Method and apparatus for determining defensive driving strategy

Abstract

This application provides a method and apparatus for determining a defensive driving strategy. The method includes: collecting multi-source data corresponding to a vehicle; determining the vehicle's driving scenario based on the multi-source data; obtaining risk identification rules corresponding to the driving scenario, and extracting features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector; analyzing the state feature vector based on the risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario; and determining the corresponding defensive driving strategy when the vehicle is in a dangerous state corresponding to the driving scenario. By determining the driving scenario based on multi-source data and combining it with risk identification rules to determine whether the vehicle is currently in a dangerous state, and then executing the corresponding defensive driving strategy, the accuracy of dangerous state identification and the practicality of the defensive strategy output can be improved.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, and in particular to a method and apparatus for determining a defensive driving strategy. Background Technology

[0002] With the rapid development of automotive intelligence technology, driver assistance functions have become industry standard features. Defensive driving technology for intelligent vehicles, as one of the core technologies for improving driving safety, has become an important component of intelligent cockpit and autonomous driving technology systems. The core of defensive driving technology lies in determining defensive driving strategies.

[0003] In related technologies, existing methods for determining defensive driving strategies mainly rely on environmental perception data acquired by lidar, millimeter-wave radar, and cameras, and then use this data to determine the defensive driving strategy. However, these methods have significant shortcomings in perception technology: lidar is prone to signal scattering in adverse weather conditions such as rain, fog, and snow, resulting in a significant decrease in the quality of the collected point cloud data; cameras are highly susceptible to target recognition failures in complex lighting environments such as backlight, strong light, or low light at night; while millimeter-wave radar can detect target distance and speed information, it struggles to distinguish between different types of static objects (such as plastic bags, stones, or road shoulders), easily leading to false alarms. These shortcomings in perception technology directly result in insufficient accuracy and reliability of environmental perception data, thereby affecting the accuracy of determining defensive driving strategies. Summary of the Invention

[0004] The purpose of this application is to provide a method and apparatus for determining defensive driving strategies, thereby solving the technical problem that existing methods for determining defensive driving strategies rely entirely on perception technology, resulting in insufficient accuracy and reliability of environmental perception data, which in turn affects the accuracy of the determined defensive driving strategies. The specific technical solution is as follows:

[0005] In a first aspect of this application, a method for determining a defensive driving strategy is provided, the method comprising:

[0006] Collect multi-source data corresponding to vehicles;

[0007] Based on the multi-source data corresponding to the vehicle, the driving scenario of the vehicle is determined;

[0008] Obtain the risk identification rules corresponding to the driving scenario, and extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector;

[0009] Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario.

[0010] When the vehicle is in a dangerous state corresponding to the driving scenario, a defensive driving strategy corresponding to the vehicle is determined.

[0011] In one optional implementation, the multi-source data includes vehicle status data and map data. Based on the multi-source data corresponding to the vehicle, the driving scenario of the vehicle is determined, including: obtaining road condition information from the map data.

[0012] Obtain slope information and vehicle speed information from the vehicle status data; obtain vehicle type information of the vehicle, the vehicle type information including at least one of vehicle size parameters, vehicle mass parameters, and vehicle center of gravity height parameters;

[0013] Based on the vehicle type information, construct a vehicle model adaptation transformation matrix;

[0014] Based on the vehicle model adaptation transformation matrix, the road condition information, the slope information, and the vehicle speed information, a multi-dimensional scene feature vector is determined;

[0015] The driving scenario of the vehicle is determined based on the multi-dimensional scene feature vector.

[0016] In an optional implementation, determining the vehicle's driving scenario based on the multi-dimensional scene feature vector includes:

[0017] Obtain a preset scene classification model, which includes weight vectors and bias terms corresponding to multiple candidate driving scenarios; the candidate driving scenario is one of the following: uphill parking scenario, intersection waiting scenario, traffic light start scenario, low-speed parallel scenario, or curved road condition scenario.

[0018] The multidimensional scene feature vector is multiplied by the weight vector of each candidate driving scene in the scene classification model to obtain the scene score corresponding to each candidate driving scene.

[0019] The scene score corresponding to each candidate driving scenario is added to the corresponding bias term to obtain the matching score corresponding to each candidate driving scenario.

[0020] The driving scenario of the vehicle is determined based on the matching score corresponding to each candidate driving scenario.

[0021] In one optional implementation, the vehicle's driving scenario is an uphill parking scenario. Based on the risk identification rule, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including:

[0022] Extract the backward slip judgment conditions and uphill scene features from the risk identification rules;

[0023] Extract the slope feature, vehicle state feature, and first vehicle motion feature from the state feature vector;

[0024] If the slope feature satisfies the uphill scene feature, the vehicle state feature satisfies the preset parking conditions, and the first vehicle motion feature satisfies the backward roll determination condition, then the vehicle is determined to be in a dangerous state corresponding to the uphill parking scene.

[0025] In one optional implementation, the vehicle's driving scenario is a waiting scenario at an intersection. Based on the risk identification rule, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including:

[0026] Extract parking status features, second vehicle type features, second vehicle motion features, and second vehicle distance features from the state feature vector;

[0027] Extract intersection waiting scenario features and second vehicle hazard determination conditions from the risk identification rules;

[0028] If the parking status feature satisfies the intersection waiting scenario feature and the second vehicle type feature, the second vehicle movement feature, and the second vehicle distance feature satisfy the second vehicle danger determination condition, then the vehicle is determined to be in a dangerous state corresponding to the intersection waiting scenario.

[0029] In one optional implementation, the vehicle's driving scenario is a traffic light start-up scenario. Based on the risk identification rule, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including:

[0030] Traffic state features, first vehicle type features, third vehicle type features, and location features are extracted from the state feature vector.

[0031] Extract traffic light start-up scenario features and line-of-sight obstruction hazard judgment conditions from the aforementioned risk identification rules;

[0032] If the traffic state features satisfy the traffic light start-up scenario features, and the first vehicle type features, the third vehicle type features, and the position features satisfy the line-of-sight obstruction hazard determination conditions, then the vehicle is determined to be in a dangerous state corresponding to the traffic light start-up scenario.

[0033] In an optional implementation, the vehicle's driving scenario is a low-speed parallel scenario. Based on the risk identification rule, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including:

[0034] Vehicle speed features, third vehicle type features, position features, lane features, and parallel duration features are extracted from the state feature vector.

[0035] Extract low-speed parallel scenario features and parallel hazard determination conditions from the risk identification rules;

[0036] If the vehicle speed feature satisfies the low-speed parallel scenario feature, and the third vehicle type feature, the location feature, the lane feature, and the parallel duration feature satisfy the parallel danger determination condition, then the vehicle is determined to be in a dangerous state corresponding to the low-speed parallel scenario.

[0037] In an optional implementation, the vehicle's driving scenario is a curved road scenario. Based on the risk identification rule, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including:

[0038] Extract curve features and road condition hazard determination conditions from the risk identification rules;

[0039] Vehicle speed features, road curvature features, and obstacle features are extracted from the state feature vector;

[0040] If the vehicle speed characteristic satisfies the curve characteristic, and the road curvature characteristic and the obstacle characteristic satisfy the road condition hazard determination conditions, the vehicle is determined to be in a dangerous state corresponding to the curve road condition scenario.

[0041] In an optional implementation, the method further includes:

[0042] Obtain user feedback data on the defensive driving strategy corresponding to the vehicle;

[0043] The risk identification rules corresponding to the driving scenario are updated based on the feedback data.

[0044] In a second aspect of this application, a device for determining a defensive driving strategy is also provided, the device comprising:

[0045] The data acquisition module is used to collect multi-source data corresponding to the vehicle.

[0046] The scenario determination module is used to determine the driving scenario of the vehicle based on the multi-source data corresponding to the vehicle.

[0047] The feature extraction module is used to obtain the risk identification rules corresponding to the driving scenario, and to extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector.

[0048] The state recognition module is used to analyze the state feature vector based on the risk recognition rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario.

[0049] The defensive driving strategy determination module is used to determine the corresponding defensive driving strategy for the vehicle when the vehicle is in a dangerous state corresponding to the driving scenario.

[0050] In a third aspect of the embodiments of this application, a vehicle is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0051] Memory, used to store computer programs;

[0052] The processor, when executing a program stored in memory, implements the method for determining the defensive driving strategy described in any one of the first aspects above.

[0053] In a fourth aspect of the embodiments of this application, a storage medium is also provided, the storage medium storing instructions that, when executed on a computer, cause the computer to execute the method for determining the defensive driving strategy described in any of the first aspects above.

[0054] In a fifth aspect of the embodiments of this application, a computer program product including instructions is also provided, which, when run on a computer, causes the computer to perform the method for determining the defensive driving strategy described in any of the first aspects above.

