meets the standards, by re-collecting historical AEB warning event misjudgment risk parameters triggered by the perception data, it can obtain the position fusion error of multi-source targets and the number of trajectory prediction interruptions to analyze the misjudgment risk feature values, thereby accurately determining whether the AEB warning at the target time is abnormal. When an abnormal state occurs, the cause of the abnormality and the corresponding handling strategy are determined based on the difference between the misjudgment risk feature value and the predetermined misjudgment risk feature threshold. This overcomes the problem of low accuracy in cooperative emergency braking decisions in complex network environments caused by the lack of network communication quality and misjudgment of actual vehicle trajectory in multi-intelligent connected vehicle cooperative AEB systems in autonomous driving scenarios.

[0078] Specifically, the process of calculating the perceived quality feature representation value based on the perceived quality parameter information includes:

[0079] Extract the clock deviation of the sensors and the variance of the receiving interval of communication messages of the target vehicle-road cooperative perception network within the historical period;

[0080] The first perceived quality factor is determined by calculating the ratio of a predetermined clock skew threshold for a single cycle to the clock skew.

[0081] The second sensing quality factor is determined by calculating the ratio of a predetermined receive interval variance threshold for a single period to the receive interval variance.

[0082] The sum of the first perceived quality factor and the second perceived quality factor is determined as the perceived quality characteristic value.

[0083] In this embodiment, the predetermined clock deviation threshold is preset, wherein clock deviation samples within 15 historical periods (i.e., 15 minutes) are predetermined, and the predetermined clock deviation threshold is determined based on the average value of the clock deviation samples.

[0084] In this embodiment, the predetermined receiving interval variance threshold is preset, wherein receiving interval variance samples over 15 historical periods (i.e., 15 minutes) are predetermined, and the predetermined receiving interval variance threshold is determined based on the average value of the receiving interval variance samples.

[0085] In this embodiment, the single cycle is preset, and the preferred single cycle is 1 minute.

[0086] This invention introduces a method for calculating perceived quality feature values ​​based on dynamically set and normalized historical data, making the system evaluation standard more closely reflect the current actual network operating state. The real-time collected clock deviation and receiving interval variance are compared with thresholds derived from recent historical data statistics, unifying parameters of different dimensions and orders of magnitude into a dimensionless perceived quality factor. The sum of these two factors serves as a comprehensive characterization value, allowing the judgment of perceived quality to reflect the combined impact of multi-dimensional network conditions. By setting a period of 1 minute and dynamically determining the threshold based on samples from the previous 15 minutes, the evaluation benchmark can smoothly follow the slow changes in the network environment, avoiding the rigidity or slow response to environmental changes that can occur with fixed thresholds, thus improving the adaptability and rationality of the evaluation.

[0087] Please see Figure 2 As shown, this is a logic diagram for determining whether the vehicle-road cooperative sensing data at a target time conforms to the standard based on the perceived quality feature characterization value according to an embodiment of the present invention. The process of determining whether the vehicle-road cooperative sensing data at a target time conforms to the standard based on the perceived quality feature characterization value includes:

[0088] If the perceived quality feature value is less than or equal to the predetermined perceived quality feature threshold, then the vehicle-road cooperative perception data at the target time is determined to be non-compliant with the standard.

[0089] If the perceived quality feature value is greater than the predetermined perceived quality feature threshold, then the vehicle-road cooperative perception data at the target time is determined to meet the standard.

[0090] In this embodiment, the predetermined threshold for perceiving quality characteristics is set in advance. Specifically, the average value of the perceiving quality characteristics within 30 cycles (30 minutes) is predetermined. The predetermined threshold for perceiving quality characteristics is determined based on the product of the average value of the perceiving quality characteristics and the deviation coefficient. The deviation coefficient is determined based on the analysis of statistical data on the stable operation status of the collaborative sensing system in engineering practice and is selected within the range of [0.56, 1.33]. In this embodiment, the preferred deviation coefficient is 1.25.

