Vehicle traffic risk prediction and safety decision method

Abstract

The invention provides a vehicle traffic risk prediction and safety decision method which is applied to a cloud control platform. The method comprises the following steps: performing vehicle traffic risk assessment by using a trajectory prediction method according to traffic vehicle state information, and determining the risk influence of a corresponding risk point on a main vehicle; and performing predictive active safety decision according to the determination result of the traffic risk and the safety influence threshold on the main vehicle. According to the invention, global perception, dynamic risk assessment and network connection vehicle collaborative safety decision-making of the regional traffic situation can be realized, the overall safety and reliability of the intelligent network connection traffic system are improved, and the optimization of the overall operation efficiency of regional traffic is facilitated.

Description

Technical Field

[0001] This invention relates to the field of connected vehicles and intelligent transportation technology, and in particular to a method for predicting vehicle traffic risks and making safety decisions. Background Technology

[0002] With the rapid development of intelligent transportation systems and autonomous driving technologies, vehicle traffic risk prediction and safety decision-making have become one of the core technologies for ensuring road traffic safety. Existing solutions mainly rely on a single-vehicle intelligent architecture, which involves collecting local environmental information through onboard sensors (millimeter-wave radar, visual cameras, lidar, etc.) and completing risk identification and driving decisions based on onboard computing units. However, such solutions have significant limitations: first, the perception range is constrained by physical sensors, resulting in blind spots and an inability to acquire beyond-line-of-sight traffic information; second, they lack the ability to comprehensively analyze the overall situation of regional traffic flow, making it difficult to predict the dynamic interaction and evolution trends of multiple vehicles; and third, the independent decision-making mode of a single vehicle is prone to causing conflicts in the intentions of multiple vehicles, potentially leading to system-level safety risks in complex mixed traffic scenarios. Summary of the Invention

[0003] To address the technical problems existing in the prior art, this invention provides a method for predicting vehicle traffic risks and making safe decisions. This method enables comprehensive perception of regional traffic conditions, dynamic risk assessment, and collaborative safety decisions for connected vehicles, thereby improving the overall safety and reliability of intelligent connected transportation systems and helping to optimize the overall operational efficiency of regional traffic.

[0004] This invention provides a method for predicting vehicle traffic risks and making safety decisions, applied to a cloud control platform. The method includes: assessing vehicle traffic risks using trajectory prediction methods based on traffic vehicle status information, and determining the risk impact of corresponding risk points on the main vehicle; and making predictive proactive safety decisions based on the determination results of the risk impact and the safety impact threshold on the main vehicle.

[0005] According to the present invention, a vehicle traffic risk prediction and safety decision-making method is provided. Before performing vehicle traffic risk assessment using trajectory prediction based on traffic vehicle status information, the method further includes: obtaining the status information of all traffic participants within a preset distance range ahead of the current lane based on the real-time position and speed of the main vehicle, as the traffic vehicle status information, to construct a local traffic scene; performing traffic trajectory preprocessing based on the traffic vehicle status information, and performing risk assessment based on the traffic vehicle status information after traffic trajectory preprocessing; the preprocessing method is one or more combinations of temporal filtering, frame loss completion, and abnormal target merging.

[0006] According to the present invention, a vehicle traffic risk prediction and safety decision-making method is provided, wherein the vehicle traffic risk assessment is performed based on the vehicle state information and using a trajectory prediction method, the method includes: predicting the motion trajectory of vehicles in a local traffic scene using an adaptive hybrid speed prediction method based on the vehicle state information; constructing an interaction matrix based on time-based collision warning based on the trajectory prediction results; and performing vehicle traffic risk assessment based on the interaction matrix based on time-based collision warning.

[0007] According to a vehicle traffic risk prediction and safety decision-making method provided by the present invention, the adaptive hybrid vehicle speed prediction method includes: performing acceleration anomaly detection on each vehicle in a local traffic scene; detecting that the acceleration of a rapidly decelerating vehicle is less than a preset abnormal deceleration threshold; when the rapidly decelerating vehicle is detected, using a constant acceleration model to extrapolate the speed / position of the rapidly decelerating vehicle, and using an intelligent driver model to predict the following vehicles of the rapidly decelerating vehicle; when the rapidly decelerating vehicle is not detected, using an intelligent driver model to uniformly predict the trajectory of vehicles in the scene.

[0008] According to the present invention, a method for predicting vehicle traffic risks and making safety decisions is provided. The step of constructing an interaction matrix based on time-based collision warning based on trajectory prediction results includes: calculating the collision risk of adjacent vehicle pairs in a longitudinal platoon at the prediction time based on the trajectory prediction results; and traversing multiple vehicles longitudinally arranged in the lane based on the calculated collision risk value, and constructing the interaction matrix based on time-based collision warning by combining the discretized prediction time step.

