Emergency braking control method and system

Abstract

The invention provides an emergency braking control method and system, and the method comprises the steps: carrying out the time-space alignment processing of multi-source sensor data collected by a preset sensor array in a vehicle, so as to generate a corresponding initial data set; the initial data set is optimized through a preset DBSCAN clustering algorithm to generate a corresponding target data set, and the confidence coefficient weight of each sensor is dynamically calculated based on the actual environment condition of the vehicle and the target attribute of each sensor; based on the D-S theory, calculating a dynamic risk coefficient corresponding to the vehicle according to the target data set and the confidence coefficient weight, and correspondingly detecting a target coefficient level corresponding to the dynamic risk coefficient; and dynamically generating an adaptive emergency braking strategy according to the target coefficient grade, and issuing the emergency braking strategy to a vehicle controller to complete corresponding emergency braking. The emergency braking efficiency can be improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle engineering technology, and in particular to an emergency braking control method and system. Background Technology

[0002] With the advancement of technology and the rapid development of productivity, existing cars are also moving towards intelligence. Among them, advanced driver assistance systems have been applied in cars and are gradually gaining people's recognition, thereby improving people's driving experience.

[0003] Among them, the automatic emergency braking system, as one of the core functions of advanced driver assistance systems, is mainly intended to avoid or mitigate collisions by automatically braking when the driver fails to react in time, thereby improving traffic safety.

[0004] Furthermore, in practical applications, most existing technologies use millimeter-wave radar to calculate the time difference of arrival of targets and assess collision risks. This method has low sensitivity to static targets and is prone to missed detections. Therefore, to compensate for the limitations of single sensors, existing technologies also employ multi-sensor detection techniques. Specifically, existing technologies typically use weighted averaging or adding weights to fuse data collected by multiple sensors. However, this fusion method cannot effectively solve the problems of data heterogeneity and confidence differences between sensors, leading to conflicting output decisions and consequently reducing the efficiency of automatic braking. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide an emergency braking control method and system to solve the problem that the prior art cannot effectively solve the data heterogeneity and confidence difference between sensors.

[0006] The first aspect of the present invention proposes: An emergency braking control method specifically includes the following steps: The multi-source sensor data collected by the preset sensor array inside the vehicle is spatiotemporally aligned to generate the corresponding initial dataset. The initial dataset is optimized using a preset DBSCAN clustering algorithm to generate a corresponding target dataset. Based on the actual environmental conditions of the vehicle and the target attributes of each sensor, the confidence weight of each sensor is dynamically calculated. Based on the DS theory, the dynamic risk coefficient corresponding to the vehicle is calculated according to the target dataset and the confidence weight, and the target coefficient level corresponding to the dynamic risk coefficient is detected accordingly. Based on the target coefficient level, an appropriate emergency braking strategy is dynamically generated and sent to the vehicle controller to complete the corresponding emergency braking.

[0007] The beneficial effects of this invention are as follows: by performing spatiotemporal alignment processing on the collected multi-source sensor data, a standard initial dataset can be generated. Based on this, further optimization processing is performed to generate a target dataset that can be used for subsequent analysis. Based on this, the dynamic risk coefficient corresponding to the vehicle can be recalculated according to the current target dataset and the calculated confidence weight, and the corresponding target coefficient level can be detected. Finally, the emergency braking strategy required by the current vehicle can be determined according to the target coefficient level to complete reasonable emergency braking. This effectively solves the problems of data heterogeneity and confidence differences between sensors, thereby improving the efficiency of emergency braking.

[0008] Furthermore, the step of optimizing the initial dataset using a preset DBSCAN clustering algorithm to generate the corresponding target dataset includes: The initial dataset is preprocessed for outliers, and isolated noise points in the sensor data are identified and removed using the 3σ criterion to generate an intermediate dataset. The neighborhood radius and minimum number of core points of the preset DBSCAN clustering algorithm are dynamically adjusted according to the sensor's acquisition frequency and data accuracy to form a suitable set of algorithm parameters. Based on the algorithm parameter set, the intermediate dataset is clustered using the preset DBSCAN clustering algorithm to select effective clusters whose data density meets the preset density threshold, and the core point data of each effective cluster are merged to generate the target dataset.

[0009] Furthermore, the step of clustering the intermediate dataset using the preset DBSCAN clustering algorithm based on the algorithm parameter set to filter out effective clusters whose data density meets a preset density threshold includes: Based on the vehicle's real-time driving speed and steering angle, the intermediate dataset is divided into multiple driving state data subsets, each of which corresponds to a preset vehicle driving condition interval. The algorithm parameter group is called to perform DBSCAN clustering operation on each subset of driving state data. During the clustering process, the rate of change of the number of core points of each potential cluster is counted in real time, and the merging threshold of the cluster is dynamically adjusted according to the rate of change of the number of core points. Calculate the average data density of each cluster that has completed the clustering operation, compare the average data density with the preset density threshold, retain the clusters with an average data density greater than or equal to the preset density threshold as candidate clusters, and then remove the clusters in the candidate clusters whose core point ratio is lower than a preset ratio threshold to obtain the effective clusters.

[0010] Furthermore, the step of dynamically calculating the confidence weight of each sensor based on the actual environmental conditions of the vehicle and the target attributes of each sensor includes: Environmental interference parameters and road condition parameters are extracted from the actual environmental conditions. Simultaneously, the perception accuracy parameters and environmental adaptation parameters are extracted from the target attributes of each sensor. The environmental interference parameters, road condition parameters, perception accuracy parameters, and environmental adaptation parameters are quantized to obtain the corresponding environmental quantization values ​​and sensor quantization values. Construct a weight mapping matrix between the target attribute and the actual environmental conditions, and input the environmental quantization value and the sensor quantization value into the weight mapping matrix to obtain the initial confidence weight of each sensor through matrix operations; The sensor stability index during the vehicle's operation is collected, and the initial confidence weight is dynamically corrected based on the sensor stability index to obtain the confidence weight of each sensor.