[0091] This invention balances decision-making security and data availability by employing a method that dynamically sets thresholds based on historical operational data and deviation coefficients. The threshold is set as an adjustable parameter, rather than a fixed value. By analyzing historical data, a baseline level can be found, and the predetermined perceived quality characteristic representation threshold can be fine-tuned based on the tolerance for system misjudgment risk. The preferred embodiment ensures that only data with perceived quality above the recent average level can be used for AEB (Automatic Emergency Response) decisions, thus providing a higher confidence base for subsequent safety decisions at the data source.

[0092] Specifically, the process of calculating the misjudgment risk characteristic value based on the misjudgment risk parameter information includes:

[0093] Extract the location fusion error of multi-source targets and the number of interruptions in trajectory prediction for historical AEB early warning events triggered by sensing data within the historical period;

[0094] The first misjudgment risk factor is determined by calculating the ratio of a predetermined location fusion error threshold for a single period to the location fusion error.

[0095] The ratio of a predetermined threshold number of interruptions per cycle to the number of interruptions is used to determine a second misjudgment risk factor.

[0096] The sum of the first misjudgment risk factor and the second misjudgment risk factor is determined as the misjudgment risk characteristic value.

[0097] In this embodiment, the predetermined location fusion error threshold is preset, wherein location fusion error samples within 15 historical periods (i.e., 15 minutes) are predetermined, and the predetermined location fusion error threshold is determined based on the average value of the location fusion error samples.

[0098] In this embodiment, the predetermined interruption threshold is preset, wherein the interruption count sample within 15 historical periods (i.e., 15 minutes) is predetermined, and the predetermined interruption threshold is determined based on the average value of the interruption count sample.

[0099] It is understandable that the two parameters, location fusion error and number of interruptions, constitute a diagnostic combination revealing the reliability of perceived data at both the physical and application levels. Location fusion error directly exposes problems with the temporal alignment of multi-source data, i.e., the instantaneous accuracy problem of data fusion; while the number of interruptions reflects the break in the temporal continuity of the data stream, i.e., the long-term problem of temporal consistency. In the intersection cooperative AEB scenario, the former can lead to instantaneous misjudgment of obstacle positions, resulting in unnecessary emergency braking; the latter can disrupt the prediction of vehicle movement trends, leading to delayed braking. This embodiment of the invention uses these two parameters as a comprehensive representation of misjudgment risk, precisely because they have a direct, measurable, and clearly causally related indicative role in the key dimensions of spatial accuracy and temporal continuity, thereby accurately pinpointing the root cause of the decline in warning confidence.

[0100] This invention employs a normalized quantification method parallel to perception quality assessment to calculate the characteristic value of misjudgment risk. This transforms the abstract risk of misjudgment into a concrete numerical value. Two parameters reflecting different levels of data quality issues—location fusion error and the number of interruptions—are compared with thresholds dynamically set based on recent historical performance, and normalized as risk factors in the form of ratios. Adding these two values ​​allows the comprehensive risk level of a warning event to be characterized by a single numerical value, providing a clear quantitative basis for subsequent risk level classification and differentiated processing. This enables the system to make standardized judgments on the reliability of current decisions based on historical operational experience, thereby supporting the transition from qualitative decision-making to quantitative assessment.

[0101] Please see Figure 3 As shown, this is a logic diagram for determining whether an AEB warning at a target time is abnormal based on the false positive risk characteristic value in an embodiment of the present invention. The process of determining whether an AEB warning at a target time is abnormal based on the false positive risk characteristic value in the present invention includes:

[0102] If the value of the misjudged risk feature is less than or equal to the predetermined misjudged risk feature threshold, then the AEB warning at the target time is determined to be abnormal.

[0103] If the value of the misjudged risk feature is greater than the predetermined threshold for misjudged risk feature, then the AEB warning at the target time is determined to be normal.