[0055] The technical solution provided in this application involves collecting multi-source data corresponding to a vehicle; determining the vehicle's driving scenario based on the multi-source data; obtaining risk identification rules corresponding to the driving scenario; extracting features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector; analyzing the state feature vector based on the risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario; and determining the corresponding defensive driving strategy when the vehicle is in a dangerous state corresponding to the driving scenario. By determining the vehicle's driving scenario based on multi-source data and combining this with risk identification rules to analyze the state feature vector to determine whether the vehicle is currently in a dangerous state, and then executing the corresponding defensive driving strategy, the accuracy of dangerous state identification and the practicality of the defensive strategy output can be improved. This solves the technical problem that existing methods for determining defensive driving strategies rely entirely on perception technology, resulting in insufficient accuracy and reliability of environmental perception data, which in turn affects the accuracy of the defensive driving strategy determination. Attached Figure Description

[0056] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0057] 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, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0059] Figure 1 A schematic diagram illustrating the implementation process of a method for determining a defensive driving strategy provided in an embodiment of this application;

[0060] Figure 2 A schematic diagram illustrating the implementation process of another method for determining a defensive driving strategy provided in this application embodiment;

[0061] Figure 3 A schematic diagram illustrating the implementation process of a method for determining a vehicle's driving scenario, provided in an embodiment of this application;

[0062] Figure 4 A schematic diagram of the structure of a device for determining a defensive driving strategy provided in an embodiment of this application;

[0063] Figure 5 This is a structural schematic diagram of a vehicle provided in an embodiment of this application. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0065] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0066] To address the technical problem that existing technologies rely entirely on perception technology, leading to insufficient accuracy and reliability of environmental perception data and consequently affecting the accuracy of defensive driving strategy determination, this application provides a method and apparatus for determining defensive driving strategies. The method involves: collecting multi-source data corresponding to the vehicle; determining the vehicle's driving scenario based on the multi-source data; acquiring risk identification rules corresponding to the driving scenario; extracting features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector; analyzing the state feature vector based on the risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario; and determining the corresponding defensive driving strategy when the vehicle is in a dangerous state. This method, which determines the vehicle's driving scenario based on multi-source data and combines this with risk identification rules to analyze the state feature vector, thereby determining whether the vehicle is currently in a dangerous state and executing the corresponding defensive driving strategy, improves the accuracy of dangerous state identification and the practicality of the defensive strategy output.

[0067] like Figure 1 The diagram shown illustrates the implementation flow of a method for determining a defensive driving strategy according to an embodiment of this application, which may specifically include the following steps:

[0068] S101 collects multi-source data corresponding to the vehicle.

[0069] In this embodiment of the application, multi-source data corresponding to the vehicle is collected. The multi-source data includes vehicle status data, surrounding environment data, and map data. Among them, the multi-source data is used to characterize information such as the environment in which the vehicle is located and the current vehicle status.

[0070] Vehicle status data refers to information reflecting the vehicle's own operating status, used to describe the vehicle's real-time motion status and driving intentions, including but not limited to: vehicle speed (which can be obtained through wheel speed sensors or CAN bus), gear information (such as P / R / N / D gear information), steering wheel angle, vehicle acceleration (longitudinal / lateral), vehicle attitude (such as pitch angle and roll angle, which can be obtained through inertial measurement unit), slope value (which can be obtained through chassis domain controller or vehicle attitude sensor), brake pedal status, accelerator pedal status, etc.

[0071] Surrounding environment data refers to vehicle external environment perception information used to perceive the environmental situation around the vehicle. This includes, but is not limited to: raw video stream data collected by dashcams and panoramic surround view cameras; information such as distance to targets (e.g., vehicles in front, vehicles behind), relative speed, and azimuth angle obtained by sensors such as millimeter-wave radar and ultrasonic radar; real-time weather information (e.g., rain, snow, fog) and lighting conditions obtained through vehicle-to-everything (V2X) networks or API interfaces; road surface conditions ahead (e.g., slippery surfaces, epoxy flooring, rough surfaces) obtained through road preview systems or high-precision maps; and information such as traffic signs, lane lines, pedestrians, and vehicle types (e.g., large trucks, tricycles) extracted through visual recognition algorithms.

[0072] Map data refers to information from navigation systems or high-precision maps, including but not limited to: real-time traffic conditions (congestion, slow traffic, smooth traffic), road curvature, slope, lane line type (solid / dashed lines), traffic light location and status (red / green / countdown), intersection type, ramps, construction areas, speed limit signs, etc. Map data provides contextual information for understanding the macroscopic scenario of vehicle movement.

[0073] The data collection method can be periodic real-time collection or trigger-based collection based on scenario requirements. Data sources can include onboard sensors, vehicle networks, cloud services, etc., and this application embodiment does not limit these sources.

[0074] S102, based on the multi-source data corresponding to the vehicle, determine the vehicle's driving scenario.

[0075] The above driving scenarios refer to typical driving situations in which the vehicle is currently located, such as: starting on a slope / following another vehicle, waiting at a red light at an intersection, driving on a curve, following another vehicle in a congested area, or driving alongside a large vehicle on the side.

[0076] In this embodiment of the application, the driving scenario of the vehicle is determined based on the multi-source data corresponding to the vehicle. The driving scenario of the vehicle can be determined based on the vehicle status data and map data obtained in the above steps.

[0077] S103, obtain the risk identification rules corresponding to the driving scenario, and extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain the state feature vector.

[0078] The aforementioned risk identification rules are predefined dangerous state discrimination logics for specific driving scenarios. Based on the driving scenario determined in step S102, the corresponding risk identification rules can be retrieved from the preset rule base. For example, for the scenario of parking on a slope, the risk identification rule is the risk identification rule of the vehicle in front rolling backward; for the scenario of waiting at a red light, the risk identification rule is the risk identification rule of a large vehicle rapidly approaching from behind.

[0079] In this embodiment, risk identification rules corresponding to the driving scenario can be obtained, and based on the risk identification rules, features can be extracted from the vehicle state data and surrounding environment data in the multi-source data corresponding to the vehicle to obtain a state feature vector. The state feature vector is used to characterize the vehicle's driving state and surrounding environment information at the current moment, including feature parameters of multiple dimensions required for vehicle hazard identification, such as vehicle's own state features, surrounding vehicle features, road environment features, and traffic state features. The vehicle's own status characteristics include, but are not limited to, slope characteristics, speed characteristics, parking status characteristics, gear status, and steering wheel angle; surrounding vehicle characteristics include, but are not limited to, vehicle type characteristics (used to characterize whether it is a large vehicle, tricycle, etc.), vehicle motion characteristics (such as characteristics characterizing speed, acceleration, driving direction, and backward tendency), vehicle distance characteristics (characteristics corresponding to the longitudinal / lateral distance to the vehicle), and relative position characteristics (used to characterize whether it is in front of, to the side of, or behind the vehicle); road environment characteristics include, but are not limited to, road curvature characteristics, lane characteristics (characterizing lane line type and variable lane information), and obstacle characteristics (characterizing whether there are vehicles / pedestrians / fixed obstacles in front of the vehicle); traffic status characteristics include, but are not limited to, traffic light status, countdown information, navigation road conditions (such as road congestion / smoothness), weather conditions, and road surface conditions (such as slippery / epoxy floor paint).

[0080] S104. Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario.

[0081] In this embodiment, the state feature vector can be analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. A dangerous state refers to an immediate or potential risk situation that may lead to a traffic accident, identified by analyzing the state feature vector based on risk identification rules within a driving scenario. A dangerous state is relative to the current driving scenario and is scenario-dependent.

[0082] For example: in a parking scenario on a slope, the danger could be that the vehicle in front is at risk of rolling backward; in a red light scenario at an intersection, the danger could be that a large vehicle is approaching quickly from behind without slowing down; in a red light turning green scenario, the danger could be that a large truck on the side or in front is obstructing the view, and there may be pedestrians crossing in the blind spot; in a low-speed driving scenario, the danger could be that a large truck is driving alongside on the side for a long time, creating a blind spot; in a curve scenario, the danger could be that the curve ahead has a large curvature, the road conditions are unclear, and there is a blind spot.

[0083] S105, when the vehicle is in a dangerous state corresponding to the driving scenario, determine the corresponding defensive driving strategy for the vehicle.

[0084] In this embodiment, when a vehicle is in a dangerous state corresponding to a driving scenario, a corresponding defensive driving strategy is determined. The defensive driving strategy refers to specific suggestions or warnings provided to the driver for the identified dangerous state to prevent accidents. The content, form, and urgency of the defensive driving strategy are matched to the type and level of the dangerous state. For example, for a dangerous state where the vehicle in front is rolling backward, the defensive driving strategy could be a quantitative reminder to maintain a safe following distance; for a dangerous state where a large vehicle is rapidly approaching from behind, the defensive driving strategy could be a warning to pay attention to vehicles approaching from behind, supplemented by visual reinforcement; for a dangerous state where a large truck obstructs the view from the side and front, the defensive driving strategy could be a warning to start cautiously and to pay attention to pedestrians in the blind spot.

[0085] In addition, the execution of the defensive driving strategy can also follow a flexible mechanism: for example, within a power-on cycle, the same reminder for the same scenario is only triggered once (without repetition within 45 seconds) to avoid frequent interruptions; when the battery is low or memory is insufficient, priority is given to ensuring key reminders such as voice, and non-urgent image rendering or animation effects are suspended.

[0086] Based on the above description of the technical solution provided in the embodiments of this application, multi-source data corresponding to the vehicle is collected; based on the multi-source data corresponding to the vehicle, the driving scenario of the vehicle is determined; risk identification rules corresponding to the driving scenario are obtained, and feature extraction is performed on the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector; the state feature vector is analyzed based on the risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario; if the vehicle is in a dangerous state corresponding to the driving scenario, the corresponding defensive driving strategy of the vehicle is determined.

[0087] By determining the vehicle's driving scenario based on multi-source data corresponding to the vehicle and analyzing the state feature vector in conjunction with risk identification rules to determine whether the vehicle is currently in a dangerous state, and in order to execute the corresponding defensive driving strategy, the accuracy of dangerous state identification and the practicality of the defensive strategy output can be improved. This solves the technical problem that the existing methods for determining defensive driving strategies rely entirely on perception technology, resulting in insufficient accuracy and reliability of environmental perception data, which in turn affects the accuracy of the determination of defensive driving strategies.

[0088] like Figure 2 The diagram shown illustrates the implementation flow of another method for determining a defensive driving strategy provided in this application, which may specifically include the following:

[0089] S201 collects multi-source data corresponding to the vehicle, including vehicle status data, surrounding environment data, and map data.

[0090] In this embodiment of the application, this step is similar to step S101 above, and will not be described in detail here.

[0091] S202 obtains road condition information from map data.

