Vehicle anti-collision early warning method and device

Abstract

The invention discloses a vehicle anti-collision early warning method and device, and relates to the technical field of vehicle automatic control, and the method comprises the steps: determining a fusion weight corresponding to each sensor based on output data and a multi-dimensional evaluation index corresponding to each sensor in a plurality of sensors, and obtaining a fusion result, integrating the fusion result, the road attribute information, the external collaborative perception data and the meteorological early warning information to obtain comprehensive data; predicting an environment change trend in a future preset duration based on the comprehensive data, outputting an environment situation level corresponding to the environment change trend, and determining a detection strategy; when it is determined that the surrounding environment of the vehicle meets the preset condition of the environment situation level, a detection strategy is executed, and a detection result is obtained; and determining an anti-collision early warning measure according to the detection result and the comprehensive data, and outputting the anti-collision early warning measure. Therefore, the anti-collision effect is improved.

Description

Technical Field

[0001] This application relates to the field of vehicle automatic control technology, and in particular to a vehicle collision avoidance warning method and device. Background Technology

[0002] Among related technologies, existing collision avoidance warning methods for autonomous driving systems have some shortcomings. In terms of warning accuracy, most rely on analysis and judgment based on data from a single sensor, which is easily affected by the sensor's own accuracy and environmental factors, resulting in low warning accuracy and frequent false alarms or missed alarms. At the same time, in rainy weather and extreme environments, such as heavy rain, heavy snow, dense fog, and strong sunlight, the performance of sensors will be severely affected. Cameras are easily obscured by rain in rainy weather, resulting in blurred images and inability to accurately identify obstacles ahead. In dense fog, the propagation of electromagnetic waves by radar is interfered with, and the detection accuracy and distance are greatly reduced, making the collision avoidance warning effect poor. Summary of the Invention

[0003] In view of this, this application provides a vehicle collision avoidance warning method and device to solve the problem of poor collision avoidance effect in the prior art.

[0004] The objective of this application can be achieved through the following technical solutions: The first aspect of this application is to provide a vehicle collision avoidance warning method, including: Acquire the output data and multi-dimensional evaluation metrics for each of the multiple sensors; Based on multi-dimensional evaluation indicators, the real-time reliability coefficient of each sensor is calculated using fuzzy logic algorithm. Based on the real-time reliability coefficient, a dynamic weight allocation model is constructed, and the fusion weights corresponding to each sensor are output through the dynamic weight allocation model. The output data of each sensor are fused based on the fusion weights to obtain the fusion result. Acquire road attribute information, external collaborative perception data, and weather warning information of the road where the vehicle is currently located; By integrating and fusing the results, road attribute information, external collaborative sensing data, and meteorological early warning information, comprehensive data is obtained. Based on comprehensive data, a spatiotemporal sequence prediction algorithm is used to predict the environmental change trend within a preset time period in the future, and output the environmental situation level corresponding to the environmental change trend. Determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed to obtain the detection results, which include the target's speed information and position information. Collision risk indicators are determined based on test results and comprehensive data; The collision risk level is determined based on collision risk indicators; Based on the collision risk level, determine the collision avoidance warning measures and output the collision avoidance warning measures.

[0005] In one optional embodiment, the fusion weights corresponding to each sensor are output through a dynamic weight allocation model, including: If there is no conflict in the output data of each sensor, the fusion weight of each sensor is determined according to the ratio of the real-time reliability coefficient of each sensor, and the first fusion weight is obtained. If there are conflicts in the output data of each sensor, the conflicting data is determined based on the output data; the confidence level of the conflicting data is calculated based on the real-time reliability coefficient and the data consistency test results; and the fusion weight of each sensor is determined based on the real-time reliability coefficient and the confidence level.

[0006] In one optional embodiment, the fusion weights of each sensor are determined based on a real-time reliability coefficient and a confidence level, including: The fusion weight of sensors whose output data is conflicting data, whose output data confidence is higher than a preset confidence threshold, and whose real-time reliability coefficient is higher than a preset real-time reliability coefficient threshold is set as the second fusion weight; the fusion weight of sensors whose output data is conflicting data and whose real-time reliability coefficient is not higher than a preset real-time reliability coefficient threshold, and / or whose output data is conflicting data and whose output data confidence is not higher than a preset confidence threshold, is set as the third fusion weight; the fusion weight of sensors whose output data is not conflicting data is determined based on the second fusion weight and the third fusion weight, wherein the second fusion weight is greater than the first fusion weight, and the third fusion weight is less than the first fusion weight.

[0007] In an optional embodiment, after executing a detection strategy and obtaining the detection result when it is determined that the vehicle's surrounding environment meets preset conditions for an environmental situation level, the method further includes: Determine the deviation between the test results and the actual test results; The detection strategy is adjusted based on the detection bias.

[0008] In one optional embodiment, the collision risk index includes collision time and collision distance. The collision risk index is determined based on detection results and comprehensive data, including: Based on the test results and the comprehensive data, the collision risk index is calculated using the following formula:

[0009]

[0010] The target distance is the distance between the target vehicle and the target.