[0011] Furthermore, the step of dynamically correcting the initial confidence weights based on the sensor stability index to obtain the confidence weights for each sensor includes: The stability index of each sensor is collected within a preset time window. The stability index includes data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value. The data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value are processed by normalization algorithm to obtain the corresponding standardized values. Based on the standardized values, a comprehensive stability evaluation index is constructed. The comprehensive stability evaluation index is compared with a preset multi-level evaluation threshold to determine the stability level of the corresponding sensor. Then, a preset weight correction ratio is matched according to the stability level. The initial confidence weight is weighted and the weight correction ratio is weighted to obtain the corrected intermediate confidence weight. At the same time, the current ambient light intensity parameter of the vehicle is collected, and the intermediate confidence weight is fine-tuned according to the ambient light intensity parameter. After the fine-tuning is completed, the confidence weight of each sensor is obtained.

[0012] Furthermore, the step of calculating the dynamic risk coefficient corresponding to the vehicle based on the DS theory, according to the target dataset and the confidence weights, includes: The sensor data in the target dataset are classified according to preset risk event types, and based on the confidence weight of the corresponding sensor, a corresponding basic probability allocation value is assigned to the sensor data corresponding to each type of risk event. Based on the evidence synthesis rules of the DS theory, the basic probability allocation values ​​corresponding to various risk events are calculated by multi-source evidence fusion to obtain the fusion probability allocation value of each type of risk event. Based on the preset risk quantification mapping relationship, the fusion probability allocation values ​​of each type of risk event are weighted and summed to generate the dynamic risk coefficient corresponding to the vehicle.

[0013] Furthermore, the step of performing a weighted summation operation on the fusion probability allocation values ​​of each type of risk event based on a preset risk quantification mapping relationship to generate the dynamic risk coefficient corresponding to the vehicle includes: According to the preset risk quantification mapping relationship, a corresponding dynamic weight coefficient is assigned to each type of risk event; For each type of risk event, the corresponding fusion probability allocation value and dynamic weight coefficient are multiplied to obtain the risk contribution value of each type of risk event; The risk contribution values ​​of all risk events are summed, and the summation result is normalized according to a preset numerical range to generate the dynamic risk coefficient corresponding to the vehicle.

[0014] The second aspect of the present invention proposes: An emergency braking control system, wherein the system comprises: The acquisition module is used to perform spatiotemporal alignment processing on the multi-source sensor data collected by the preset sensor array inside the vehicle to generate the corresponding initial dataset. The optimization module is used to optimize the initial dataset using a preset DBSCAN clustering algorithm to generate a corresponding target dataset, and dynamically calculate the confidence weight of each sensor based on the actual environmental conditions of the vehicle and the target attributes of each sensor. The calculation module is used to calculate the dynamic risk coefficient corresponding to the vehicle based on the DS theory, the target dataset, and the confidence weight, and to detect the target coefficient level corresponding to the dynamic risk coefficient. The generation module is used to dynamically generate an adapted emergency braking strategy based on the target coefficient level, and send the emergency braking strategy to the vehicle controller to complete the corresponding emergency braking.

[0015] Furthermore, the optimization module is specifically used for: The initial dataset is preprocessed for outliers, and isolated noise points in the sensor data are identified and removed using the 3σ criterion to generate an intermediate dataset. The neighborhood radius and minimum number of core points of the preset DBSCAN clustering algorithm are dynamically adjusted according to the sensor's acquisition frequency and data accuracy to form a suitable set of algorithm parameters. Based on the algorithm parameter set, the intermediate dataset is clustered using the preset DBSCAN clustering algorithm to select effective clusters whose data density meets the preset density threshold, and the core point data of each effective cluster are merged to generate the target dataset.

[0016] Furthermore, the optimization module is specifically used for: Based on the vehicle's real-time driving speed and steering angle, the intermediate dataset is divided into multiple driving state data subsets, each of which corresponds to a preset vehicle driving condition interval. The algorithm parameter group is called to perform DBSCAN clustering operation on each subset of driving state data. During the clustering process, the rate of change of the number of core points of each potential cluster is counted in real time, and the merging threshold of the cluster is dynamically adjusted according to the rate of change of the number of core points. Calculate the average data density of each cluster that has completed the clustering operation, compare the average data density with the preset density threshold, retain the clusters with an average data density greater than or equal to the preset density threshold as candidate clusters, and then remove the clusters in the candidate clusters whose core point ratio is lower than a preset ratio threshold to obtain the effective clusters.

[0017] Furthermore, the optimization module is specifically used for: Environmental interference parameters and road condition parameters are extracted from the actual environmental conditions. Simultaneously, the perception accuracy parameters and environmental adaptation parameters are extracted from the target attributes of each sensor. The environmental interference parameters, road condition parameters, perception accuracy parameters, and environmental adaptation parameters are quantized to obtain the corresponding environmental quantization values ​​and sensor quantization values. Construct a weight mapping matrix between the target attribute and the actual environmental conditions, and input the environmental quantization value and the sensor quantization value into the weight mapping matrix to obtain the initial confidence weight of each sensor through matrix operations; The sensor stability index during the vehicle's operation is collected, and the initial confidence weight is dynamically corrected based on the sensor stability index to obtain the confidence weight of each sensor.