[0009] According to the present invention, a vehicle traffic risk prediction and safety decision-making method is provided, wherein the time-based collision warning interaction matrix includes an environmental vehicle collision risk assessment matrix and a host vehicle safety risk assessment matrix; the environmental vehicle collision risk assessment matrix is ​​used to assess the collision risk of different interactive vehicles during the prediction period, and the host vehicle safety risk assessment matrix is ​​used to assess the collision risk between the host vehicle and the interactive vehicles during the prediction period.

[0010] According to the present invention, a method for predicting vehicle traffic risks and making safety decisions, the step of determining the risk impact of corresponding risk points on the main vehicle includes: extracting feature values ​​based on the interaction matrix of the time-based collision warning; and determining the deterministic and indeterminate risk points on the main vehicle based on the extracted feature values ​​and a multi-dimensional feature fusion traffic risk classification and determination mechanism.

[0011] According to the vehicle traffic risk prediction and safety decision-making method provided by the present invention, the extracted feature values ​​include the minimum collision risk value of each pair of interactive vehicles during the prediction period, the time point when the vehicle first falls into the dangerous time collision warning zone, and the proportion of time during the prediction period when the calculated collision risk value is lower than a preset risk threshold.

[0012] According to the vehicle traffic risk prediction and safety decision-making method provided by the present invention, the traffic risk classification and determination mechanism based on multi-dimensional feature fusion includes: when there is an abnormal target ahead of the main vehicle's driving path and the speed of the abnormal target is less than a preset static determination threshold, the abnormal target is determined as a deterministic risk target; or, when there is an abnormal target ahead of the main vehicle's driving path and the speed of the abnormal target is not less than a preset static determination threshold, and the extracted feature value does not meet the preset feature threshold, the abnormal target is determined as a non-deterministic risk target.

[0013] According to a vehicle traffic risk prediction and safety decision-making method provided by the present invention, the step of making predictive proactive safety decisions based on the determination result of risk impact and the safety impact threshold of the main vehicle includes: constructing a risk candidate set based on the determination result of risk impact; selecting a main risk target based on the risk candidate set; and making predictive proactive safety decisions based on the main risk target and a warning and braking triggering strategy; wherein the safety impact threshold of the main vehicle is used in the warning and braking triggering strategy.

[0014] According to the present invention, a method for predicting vehicle traffic risks and making safety decisions includes a risk candidate set comprising a deterministic risk candidate set and a non-deterministic risk candidate set. The step of selecting a primary risk objective based on the risk candidate set includes: selecting the objective that minimizes urgency from the deterministic risk candidate set as the primary risk objective when the deterministic risk candidate set is not empty; or selecting the objective that minimizes urgency from the non-deterministic risk candidate set as the primary risk objective when the deterministic risk candidate set is empty and the non-deterministic risk candidate set is not empty.

[0015] According to the vehicle traffic risk prediction and safety decision-making method provided by the present invention, the warning and braking triggering strategy includes: using time collision warning threshold criteria and distance domain threshold criteria to perform collaborative safety determination, and combining braking triggering logic to determine whether to perform parking braking.

[0016] The vehicle traffic risk prediction and safety decision-making method provided by the present invention further includes: switching the main decision target when a risk point shift is triggered.

[0017] The present invention provides a vehicle traffic risk prediction and safety decision-making method, which can realize the full-domain perception of regional traffic situation, dynamic risk assessment and collaborative safety decision-making of connected vehicles, improve the overall safety and reliability of intelligent connected transportation system, and also help optimize the overall operation efficiency of regional traffic. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the principle of a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0021] Figure 3 This is one of the schematic diagrams of a specific process for a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0022] Figure 4 This is the second schematic diagram of the specific process of a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0023] Figure 5 This is a schematic diagram of the structure of a vehicle traffic risk prediction and safety decision-making system provided by the present invention.

[0024] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] Current technologies for traffic risk warning and safety decision-making typically rely on onboard sensors (such as radar and cameras) to collect information about the surrounding environment, which is then combined with onboard computing units for risk identification and decision-making. For example, some technologies analyze parameters such as the distance and relative speed between the vehicle and obstacles ahead to assess collision risk and provide braking or avoidance suggestions. However, existing methods are insufficient to meet the demands of efficient and precise collaboration based on cloud control platforms.

[0027] In the field of intelligent transportation and connected vehicle technology, with the development of technologies such as vehicle-road-cloud integration, next-generation mobile communication, and cloud computing, assessing traffic risks through cloud control platforms and their associated roadside perception devices, and providing safety warnings and decision-making for connected vehicles, has become an important direction for research and application.

[0028] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0029] This invention provides a method for predicting vehicle traffic risks and making safety decisions, applied to a cloud control platform; the method includes: 101: Based on traffic vehicle status information, use trajectory prediction methods to assess vehicle traffic risks and determine the risk impact of corresponding risk points on the main vehicle.