[0011] In one optional embodiment, determining the collision risk level based on a collision risk index includes: When the collision time falls within the first preset time range or the collision distance falls within the first preset distance range, the current collision risk level is determined to be a Level 1 collision risk level. When the collision time falls within the second preset time range or the collision distance falls within the second preset distance range, the current collision risk level is determined to be a level two collision risk level. If the collision time falls within the third preset time range or the collision distance falls within the third preset distance range, the current collision risk level is determined to be a level three collision risk level.

[0012] In one optional embodiment, collision avoidance warning measures are determined based on the collision risk level, including: If the current collision risk level is Level 1, the driver will be alerted via indicator lights and a buzzer on the dashboard. If the forward collision risk level is Level 2, control the vehicle's braking and reduce its speed; If the current collision risk level is level three, emergency braking and steering avoidance measures will be taken. A second aspect of this application provides a vehicle collision avoidance warning device, comprising: The first acquisition module is used to acquire the output data and multi-dimensional evaluation indicators corresponding to each of the multiple sensors. The calculation module is used to calculate the real-time reliability coefficient of each sensor based on multi-dimensional evaluation indicators and using fuzzy logic algorithms. The module is used to build a dynamic weight allocation model based on the real-time reliability coefficient, and output the fusion weights corresponding to each sensor through the dynamic weight allocation model; The fusion module is used to fuse the output data of each sensor based on fusion weights to obtain the fusion result. The second acquisition module is used to acquire road attribute information of the road where the vehicle is currently located, external collaborative perception data, and weather warning information. The integration module is used to integrate the fusion results, road attribute information, external collaborative sensing data, and meteorological early warning information to obtain comprehensive data; The prediction module is used to predict the environmental change trend within a preset time period based on comprehensive data and spatiotemporal sequence prediction algorithms, and output the environmental situation level corresponding to the environmental change trend. The first determining module is used to determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; The execution module is used to execute the detection strategy and obtain the detection results when it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level. The detection results include the target's speed information and position information. The second determination module is used to determine collision risk indicators based on the detection results and comprehensive data; The third determination module is used to determine the collision risk level based on collision risk indicators; The fourth determination module is used to determine collision avoidance warning measures based on the collision risk level and output the collision avoidance warning measures.

[0013] A third aspect of this application is to provide an electronic device, 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 perform the method as described in the first aspect.

[0014] A fourth aspect of this application is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the method as described in the first aspect.

[0015] Compared with existing technologies, this application provides a vehicle collision avoidance warning method that determines the fusion weights for each sensor based on the output data and multi-dimensional evaluation indicators of multiple sensors, thereby obtaining a fusion result. This fusion result, along with road attribute information, external collaborative perception data, and meteorological warning information, is then integrated to obtain comprehensive data. Based on this comprehensive data, the method predicts environmental change trends within a preset time period and outputs the corresponding environmental situation level, thereby determining a detection strategy. When the vehicle's surrounding environment meets the preset conditions for the environmental situation level, the detection strategy is executed, yielding a detection result. Based on the detection result and the comprehensive data, collision avoidance warning measures are determined and output. This improves the collision avoidance effect. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart of a vehicle collision avoidance warning method provided in an embodiment of this application; Figure 2 A structural block diagram of a vehicle collision avoidance warning device provided in an embodiment of this application; Figure 3This is a structural block diagram of an electronic device for implementing a vehicle collision avoidance warning method, provided in an embodiment of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] It should be understood that in the embodiments of this application, "at least one" means one or more, and "more than one" means two or more. "And / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the related objects before and after it are in an "or" relationship. "Contains A, B and / or C" means containing any one, two, or three of A, B, and C.

[0021] It should be understood that in the embodiments of this application, "B corresponding to A", "B corresponding to A", "A corresponds to B" or "B corresponds to A" means that B is associated with A, and B can be determined based on A. Determining B based on A does not mean that B is determined solely based on A; B can also be determined based on A and / or other information.

[0022] To address the technical problems existing in related technologies, this application provides a vehicle collision avoidance warning method and device.

[0023] The vehicle collision avoidance warning method provided in this application can be executed by an electronic device, which can be a terminal or a server. The terminal can be a smartphone, tablet, laptop, or other similar device. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. It is understood that this application does not limit the specific entity executing the vehicle collision avoidance warning method.

[0024] The technical solution of this application will be described in detail below through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments described below are used to explain the technical solution of this application and are not intended to limit actual use.

[0025] To address the technical problems existing in related technologies, embodiments of this application provide a vehicle collision avoidance warning method, such as... Figure 1 As shown, Figure 1 This is an example flowchart of a vehicle collision avoidance warning method provided in an embodiment of this application. It should be noted that the steps shown may be executed in a different logical order than that shown in the flowchart. The method may include the following steps S101 to S112.

[0026] Step S101: Obtain the output data and multi-dimensional evaluation indicators for each of the multiple sensors.

[0027] In one alternative embodiment, the multiple sensors include, but are not limited to, at least two of the following: lidar, millimeter-wave radar, camera, ultrasonic sensor, humidity sensor, light sensor, and rain sensor.