[0018] Furthermore, the optimization module is specifically used for: The stability index of each sensor is collected within a preset time window. The stability index includes data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value. The data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value are processed by normalization algorithm to obtain the corresponding standardized values. Based on the standardized values, a comprehensive stability evaluation index is constructed. The comprehensive stability evaluation index is compared with a preset multi-level evaluation threshold to determine the stability level of the corresponding sensor. Then, a preset weight correction ratio is matched according to the stability level. The initial confidence weight is weighted and the weight correction ratio is weighted to obtain the corrected intermediate confidence weight. At the same time, the current ambient light intensity parameter of the vehicle is collected, and the intermediate confidence weight is fine-tuned according to the ambient light intensity parameter. After the fine-tuning is completed, the confidence weight of each sensor is obtained.

[0019] Furthermore, the calculation module is specifically used for: The sensor data in the target dataset are classified according to preset risk event types, and based on the confidence weight of the corresponding sensor, a corresponding basic probability allocation value is assigned to the sensor data corresponding to each type of risk event. Based on the evidence synthesis rules of the DS theory, the basic probability allocation values ​​corresponding to various risk events are calculated by multi-source evidence fusion to obtain the fusion probability allocation value of each type of risk event. Based on the preset risk quantification mapping relationship, the fusion probability allocation values ​​of each type of risk event are weighted and summed to generate the dynamic risk coefficient corresponding to the vehicle.

[0020] Furthermore, the calculation module is specifically used for: According to the preset risk quantification mapping relationship, a corresponding dynamic weight coefficient is assigned to each type of risk event; For each type of risk event, the corresponding fusion probability allocation value and dynamic weight coefficient are multiplied to obtain the risk contribution value of each type of risk event; The risk contribution values ​​of all risk events are summed, and the summation result is normalized according to a preset numerical range to generate the dynamic risk coefficient corresponding to the vehicle.

[0021] The third aspect of the present invention proposes: A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the emergency braking control method as described above.

[0022] The fourth aspect of the present invention proposes: A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the emergency braking control method as described above.

[0023] Additional aspects and advantages 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

[0024] Figure 1 A flowchart of the emergency braking control method provided in the first embodiment of the present invention; Figure 2 This is a structural block diagram of an emergency braking control system provided in the third embodiment of the present invention.

[0025] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0027] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0028] Unless otherwise defined, 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. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0029] Please see Figure 1 The figure shows an emergency braking control method provided in the first embodiment of the present invention. The emergency braking control method provided in this embodiment can reasonably and effectively complete the emergency braking of the vehicle, thereby improving the control efficiency of emergency braking.

[0030] Specifically, this embodiment provides: An emergency braking control method specifically includes the following steps: Step S10: Perform spatiotemporal alignment processing on the multi-source sensor data collected by the preset sensor array inside the vehicle to generate the corresponding initial dataset. It's important to note that, firstly, the vehicle's internal sensor array (such as millimeter-wave radar, cameras, lidar, wheel speed sensors, etc.) collects multi-dimensional data, including obstacle distance, relative speed, vehicle speed, and steering angle. Because different sensors may have different sampling frequencies and time bases (e.g., camera frame rate 30Hz, radar 10Hz), spatiotemporal alignment is necessary. Specifically, timestamps are unified and spatial coordinates are calibrated (e.g., mapping obstacle coordinates from radar and cameras to the same vehicle coordinate system), generating an initial dataset containing time, space, and physical quantities. This provides a consistent data foundation for subsequent analysis and facilitates processing.

[0031] Step S20: The initial dataset is optimized using a preset DBSCAN clustering algorithm to generate a corresponding target dataset. Based on the actual environmental conditions of the vehicle and the target attributes of each sensor, the confidence weight of each sensor is dynamically calculated. It's worth noting that the initial dataset is then optimized using the pre-defined DBSCAN clustering algorithm. DBSCAN excels at handling noisy datasets, eliminating outliers (such as false obstacle detections from sensors) and filtering out high-density, high-confidence data clusters to generate the target dataset. Simultaneously, considering the vehicle's actual environmental conditions (such as rain or strong light) and sensor target attributes (e.g., camera accuracy decreases in strong light, while radar is more stable in rain), the confidence weight of each sensor is dynamically calculated. Specifically, a higher weight indicates more reliable data from that sensor in the current environment, providing a reference for data importance in subsequent risk calculations. This facilitates subsequent processing.

[0032] Step S30: Based on DS theory, calculate the dynamic risk coefficient corresponding to the vehicle according to the target dataset and the confidence weight, and detect the target coefficient level corresponding to the dynamic risk coefficient. It should be noted that, based on DS evidence theory (a multi-source information fusion method), the dynamic risk coefficient of the vehicle is calculated using the target dataset and confidence weights. DS theory can effectively fuse sensor data with different levels of confidence to quantify the collision risks faced by the vehicle (such as the probability of collision with the vehicle in front and the collision time). Then, the dynamic risk coefficient is used to correspond to the target coefficient level (such as low risk, medium risk, and high risk) to facilitate subsequent processing.

[0033] Step S40: Dynamically generate an adapted emergency braking strategy based on the target coefficient level, and send the emergency braking strategy to the vehicle controller to complete the corresponding emergency braking.