[0092] In this embodiment, traffic information can be obtained from map data. Traffic information refers to information describing the current road traffic status, including but not limited to: real-time traffic flow status: such as smooth flow, slow traffic, congestion, severe congestion, etc., usually marked on the map with different colors (green, yellow, red, dark red); event-related information: such as traffic accidents, road construction, temporary traffic control, road closures, etc.; predictive information: such as traffic condition changes predicted over a future period based on historical data or real-time trends; and road attribute information: such as road type (expressway, urban expressway, ordinary road), number of lanes, speed limit, etc.

[0093] Specifically, traffic information requests can be sent to map service providers in real time via in-vehicle navigation modules or vehicle-to-everything (V2X) communication units. The information can be retrieved periodically (e.g., updated every 30 seconds) or triggered by location-based events (e.g., when a vehicle enters a new road segment). The retrieved information includes road segment identification, traffic condition level, average vehicle speed, and estimated travel time.

[0094] S203 obtains slope and speed information from vehicle status data.

[0095] In this embodiment, slope information and vehicle speed information can be obtained from vehicle status data. Slope information refers to the longitudinal tilt angle of the road the vehicle is currently traveling on, used to identify uphill and downhill scenarios, and to assess risks such as the vehicle ahead rolling backward. Slope information can be obtained by directly measuring the vehicle's pitch angle using chassis-domain sensors, such as vehicle attitude sensors or inertial measurement units, and then converting it into slope information. Vehicle speed information refers to the vehicle's current speed, used to determine the driving state (e.g., parked, low speed, high speed). Vehicle speed information can be obtained through wheel speed sensors, CAN bus, or Global Positioning System (GPS).

[0096] Specifically, the vehicle reads the vehicle pitch angle data reported by the electronic stability control system or chassis domain controller via the CAN bus, and obtains the current road slope value (i.e., slope information) after filtering and calibration. At the same time, it reads the wheel speed pulse signal from the CAN bus and calculates the current vehicle speed (i.e., vehicle speed information). If the vehicle is equipped with a high-precision positioning module, it can also combine GPS speed for fusion correction to improve accuracy at low speeds.

[0097] In another embodiment of this application, the design slope of the current road segment can be queried from map data as slope information.

[0098] In another embodiment of this application, the slope information obtained from the chassis domain sensor and the slope information obtained from the map data can be fused to improve accuracy.

[0099] S204, Obtain vehicle type information, which includes at least one of vehicle size parameters, vehicle mass parameters, and vehicle center of gravity height parameters.

[0100] In this embodiment, vehicle type information is obtained, which includes at least one of vehicle size parameters, vehicle mass parameters, and vehicle center of gravity height parameters. The vehicle type information is used to characterize the vehicle's physical properties and dynamic characteristics.

[0101] Vehicle dimensions, including length, width, height, and wheelbase, characterize the vehicle's external profile. Vehicles of different sizes exhibit significant differences in cornering ability, blind spot range, and driving risk on narrow roads.

[0102] Vehicle mass parameters, such as curb weight and maximum permissible gross weight, are used to characterize the vehicle's inertial characteristics. Larger vehicles (such as commercial vehicles and trucks) are at higher risk of rolling backward when parked on slopes and have longer braking distances when driving on curves, requiring more stringent risk assessment thresholds.

[0103] The vehicle center of gravity height parameter refers to the vertical height of the vehicle's center of gravity from the ground, and is used to characterize the vehicle's rollover stability. Vehicles with a higher center of gravity (such as large SUVs, MPVs, and commercial vehicles) have a significantly increased risk of rollover when cornering and are more affected by crosswinds when driving at low speeds.

[0104] S205, construct a vehicle model adaptation transformation matrix based on vehicle type information.

[0105] In this embodiment, a vehicle model adaptation transformation matrix is ​​constructed based on vehicle type information. This matrix is ​​used to perform a linear transformation on the original multi-source data (such as road condition information, slope information, and vehicle speed information).

[0106] Specifically, the vehicle model adaptation transformation matrix is ​​a 3×3 matrix. The first row of the matrix corresponds to the feature transformation of vehicle size parameters (vehicle width W, vehicle height H, wheelbase L), the second row corresponds to the feature transformation of vehicle mass parameters (curb weight m), and the third row corresponds to the transformation of vehicle center of gravity height parameters.

[0107] For the construction of the first row of elements in the matrix [a1, a2, a3], a1 = 1 + k1 × (W - W0) + k2 × (H - H0), where W is the vehicle width in the vehicle size parameters, W0 is the baseline vehicle width (e.g., 1.8m), H is the vehicle height in the vehicle size parameters, k1 is the width influence coefficient (e.g., 0.3), k1 is the height influence coefficient (e.g., 0.15); a2 can be a preset parameter, such as 0; a3 = k3 × (L - L0) / L0, where k3 is the wheelbase influence coefficient (e.g., 0.2), L is the wheelbase in the vehicle size parameters, and L is the baseline wheelbase (e.g., 2.5m).

[0108] For the construction of the second row of elements in the matrix [b1, b2, b3], b1 can be a preset parameter, such as 0; b2=1+k4×(m-m0) / m0, where k4 is the mass influence coefficient (such as 0.5), m is the curb weight in the vehicle mass parameter, and m0 is the base weight (such as 1.5t); b3=k5, where k5 is a preset slope coefficient (such as 0.25).

[0109] For the construction of the third row of elements in the matrix [c1, c2, c3], c1 can be a preset parameter, such as 0, c2=k6×(H-H0), where H is the vehicle center of gravity height parameter, H0 is the reference center of gravity height (such as 0.5m), and c3=preset center of gravity coefficient (such as 0.25).

[0110] It should be noted that when a certain parameter is missing in the vehicle type information, the matrix elements of the corresponding row will use the default value: if the vehicle size parameter is missing, the first row will be set to [1, 0, 0]; if the vehicle mass parameter is missing, the second row will be set to [0, 1, 0]; if the vehicle center of gravity parameter is missing, the third row will be set to [0, 0, 1].

[0111] S206 determines multi-dimensional scene feature vectors based on vehicle model adaptation transformation matrix, road condition information, slope information, and vehicle speed information.

[0112] In this embodiment, a multi-dimensional scene feature vector is determined based on the vehicle model adaptation transformation matrix, road condition information, slope information, and vehicle speed information.

[0113] Specifically, road condition information, slope information, and vehicle speed information can be characterized to generate an initial feature vector. The initial feature vector is then multiplied with the vehicle model adaptation transformation matrix to obtain a multidimensional scene feature vector.

[0114] To characterize traffic information, traffic conditions (such as "smooth traffic", "slow traffic", "congested", and "severe congestion") can be mapped to numerical variables. One mapping method is one-hot encoding: converting various traffic conditions into an n-dimensional binary vector (n is a positive integer, such as 4 or 8), for example, [1, 0, 0, 0] represents "smooth traffic", and [0, 1, 0, 0] represents "slow traffic". Another mapping method is ordered encoding: assigning continuous values ​​based on the degree of congestion, for example, "smooth traffic" = 0, "slow traffic" = 0.33, "congested" = 0.67, and "severe congestion" = 1.0.

[0115] For feature processing of vehicle speed information, the average vehicle speed contained in the road condition information can be directly used as the feature value. Alternatively, the vehicle speed value (e.g., 30 km / h) can be segmented or normalized.

[0116] To characterize slope information, slope values ​​(e.g., 12°) can be normalized or segmented. Specifically, normalization can be performed directly: divide the slope value in the slope information by a preset maximum slope (e.g., 30°) to obtain values ​​within the range [-1, 1]. Alternatively, segmented characterization can be used: classify slopes into five categories: large uphill (>9°), small uphill (3°~9°), flat (-3°~3°), small downhill (-9°~-3°), and large downhill (<-9°), and use one-hot encoding.

[0117] After obtaining the multi-dimensional scene feature vector, the following can be further performed: acquire the vehicle's historical behavior data; determine the driving style type based on the historical behavior data, and update the multi-dimensional scene feature vector using the driving style type. Historical behavior data refers to the vehicle's driving data within a preset historical time period (e.g., within one month), including but not limited to: historical speed, historical acceleration, historical braking frequency, steering angle, historical following distance, historical lane change frequency, historical hill driving habits, curve speed, and intersection start-stop behavior, etc., used to objectively reflect the driver's long-term driving habits. The driving style type characterizes the driver's driving habit type and can include three categories: conservative, standard, and aggressive.

[0118] Specifically, determining driving style type based on historical behavior data can involve statistical and cluster analysis of the historical behavior data. Parameters such as average vehicle speed, average following distance, average braking intensity, average lane change frequency, and average cornering speed obtained from the statistics are compared with preset style thresholds to classify drivers into conservative, standard, or aggressive types, thus obtaining the corresponding driving style type. Based on the determined driving style type, the corresponding style weight coefficient is retrieved. This style weight coefficient is then used to perform weighted calculations on the road condition feature components, slope feature components, and vehicle speed feature components in the multi-dimensional scene feature vector, resulting in an updated multi-dimensional scene feature vector.

[0119] S207 determines the vehicle's driving scenario based on multi-dimensional scene feature vectors.

[0120] In this embodiment of the application, the driving scenario of the vehicle can be determined based on multi-dimensional scene feature vectors.

[0121] For details on how to determine the vehicle's driving scenario using multi-dimensional scene feature vectors, please refer to... Figure 3 The method shown. (As illustrated) Figure 3 The diagram shown illustrates the implementation flow of a method for determining a vehicle's driving scenario according to an embodiment of this application. Specifically, it may include the following steps:

[0122] S301, Obtain a preset scene classification model. The scene classification model contains weight vectors and bias terms corresponding to multiple candidate driving scenarios. The candidate driving scenario is one of the following: uphill parking scenario, intersection waiting scenario, traffic light start scenario, low-speed parallel scenario, or curved road condition scenario.