[0028] In one specific embodiment, the multi-sensor sensing system consists of a 128-line lidar, a 77GHz millimeter-wave radar, a high-definition wide-angle camera, and an ultrasonic sensor. The lidar has a detection range of up to 200 meters and an angular resolution of 0.1 degrees, enabling it to acquire three-dimensional information of the target with high precision. The 77GHz millimeter-wave radar has a detection range of up to 300 meters and a speed measurement range of -100km / h to +200km / h, and can operate stably in adverse weather conditions. The high-definition wide-angle camera has a resolution of 1920×1080 and a frame rate of 30fps, used to identify detailed features of the target. The ultrasonic sensor has a detection range of 0.1-5 meters, suitable for close-range detection.

[0029] In one optional embodiment, the output data includes: three-dimensional information of the target acquired by lidar; relative distance, relative velocity, and azimuth of the target acquired by millimeter-wave radar; detailed features of the target acquired by a high-definition wide-angle camera; relative distance of the target acquired by an ultrasonic sensor; air humidity value acquired by a humidity sensor; illumination information acquired by a light sensor; and rainfall amount acquired by a rain sensor.

[0030] In one specific embodiment, the target refers to vehicles, pedestrians, bicycles, curbs, and stationary or moving objects in front. Three-dimensional information includes the target's spatial location, shape, and size. Detailed features include the target's texture, color, outline, and semantic information.

[0031] In another alternative embodiment, a combination of hardware trigger signals and a precise time protocol is used to align the sampling times of multiple source sensors to a unified time reference, ensuring that the timestamp error of the data collected by each sensor at any time does not exceed a preset time.

[0032] In another optional embodiment, the coordinate systems of each sensor are integrated into the vehicle coordinate system, with the vehicle's center of mass as the origin, to establish a unified spatial coordinate system and ensure the consistency of data in space.

[0033] In one specific embodiment, in terms of data synchronization calibration, the acquisition time error of each sensor is strictly controlled within 1ms to ensure that the data acquired by different sensors at the same time point can correspond to each other. The coordinate system of each sensor is integrated into the vehicle coordinate system, and a unified spatial coordinate system is established with the center of mass of the vehicle as the origin to ensure the consistency of data in space.

[0034] In another alternative embodiment, the multi-dimensional evaluation metrics may include: data stability, environmental fit, and historical error rate.

[0035] It should be noted that data stability can be defined as the variance of the output data in consecutive preset frames; the smaller the variance, the better the data stability. Environmental matching includes the image clarity score of the camera in rainy weather and the point cloud density compliance rate of the LiDAR; a higher score indicates a clearer image. The historical error rate is the average deviation between the output data and the first fusion result in the last 100 detections; a smaller deviation indicates a lower historical error rate.

[0036] Step S102: Based on multi-dimensional evaluation indicators, the real-time reliability coefficients of each sensor are calculated using fuzzy logic algorithms.

[0037] It should be noted that fuzzy logic control algorithms are a type of computer digital control technology based on fuzzy set theory, fuzzy linguistic variables, and fuzzy logic reasoning. The real-time reliability coefficient ranges from 0 to 1.

[0038] In one alternative embodiment, the real-time reliability coefficient of each sensor is dynamically changing. For example, in rainy weather, the point cloud density compliance rate of lidar decreases, and its reliability coefficient may drop to 0.3, while millimeter-wave radar is less affected by weather, and its reliability coefficient may remain at around 0.8.

[0039] Step S103: Based on the real-time reliability coefficient, construct a dynamic weight allocation model, and output the fusion weights corresponding to each sensor through the dynamic weight allocation model.

[0040] It should be noted that the dynamic weight allocation model is a mathematical or algorithmic model that adjusts the weight of a sensor in the fusion result in real time based on the sensor's current working state and environmental conditions.

[0041] In one optional embodiment, the fusion weights corresponding to each sensor are output through a dynamic weight allocation model, specifically including the following steps: if there is no conflict in the output data of each sensor, the fusion weight of each sensor is determined according to the proportion of the real-time reliability coefficient of each sensor to obtain the first fusion weight; if there is a conflict in the output data of each sensor, the conflicting data is determined based on the output data; the confidence level of the conflicting data is calculated based on the real-time reliability coefficient and the data consistency test result; and the fusion weight of each sensor is determined based on the real-time reliability coefficient and the confidence level.

[0042] In one specific embodiment, the fusion weight of each sensor is determined based on the real-time reliability coefficient and confidence level, specifically including the following steps: setting the fusion weight of the sensor whose output data is conflicting data, whose output data confidence level is higher than a preset confidence threshold, and whose real-time reliability coefficient is higher than a preset real-time reliability coefficient threshold as the second fusion weight; setting the fusion weight of the sensor whose output data is conflicting data and whose real-time reliability coefficient is not higher than the preset real-time reliability coefficient threshold, and / or whose output data is conflicting data and whose output data confidence level is not higher than the preset confidence threshold as the third fusion weight; determining the fusion weight of the sensor whose output data is not conflicting data based on the second fusion weight and the third fusion weight, wherein the second fusion weight is greater than the first fusion weight, and the third fusion weight is less than the first fusion weight.