[0034] It's worth noting that, finally, an appropriate emergency braking strategy is dynamically generated based on the risk level. Specifically, for example, a warning is only issued for low-risk situations, light braking is applied for medium-risk situations, and full braking and seatbelt pretensioning are triggered for high-risk situations. This strategy is then executed after being sent to the vehicle controller, achieving "risk-level classification and precise braking," ensuring safety while avoiding unnecessary sudden braking that could negatively impact the driving experience. This facilitates subsequent processing.

[0035] Second Embodiment Furthermore, the step of optimizing the initial dataset using a preset DBSCAN clustering algorithm to generate the corresponding target dataset includes: The initial dataset is preprocessed for outliers, and isolated noise points in the sensor data are identified and removed using the 3σ criterion to generate an intermediate dataset. The neighborhood radius and minimum number of core points of the preset DBSCAN clustering algorithm are dynamically adjusted according to the sensor's acquisition frequency and data accuracy to form a suitable set of algorithm parameters. Based on the algorithm parameter set, the intermediate dataset is clustered using the preset DBSCAN clustering algorithm to select effective clusters whose data density meets the preset density threshold, and the core point data of each effective cluster are merged to generate the target dataset.

[0036] It should be noted that the first step is outlier preprocessing: the 3σ criterion (in a normal distribution, 99.7% of the data falls within the mean ± 3σ range, and those exceeding this range are considered outliers) is used to identify isolated noise points in the sensor data. Specifically, for example, the occasional "false obstacles at a distance" data in radar data are removed and an intermediate dataset is generated to reduce the interference of noise on subsequent analysis.

[0037] The second step is to dynamically adjust the DBSCAN algorithm parameters: The core parameters of DBSCAN are the "neighborhood radius" (the distance range around a point that is included in the cluster) and the "minimum number of core points" (the minimum number of points required to form a cluster). Since the sensor acquisition frequency (higher frequency data is needed when traveling at high speeds) and data accuracy (more accurate data for nearby obstacles) will change with the scene, these two parameters need to be dynamically adjusted. Specifically, for example, the neighborhood radius is reduced at low speeds (to ensure the accuracy of near-field data), and the minimum number of core points is increased at high speeds (to avoid false alarms from a single sensor forming a cluster), thus forming an algorithm parameter set adapted to the current scene.

[0038] The third step is clustering and target dataset generation: Based on the algorithm parameter set, DBSCAN clustering is performed on the intermediate dataset to aggregate spatially and temporally similar data points into clusters (such as consecutive frames of data about the same obstacle). Valid clusters with a data density (number of data points per unit space) that meets a preset density threshold are selected. Specifically, low-density clusters may be noise or transient interference and need to be removed. Finally, the core point data of each valid cluster (the most representative points in the cluster, such as the average distance and speed of obstacles) are merged to generate the target dataset. This dataset retains key information while removing redundancy and noise, providing high-quality input for subsequent risk calculations and facilitating subsequent processing.

[0039] Furthermore, the step of clustering the intermediate dataset using the preset DBSCAN clustering algorithm based on the algorithm parameter set to filter out effective clusters whose data density meets a preset density threshold includes: Based on the vehicle's real-time driving speed and steering angle, the intermediate dataset is divided into multiple driving state data subsets, each of which corresponds to a preset vehicle driving condition interval. The algorithm parameter group is called to perform DBSCAN clustering operation on each subset of driving state data. During the clustering process, the rate of change of the number of core points of each potential cluster is counted in real time, and the merging threshold of the cluster is dynamically adjusted according to the rate of change of the number of core points. Calculate the average data density of each cluster that has completed the clustering operation, compare the average data density with the preset density threshold, retain the clusters with an average data density greater than or equal to the preset density threshold as candidate clusters, and then remove the clusters in the candidate clusters whose core point ratio is lower than a preset ratio threshold to obtain the effective clusters.

[0040] It should be noted that, firstly, based on the vehicle's real-time driving speed and steering angle, the intermediate dataset is divided into multiple driving state data subsets: Under different driving conditions (such as straight high-speed driving, low-speed turning, idling), the distribution characteristics of sensor data differ significantly (e.g., obstacles are mostly distributed laterally when turning, and mostly in front when driving straight). Therefore, subsets are divided according to preset driving condition intervals (e.g., speed > 60 km / h and steering angle < 5° is the "high-speed straight driving" interval) to ensure that the data in each subset has similar distribution patterns, which facilitates accurate processing by clustering algorithms.

[0041] Secondly, the algorithm parameter set is used to perform DBSCAN clustering operations on each subset of driving state data. During the clustering process, the rate of change of the number of core points in each potential cluster is statistically analyzed in real time. Specifically, if the number of core points in a cluster increases rapidly over time (e.g., the number of data points doubles for three consecutive frames), it indicates that the obstacle is rapidly approaching, and the merging threshold of the cluster needs to be dynamically adjusted (lowering the merging threshold to avoid the same obstacle being divided into multiple clusters); if the number of core points drops sharply, it may be noise, and the merging threshold needs to be increased (to avoid invalid clustering). Through dynamic adjustment, it is ensured that the clusters accurately reflect the true motion state of the obstacle.

[0042] Finally, effective clusters are selected: the average data density of each completed cluster is calculated, and candidate clusters with a density ≥ a preset threshold are retained (to ensure data reliability); then, clusters with a core point ratio (number of core points / total number of points) lower than a preset threshold (e.g., 50%) are removed. Specifically, a low core point ratio indicates a high number of noisy points in the cluster, resulting in poor reliability. The effective clusters obtained through this double selection process accurately represent real-world obstacles or risk sources, providing core data for the generation of the target dataset, thus facilitating subsequent processing.