[0123] In this embodiment, a preset scene classification model is obtained. This model includes weight vectors and bias terms corresponding to multiple candidate driving scenarios. The candidate driving scenarios are one of the following: uphill parking, intersection waiting, traffic light start, low-speed parallel driving, or curved road conditions. The scene classification model refers to a pre-trained linear classifier (such as logistic regression or Softmax regression) or multilayer perceptron, used to map multidimensional scene feature vectors to specific driving scenario categories. The parameters (weight vectors and bias terms) of the scene classification model can be obtained through offline training, using a large amount of labeled real-world driving data. The weight vector refers to a weight vector with the same dimension as the multidimensional scene feature vector for each candidate scenario. Each weight value represents the importance (positive or negative correlation) of the corresponding feature for identifying the scenario. For example, for an uphill parking scenario, the weight of the slope feature will be high, while the weight of the vehicle speed feature will tend to favor the parking state. The bias term refers to a constant term corresponding to each scenario, used to adjust the classification threshold so that the preset scene classification model can better fit the training data.

[0124] Uphill parking refers to a vehicle waiting at a standstill on an uphill section of road (such as waiting at a red light or in traffic jams).

[0125] The scenario of waiting at an intersection refers to vehicles waiting at a red light at a straight intersection (with no slope or a very small slope).

[0126] The traffic light starting scenario refers to a vehicle preparing to start when the light turns from red to green, which may face risks such as obstruction of view by vehicles on the side.

[0127] Low-speed parallel scenario refers to a situation where a vehicle is traveling at low speed and is parallel to a large vehicle on the side for an extended period of time, which is in the blind spot of the driver's field of vision.

[0128] The scenario of a curved road refers to a vehicle driving on a curve where the road conditions ahead are unclear and there are blind spots.

[0129] S302, perform dot product operation between the multidimensional scene feature vector and the weight vector of each candidate driving scene in the scene classification model to obtain the scene score corresponding to each candidate driving scene.

[0130] In this embodiment of the application, the multidimensional scene feature vector can be multiplied by the weight vector of each candidate driving scene in the scene classification model to obtain the scene score corresponding to each candidate driving scene.

[0131] For example, given a multidimensional scene feature vector X = [road condition code = 1 (congestion), gradient = 0.4 (normalized), vehicle speed = 0 (parking)], the weights for the uphill parking scene W_up = [0.1, 0.8, 0.5], and the weights for the intersection waiting scene W_wait = [0.3, 0.1, 0.6], then: raw_score_up = 1 0.1 + 0.4 0.8+0 0.5 = 0.42, raw_score_wait = 1 0.3 + 0.4 0.1+0 0.6 = 0.34. Therefore, the scenario score for the uphill parking scenario is 0.42, and the scenario score for the intersection waiting scenario is 0.34.

[0132] S303, add the scenario score corresponding to each candidate driving scenario to the corresponding bias term to obtain the matching score corresponding to each candidate driving scenario.

[0133] In this embodiment of the application, the scenario score corresponding to each candidate driving scenario is added to the corresponding bias term to obtain the matching score corresponding to each candidate driving scenario.

[0134] For example, in step S303 above, the scenario score for the uphill parking scenario is 0.42, and the scenario score for the intersection waiting scenario is 0.34. The bias term for the uphill parking scenario is -0.1, and the bias term for the intersection waiting scenario is 0.2. Therefore, the matching score for the intersection waiting scenario is 0.32, and the matching score for the uphill parking scenario is 0.54.

[0135] S304. Based on the matching scores corresponding to each candidate driving scenario, the driving scenario of the vehicle is determined.

[0136] In this embodiment, the vehicle's driving scenario is determined based on the matching score corresponding to each candidate driving scenario. A maximum value method can be used: directly selecting the candidate scenario with the highest matching score from among all candidate driving scenarios as the vehicle's driving scenario.

[0137] S208: Obtain the risk identification rules corresponding to the driving scenario, and extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain the state feature vector.

[0138] In this embodiment of the application, this step is similar to step S103 above, and will not be described in detail here.

[0139] S209, based on risk identification rules, analyzes the state feature vector to determine whether the vehicle is in a dangerous state corresponding to the driving scenario.

[0140] In this embodiment of the application, the state feature vector can be analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario.

[0141] For a vehicle driving scenario involving uphill parking, the state feature vector is analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. This can specifically include the following steps:

[0142] Step 11: Extract the backward slip judgment conditions and uphill scene features from the risk identification rules.

[0143] In this embodiment, backward slip determination conditions and uphill scenario features are extracted from the risk identification rules. The backward slip determination conditions are quantitative standards used to determine whether the vehicle ahead poses a risk of backward slip, and may include at least one of the following: vehicle distance threshold: a critical value used to determine whether the current vehicle distance is safe; speed threshold: used to determine whether the backward movement speed of the vehicle ahead exceeds a safe range; displacement threshold: used to determine whether the displacement of the vehicle ahead relative to the ground exceeds a safe range; time threshold: used to confirm the duration of the backward slip trend to avoid misjudgments caused by momentary vibrations; wheel rotation direction features: used to identify whether the wheels of the vehicle ahead are rotating in the opposite direction.

[0144] Uphill scene features refer to fixed feature parameters used to assist in identifying uphill parking scenes, which may include at least one of the following: slope threshold: the minimum slope value used to determine an uphill scene; slope range feature: used to define the slope range of the uphill scene; slope change rate threshold: used to identify uphill road sections with abrupt slope changes.

[0145] The vehicle distance threshold is calculated as follows: Braking time × Vehicle speed + User reaction time × Vehicle speed. The braking time can be set based on vehicle speed and vehicle type, such as 0.6s or 1s. The user reaction time can also be set based on vehicle speed and vehicle type, such as 0.5s. Correspondingly, in low-speed scenarios, the vehicle distance threshold can be 9m at 30km / h, 6m at 20km / h, and 3m at 10km / h. In high-speed scenarios, the vehicle distance threshold can be 69m at 100km / h, 56m at 80km / h, and 42m at 60km / h.

[0146] Step 12: Extract slope features, vehicle state features, and first vehicle motion features from the state feature vector.

[0147] In this embodiment of the application, slope features, vehicle state features, and first vehicle motion features can be extracted from the state feature vector. Among them, slope features refer to feature information extracted from the state feature vector that reflects the longitudinal inclination of the current road, and may include at least one of the following: a slope value representing the current road, a slope grade, and a slope change rate (representing the rate of change of the slope over time, used to identify slope transition sections or special road conditions).

[0148] Vehicle status characteristics refer to the characteristic information that reflects the current driving status of the vehicle and is used to determine whether the vehicle is in a parked state. These characteristics include at least one of the following: vehicle speed characteristics (such as whether the vehicle speed is 0 km / h or below the parking speed threshold), gear characteristics (such as whether it is in D or N gear), and parking duration characteristics (such as the duration of the vehicle speed being 0).

[0149] The first vehicle motion feature refers to the feature information reflecting the motion state of the preceding vehicle extracted from the state feature vector for uphill parking scenarios, in order to determine whether the preceding vehicle is in a stable parking state or has a tendency to roll backward. It may include at least one of the following: the rotation direction feature of the preceding vehicle's wheels (the rotation direction of the preceding vehicle's wheels can be identified by visual analysis algorithms), the displacement feature of the preceding vehicle relative to the ground (i.e., the displacement change of the preceding vehicle relative to the ground lane lines and road texture, used to quantify the backward rolling distance and speed), the backward movement speed feature of the preceding vehicle (referring to the instantaneous speed of the preceding vehicle moving backward), the backward rolling duration feature of the preceding vehicle (referring to the length of time that the backward rolling trend of the preceding vehicle continues to exist), the brake light status feature of the preceding vehicle (referring to whether the brake lights of the preceding vehicle are lit, used to assist in judging the braking intention of the preceding vehicle driver), and the vehicle body posture feature of the preceding vehicle (i.e., the change in the pitch angle of the preceding vehicle, used to assist in judging whether the vehicle body dynamics are caused by the uphill start operation).

[0150] Step 13: If the slope feature meets the uphill scene feature, the vehicle state feature meets the preset parking conditions, and the first vehicle motion feature meets the backward roll determination condition, then determine that the vehicle is in a dangerous state corresponding to the uphill parking scene.

[0151] In this embodiment of the application, when the slope feature meets the uphill scene feature, the vehicle state feature meets the preset parking conditions, and the first vehicle motion feature meets the backward roll determination condition, it indicates that the vehicle in front has a backward roll tendency. At this time, it is determined that the vehicle is in a dangerous state of backward roll of the vehicle in front, that is, the dangerous state corresponding to the uphill parking scene.

[0152] Specifically, the slope feature must satisfy at least one of the following uphill scene features: the slope value is greater than the slope threshold in the uphill scene features; the slope level belongs to the slope range feature in the uphill scene features; the slope change rate is greater than the slope change rate threshold in the uphill scene features, and the duration exceeds the preset duration.

[0153] The vehicle status characteristics meet the parking conditions, including at least one of the following: the vehicle speed characteristic corresponding to the vehicle status characteristics indicates that the vehicle speed is 0 km / h or lower than the preset parking speed threshold; the gear characteristic corresponding to the vehicle status characteristics indicates that the vehicle is in D gear or N gear; the parking duration characteristic corresponding to the vehicle status characteristics indicates that the vehicle parking state continues to exceed the preset parking duration threshold.

[0154] The first vehicle motion characteristic satisfies the backward slip determination condition, including at least one of the following: the front vehicle's wheels are rotating in the opposite direction; the front vehicle's backward speed is greater than the speed threshold in the backward slip determination condition; the front vehicle's displacement relative to the ground exceeds the displacement threshold in the backward slip determination condition; the duration of the front vehicle's backward slip trend exceeds the time threshold in the backward slip determination condition; and the distance between the front vehicle and the vehicle itself is less than the distance threshold in the backward slip determination condition.