[0043] In a more specific embodiment, under normal weather conditions, assuming the real-time reliability coefficient of the lidar is 0.8, the millimeter-wave radar is 0.7, the camera is 0.6, and the ultrasonic sensor is 0.5, their fusion weights can be proportionally allocated as 0.32, 0.28, 0.24, and 0.16. When data conflicts occur, such as the millimeter-wave radar detecting an obstacle ahead while the camera fails to recognize it due to rain or fog, a conflict resolution mechanism is activated. The confidence level of the conflicting data is calculated, and combined with the sensor's own reliability and data consistency checks, sensors with high reliability coefficients and high data confidence are assigned higher weights. For example, in a rainstorm, the real-time reliability coefficient of the lidar is 0.3, and the real-time reliability coefficient of the millimeter-wave radar is 0.8. In this case, the fusion weight of the millimeter-wave radar is increased from the usual 0.28 to 0.6, the fusion weight of the lidar is reduced from 0.32 to 0.1, and the remaining weight of 0.3 is allocated to the camera and the ultrasonic sensor according to their real-time reliability coefficients of 0.5 and 0.7, respectively. That is, the fusion weight of the camera is 0.1, and the fusion weight of the ultrasonic sensor is 0.2.

[0044] In another optional embodiment, the fusion result is compared with the actual road conditions after the vehicle has traveled a preset distance to calculate the fusion error. Using a reinforcement learning algorithm, the parameters of the weight allocation model are corrected in reverse based on the magnitude of the error, making the weight adjustment more closely reflect actual environmental changes, thus forming a complete closed loop of "evaluation-allocation-feedback-optimization".

[0045] Step S104: Perform fusion processing on the output data of each sensor based on the fusion weight to obtain the fusion result.

[0046] In one alternative embodiment, the output data is weighted and summed according to the fusion weights to obtain the fusion result.

[0047] In another alternative embodiment, the fusion result is filtered to remove noise and interference signals.

[0048] In one specific embodiment, for LiDAR data, a voxel filtering algorithm is used to divide the 3D point cloud space into voxels of a certain size, with each voxel retaining a representative point, removing redundant and noise points to improve data processing efficiency. For millimeter-wave radar data, an extended Kalman filter algorithm is used to estimate the target's position and velocity by establishing system state equations and observation equations, reducing the impact of random noise. For camera images, a Gaussian filtering algorithm is first used for smoothing to reduce image noise, and then the Canny edge detection algorithm is used to extract the target's contour features, providing a foundation for subsequent target recognition.

[0049] In another specific embodiment, targeted processing algorithms are employed for special data in extreme environments. For example, special data processing algorithms for extreme environments include rain removal algorithms for rainy weather. In rainy weather, a deep learning-based rain removal algorithm is used on images captured by the camera to remove raindrops and water mist from the images, thereby enhancing image clarity.

[0050] In another specific embodiment, in rainy weather, the detection frequency and power of millimeter-wave radar and lidar are increased, and the images captured by the camera are processed to remove rain. The braking and steering parameters of the vehicle are adjusted according to the slipperiness of the road surface. In low visibility environments such as heavy fog and heavy snow, the detection is mainly carried out by millimeter-wave radar and lidar, with infrared cameras assisting in detection. In strong light environments, the camera is automatically exposed and white balance is corrected, and strong light interference is filtered out by polarizers.

[0051] Step S105: Obtain road attribute information, external collaborative perception data, and weather warning information of the road where the vehicle is currently located.

[0052] In one optional embodiment, the road attributes of the road where the vehicle is currently located are obtained through a high-precision map; external collaborative perception information and weather warning information are obtained through vehicle-to-everything (V2X) communication.

[0053] In a more specific embodiment, road attributes include at least one of the following attributes: lane line type, road width, number of roads, curve curvature, altitude, road slope, traffic signs, intersection topology, curb, guardrail, median strip, and road functional attributes.

[0054] In a more specific embodiment, external collaborative sensing data and weather warning information are acquired via V2X.

[0055] In a more specific embodiment, the external collaborative sensing data includes information such as the position, speed, acceleration, and size of other vehicles, as well as target information detected by other vehicles. The weather warning information includes current weather type, road condition information, visibility and illumination information, and warning spatiotemporal range information. The current weather type includes rain, snowfall, and fog, etc.; the road condition information includes road slippage index, icing warning, water depth, and road temperature, etc.; and the warning spatiotemporal range information includes affected road sections, expected duration, and warning level, etc. This application does not limit the specific content of the external collaborative sensing data and the weather warning information.

[0056] Step S106: Integrate the fusion results, road attribute information, external collaborative perception data, and meteorological early warning information to obtain comprehensive data.

[0057] In one alternative embodiment, the first fusion result, road attribute information, external collaborative perception data, and meteorological early warning information are integrated to construct a comprehensive data pool for storing the comprehensive data.

[0058] Step S107: Based on comprehensive data, use a spatiotemporal sequence prediction algorithm to predict the environmental change trend within a preset time period in the future, and output the environmental situation level corresponding to the environmental change trend.

[0059] In one optional embodiment, the environmental situation levels include normal, slightly adverse, moderately adverse, and severely adverse. It should be noted that the specific classification of these levels is based on actual needs, and this application does not impose any limitations on this classification.

[0060] In one specific embodiment, an improved Transformer spatiotemporal sequence prediction algorithm is used, taking historical data from the environmental information pool over the past 3 minutes (such as changes in parameters like rainfall, visibility, and light intensity) and real-time data as input, to predict environmental change trends within the next 30 seconds. For example, it predicts that the rainfall intensity will escalate from the current moderate rain to heavy rain within the next 30 seconds, and the visibility will decrease from 500 meters to 300 meters, outputting an environmental situation level of "moderately severe".