[0043] Furthermore, the step of dynamically calculating the confidence weight of each sensor based on the actual environmental conditions of the vehicle and the target attributes of each sensor includes: Environmental interference parameters and road condition parameters are extracted from the actual environmental conditions. Simultaneously, the perception accuracy parameters and environmental adaptation parameters are extracted from the target attributes of each sensor. The environmental interference parameters, road condition parameters, perception accuracy parameters, and environmental adaptation parameters are quantized to obtain the corresponding environmental quantization values ​​and sensor quantization values. Construct a weight mapping matrix between the target attribute and the actual environmental conditions, and input the environmental quantization value and the sensor quantization value into the weight mapping matrix to obtain the initial confidence weight of each sensor through matrix operations; The sensor stability index during the vehicle's operation is collected, and the initial confidence weight is dynamically corrected based on the sensor stability index to obtain the confidence weight of each sensor.

[0044] It's important to note that the first step is parameter extraction and quantification: This involves extracting environmental interference parameters (such as light intensity, rainfall, and haze concentration) and road condition parameters (such as road surface smoothness and slope) from actual environmental conditions; simultaneously, extracting perception accuracy parameters (such as ranging error range and angular resolution) and environmental adaptation parameters (such as the camera's backlight suppression capability and the radar's rain and fog penetration rate) from the sensor's target attributes. These parameters are then quantified (e.g., converting light intensity from 0-10000 lux to an environmental quantization value of 0-1, and converting radar rain and fog penetration rate of 80% to a sensor quantization value of 0.8), transforming qualitative descriptions into calculable numerical values.

[0045] The second step is to construct a weight mapping matrix and calculate the initial weights: the rows of the matrix represent the sensor target attributes (accuracy, adaptability), the columns represent the actual environmental conditions (lighting, rainfall, etc.), and the matrix elements are the correlation coefficients between the attributes and the environment (e.g., the correlation coefficient between camera accuracy and light intensity is -0.8, meaning the stronger the light, the worse the accuracy). The environmental quantization value and the sensor quantization value are input into the matrix, and the initial confidence weight of each sensor is obtained through matrix multiplication (e.g., sensor quantization value × correlation coefficient × environmental quantization value). Specifically, for example, in rainy weather, the initial weight of the radar (0.8) is higher than that of the camera (0.3), which reflects the differences in sensor performance under actual environmental conditions.

[0046] The third step is to adjust the initial weights based on stability indicators: Stability indicators of the sensors during operation are collected (such as data packet loss rate and sampling frequency fluctuations). If a sensor frequently loses packets, it indicates poor stability, and its weight should be reduced. Through stability adjustment, the final confidence weights reflect both the inherent performance of the sensors in the current environment and their real-time operating status, providing a reliable weighting basis for subsequent risk fusion and facilitating subsequent processing.

[0047] Furthermore, the step of dynamically correcting the initial confidence weights based on the sensor stability index to obtain the confidence weights for each sensor includes: The stability index of each sensor is collected within a preset time window. The stability index includes data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value. The data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value are processed by normalization algorithm to obtain the corresponding standardized values. Based on the standardized values, a comprehensive stability evaluation index is constructed. The comprehensive stability evaluation index is compared with a preset multi-level evaluation threshold to determine the stability level of the corresponding sensor. Then, a preset weight correction ratio is matched according to the stability level. The initial confidence weight is weighted and the weight correction ratio is weighted to obtain the corrected intermediate confidence weight. At the same time, the current ambient light intensity parameter of the vehicle is collected, and the intermediate confidence weight is fine-tuned according to the ambient light intensity parameter. After the fine-tuning is completed, the confidence weight of each sensor is obtained.

[0048] It should be noted that, firstly, sensor stability indicators are collected within a preset time window (e.g., the last 5 seconds): data packet loss rate (number of packet losses / total number of transmissions) reflects data integrity; sampling frequency fluctuation value (deviation between actual frequency and nominal frequency) reflects sampling stability; and measurement error deviation value (average deviation from the true value) reflects data accuracy stability. These indicators are processed using a normalization algorithm (mapping indicator values ​​to 0-1, with smaller values ​​indicating better stability) to obtain standardized values. Specifically, for example, a packet loss rate of 10% is normalized to 0.1, and 50% is normalized to 0.5.

[0049] Secondly, a comprehensive stability evaluation index is constructed: three standardized values ​​are weighted and summed according to preset weights (e.g., packet loss rate 40%, frequency fluctuation 30%, error deviation 30%) to obtain the comprehensive index. The index is compared with preset multi-level evaluation thresholds (e.g., ≤0.2 is "excellent", 0.2-0.5 is "good", >0.5 is "poor") to determine the stability level of the sensor; then, a preset weight adjustment ratio is matched according to the level (e.g., +10% for "excellent", -30% for "poor").

[0050] Finally, weight correction and fine-tuning are performed: the initial confidence weight is weighted and calculated with the correction ratio to obtain the intermediate confidence weight (e.g., if the initial weight is 0.8, and the "good" correction ratio is increased by 5%, then the intermediate weight is 0.84). Simultaneously, the current ambient light intensity parameter is collected. Specifically, for example, the intermediate weight of the camera under strong light needs to be reduced by 5%, and under backlight by 10%, further adapting to instantaneous environmental changes through fine-tuning of lighting. The final confidence weight comprehensively reflects the sensor's environmental adaptability and real-time stability, ensuring the reliability of data fusion and facilitating subsequent processing.