[0104] In this embodiment, the predetermined threshold for the false alarm risk feature is set in advance. Specifically, the average value of the false alarm risk feature features over 30 historical periods (30 minutes) is predetermined. The predetermined threshold for the false alarm risk feature is determined based on the product of the average value and the offset coefficient. The offset coefficient is determined through regression analysis of a large amount of offline simulation and real vehicle test data, based on prior information such as the average traffic flow of the target intersection, historical accident data, and the system's tolerance level for false alarms. It is selected within the interval [1.37.3.46]. In this embodiment, the preferred offset coefficient is 1.92.

[0105] This invention, through a correlation between a risk assessment threshold and the average historical performance, introduces an offset coefficient that can be flexibly adjusted according to scenario safety requirements, enabling the high-confidence warning standard to possess scenario adaptability and configurability. Based on the system's recent historical actual risk level, a more stringent confidence standard than the recent average risk level is set by multiplying this by an offset coefficient greater than 1. The value of the offset coefficient is determined based on engineering experience and scenario characteristics, allowing the same core algorithm to adjust the severity of its warnings according to the traffic complexity and safety level requirements of different intersections by configuring different coefficients. This balances system sensitivity and specificity, improving the adaptability and engineering practicality of the solution in different application scenarios.

[0106] Please see Figure 4 As shown, this is a logical decision diagram of the method for determining the cause of an AEB warning anomaly at a target time and its corresponding processing strategy based on the difference in an embodiment of the present invention. The process of determining the cause of an AEB warning anomaly at a target time based on the difference in the present invention includes:

[0107] If the difference is less than or equal to a predetermined difference threshold, it is determined to be the first cause label;

[0108] If the difference is greater than a predetermined difference threshold, it is determined to be a second cause label.

[0109] In this embodiment, the predetermined difference threshold is preset. Specifically, the average difference between the value of the misjudgment risk feature and the predetermined misjudgment risk feature is predetermined over 30 historical periods (30 minutes). The predetermined difference threshold is determined based on the product of the average difference and a tolerance coefficient. The tolerance coefficient is determined based on engineering experience in network maintenance and is selected within the range of [1.21, 3.66]. In this embodiment, the preferred tolerance coefficient is 1.58.

[0110] Understandably, the primary cause is the failure of multi-sensor data time alignment, with the main problem being poor clock synchronization quality.

[0111] Understandably, the second reason is that unstable data flow leads to frequent trajectory breaks, and the main problem lies in the poor quality of the communication link.

[0112] This invention, by associating a cause determination threshold with the degree of historical risk deviation and introducing a configurable tolerance coefficient, enables the system to distinguish between normal risk fluctuations and abnormal performance degradation, achieving accurate attribution and differentiated processing. By setting a dynamic threshold based on the recent average deviation level and the tolerance coefficient, it can differentiate between risks exceeding the historical fluctuation range and abnormal risks exceeding the normal fluctuation range, allowing the system to make more refined responses: implementing algorithmic fault-tolerant adjustments or triggering communication link switching. The layered response mechanism optimizes the allocation of operational resources and improves the efficiency and targeting of problem handling.

[0113] Specifically, the process of determining the corresponding processing strategy for why the AEB warning at the target time does not meet the high-confidence warning standard based on the difference includes:

[0114] If it is the first cause label, then determine the increase in the fault tolerance threshold for time alignment in the fusion algorithm;

[0115] If the second cause label is selected, then the link quality test of the communication module is triggered to switch to the backup communication frequency band.

[0116] It is understandable that determining the increase in the time alignment tolerance threshold in the fusion algorithm specifically refers to determining the increase in the time alignment tolerance threshold in the fusion algorithm, so as to avoid frequently discarding valid data due to temporary high multi-source target position fusion errors.

[0117] In this embodiment, the increase in the time alignment fault tolerance threshold is preset, and the preferred increase in this embodiment is 20%.

[0118] Understandably, by increasing the fault tolerance threshold for time alignment, the problem of instantaneous decrease in fusion accuracy caused by poor clock synchronization quality can be temporarily alleviated without changing the underlying clock synchronization hardware and its protocol. This allows the perception fusion algorithm to retain more data points for comprehensive judgment during the current unstable period, thereby reducing the delay in perception information caused by data filtering and ensuring the continuity and integrity of the information required for AEB decision-making.