[0155] For example, as an optional implementation, when the slope value is greater than the slope threshold, the vehicle speed feature indicates that the vehicle speed is 0 km / h, the gear feature indicates that the vehicle is in D gear, and the first vehicle motion feature simultaneously satisfies the following conditions: the front vehicle's wheels are rotating in the opposite direction, the front vehicle's backward movement speed exceeds the speed threshold, the front vehicle's backward rolling duration exceeds the time threshold, and the current vehicle distance is less than the vehicle distance threshold, then the vehicle is determined to be in the dangerous state of the front vehicle rolling backward corresponding to the uphill parking scenario.

[0156] For a vehicle's driving scenario of waiting at an intersection, the state feature vector is analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. This can specifically include the following steps:

[0157] Step 21: Extract parking status features, second vehicle type features, second vehicle motion features, and second vehicle distance features from the state feature vector.

[0158] In this embodiment, parking state features, second vehicle type features, second vehicle motion features, and second vehicle distance features are extracted from the state feature vector. Parking state features refer to the feature information extracted from the state feature vector that reflects whether the vehicle is currently in a stable parking waiting state, used to confirm whether the vehicle is waiting at a red light at an intersection. Parking state features may include vehicle speed features (current instantaneous speed of the vehicle, vehicle speed duration features (the continuous duration of the vehicle speed remaining at 0, used to confirm that the vehicle is stably parked rather than a short pause), gear position features (i.e., the current gearbox gear, such as P=0, R=1, N=2, D=3), brake pedal position features (reflecting whether the brake pedal is depressed, used to assist in determining the driver's parking intention), lane departure features (i.e., the alignment deviation between the vehicle body and the lane lines, used to confirm whether the vehicle is stably stopped within the lane), and traffic light position features (used to assist in confirming the waiting scenario).

[0159] The second vehicle type feature refers to the feature information extracted from the state feature vector that reflects the type of the target vehicle behind (or to the side). It is used to identify whether a large vehicle (such as a truck or bus) is approaching. This can include vehicle size features (i.e., the physical dimensions of the vehicle behind, including length, width, and height) and vehicle type labels (such as truck, bus, sedan, SUV, etc.). The second vehicle motion feature refers to the feature information extracted from the state feature vector that reflects the motion state of the vehicle behind. It is used to analyze its approaching speed and deceleration trend. This can include rear vehicle speed features (characterizing the instantaneous speed of the vehicle behind), rear vehicle acceleration features (characterizing the acceleration of the vehicle behind), rear vehicle speed change rate features (characterizing the change in speed per unit time), rear vehicle relative speed features (characterizing the relative speed between the rear vehicle and the vehicle behind), and rear vehicle lateral offset features (characterizing the lateral position of the rear vehicle relative to the vehicle's lane, determining whether they are in the same lane). The second vehicle distance feature refers to the feature information extracted from the state feature vector that reflects the distance between the vehicle behind and the vehicle behind, used to determine the degree of proximity.

[0160] Step 22: Extract the intersection waiting scenario features and the second vehicle hazard determination conditions from the risk identification rules.

[0161] In this embodiment, intersection waiting scenario features and second vehicle hazard determination conditions are extracted from risk identification rules. The intersection waiting scenario features are scenario feature parameters used to identify whether a vehicle is currently waiting at an intersection, including at least one of the following: parking speed threshold: the upper speed limit for determining whether a vehicle is in a parking state, such as a speed of 0 km / h or less than 3 km / h; parking duration threshold: the minimum duration for confirming stable parking of a vehicle, such as more than 3 seconds; gear condition feature: the gear requirement for confirming that the vehicle is in a waiting state, such as being in D or N gear; lane departure threshold: the maximum deviation of a vehicle from the lane, such as a deviation of the vehicle body from the lane line ≤ 30 cm; traffic light status feature: whether the current state is a red light waiting state.

[0162] The second vehicle hazard assessment criteria are quantitative standards used to determine whether there is a risk of rear-end collision with a large vehicle behind, including at least one of the following: Vehicle type condition: used to identify whether the target vehicle is a large vehicle, including vehicle size thresholds (e.g., length ≥ 6m, width ≥ 2.5m, height ≥ 3m) or vehicle type labels (e.g., large truck, large bus); Lane co-lane assessment condition: used to confirm whether the vehicle behind is in the same lane as the vehicle, including lateral deviation thresholds; Distance threshold: used to determine whether the distance between the vehicle behind and the vehicle has entered the danger zone, such as ≤ 50m; Speed ​​threshold: used to determine whether the speed of the vehicle behind poses a threat, such as ≥ 60km / h; Deceleration trend assessment condition: used to determine whether the vehicle behind has a clear intention to decelerate, including acceleration thresholds or speed change rate thresholds, such as a speed decrease of ≤ 5km / h / 3s is considered as no significant deceleration; Dangerous state duration threshold: used to confirm the duration of the dangerous state to avoid misjudgment caused by instantaneous fluctuations, such as duration ≥ 1.5s.

[0163] Step 23: If the parking status features satisfy the intersection waiting scenario features and the second vehicle type features, second vehicle movement features, and second vehicle distance features satisfy the second vehicle danger determination conditions, then determine that the vehicle is in a dangerous state corresponding to the intersection waiting scenario.

[0164] In this embodiment of the application, if the parking status features satisfy the intersection waiting scenario features and the second vehicle type features, the second vehicle movement features, and the second vehicle distance features satisfy the second vehicle danger determination conditions, the vehicle is determined to be in a dangerous state corresponding to the intersection waiting scenario.

[0165] Among them, the parking status feature satisfies the intersection waiting scenario feature, including at least one of the following: vehicle speed feature indicates that the vehicle speed is lower than the parking speed threshold in the intersection waiting scenario feature; vehicle speed duration feature indicates that the vehicle parking status exceeds the parking duration threshold in the intersection waiting scenario feature; gear status feature indicates that the vehicle is in the gear condition feature (such as D gear or N gear) in the intersection waiting scenario feature; lane deviation feature indicates that the deviation of the vehicle from the lane line is less than the lane deviation threshold in the intersection waiting scenario feature; traffic light status feature indicates that the current state is red light waiting. The second vehicle type characteristics, second vehicle motion characteristics, and second vehicle distance characteristics satisfy the second vehicle hazard determination conditions, including at least one of the following: the vehicle size characteristics or vehicle type label in the second vehicle type characteristics satisfy the vehicle type conditions in the second vehicle hazard determination conditions (e.g., vehicle length ≥ 6m, vehicle width ≥ 2.5m, vehicle height ≥ 3m, or labeled as a large truck or bus); the lateral deviation characteristics of the rear vehicles in the second vehicle motion characteristics satisfy the same lane determination conditions in the second vehicle hazard determination conditions (e.g., in the same lane as the vehicle); the vehicle distance characteristics in the second vehicle distance characteristics are less than the distance threshold in the second vehicle hazard determination conditions; the speed characteristics of the rear vehicles in the second vehicle motion characteristics are greater than the speed threshold in the second vehicle hazard determination conditions; the acceleration characteristics or rate of change of speed characteristics of the rear vehicles in the second vehicle motion characteristics satisfy the deceleration trend determination conditions in the second vehicle hazard determination conditions (e.g., speed decrease ≤ 5km / h / 3s, determined as no significant deceleration); and the duration of the above-mentioned dangerous state exceeds the dangerous state duration threshold in the second vehicle hazard determination conditions.

[0166] For example, as an optional implementation, when the vehicle speed feature indicates that the vehicle speed is 0 km / h, the vehicle speed duration feature indicates that the parking time is ≥3s, the gear status feature indicates that it is in D gear, the lane departure feature indicates that the lane departure amount is ≤30cm, the second vehicle type feature indicates that the vehicle behind is a large truck, the second vehicle motion feature indicates that the vehicle behind has a speed ≥60km / h and no obvious deceleration trend (speed decrease ≤5km / h / 3s), and the second vehicle distance feature indicates that the distance between vehicles is ≤50m, and the above state lasts for ≥1.5s, then it is determined that the vehicle is in a rear-end collision danger state corresponding to the intersection waiting scenario.

[0167] For a vehicle's driving scenario involving starting at a traffic light, the state feature vector is analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. This can specifically include the following steps:

[0168] Step 31: Extract traffic state features, first vehicle type features, third vehicle type features, and location features from the state feature vector.

[0169] In this embodiment, traffic state features, a first vehicle type feature, a third vehicle type feature, and location features are extracted from the state feature vector. The traffic state feature refers to the feature information extracted from the state feature vector that reflects the current traffic light status and the vehicle's starting intention, used to confirm whether the vehicle is in the critical state of a red light turning green. The first vehicle type feature refers to the feature information extracted from the state feature vector that reflects the type and attributes of obstructing vehicles (i.e., large trucks located to the left or right front of the vehicle), used to identify whether large vehicles are obstructing the view. The third vehicle type feature refers to the feature information extracted from the state feature vector that reflects the type and attributes of vehicles that may appear in the blind spot, used to identify potential crossing risks. Positional features refer to the feature information extracted from the state feature vector that reflects the relative positional relationship between the occluding vehicle and the own vehicle. These features may include occluding vehicle distance features (straight-line distance between the occluding vehicle and the own vehicle), occluding vehicle angle features (azimuth angle of the occluding vehicle relative to the driving direction of the own vehicle), occlusion range features (angular range or blind spot area occupied by the occluding vehicle in the own vehicle's field of vision, used to quantify the degree of occlusion), and lane alignment features (whether the occluding vehicle is in the same lane as the own vehicle, or in an adjacent lane but close to the own vehicle's lane line, used to confirm the effectiveness of the occlusion).