[0051] Furthermore, the step of calculating the dynamic risk coefficient corresponding to the vehicle based on the DS theory, according to the target dataset and the confidence weights, includes: The sensor data in the target dataset are classified according to preset risk event types, and based on the confidence weight of the corresponding sensor, a corresponding basic probability allocation value is assigned to the sensor data corresponding to each type of risk event. Based on the evidence synthesis rules of the DS theory, the basic probability allocation values ​​corresponding to various risk events are calculated by multi-source evidence fusion to obtain the fusion probability allocation value of each type of risk event. Based on the preset risk quantification mapping relationship, the fusion probability allocation values ​​of each type of risk event are weighted and summed to generate the dynamic risk coefficient corresponding to the vehicle.

[0052] It should be noted that the first step is risk event classification and basic probability allocation: the sensor data in the target dataset are classified according to preset risk event types (such as "frontal collision risk", "lateral pedestrian risk", "road obstacle risk"); combined with the confidence weight of the corresponding sensor, a basic probability allocation value (BPA) is assigned to each type of risk event. Specifically, BPA represents the probability that a certain sensor data supports the occurrence of the risk event. The higher the weight of the sensor, the greater the BPA weight of its data (for example, radar data with a confidence of 0.8 has a BPA of 0.7 for supporting "frontal collision risk", and camera data with a confidence of 0.3 has a BPA of 0.2).

[0053] The second step is multi-source evidence fusion: Based on evidence synthesis rules of DS theory (such as Dempster's combination rule), the BPA (Best Probability Aspect) of different sensors and different types of risk events is fused and calculated. Specifically, for example, if both radar and camera detect the approaching vehicle, the BPA of "forward collision risk" will be significantly improved after fusion. If there is conflict in the sensor data (e.g., radar detects an obstacle, but the camera does not), the rule will redistribute the probability through the conflict coefficient to avoid misjudgment by a single sensor. After fusion, the fused probability allocation value for each type of risk event is obtained, representing the comprehensive probability that multiple sources of evidence jointly support that risk.

[0054] The third step is the generation of dynamic risk coefficients: Based on the preset risk quantification mapping relationship (e.g., a fusion probability of 0.8 for "forward collision risk" corresponds to a risk value of 80, and a fusion probability of 0.6 for "lateral pedestrian risk" corresponds to a risk value of 60), the fusion probability allocation values ​​of each type of risk event are weighted and summed according to their impact weight on vehicle safety (e.g., collision risk weight 0.6, pedestrian risk weight 0.4) to obtain the overall dynamic risk coefficient of the vehicle. Specifically, this coefficient quantifies the comprehensive risk currently faced by the vehicle, providing a quantitative basis for subsequent braking strategies to facilitate subsequent processing.

[0055] Furthermore, the step of performing a weighted summation operation on the fusion probability allocation values ​​of each type of risk event based on a preset risk quantification mapping relationship to generate the dynamic risk coefficient corresponding to the vehicle includes: According to the preset risk quantification mapping relationship, a corresponding dynamic weight coefficient is assigned to each type of risk event; For each type of risk event, the corresponding fusion probability allocation value and dynamic weight coefficient are multiplied to obtain the risk contribution value of each type of risk event; The risk contribution values ​​of all risk events are summed, and the summation result is normalized according to a preset numerical range to generate the dynamic risk coefficient corresponding to the vehicle.

[0056] It should be noted that, firstly, dynamic weight coefficients are assigned to each type of risk event according to a preset risk quantification mapping relationship: the dynamic weight is determined based on the urgency and severity of the risk event, for example, the dynamic weight of the collision risk of "distance to the vehicle in front <50m and relative speed >20km / h" is 0.7 (highest), and the dynamic weight of "minor obstacles on the road" is 0.2 (lower); at the same time, the dynamic weight will be adjusted according to the vehicle status (e.g., the weight of the side pedestrian risk increases at low speeds, and the weight of the frontal collision risk increases at high speeds).

[0057] Secondly, calculate the risk contribution value for each type of risk event: multiply the fusion probability allocation value of this type of event (e.g., 0.9) by its dynamic weight coefficient (e.g., 0.7) to obtain the risk contribution value (0.9 × 0.7 = 0.63). Specifically, the higher the contribution value, the greater the impact of the event on the overall risk. For example, the "risk of emergency collision with the vehicle ahead" with a high fusion probability and a large dynamic weight will become a major component of the overall risk.

[0058] Finally, a dynamic risk coefficient is generated: the risk contribution values ​​of all risk events are summed (e.g., 0.63 + 0.12 + 0.05 = 0.8), and then normalized according to a preset value range (e.g., 0-100) (0.8 × 100 = 80), resulting in the final dynamic risk coefficient (e.g., 80). This provides a unified and intuitive quantitative standard for the subsequent classification of target coefficient levels and the generation of braking strategies, ensuring the accuracy and consistency of emergency braking control, thus facilitating subsequent processing.

[0059] Please see Figure 2 The third embodiment of the present invention provides: An emergency braking control system, wherein the system comprises: The acquisition module is used to perform spatiotemporal alignment processing on the multi-source sensor data collected by the preset sensor array inside the vehicle to generate the corresponding initial dataset. The optimization module is used to optimize the initial dataset using a preset DBSCAN clustering algorithm to generate a corresponding target dataset, and dynamically calculate the confidence weight of each sensor based on the actual environmental conditions of the vehicle and the target attributes of each sensor. The calculation module is used to calculate the dynamic risk coefficient corresponding to the vehicle based on the DS theory, the target dataset, and the confidence weight, and to detect the target coefficient level corresponding to the dynamic risk coefficient. The generation module is used to dynamically generate an adapted emergency braking strategy based on the target coefficient level, and send the emergency braking strategy to the vehicle controller to complete the corresponding emergency braking.