[0119] It is understandable that determining the trigger for the communication module's link quality detection to switch to the backup communication frequency band specifically means that if the time during which the message reception interval variance exceeds a predetermined variance threshold is greater than or equal to a predetermined time threshold, then the switch to the backup communication frequency band will occur.

[0120] In this embodiment, the predetermined time threshold is preset, and the preferred predetermined time threshold is 30 minutes.

[0121] In this embodiment, the predetermined variance threshold is preset, and preferably, the predetermined variance threshold is 2.5 times the average variance of the communication message reception interval over a historical 30-minute period.

[0122] Understandably, by switching to a backup communication frequency band, the processing can be upgraded from application-layer algorithm adjustment to physical communication layer intervention. When the communication link is diagnosed as the root cause of the problem, the system proactively initiates a link health check and performs fault confirmation and switching based on predetermined time thresholds. This aims to restore stable data link transmission at the physical level, fundamentally resolving the trajectory tracking and prediction interruption issues caused by communication instability, and providing a more reliable data transmission environment for upper-layer applications.

[0123] Understandably, the two processing strategies work together to form a progressive system self-healing solution, from algorithmic fault tolerance to physical repair, progressing from simple to complex. The former prioritizes attempting to compensate for transient disturbances caused by clock problems in software to maintain system functionality; while the latter takes physical layer reconstruction measures when communication problems persist, avoiding the limitations of a single strategy. It can cope with both short-term network fluctuations and persistent hardware or link failures, enabling the system to take the most appropriate measures when facing network quality degradation of different natures and severity, thereby comprehensively enhancing the system's robustness and continuous service capabilities.

[0124] This invention employs a layered closed-loop processing strategy to directly translate diagnostic results into executable operations, thereby improving the overall system's dynamic adaptability and problem-solving efficiency. When a warning is determined to be of low confidence and attributed to a clock problem, the system immediately adjusts the time alignment tolerance threshold of the fusion algorithm; when the problem is determined to originate from the communication link, link detection is initiated, and even switching to a backup frequency band is performed. This allows the system to proactively perform internal adjustments or resource reorganization when the performance of some functional units degrades, maintaining overall decision-making functionality and reducing reliance on external manual intervention.

[0125] Specifically, the method for collecting the position fusion error of multi-source targets is to detect the position of the same target vehicle at the same time through vehicle-mounted cameras and roadside radars, and calculate the Euclidean distance between the center points of the detected target vehicle positions after the two are time-aligned.

[0126] As is understandable, Euclidean distance refers to the straight-line distance between two points in a three-dimensional Cartesian coordinate system. Based on the raw detection results from the vehicle-mounted camera and roadside radar, the data is first uniformly transformed to this common coordinate system according to the timestamps and calibration parameters of their respective sensors, ensuring that their timestamps are aligned to the same decision time. Subsequently, for two detection boxes that have been confirmed to be associated with the same target, the coordinates of their geometric center points are extracted, and the Euclidean distance formula is applied for calculation.

[0127] This invention employs Euclidean distance as a metric to calculate location fusion error, directly quantifying the spatiotemporal alignment accuracy of multi-source data. The error value clearly reflects the consistency of different sensors' spatial location judgments of the same physical entity, providing a traceable numerical basis for evaluating the instantaneous accuracy and clock synchronization effect of the perception fusion algorithm, and supporting subsequent risk factor calculation and problem attribution analysis.

[0128] Specifically, the number of interruptions in trajectory prediction is determined by the number of times the predicted trajectory of the target vehicle is lost per unit time due to data loss during the acquisition of the trajectory via the communication network.

[0129] Understandably, the interruption is defined and recorded by the state output of the multi-target tracking algorithm in the roadside fusion unit. The multi-target tracking algorithm maintains a unique trajectory ID for each continuously tracked target. When it fails to successfully associate with any sensed target for multiple consecutive sensing cycles, for example, corresponding to the duration of more than two consecutive lost BSM messages, the algorithm marks its trajectory as lost and terminates its ID. Subsequently, if a new sensed target is identified as the previously lost target vehicle, but a data gap exists, the algorithm will assign it a new trajectory ID for reliability reasons. The period from the termination of the old trajectory ID to the generation of the new trajectory ID is recorded as an interruption of trajectory continuity.