[0061] It's important to note that traditional spatiotemporal series prediction often employs the Transformer model. The core of the Transformer mechanism is self-attention, which captures long-term dependencies by calculating the correlation between different time points in the sequence. However, the Transformer model has the following limitations: it is primarily designed for one-dimensional time series and cannot directly extract spatial features from sensor data (such as environmental differences between sensors in different orientations). For high-frequency, long-sequence data like environmental monitoring, the computational complexity increases quadratically with the sequence length, making it difficult to meet the real-time requirements of predicting vehicle speeds within 30 seconds. Furthermore, it struggles to effectively handle heterogeneous data streams such as humidity, illumination, rainfall, and V2X warnings simultaneously.

[0062] In one alternative embodiment, to address the need for predicting extreme environments in autonomous driving, this application improves the algorithm as follows: A spatial attention layer is added to the encoder, introducing a spatiotemporal dual-dimensional attention mechanism to encode the spatial relationships of different sources (onboard sensors, V2X, high-precision maps). This allows the model to understand the spatial logical relationship between events such as "blizzard 3 kilometers ahead" and "increased humidity in the vehicle." A structure similar to Informer or LogSparse is adopted, performing attention calculations only on key time nodes (moments with large fluctuations in feature values). This significantly improves inference speed, ensuring the model can output predictions for the next 30 seconds in real time. A heterogeneous data alignment layer is added, mapping non-serialized "road attributes" (curvature, altitude) and serialized "sensor values" to the same high-dimensional feature space, enabling the Transformer to recognize the coupling relationship between environmental degradation and geographical conditions (such as fog in mountainous areas).

[0063] Furthermore, to enable the model to predict both "numerical trends" and "situation levels," a joint loss function is constructed. This joint loss function includes a regression loss for predicting rainfall, visibility, etc., and a classification loss for predicting situation levels. The regression loss can use mean squared error to ensure the model accurately predicts specific numerical changes over the next 30 seconds. The classification loss can be cross-entropy loss, with four situation levels: "normal," "mild," "moderate," and "severe." A spatiotemporal sliding window method is used to construct samples; for example, 3 minutes of historical data is used as the input window, and the actual observation value 30 seconds later is used as the label. For low-frequency but high-risk scenarios such as heavy rain and heavy snow, offline simulated data or historical extreme weather data recorded by V2X are introduced into the training set to address the problem of uneven sample distribution and improve the model's recall rate at the "severe" level. The predicted "situation level" is compared with the actual road conditions (ground truth) 10 meters later. If the prediction deviation exceeds 10%, gradient updates are triggered, dynamically correcting the weight parameters of the improved Transformer.

[0064] Step S108: Determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library.

[0065] In one specific embodiment, multiple adaptive strategies for sensors and vehicle control are preset for different environmental situation levels and predicted trends. For example, when it is predicted that the vehicle will enter a blizzard area in 10 seconds (situation level: "severely severe"), the detection strategy may include: increasing the detection frequency of the lidar from 10Hz to 20Hz, increasing the transmission power by 20%, and simultaneously activating a point cloud denoising enhancement algorithm; switching the millimeter-wave radar to "penetration mode" to expand the detection range to 350 meters; activating the camera's heating and defogging function, switching to infrared imaging mode, and loading a target recognition model for snow scenes; increasing the vehicle braking distance coefficient from 1.2 times (rainy weather) to 1.8 times, and increasing the steering assist gain by 10%.

[0066] In another optional embodiment, after determining that the vehicle's surrounding environment meets the preset conditions for an environmental situation level, the detection strategy is executed, and the detection result is obtained, the method further includes: Determine the detection deviation between the detection results and the actual detection results; adjust the detection strategy based on the detection deviation.

[0067] In one specific embodiment, the deviation between actual environmental changes and predictions is monitored in real time. If the deviation exceeds 10%, the strategy parameters are dynamically adjusted. For example, if the actual rate of visibility decline is 15% faster than predicted, the lidar detection frequency is further increased to 25Hz, and the millimeter-wave radar detection range is expanded to 380 meters.

[0068] Step S109: When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed and the detection result is obtained.

[0069] It should be noted that the detection results include the target's speed and position information.

[0070] In one optional embodiment, a grid map is used to construct a dynamic environment model, dividing the environment around the vehicle into grids of a preset size; a long short-term memory network is used to predict the target's trajectory to obtain the target's speed and position information.

[0071] In one specific embodiment, each grid cell is labeled based on the presence and type of target (e.g., vehicle, pedestrian, obstacle, etc.). For example, if a car is present in a grid cell, it is labeled as "vehicle," and its relevant characteristic parameters are recorded. This model can be updated in real time, accurately reflecting the dynamic changes in the vehicle's surrounding environment, including road conditions in extreme environments such as flooded or snow-covered roads.

[0072] In another specific embodiment, a Long Short-Term Memory (LSTM) network is used to predict the target's velocity and position information. During the prediction process, a dynamic environment model and historical data are combined to fully consider the impact of extreme environments on the target's motion. For example, on snow-covered roads, pedestrians may walk slower, and vehicles may brake more slowly, thus improving the accuracy of the prediction.

[0073] Step S110: Determine the collision risk index based on the detection results and comprehensive data.