[0060] Furthermore, the optimization module is specifically used for: The initial dataset is preprocessed for outliers, and isolated noise points in the sensor data are identified and removed using the 3σ criterion to generate an intermediate dataset. The neighborhood radius and minimum number of core points of the preset DBSCAN clustering algorithm are dynamically adjusted according to the sensor's acquisition frequency and data accuracy to form a suitable set of algorithm parameters. Based on the algorithm parameter set, the intermediate dataset is clustered using the preset DBSCAN clustering algorithm to select effective clusters whose data density meets the preset density threshold, and the core point data of each effective cluster are merged to generate the target dataset.

[0061] Furthermore, the optimization module is specifically used for: Based on the vehicle's real-time driving speed and steering angle, the intermediate dataset is divided into multiple driving state data subsets, each of which corresponds to a preset vehicle driving condition interval. The algorithm parameter group is called to perform DBSCAN clustering operation on each subset of driving state data. During the clustering process, the rate of change of the number of core points of each potential cluster is counted in real time, and the merging threshold of the cluster is dynamically adjusted according to the rate of change of the number of core points. Calculate the average data density of each cluster that has completed the clustering operation, compare the average data density with the preset density threshold, retain the clusters with an average data density greater than or equal to the preset density threshold as candidate clusters, and then remove the clusters in the candidate clusters whose core point ratio is lower than a preset ratio threshold to obtain the effective clusters.

[0062] Furthermore, the optimization module is specifically used for: Environmental interference parameters and road condition parameters are extracted from the actual environmental conditions. Simultaneously, the perception accuracy parameters and environmental adaptation parameters are extracted from the target attributes of each sensor. The environmental interference parameters, road condition parameters, perception accuracy parameters, and environmental adaptation parameters are quantized to obtain the corresponding environmental quantization values ​​and sensor quantization values. Construct a weight mapping matrix between the target attribute and the actual environmental conditions, and input the environmental quantization value and the sensor quantization value into the weight mapping matrix to obtain the initial confidence weight of each sensor through matrix operations; The sensor stability index during the vehicle's operation is collected, and the initial confidence weight is dynamically corrected based on the sensor stability index to obtain the confidence weight of each sensor.

[0063] Furthermore, the optimization module is specifically used for: The stability index of each sensor is collected within a preset time window. The stability index includes data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value. The data packet loss rate, acquisition frequency fluctuation value and measurement error deviation value are processed by normalization algorithm to obtain the corresponding standardized values. Based on the standardized values, a comprehensive stability evaluation index is constructed. The comprehensive stability evaluation index is compared with a preset multi-level evaluation threshold to determine the stability level of the corresponding sensor. Then, a preset weight correction ratio is matched according to the stability level. The initial confidence weight is weighted and the weight correction ratio is weighted to obtain the corrected intermediate confidence weight. At the same time, the current ambient light intensity parameter of the vehicle is collected, and the intermediate confidence weight is fine-tuned according to the ambient light intensity parameter. After the fine-tuning is completed, the confidence weight of each sensor is obtained.

[0064] Furthermore, the calculation module is specifically used for: The sensor data in the target dataset are classified according to preset risk event types, and based on the confidence weight of the corresponding sensor, a corresponding basic probability allocation value is assigned to the sensor data corresponding to each type of risk event. Based on the evidence synthesis rules of the DS theory, the basic probability allocation values ​​corresponding to various risk events are calculated by multi-source evidence fusion to obtain the fusion probability allocation value of each type of risk event. Based on the preset risk quantification mapping relationship, the fusion probability allocation values ​​of each type of risk event are weighted and summed to generate the dynamic risk coefficient corresponding to the vehicle.

[0065] Furthermore, the calculation module is specifically used for: According to the preset risk quantification mapping relationship, a corresponding dynamic weight coefficient is assigned to each type of risk event; For each type of risk event, the corresponding fusion probability allocation value and dynamic weight coefficient are multiplied to obtain the risk contribution value of each type of risk event; The risk contribution values ​​of all risk events are summed, and the summation result is normalized according to a preset numerical range to generate the dynamic risk coefficient corresponding to the vehicle.

[0066] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the emergency braking control method as described above.

[0067] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the emergency braking control method as described above.

[0068] In summary, the emergency braking control method and system provided by the above embodiments of the present invention can reasonably and effectively complete the emergency braking of the vehicle, thereby improving the control efficiency of emergency braking.

[0069] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0070] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0071] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0072] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0073] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0074] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. An emergency brake control method characterized by, The method comprises: The multi-source sensor data collected by the preset sensor array in the vehicle interior is processed by space-time alignment to generate a corresponding initial data set; The initial data set is optimized by a preset DBSCAN clustering algorithm to generate a corresponding target data set, and the confidence weight of each sensor is dynamically calculated based on the actual environmental conditions of the vehicle and the target attributes of each sensor; Based on the D-S theory, the dynamic risk coefficient corresponding to the vehicle is calculated according to the target data set and the confidence weight, and the target coefficient level corresponding to the dynamic risk coefficient is detected; According to the target coefficient level, an adaptive emergency braking strategy is dynamically generated, and the emergency braking strategy is sent to the vehicle controller to complete the corresponding emergency braking.