[0170] Step 32: Extract traffic light starting scene features and line-of-sight occlusion hazard judgment conditions from the risk identification rules.

[0171] In this embodiment, traffic light start-up scenario features and line-of-sight obstruction hazard judgment conditions are extracted from risk identification rules. The traffic light start-up scenario features are scenario feature parameters used to identify whether the current vehicle is in the critical state of starting from a red light turning green, including at least one of the following: vehicle speed condition feature: used to determine the upper speed limit of the vehicle in a non-starting state, such as a vehicle speed of 0 km / h; gear condition feature: used to confirm the gear requirement for the vehicle to start, such as being in D gear; accelerator pedal status feature: used to confirm that the driver has not pressed the accelerator, such as the accelerator pedal opening being 0; traffic light status feature: used to confirm that the current light is red and the countdown is less than or equal to a preset time threshold (such as ≤3s), or that the light has turned green but the vehicle has not yet started; starting intention feature: used to confirm whether the driver has a starting intention, such as a brake pedal release signal.

[0172] The criteria for determining the risk of collision due to obstructed vision are quantitative standards used to assess the potential collision risk caused by a large vehicle obstructing vision from the left or right front. These criteria include at least one of the following: Vehicle type condition: used to identify whether the obstructing vehicle is a large vehicle, including vehicle size thresholds (e.g., length ≥ 6m, width ≥ 2.5m, height ≥ 3m) or vehicle type labels (e.g., large truck, large bus); Obstruction location condition: used to confirm whether the large vehicle is located in the left or right front of the vehicle, forming a blind spot, including the azimuth angle range (e.g., within 30° of the left or right front of the vehicle) and distance range (e.g., 2-5m); Obstructing vehicle status condition: used to confirm the obstruction... Whether the vehicle is stationary or waiting, such as at a speed of 0 km / h; Starting synchronization condition: used to confirm whether the vehicle and the obstructing vehicle start at the same time, including the starting speed threshold of the obstructing vehicle (e.g., a speed ≥ 5 km / h is considered starting first) and the starting time difference threshold; Blind spot risk target condition: used to identify potential risk targets in the blind spot (e.g., pedestrians, non-motorized vehicles, vehicles approaching from the side), including target type characteristics, target position characteristics, and target movement characteristics (e.g., speed of movement, direction of movement); Dangerous state duration threshold: used to confirm the duration of the dangerous state to avoid misjudgment caused by instantaneous fluctuations, such as duration ≥ 1 second.

[0173] Step 33: If the traffic state features meet the traffic light start-up scenario features, and the first vehicle type features, the third vehicle type features, and the position features meet the line-of-sight obstruction hazard determination conditions, then determine that the vehicle is in a dangerous state corresponding to the traffic light start-up scenario.

[0174] In this embodiment of the application, if the traffic state features meet the traffic light start-up scenario features, and the first vehicle type features, the third vehicle type features, and the position features meet the line-of-sight obstruction hazard determination conditions, the vehicle is determined to be in a dangerous state corresponding to the traffic light start-up scenario.

[0175] Specifically: The traffic state characteristics satisfy the traffic light start-up scenario characteristics, including at least one of the following: vehicle speed characteristics indicating that the vehicle speed is lower than the vehicle speed condition characteristics in the traffic light start-up scenario characteristics (e.g., vehicle speed is 0km / h); gear characteristics indicating that the vehicle is in the gear condition characteristics in the traffic light start-up scenario characteristics (e.g., D gear); accelerator pedal state characteristics indicating that the vehicle is not pressing the accelerator; traffic light state characteristics indicating that the current light is red and the countdown is less than or equal to a preset time threshold, or the light has turned green but the vehicle has not yet started; brake pedal state characteristics indicating that the brake has been released and there is an intention to start.

[0176] The first vehicle type feature, the third vehicle type feature, and the location feature meet the conditions for determining the danger of obstructing vision, including at least one of the following: the vehicle size feature or vehicle type label in the first vehicle type feature meets the vehicle type conditions in the conditions for determining the danger of obstructing vision (such as vehicle length ≥ 6m, vehicle width ≥ 2.5m, vehicle height ≥ 3m, or the label is a large truck or bus).

[0177] The location features, specifically the distance and angle features of the obstructing vehicle, satisfy the obstruction location condition in the hazard determination criteria for obstruction of vision (e.g., located within 30° to the left or right front of the vehicle, at a distance of 2-5m); the obstructing vehicle state feature in the first vehicle type feature satisfies the obstructing vehicle state condition in the hazard determination criteria for obstruction of vision (e.g., vehicle speed is 0km / h, in a stationary waiting state); the obstructing vehicle starting feature in the first vehicle type feature satisfies the starting synchronization condition in the hazard determination criteria for obstruction of vision (e.g., the vehicle and the obstructing vehicle start simultaneously, or the obstructing vehicle starts first); the blind spot risk target feature in the third vehicle type feature satisfies the blind spot risk target condition in the hazard determination criteria for obstruction of vision (e.g., pedestrians, non-motorized vehicles, or vehicles approaching from the side are identified as having a risk of crossing); and the duration of the above-mentioned dangerous state exceeds the dangerous state duration threshold in the hazard determination criteria for obstruction of vision.

[0178] For example, as an optional implementation, when the traffic state characteristics indicate that the current light is red to green (red light countdown ≤ 3s or has already turned green), the vehicle speed is 0km / h, the gear is in D, the brake has been released, and the first vehicle type characteristics indicate that there is a large truck (length ≥ 6m, width ≥ 2.5m, height ≥ 3m) in front of the left, the position characteristics indicate that the large truck is within 30° to the left of the vehicle and 2-5m away, the first vehicle type characteristics also indicate that the large truck starts at the same time as the vehicle, and the third vehicle type characteristics indicate that there is a pedestrian crossing in the blind spot, and the above state lasts for ≥ 1s, then the vehicle is determined to be in a dangerous state of obstructed vision corresponding to the traffic light starting scenario.

[0179] For vehicles operating in a low-speed parallel scenario, the state feature vector is analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. This can specifically include the following steps:

[0180] Step 41: Extract vehicle speed features, third vehicle type features, position features, lane features, and parallel duration features from the state feature vector.

[0181] In this embodiment, vehicle speed features, third vehicle type features, position features, lane features, and parallel duration features are extracted from the state feature vector. Vehicle speed features refer to the feature information extracted from the state feature vector reflecting the current speed of the vehicle, used to confirm whether the vehicle is in a low-speed driving state. Third vehicle type features refer to the feature information extracted from the state feature vector reflecting the type and attributes of parallel vehicles (such as large vehicles, such as trucks and buses), used to identify whether large vehicles are traveling alongside the vehicle for an extended period. Position features refer to the feature information extracted from the state feature vector reflecting the relative positional relationship between the parallel vehicles and the vehicle. Lane features refer to the features extracted from the state feature vector reflecting the current lane and adjacent lane information, used to determine whether there are variable lanes and lane-changing conditions. Parallel duration features refer to the feature information extracted from the state feature vector reflecting the duration for which the parallel vehicles maintain a parallel state with the vehicle.

[0182] Step 42: Extract low-speed parallel scenario features and parallel hazard determination conditions from the risk identification rules.

[0183] In this embodiment, low-speed parallel scenario features and parallel hazard determination conditions are extracted from the risk identification rules. The low-speed parallel scenario features are scenario feature parameters used to identify whether the current vehicle is traveling at a low speed and is in a parallel state with vehicles on the side, including at least one of the following: speed range feature: used to determine the speed range in which the vehicle is traveling at a low speed, such as a speed of 5-30 km / h; driving stability condition: used to confirm whether the vehicle's driving state is stable, such as a steering wheel rotation angle ≤10° / s and no sudden acceleration or deceleration; gear condition feature: used to confirm that the vehicle is in a driving gear, such as in D gear.

[0184] The parallel hazard assessment criteria are quantitative standards used to determine whether there is a collision risk when large vehicles are parallel to each other for an extended period of time. These criteria include at least one of the following: Vehicle type criteria: used to identify whether the vehicle is large, including vehicle size thresholds (e.g., length ≥ 6m, width ≥ 2.5m, height ≥ 3m) or vehicle type labels (e.g., large truck, large bus); Parallel position criteria: used to confirm whether the vehicle is within the blind spot of the vehicle, including lateral distance range (e.g., 1.5–3m) and longitudinal overlap range (e.g., overlap area from front to rear of the vehicle); Parallel speed criteria: used to confirm whether the vehicle is traveling in sync with the vehicle, including speed difference thresholds (e.g., ≤ 5km / h); Parallel duration thresholds: used to confirm... The duration of parallel driving between the vehicle on the side and the vehicle itself, such as ≥3s; Lane change condition judgment conditions: used to determine whether a reversible lane exists and the feasibility of lane changing, including reversible lane existence conditions (e.g., no obstacles in adjacent lanes, no lane change prohibition signs), forward space conditions (e.g., no slow vehicles within 50m in front of adjacent lanes), and rear space conditions (e.g., no fast-approaching vehicles within 30m behind adjacent lanes); Rear approach warning conditions: used to identify whether there are fast-approaching vehicles behind adjacent lanes, including rear vehicle distance threshold (e.g., ≤30m) and rear vehicle relative speed threshold (e.g., vehicle speed ≥ the vehicle's speed 10km / h); Dangerous state duration threshold: used to confirm the duration of dangerous state, such as ≥3s.

[0185] Step 43: If the vehicle speed feature satisfies the low-speed parallel scenario feature, and the third vehicle type feature, position feature, lane feature, and parallel duration feature satisfy the parallel danger determination condition, then determine that the vehicle is in a dangerous state corresponding to the low-speed parallel scenario.