[0074] In one optional embodiment, the collision risk index includes collision time and collision distance. The collision risk index is determined based on detection results and comprehensive data, including: Based on the test results and comprehensive data, the collision risk index is calculated using the following formula:

[0075]

[0076] It should be noted that the target distance is the distance between the target and the vehicle, the vehicle braking acceleration is the maximum deceleration of the vehicle braking system, and the reaction time is the driver's reaction time or the system response delay time.

[0077] In one optional embodiment, the safety distance and braking acceleration parameters are adjusted based on changes in vehicle braking distance and steering performance under extreme conditions. For example, on a dry road surface, the safety distance is set to 50 meters and the braking acceleration to 8 m / s²; in rainy weather, due to the slippery road surface, the safety distance is adjusted to 80 meters and the braking acceleration to 5 m / s².

[0078] Step S111: Determine the collision risk level based on the collision risk index.

[0079] In one optional embodiment, determining the collision risk level based on a collision risk index includes: When the collision time falls within the first preset time range or the collision distance falls within the first preset distance range, the current collision risk level is determined to be a Level 1 collision risk level. When the collision time falls within the second preset time range or the collision distance falls within the second preset distance range, the current collision risk level is determined to be a level two collision risk level. If the collision time falls within the third preset time range or the collision distance falls within the third preset distance range, the current collision risk level is determined to be a level three collision risk level.

[0080] In one specific embodiment, the system issues a Level 1 warning when the collision time is less than 3 seconds or the collision distance is less than the safe distance; a Level 2 warning when the collision time is less than 1.5 seconds or the collision distance is less than the emergency braking distance; and a Level 3 warning when the collision time is less than 0.5 seconds or the collision distance is less than the collision distance.

[0081] Step S112: Determine collision avoidance warning measures based on the collision risk level and output the collision avoidance warning measures.

[0082] In one optional embodiment, collision avoidance warning measures are determined based on the collision risk level, including: If the current collision risk level is Level 1, the driver will be alerted via indicator lights and a buzzer on the instrument panel; if the forward collision risk level is Level 2, partial braking will be applied to reduce the vehicle's speed; if the current collision risk level is Level 3, emergency braking and steering evasive action will be taken.

[0083] In this embodiment, multi-sensor fusion sensing comprehensively utilizes the advantages of each sensor to compensate for the shortcomings of a single sensor. Furthermore, the weights are dynamically adjusted based on the real-time reliability of the sensors, improving the accuracy of target detection and identification. This provides an accurate basis for collision risk assessment, enhances prediction accuracy, and thus improves the precision of collision warning. Moreover, the combined use of multiple sensors fully leverages the detection advantages of different sensors at different distances. LiDAR and millimeter-wave radar enable long-range detection, while ultrasonic sensors enable short-range detection. The environmental situation prediction module can adjust the sensor detection range in advance, thereby expanding the scope of collision avoidance warning and enabling timely detection of potential collision hazards at long distances and in complex road conditions. Combined with environmental prediction and the early activation of targeted adjustment strategies, the system's stability and reliability in extreme environments are improved, ensuring the effectiveness of collision avoidance warning.

[0084] Corresponding to the vehicle collision avoidance warning method provided in the embodiments of this application, the embodiments of this application also provide a vehicle collision avoidance warning device, such as... Figure 2 As shown, the vehicle collision avoidance warning device includes: The first acquisition module 201 is used to acquire the output data and multi-dimensional evaluation indicators corresponding to each of the multiple sensors; The calculation module 202 is used to calculate the real-time reliability coefficient of each sensor based on multi-dimensional evaluation indicators and using fuzzy logic algorithm. Module 203 is used to build a dynamic weight allocation model based on the real-time reliability coefficient, and output the fusion weights corresponding to each sensor through the dynamic weight allocation model; The fusion module 204 is used to fuse the output data of each sensor based on the fusion weight to obtain the fusion result; The second acquisition module 205 is used to acquire road attribute information of the road where the vehicle is currently located, external collaborative perception data, and meteorological early warning information. The integration module 206 is used to integrate the fusion results, road attribute information, external collaborative sensing data, and meteorological early warning information to obtain comprehensive data. The prediction module 207 is used to predict the environmental change trend within a preset time period based on comprehensive data and a spatiotemporal sequence prediction algorithm, and output the environmental situation level corresponding to the environmental change trend. The first determining module 208 is used to determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; The execution module 209 is used to execute a detection strategy and obtain detection results when it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level. The detection results include the target's speed information and position information. The second determination module 210 is used to determine the collision risk index based on the detection results and comprehensive data; The third determination module 211 is used to determine the collision risk level based on the collision risk index; The fourth determination module 212 is used to determine the anti-collision warning measures based on the collision risk level and output the anti-collision warning measures.