[0130] This invention quantifies the number of interruptions by statistically analyzing the discontinuous switching frequency of trajectory IDs. It characterizes the stability of communication message reception from the perspective of trajectory continuity, avoiding the difficulty of directly monitoring complex physical layer communication processes. Instead, it reflects the continuity and reliability of network data transmission through the operational status.

[0131] Specifically, embodiments of the present invention also provide an AEB braking control system for intelligent connected vehicles based on multiple intersections, including:

[0132] The data acquisition module is used to collect perception quality parameters of the vehicle-road cooperative perception network and misjudgment risk parameters of historical AEB warning events triggered by perception data.

[0133] The analysis module is used to calculate the perceived quality feature representation value based on the perceived quality parameter information, and to calculate the misjudgment risk feature representation value based on the misjudgment risk parameter information;

[0134] The monitoring module is used to determine whether the vehicle-road cooperative perception data at the target time meets the standard based on the perception quality feature characterization value, and to determine whether the AEB warning at the target time is abnormal based on the misjudgment risk feature characterization value.

[0135] The diagnostic module is used to determine the cause of the AEB warning anomaly at the target time and its corresponding handling strategy based on the difference between the misjudged risk feature characterization value and the predetermined misjudged risk feature characterization threshold: determining the increase of the fault tolerance threshold for time alignment in the fusion algorithm, and triggering the link quality detection of the communication module to switch to the backup communication frequency band.

[0136] The perception quality parameters include the clock deviation of the sensor and the variance of the reception interval of the communication message, while the misjudgment risk parameters include the position fusion error of the multi-source target and the number of interruptions in trajectory prediction.

[0137] This invention uses modular design to solidify the logical flow of the method into the system architecture. The acquisition module is responsible for uniformly extracting key parameters from heterogeneous network and sensor data sources, providing standardized input for subsequent analysis. The analysis module and monitoring module are respectively responsible for quantitative calculation and threshold determination. The diagnosis module converts the abnormal signals output by the monitoring module into specific executable control instructions, making the functional boundaries of data acquisition, status assessment, problem diagnosis and strategy execution clear, reducing the internal coupling of the system, and improving the independence and maintainability of each part.

[0138] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for AEB braking control of intelligent connected vehicles based on multiple intersections, characterized in that, include: Collect perception quality parameter information of the vehicle-road cooperative perception network; Calculate the perceived quality feature representation value based on the perceived quality parameter information; Based on the perceived quality characteristic values, determine whether the vehicle-road cooperative perception data at the target time meets the standard; When the vehicle-road cooperative perception data meets the standards, collect the misjudgment risk parameter information of historical AEB warning events triggered by the perception data; Calculate the characteristic value of misjudgment risk based on the aforementioned misjudgment risk parameter information; Based on the aforementioned misjudgment risk characteristic value, determine whether the AEB warning at the target time is abnormal; When an AEB warning anomaly is detected at a target time, the difference between the misjudged risk feature characterization value and the predetermined misjudged risk feature characterization threshold is calculated. Based on the difference, determine the cause of the AEB warning anomaly at the target time and the corresponding handling strategy; Determine the increase in the fault tolerance threshold for time alignment in the fusion algorithm, and determine whether to trigger the link quality detection of the communication module to switch to the backup communication frequency band; The perception quality parameters include the sensor's clock deviation and the variance of the communication message reception interval; the misjudgment risk parameters include the position fusion error of multi-source targets and the number of trajectory prediction interruptions.

2. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The process of calculating the perceived quality feature representation value based on the perceived quality parameter information includes: Extract the clock deviation of the sensors and the variance of the receiving interval of communication messages of the target vehicle-road cooperative perception network within the historical period; The first perceived quality factor is determined by calculating the ratio of a predetermined clock skew threshold for a single cycle to the clock skew. The second sensing quality factor is determined by calculating the ratio of a predetermined receive interval variance threshold for a single period to the receive interval variance. The sum of the first perceived quality factor and the second perceived quality factor is determined as the perceived quality characteristic value.

3. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The process of determining whether the vehicle-road cooperative sensing data at the target time meets the standard based on the perceived quality feature characterization value includes: If the perceived quality feature value is less than or equal to the predetermined perceived quality feature threshold, then the vehicle-road cooperative perception data at the target time is determined to be non-compliant with the standard. If the perceived quality feature value is greater than the predetermined perceived quality feature threshold, then the vehicle-road cooperative perception data at the target time is determined to meet the standard.

4. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The process of calculating the misjudgment risk characteristic value based on the misjudgment risk parameter information includes: Extract the location fusion error of multi-source targets and the number of interruptions in trajectory prediction for historical AEB early warning events triggered by sensing data within the historical period; The first misjudgment risk factor is determined by calculating the ratio of a predetermined location fusion error threshold for a single period to the location fusion error. The ratio of a predetermined threshold number of interruptions per cycle to the number of interruptions is used to determine a second misjudgment risk factor. The sum of the first misjudgment risk factor and the second misjudgment risk factor is determined as the misjudgment risk characteristic value.

5. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The process of determining whether an AEB warning at a target time is abnormal based on the aforementioned misjudgment risk characteristic value includes: If the value of the misjudged risk feature is less than or equal to the predetermined misjudged risk feature threshold, then the AEB warning at the target time is determined to be abnormal. If the value of the misjudged risk feature is greater than the predetermined threshold for misjudged risk feature, then the AEB warning at the target time is determined to be normal.

6. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The process of determining the cause of the AEB warning anomaly at the target time based on the difference includes: If the difference is less than or equal to a predetermined difference threshold, it is determined to be the first cause label; If the difference is greater than a predetermined difference threshold, it is determined to be a second cause label.

7. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The process of determining the corresponding handling strategy for the cause of the AEB warning anomaly at the target time based on the difference includes: If it is the first cause label, then determine the increase in the fault tolerance threshold for time alignment in the fusion algorithm; If the second cause label is selected, then the link quality test of the communication module is triggered to switch to the backup communication frequency band.

8. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The method for collecting the position fusion error of multi-source targets is to detect the position of the same target vehicle at the same time by using vehicle-mounted cameras and roadside radar, and then calculate the Euclidean distance between the center points of the detected target vehicle positions after the two are time-aligned.

9. The AEB braking control method for intelligent connected vehicles based on multiple intersections according to claim 1, characterized in that, The number of interruptions in trajectory prediction is determined by the number of times the predicted trajectory of the target vehicle is lost per unit time due to data loss during the acquisition of the trajectory via the communication network.

10. A system using the AEB braking control method for intelligent connected vehicles at multiple intersections as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to collect perception quality parameters of the vehicle-road cooperative perception network and misjudgment risk parameters of historical AEB warning events triggered by perception data. The analysis module is used to calculate the perceived quality feature representation value based on the perceived quality parameter information, and to calculate the misjudgment risk feature representation value based on the misjudgment risk parameter information; The monitoring module is used to determine whether the vehicle-road cooperative perception data at the target time meets the standard based on the perception quality feature characterization value, and to determine whether the AEB warning at the target time is abnormal based on the misjudgment risk feature characterization value. The diagnostic module is used to determine the cause of the AEB warning anomaly at the target time and its corresponding handling strategy based on the difference between the misjudged risk feature characterization value and the predetermined misjudged risk feature characterization threshold: determining the increase of the fault tolerance threshold for time alignment in the fusion algorithm, and triggering the link quality detection of the communication module to switch to the backup communication frequency band. The perception quality parameters include the clock deviation of the sensor and the variance of the reception interval of the communication message, while the misjudgment risk parameters include the position fusion error of the multi-source target and the number of interruptions in trajectory prediction.

Images