[0186] In this embodiment of the application, when the vehicle speed feature satisfies the low-speed parallel scenario feature, and the third vehicle type feature, position feature, lane feature, and parallel duration feature satisfy the parallel danger determination condition, it indicates that there is a large vehicle paralleling on the side for a long time, and there is a risk of scraping or collision. At this time, it is determined that the vehicle is in a side parallel danger state, that is, the vehicle is in a dangerous state corresponding to the low-speed parallel scenario.

[0187] Specifically: The vehicle speed feature satisfies the characteristics of a low-speed parallel scenario, including at least one of the following: the vehicle speed feature represents the speed range of the vehicle in the low-speed parallel scenario (e.g., 5-30 km / h); the driving state feature represents the vehicle's driving stability and meets the driving stability conditions in the low-speed parallel scenario (e.g., steering wheel rotation angle ≤ 10° / s, no rapid acceleration or deceleration); the gear feature represents the gear condition characteristics of the vehicle in the low-speed parallel scenario (e.g., D gear).

[0188] The third vehicle type feature, position feature, lane feature, and parallel duration feature meet the parallel hazard determination criteria, including at least one of the following: The vehicle size feature or vehicle type label in the third vehicle type feature meets the vehicle type conditions in the parallel hazard determination criteria (e.g., vehicle length ≥ 6m, vehicle width ≥ 2.5m, vehicle height ≥ 3m, or labeled as a large truck or bus); the lateral distance and longitudinal overlap range in the position feature meet the parallel position conditions in the parallel hazard determination criteria (e.g., lateral distance 1.5–3m, within the blind spot of the vehicle on the side); the speed feature of the lateral vehicles in the third vehicle type feature meets the parallel hazard determination criteria. The following conditions must be met: parallel speed (e.g., speed difference with the vehicle ≤ 5 km / h); parallel duration characteristics, satisfying the parallel duration threshold in the parallel hazard determination conditions (e.g., ≥ 3s); lane characteristics, satisfying the lane change condition determination conditions in the parallel hazard determination conditions (e.g., existence of a reversible lane, no slow vehicles within 50m ahead, no fast-moving vehicles within 30m behind); rear vehicle information in lane characteristics, satisfying the rear vehicle warning conditions in the parallel hazard determination conditions (e.g., distance of the vehicle behind ≤ 30m, and vehicle speed ≥ 10 km / h of the vehicle); and the duration of the above-mentioned dangerous state must exceed the dangerous state duration threshold in the parallel hazard determination conditions.

[0189] For example, as an optional implementation, when the vehicle speed feature indicates that the vehicle's speed is 20 km / h, the third vehicle type feature indicates that there is a large truck (length ≥ 6m, width ≥ 2.5m, height ≥ 3m) on the side, the position feature indicates that the lateral distance between the large truck and the vehicle is 2m and it is within the side blind spot, the third vehicle type feature also indicates that the speed difference between the large truck and the vehicle is 3 km / h (driving synchronously), the parallel duration feature indicates that the parallel state lasts for ≥ 4s, and the lane feature indicates that the current adjacent lane is a reversible lane, there are no slow vehicles within 50m ahead, and no fast-approaching vehicles within 30m behind, then the vehicle is determined to be in a dangerous side parallel state corresponding to a low-speed parallel scenario, triggering a lane change avoidance warning. If the lane feature also indicates that there is a fast-approaching vehicle within 30m behind in the adjacent lane (distance ≤ 30m, and speed ≥ 10 km / h of the vehicle's speed), then a supplementary warning is issued: "There is a vehicle approaching from behind, do not change lanes."

[0190] For vehicles traveling on curved roads, the state feature vector is analyzed based on risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. This can specifically include the following steps:

[0191] Step 51: Extract curve features and road condition hazard judgment conditions from the risk identification rules.

[0192] In this embodiment, curve scene features and road condition hazard judgment conditions are extracted from risk identification rules. The curve scene features are characteristic parameters used to identify whether the current vehicle is in a curve requiring attention, including at least one of the following: speed range features: used to determine whether the vehicle is in a speed range requiring attention, such as a speed of 5-30 km / h; driving stability conditions: used to confirm whether the vehicle's driving state is stable, such as a steering wheel rotation angle ≤10° / s and no sudden acceleration or deceleration; gear position conditions: used to confirm that the vehicle is in a driving gear, such as in D gear.

[0193] The road condition hazard assessment criteria are quantitative standards used to determine whether there is a risk of obstructed visibility on curved roads, including at least one of the following: Road curvature threshold: the minimum curvature value used to determine whether the current road is a curve, such as a curvature ≥ 50° (based on the vehicle's direction of travel, the angle of the road turning to the left or right); Curve warning sign condition: used to exclude duplicate warnings from existing curve warning signs, such as if there is no "curve warning" traffic sign ahead; Ahead interference elimination condition: used to confirm whether there are interfering vehicles or obstacles directly in front of the vehicle's direction of travel to avoid duplicate warnings in the following scenario, including no vehicles, pedestrians, non-motorized vehicles, or fixed obstacles (such as roadblocks or construction areas) ahead, within a distance range such as within 50m; Dangerous state duration threshold: used to confirm the duration of the dangerous state to avoid false alarms caused by the vehicle briefly passing through a small-angle curve, such as a duration ≥ 2s; Deceleration trigger cancellation condition: used to confirm that the driver has actively taken deceleration measures, such as canceling the trigger when the vehicle speed drops to ≤ 10km / h.

[0194] Step 52: Extract vehicle speed features, road curvature features, and obstacle features from the state feature vector.

[0195] In this embodiment, vehicle speed features, road curvature features, and obstacle features are extracted from the state feature vector. Vehicle speed features refer to the feature information extracted from the state feature vector reflecting the vehicle's current speed, used to confirm whether the vehicle is within the speed range requiring attention and whether there is deceleration behavior, including at least one of the following: instantaneous vehicle speed, rate of change of vehicle speed, accelerator pedal status, and brake pedal status. Road curvature features refer to the feature information extracted from the state feature vector reflecting the current degree of road curvature, used to identify whether the vehicle is on a curved road segment, including at least one of the following: road curvature value (e.g., curve angle), rate of change of curvature, curve direction (left or right curve), and curve length. Obstacle features refer to the feature information extracted from the state feature vector reflecting whether there are obstacles (e.g., other vehicles, pedestrians, non-motorized vehicles, fixed obstacles) ahead of the vehicle's direction of travel, used to exclude following scenarios and confirm obstructed visibility on curves, including at least one of the following: features of vehicles ahead (existence of a vehicle ahead, distance to the vehicle ahead, speed of the vehicle ahead), pedestrian features, non-motorized vehicle features, and features of fixed obstacles (e.g., roadblocks, construction areas, parked vehicles).

[0196] Step 53: If the vehicle speed characteristics meet the curve characteristics, and the road curvature characteristics and obstacle characteristics meet the road condition hazard determination conditions, determine that the vehicle is in a dangerous state corresponding to the curve road condition scenario.

[0197] In this embodiment of the application, when the vehicle speed characteristics meet the curve characteristics and the road curvature characteristics and obstacle characteristics meet the road condition danger determination conditions, it indicates that the vehicle is in a curved road section and the road conditions ahead are unclear. At this time, it is determined that the vehicle is in a dangerous state of unclear curved road conditions, that is, the vehicle is in a dangerous state corresponding to the curved road condition scenario.

[0198] Specifically: The vehicle speed characteristics satisfy the characteristics of a curve scene, including at least one of the following: the vehicle speed characteristics represent the speed range of the vehicle in the curve scene characteristics (e.g., 5-30km / h); the driving state characteristics represent the driving stability of the vehicle and meet the driving stability conditions in the curve scene characteristics (e.g., steering wheel rotation angle ≤10° / s, no sudden acceleration or deceleration); the gear characteristics represent the gear conditions of the vehicle in the curve scene characteristics (e.g., D gear).

[0199] The road curvature features and obstacle features meet the road condition hazard determination conditions, including at least one of the following: the road curvature value in the road curvature features is greater than the road curvature threshold in the road condition hazard determination conditions (e.g., curvature ≥ 50°); the curve warning sign feature in the road curvature features meets the curve warning sign condition in the road condition hazard determination conditions (e.g., no "curvature warning" traffic sign ahead); the obstacle feature meets the forward interference elimination condition in the road condition hazard determination conditions (e.g., no other motor vehicles, pedestrians, non-motor vehicles, or fixed obstacles within 50m directly in front of the vehicle's direction of travel); the duration of the above-mentioned dangerous state exceeds the dangerous state duration threshold in the road condition hazard determination conditions (e.g., duration ≥ 2s); if the vehicle speed feature indicates that the driver actively decelerates during the triggering period (e.g., the vehicle speed drops to ≤ 10km / h), then the triggering conditions are not met, and the reminder is canceled.

[0200] For example, as an optional implementation, when the vehicle speed feature indicates a vehicle speed of 20 km / h, the road curvature feature indicates a road curvature of 60° (left curve), the road curvature feature also indicates no "curve warning" traffic sign ahead, and the obstacle feature indicates no vehicles, pedestrians, non-motorized vehicles, or fixed obstacles within 50m directly in front of the vehicle's direction of travel, and the above conditions persist for ≥2s, then the vehicle is determined to be in a dangerous state of unclear curve conditions corresponding to the curve condition scenario, triggering a "Unclear curve conditions ahead, please drive with caution" reminder. If the driver is detected actively decelerating to below 10 km / h during the triggering period, the triggering is canceled, and no reminder is generated.

[0201] S210 determines the corresponding defensive driving strategy for the vehicle when it is in a dangerous state corresponding to the driving scenario.

[0202] In this embodiment of the application, this step is similar to step S105 above, and will not be described in detail here.

[0203] In addition, user feedback data on the vehicle's corresponding defensive driving strategies can be obtained; based on this feedback data, the risk identification rules for the driving scenarios can be updated. The feedback data is used to evaluate the effectiveness, accuracy, and user acceptance of the defensive driving strategies.