[0085] Corresponding to the vehicle collision avoidance warning method provided in the embodiments of this application, the embodiments of this application also provide an electronic device for performing the vehicle collision avoidance warning method, such as... Figure 3 As shown, the electronic device includes: a processor 301; and a memory 302 for storing a program for a vehicle collision avoidance warning method. After the device is powered on and the program for the vehicle collision avoidance warning method is run by the processor, the following steps are performed: Acquire the output data and multi-dimensional evaluation metrics for each of the multiple sensors; Based on multi-dimensional evaluation indicators, the real-time reliability coefficient of each sensor is calculated using fuzzy logic algorithm. Based on the real-time reliability coefficient, a dynamic weight allocation model is constructed, and the fusion weights corresponding to each sensor are output through the dynamic weight allocation model. The output data of each sensor are fused based on the fusion weights to obtain the fusion result. Acquire road attribute information, external collaborative perception data, and weather warning information of the road where the vehicle is currently located; By integrating and fusing the results, road attribute information, external collaborative sensing data, and meteorological early warning information, comprehensive data is obtained. Based on comprehensive data, a spatiotemporal sequence prediction algorithm is used to predict the environmental change trend within a preset time period in the future, and output the environmental situation level corresponding to the environmental change trend. Determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed to obtain the detection results, which include the target's speed information and position information. Collision risk indicators are determined based on test results and comprehensive data; The collision risk level is determined based on collision risk indicators; Based on the collision risk level, determine the collision avoidance warning measures and output the collision avoidance warning measures.

[0086] Corresponding to the vehicle collision avoidance warning method provided in the embodiments of this application, the embodiments of this application also provide a computer-readable storage medium storing a program for the vehicle collision avoidance warning method, which is executed by a processor to perform the following steps: Acquire the output data and multi-dimensional evaluation metrics for each of the multiple sensors; Based on multi-dimensional evaluation indicators, the real-time reliability coefficient of each sensor is calculated using fuzzy logic algorithm. Based on the real-time reliability coefficient, a dynamic weight allocation model is constructed, and the fusion weights corresponding to each sensor are output through the dynamic weight allocation model. The output data of each sensor are fused based on the fusion weights to obtain the fusion result. Acquire road attribute information, external collaborative perception data, and weather warning information of the road where the vehicle is currently located; By integrating and fusing the results, road attribute information, external collaborative sensing data, and meteorological early warning information, comprehensive data is obtained. Based on comprehensive data, a spatiotemporal sequence prediction algorithm is used to predict the environmental change trend within a preset time period in the future, and output the environmental situation level corresponding to the environmental change trend. Determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed to obtain the detection results, which include the target's speed information and position information. Collision risk indicators are determined based on test results and comprehensive data; The collision risk level is determined based on collision risk indicators; Based on the collision risk level, determine the collision avoidance warning measures and output the collision avoidance warning measures.

[0087] Corresponding to the vehicle collision avoidance warning method provided in the embodiments of this application, the embodiments of this application also provide a computer program containing instructions, which, when executed by a computer, cause the computer to perform the following steps: Acquire the output data and multi-dimensional evaluation metrics for each of the multiple sensors; Based on multi-dimensional evaluation indicators, the real-time reliability coefficient of each sensor is calculated using fuzzy logic algorithm. Based on the real-time reliability coefficient, a dynamic weight allocation model is constructed, and the fusion weights corresponding to each sensor are output through the dynamic weight allocation model. The output data of each sensor are fused based on the fusion weights to obtain the fusion result. Acquire road attribute information, external collaborative perception data, and weather warning information of the road where the vehicle is currently located; By integrating and fusing the results, road attribute information, external collaborative sensing data, and meteorological early warning information, comprehensive data is obtained. Based on comprehensive data, a spatiotemporal sequence prediction algorithm is used to predict the environmental change trend within a preset time period in the future, and output the environmental situation level corresponding to the environmental change trend. Determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed to obtain the detection results, which include the target's speed information and position information. Collision risk indicators are determined based on test results and comprehensive data; The collision risk level is determined based on collision risk indicators; Based on the collision risk level, determine the collision avoidance warning measures and output the collision avoidance warning measures.

[0088] It should be noted that for a detailed description of the vehicle collision avoidance warning device, electronic device, computer-readable storage medium and computer program product provided in the embodiments of this application, please refer to the relevant description of the vehicle collision avoidance warning method embodiments provided in the embodiments of this application, which will not be repeated here.

[0089] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

[0090] In a typical configuration, an electronic device includes one or more processors (Central Processing Units), input / output interfaces, network interfaces, and memory.

[0091] Memory may include non-persistent storage in computer-readable media, such as random access memory and / or non-volatile memory, like read-only memory or flash memory. Memory is an example of computer-readable media.

[0092] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable operations, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory, static random access memory, dynamic random access memory, other types of random access memory, read-only memory, electrically erasable programmable read-only memory, flash memory or other memory technologies, compact disc read-only memory, digital video disc or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0093] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, compact disc read-only memory, optical storage, etc.) containing computer-usable program code.

[0094] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A vehicle collision avoidance warning method, characterized in that, include: Acquire the output data and multi-dimensional evaluation metrics for each of the multiple sensors; Based on the multi-dimensional evaluation indicators, the real-time reliability coefficients of each sensor are calculated using a fuzzy logic algorithm. Based on the real-time reliability coefficient, a dynamic weight allocation model is constructed, and the fusion weights corresponding to each sensor are output through the dynamic weight allocation model. Based on the fusion weights, the output data of each sensor are fused to obtain the fusion result; Acquire road attribute information, external collaborative perception data, and weather warning information of the road where the vehicle is currently located; By integrating the fusion results, road attribute information, external collaborative sensing data, and meteorological early warning information, comprehensive data is obtained. Based on the comprehensive data, a spatiotemporal sequence prediction algorithm is used to predict the environmental change trend within a preset time period in the future, and the environmental situation level corresponding to the environmental change trend is output. Determine the detection strategy corresponding to the environmental situation level from the preset detection strategy library; When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed to obtain the detection result, which includes the target's speed information and position information. The collision risk index is determined based on the test results and the comprehensive data. The collision risk level is determined based on the aforementioned collision risk indicators; Based on the collision risk level, determine the collision avoidance warning measures and output the collision avoidance warning measures.