2. The emergency brake control method according to claim 1, characterized by, The step of optimizing the initial data set by the preset DBSCAN clustering algorithm to generate the corresponding target data set comprises: The initial data set is preprocessed for outliers, and isolated noise points in the sensor data are identified and removed by the 3σ rule to generate an intermediate data set; The field radius and minimum core point number of the preset DBSCAN clustering algorithm are dynamically adjusted according to the collection frequency and data accuracy of the sensor to form an adaptive algorithm parameter group; Based on the algorithm parameter group, the intermediate data set is clustered by the preset DBSCAN clustering algorithm to screen out effective cluster families with data density meeting a preset density threshold, and the core point data of each effective cluster family is fused to generate the target data set.

3. The emergency brake control method according to claim 2, characterized by The step of clustering the intermediate data set by the preset DBSCAN clustering algorithm based on the algorithm parameter group to screen out effective cluster families with data density meeting a preset density threshold comprises: Based on the real-time driving speed and steering angle of the vehicle, the intermediate data set is divided into a plurality of driving state data subsets, each of which corresponds to a preset vehicle driving operating condition interval; The algorithm parameter group is called to perform DBSCAN clustering operation on each driving state data subset, and the core point number change rate of each potential cluster family is calculated in real time during the clustering process, and the merging threshold of the cluster family is dynamically corrected according to the core point number change rate; The average data density of each cluster family after clustering operation is calculated, the average data density is compared with the preset density threshold, the cluster family with average data density greater than or equal to the preset density threshold is reserved as a candidate cluster family, and the cluster family with core point proportion less than a preset proportion threshold is removed from the candidate cluster family to obtain the effective cluster family.

4. The emergency brake control method according to claim 1, characterized by The step of dynamically calculating the confidence weight of each sensor based on the actual environmental conditions of the vehicle and the target attributes of each sensor comprises: extracting an environmental interference parameter and a road working condition parameter in the actual environmental condition, and extracting a perception accuracy parameter and an environmental adaptation parameter in a target attribute of each sensor, and respectively quantifying the environmental interference parameter, the road working condition parameter, the perception accuracy parameter and the environmental adaptation parameter to obtain corresponding environmental quantization values and sensor quantization values; constructing a weight mapping matrix of the target attribute and the actual environmental condition, and inputting the environmental quantization values and the sensor quantization values into the weight mapping matrix to obtain an initial confidence weight of each sensor through matrix operation; acquiring a sensor stability index in the vehicle driving process, and dynamically correcting the initial confidence weight based on the sensor stability index to obtain a confidence weight of each sensor.

5. The emergency brake control method according to claim 4, characterized by The step of dynamically correcting the initial confidence weight based on the sensor stability index to obtain a confidence weight of each sensor includes: acquiring a stability index of each sensor within a preset time window, the stability index including a data packet loss rate, an acquisition frequency fluctuation value and a measurement error deviation value, and respectively processing the data packet loss rate, the acquisition frequency fluctuation value and the measurement error deviation value by a normalization algorithm to obtain corresponding standardized values; constructing a stability comprehensive evaluation index based on the standardized values, comparing the stability comprehensive evaluation index with a preset multi-level evaluation threshold to determine a stability level of the corresponding sensor, and matching a preset weight correction proportion according to the stability level; performing weighted operation on the initial confidence weight and the weight correction proportion to obtain a corrected intermediate confidence weight, simultaneously acquiring a current environmental light intensity parameter of the vehicle, and fine-tuning the intermediate confidence weight according to the environmental light intensity parameter, and obtaining a confidence weight of each sensor after fine-tuning is completed.

6. The emergency brake control method according to claim 1, characterized by The step of calculating a dynamic risk coefficient corresponding to the vehicle based on the target data set and the confidence weight according to the D-S theory includes: classifying sensor data in the target data set according to a preset risk event type, and combining the confidence weight of the corresponding sensor to assign a corresponding basic probability distribution value to the sensor data corresponding to each type of risk event; based on the evidence combination rule of the D-S theory, performing multi-source evidence fusion calculation on the basic probability distribution values corresponding to each type of risk event to obtain a fusion probability distribution value of each type of risk event; performing weighted summation operation on the fusion probability distribution value of each type of risk event according to a preset risk quantization mapping relationship to generate the dynamic risk coefficient corresponding to the vehicle.

7. The emergency brake control method according to claim 6, characterized by The step of performing weighted summation operation on the fusion probability distribution value of each type of risk event according to a preset risk quantization mapping relationship to generate the dynamic risk coefficient corresponding to the vehicle includes: assigning a corresponding dynamic weight coefficient to each type of risk event according to the preset risk quantization mapping relationship; For each type of risk event, the corresponding fusion probability distribution value and dynamic weight coefficient are multiplied to obtain the risk contribution value of each type of risk event; The risk contribution values of all risk events are summed and normalized according to a preset numerical range to generate a dynamic risk coefficient corresponding to the vehicle.

8. An emergency brake control system characterized by, The system comprises: A collection module for performing spatio-temporal alignment processing on multi-source sensor data collected by a preset sensor array inside a vehicle to generate a corresponding initial data set; An optimization module for performing optimization processing on the initial data set by a preset DBSCAN clustering algorithm to generate a corresponding target data set, and dynamically calculating a confidence weight of each sensor based on actual environmental conditions of the vehicle and target attributes of each sensor; A calculation module for calculating a dynamic risk coefficient corresponding to the vehicle based on the target data set and the confidence weight according to D-S theory, and correspondingly detecting a target coefficient level corresponding to the dynamic risk coefficient; A generation module for dynamically generating an adaptive emergency braking strategy according to the target coefficient level, and issuing the emergency braking strategy to a vehicle controller to complete corresponding emergency braking.

9. A computer comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the emergency braking control method of any one of claims 1-7.

10. A readable storage medium, having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the emergency braking control method of any one of claims 1-7.

Images