[0204] To update risk identification rules for driving scenarios based on feedback data, preset thresholds in the risk identification rules can be adjusted according to the feedback data. For example, if a large number of users report that the warning of the vehicle in front rolling backwards is too sensitive (frequent false alarms), the safe distance threshold or the backwards confirmation time can be appropriately increased; if the feedback is not timely, the threshold can be appropriately decreased or the confirmation time shortened.

[0205] Corresponding to the above method embodiments, this application also provides a device for determining a defensive driving strategy, such as... Figure 4As shown, the device may include a data acquisition module 401, a scene determination module 402, a feature extraction module 403, a state recognition module 404, and a defensive driving strategy determination module 405.

[0206] Data acquisition module 401 is used to collect multi-source data corresponding to the vehicle;

[0207] The scenario determination module 402 is used to determine the driving scenario of the vehicle based on the multi-source data corresponding to the vehicle.

[0208] The feature extraction module 403 is used to obtain the risk identification rules corresponding to the driving scenario, and to extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain the state feature vector.

[0209] The state recognition module 404 is used to analyze the state feature vector based on risk recognition rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario.

[0210] The defensive driving strategy determination module 405 is used to determine the corresponding defensive driving strategy for the vehicle when the vehicle is in a dangerous state corresponding to the driving scenario.

[0211] This application also provides a vehicle, such as... Figure 5 As shown, it includes a processor 501, a communication interface 502, a memory 503, and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other through the communication bus 504.

[0212] Memory 503 is used to store computer programs;

[0213] In one embodiment of this application, when the processor 501 executes a program stored in the memory 503, it performs the following steps:

[0214] Collect multi-source data corresponding to the vehicle; determine the vehicle's driving scenario based on the multi-source data corresponding to the vehicle; obtain risk identification rules corresponding to the driving scenario, and extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector; analyze the state feature vector based on the risk identification rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario; if the vehicle is in a dangerous state corresponding to the driving scenario, determine the corresponding defensive driving strategy for the vehicle.

[0215] The communication bus mentioned in the above vehicles can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0216] The communication interface is used for communication between the aforementioned vehicle and other devices.

[0217] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0218] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0219] In another embodiment provided in this application, a storage medium is also provided, which stores instructions that, when executed on a computer, cause the computer to execute the method for determining the defensive driving strategy described in any of the above embodiments.

[0220] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the method for determining the defensive driving strategy described in any of the above embodiments.

[0221] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted from one storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0222] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0223] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0224] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the protection scope of this application.

Claims

1. A method for determining a defensive driving strategy, characterized in that, The method includes: Collect multi-source data corresponding to vehicles; Based on the multi-source data corresponding to the vehicle, the driving scenario of the vehicle is determined, including: A preset scene classification model is obtained, which includes weight vectors and bias terms corresponding to multiple candidate driving scenarios. The candidate driving scenarios are one of the following: uphill parking scenario, intersection waiting scenario, traffic light start scenario, low-speed parallel driving scenario, or curved road condition scenario. The multi-dimensional scene feature vector is then multiplied by the weight vectors of each candidate driving scenario in the scene classification model to obtain a scene score for each candidate driving scenario. The scene score for each candidate driving scenario is then added to its corresponding bias term to obtain a matching score for each candidate driving scenario. Based on the matching scores for each candidate driving scenario, the driving scenario of the vehicle is determined. The multi-source data includes vehicle status data and map data. Based on the multi-source data corresponding to the vehicle, the driving scenario of the vehicle is determined, including: obtaining road condition information from the map data; obtaining slope information and vehicle speed information from the vehicle status data; obtaining vehicle type information, which includes at least one of vehicle size parameters, vehicle mass parameters, and vehicle center of gravity height parameters; constructing a vehicle model adaptation transformation matrix based on the vehicle type information; determining a multi-dimensional scene feature vector based on the vehicle model adaptation transformation matrix, the road condition information, the slope information, and the vehicle speed information; and determining the driving scenario of the vehicle based on the multi-dimensional scene feature vector. After obtaining the multi-dimensional scene feature vector, the method further includes: acquiring historical behavior data corresponding to the vehicle; determining the driving style type based on the historical behavior data; and updating the multi-dimensional scene feature vector based on the driving style type. Obtain the risk identification rules corresponding to the driving scenario, and extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector; Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. When the vehicle is in a dangerous state corresponding to the driving scenario, a defensive driving strategy corresponding to the vehicle is determined.

2. The method of claim 1, wherein, The vehicle's driving scenario is an uphill parking scenario. Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including: Extract the backward slip judgment conditions and uphill scene features from the risk identification rules; Extract the slope feature, vehicle state feature, and first vehicle motion feature from the state feature vector; If the slope feature satisfies the uphill scene feature, the vehicle state feature satisfies the preset parking conditions, and the first vehicle motion feature satisfies the backward roll determination condition, then the vehicle is determined to be in a dangerous state corresponding to the uphill parking scene.

3. The method of claim 1, wherein, The vehicle's driving scenario is a waiting scenario at an intersection. Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including: Extract parking status features, second vehicle type features, second vehicle motion features, and second vehicle distance features from the state feature vector; Extract intersection waiting scenario features and second vehicle hazard determination conditions from the risk identification rules; If the parking status feature satisfies the intersection waiting scenario feature and the second vehicle type feature, the second vehicle movement feature, and the second vehicle distance feature satisfy the second vehicle danger determination condition, then the vehicle is determined to be in a dangerous state corresponding to the intersection waiting scenario.

4. The method of claim 1, wherein, The vehicle's driving scenario is a traffic light start-up scenario. Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including: Traffic state features, first vehicle type features, third vehicle type features, and location features are extracted from the state feature vector; Extract traffic light start-up scenario features and line-of-sight obstruction hazard judgment conditions from the aforementioned risk identification rules; If the traffic state features satisfy the traffic light start-up scenario features, and the first vehicle type features, the third vehicle type features, and the position features satisfy the line-of-sight obstruction hazard determination conditions, then the vehicle is determined to be in a dangerous state corresponding to the traffic light start-up scenario.

5. The method of claim 1, wherein, The vehicle is traveling in a low-speed parallel scenario. Based on the risk identification rule, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including: Vehicle speed features, third vehicle type features, position features, lane features, and parallel duration features are extracted from the state feature vector. Extract low-speed parallel scenario features and parallel hazard determination conditions from the risk identification rules; If the vehicle speed feature satisfies the low-speed parallel scenario feature, and the third vehicle type feature, the location feature, the lane feature, and the parallel duration feature satisfy the parallel danger determination condition, then the vehicle is determined to be in a dangerous state corresponding to the low-speed parallel scenario.

6. The method of claim 1, wherein, The vehicle is traveling on a curved road. Based on the risk identification rules, the state feature vector is analyzed to determine whether the vehicle is in a dangerous state corresponding to the driving scenario, including: Extract curve features and road condition hazard determination conditions from the risk identification rules; Vehicle speed features, road curvature features, and obstacle features are extracted from the state feature vector; If the vehicle speed characteristic satisfies the curve characteristic, and the road curvature characteristic and the obstacle characteristic satisfy the road condition hazard determination conditions, the vehicle is determined to be in a dangerous state corresponding to the curve road condition scenario.

7. The method of claim 1, wherein, The method further includes: Obtain user feedback data on the defensive driving strategy corresponding to the vehicle; The risk identification rules corresponding to the driving scenario are updated based on the feedback data.

8. A device for determining a defense driving strategy, characterized in that The device includes: The data acquisition module is used to collect multi-source data corresponding to the vehicle. The scenario determination module is used to determine the driving scenario of the vehicle based on multi-source data corresponding to the vehicle, including: acquiring a preset scenario classification model, wherein the scenario classification model contains weight vectors and bias terms corresponding to multiple different candidate driving scenarios; the candidate driving scenarios are one of the following: uphill parking scenario, intersection waiting scenario, traffic light start scenario, low-speed parallel scenario, or curved road condition scenario; performing a dot product operation between the multi-dimensional scenario feature vector and the weight vector of each candidate driving scenario in the scenario classification model to obtain a scenario score corresponding to each candidate driving scenario; adding the scenario score corresponding to each candidate driving scenario to the corresponding bias term to obtain a matching score corresponding to each candidate driving scenario; and determining the driving scenario of the vehicle based on the matching scores corresponding to each candidate driving scenario. The multi-source data includes vehicle status data and map data. Based on the multi-source data corresponding to the vehicle, the driving scenario of the vehicle is determined, including: obtaining road condition information from the map data; obtaining slope information and vehicle speed information from the vehicle status data; obtaining vehicle type information, which includes at least one of vehicle size parameters, vehicle mass parameters, and vehicle center of gravity height parameters; constructing a vehicle model adaptation transformation matrix based on the vehicle type information; determining a multi-dimensional scene feature vector based on the vehicle model adaptation transformation matrix, the road condition information, the slope information, and the vehicle speed information; and determining the driving scenario of the vehicle based on the multi-dimensional scene feature vector. After obtaining the multi-dimensional scene feature vector, the method further includes: acquiring historical behavior data corresponding to the vehicle; determining the driving style type based on the historical behavior data; and updating the multi-dimensional scene feature vector based on the driving style type. The feature extraction module is used to obtain the risk identification rules corresponding to the driving scenario, and to extract features from the multi-source data corresponding to the vehicle based on the risk identification rules to obtain a state feature vector. The state recognition module is used to analyze the state feature vector based on the risk recognition rules to determine whether the vehicle is in a dangerous state corresponding to the driving scenario. The defensive driving strategy determination module is used to determine the defensive driving strategy corresponding to the vehicle when the vehicle is in a dangerous state corresponding to the driving scenario.

Images