2. The vehicle collision avoidance warning method according to claim 1, characterized in that, The step of outputting the fusion weights corresponding to each sensor through the dynamic weight allocation model includes: If there is no conflict in the output data of each sensor, the fusion weight of each sensor is determined according to the proportion of the real-time reliability coefficient corresponding to each sensor, and the first fusion weight is obtained. If there is a conflict in the output data of each of the sensors, the conflicting data is determined based on the output data; the confidence level of the conflicting data is calculated based on the real-time reliability coefficient and the data consistency test result; and the fusion weight of each of the sensors is determined based on the real-time reliability coefficient and the confidence level.

3. The vehicle collision avoidance warning method according to claim 2, characterized in that, The determination of the fusion weights of each sensor based on the real-time reliability coefficient and confidence level includes: The fusion weight of a sensor whose output data is conflicting data, whose output data confidence level is higher than a preset confidence threshold, and whose real-time reliability coefficient is higher than a preset real-time reliability coefficient threshold is set as the second fusion weight; the fusion weight of a sensor whose output data is conflicting data and whose real-time reliability coefficient is not higher than a preset real-time reliability coefficient threshold, and / or whose output data is conflicting data and whose output data confidence level is not higher than a preset confidence threshold, is set as the third fusion weight; the fusion weight of a sensor whose output data is not conflicting data is determined based on the second fusion weight and the third fusion weight, wherein the second fusion weight is greater than the first fusion weight, and the third fusion weight is less than the first fusion weight.

4. The vehicle collision avoidance warning method according to claim 1, characterized in that, When it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level, the detection strategy is executed, and after obtaining the detection result, the method further includes: Determine the detection deviation between the detection results and the actual detection results; The detection strategy is adjusted based on the detection bias.

5. The vehicle collision avoidance warning method according to claim 1, characterized in that, The collision risk index includes collision time and collision distance. Determining the collision risk index based on the detection results and the comprehensive data includes: Based on the detection results and the comprehensive data, the collision risk index is calculated using the following formula: The target distance is the distance between the target vehicle and the target.

6. The vehicle collision avoidance warning method according to claim 1, characterized in that, Determining the collision risk level based on the collision risk index includes: When the collision time falls within a first preset time range or the collision distance falls within a first preset distance range, the current collision risk level is determined to be a Level 1 collision risk level. When the collision time falls within a second preset time range or the collision distance falls within a second preset distance range, the current collision risk level is determined to be a level two collision risk level. When the collision time falls within a third preset time range or the collision distance falls within a third preset distance range, the current collision risk level is determined to be a level three collision risk level.

7. The vehicle collision avoidance warning method according to claim 6, characterized in that, The method for determining collision avoidance warning measures based on the collision risk level includes: If the current collision risk level is Level 1, the driver will be alerted via indicator lights and a buzzer on the dashboard. If the forward collision risk level is Level 2, control the vehicle's braking and reduce its speed; If the current collision risk level is Level 3, determine emergency braking and steering avoidance measures.

8. A vehicle collision avoidance warning device, characterized in that, include: The first acquisition module is used to acquire the output data and multi-dimensional evaluation indicators corresponding to each of the multiple sensors; The calculation module is used to calculate the real-time reliability coefficient of each sensor based on the multi-dimensional evaluation index using a fuzzy logic algorithm. The module is used to construct a dynamic weight allocation model based on the real-time reliability coefficient, and output the fusion weights corresponding to each sensor through the dynamic weight allocation model; The fusion module is used to perform fusion processing on the output data of each sensor based on the fusion weights to obtain the fusion result; The second acquisition module is used to acquire road attribute information, external collaborative perception data, and meteorological early warning information of the road where the vehicle is currently located. The integration module is used to integrate the fusion results, road attribute information, external collaborative sensing data, and meteorological early warning information to obtain comprehensive data. The prediction module is used to predict the environmental change trend within a preset time period based on the comprehensive data using a spatiotemporal sequence prediction algorithm, and output the environmental situation level corresponding to the environmental change trend. The first determining module is used to determine the detection strategy corresponding to the environmental situation level from a preset detection strategy library; An execution module is used to execute the detection strategy and obtain detection results when it is determined that the surrounding environment of the vehicle meets the preset conditions of the environmental situation level. The detection results include the target's speed information and position information. The second determining module is used to determine the collision risk index based on the detection results and the comprehensive data; The third determining module is used to determine the collision risk level based on the collision risk index; The fourth determining module is used to determine anti-collision warning measures based on the collision risk level and output the anti-collision warning measures.

9. An electronic device, characterized in that, The system 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 vehicle collision avoidance warning method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the vehicle collision avoidance warning method according to any one of claims 1-7.

Images