[0030] As a preferred embodiment, before conducting vehicle traffic risk assessment using trajectory prediction based on traffic vehicle status information, the method further includes: obtaining the status information of all traffic participants within a preset distance range ahead of the current lane based on the real-time position and speed of the main vehicle, as traffic vehicle status information, to construct a local traffic scenario; performing traffic trajectory preprocessing based on the traffic vehicle status information, and then performing risk assessment based on the traffic vehicle status information after traffic trajectory preprocessing; the preprocessing method is one or more combinations of temporal filtering, frame loss completion, and abnormal target merging.

[0031] This invention fully utilizes multi-vehicle traffic status information collected by roadside sensing devices and aggregated on a cloud control platform. This enables comprehensive perception of the overall traffic situation around vehicles, accurate prediction of multi-vehicle spatiotemporal trajectories, and improved accuracy in collision risk assessment. The cloud control platform serves as the core hub for information aggregation and processing.

[0032] In this embodiment, the cloud retrieves the status information of all traffic participants (including vehicle / obstacle IDs, positions, speeds, accelerations, heading angles, and lane affiliations) within a distance range D ahead of the current lane based on the real-time location and speed of the connected vehicle (main vehicle), in order to construct a local traffic scenario. To balance driver reaction, braking process, and necessary safety margin, this invention defines the forward query distance as: , in, Indicates the current speed of the connected vehicle; This refers to the system response time. The deceleration threshold available for connected main vehicles; This is a safety distance used to cover measurement errors, positioning uncertainties, and control execution margins. Based on the system configuration, this embodiment sets: , , .

[0033] Through the above query mechanism, the cloud forms a scene snapshot at time t in the form of "main vehicle - local traffic participants in front", providing input for trajectory processing and prediction.

[0034] This invention requires time-series filtering of trajectories based on historical data. If data frames are lost, they need to be supplemented / stitched. Furthermore, when encountering very close obstacles or abnormal targets that do not conform to physical states, they are merged. Due to limitations such as multi-source sensing errors, communication jitter, and unstable target tracking, the original trajectory sequence often exhibits problems such as position jitter, sudden velocity changes, short-term frame drops, and "shadow vehicles." This invention addresses these issues by employing the following steps: To suppress high-frequency jitter caused by positioning noise, the timing filtering method in this embodiment uses the sliding window mean filtering method. Let the window size be... For time State observations (Including location) or speed ), the filtered state The calculation is as follows: , in, The sampling index within the sliding window; This method can effectively smooth out trajectory spikes while preserving the main trend of vehicle motion.

[0035] When a target is missing for several consecutive frames but is expected to remain within the field of view, the frame-loss completion method in this embodiment uses a constant acceleration model for short-term extrapolation. To prevent excessive prediction deviation due to prolonged periods without observation, this embodiment sets a maximum completion time. .

[0036] If the current frame loss duration Then complete the equations according to the kinematic formulas: , in, To fill in the later time The predicted location, For a moment Location, For a moment speed, For a moment acceleration, The step size.

[0037] like If the target is lost, the trajectory association is terminated.

[0038] In the process of multi-source perception fusion, the same entity may be repeatedly detected as multiple neighboring targets (i.e., the "shadow car" phenomenon). To eliminate redundancy and ensure the conservatism of safety assessment, the abnormal target merging method in this embodiment is based on longitudinal distance sorting and minimum speed criterion for merging processing.

[0039] First, categorize all detected obstacle targets by their vertical position. Sort the targets. Iterate through the sorted target list and calculate adjacent targets. and Longitudinal distance difference Set the merging threshold. .

[0040] When the criterion is met When two targets are identified as duplicate detections of the same entity, a merging operation is performed. The merging logic follows the principle of "closer location, lower speed" to ensure that the more dangerous (closer) and more conservative (slower) braking distance state is retained in subsequent following risk calculations. Position merging: Retain the target position with the smaller longitudinal distance (i.e., the position closer to the main vehicle), denoted as : , Speed ​​merging: To avoid false alarms at high speeds due to sensing noise, the minimum speed of the two targets is taken as the merged speed. : , Through the above iterative traversal and elimination operations, a list of obstacles without redundancy is finally output.

[0041] This invention enables efficient fusion processing and in-depth mining of multi-source roadside perception data, accurately depicting the dynamic risk evolution patterns in complex traffic scenarios. In particular, it accurately acquires risk locations in abnormal and emergency traffic scenarios and provides effective feature information for lower-level early warning and safety decision-making modules.

[0042] As a preferred embodiment, a vehicle traffic risk assessment is performed using a trajectory prediction method based on traffic vehicle status information, including: predicting the motion trajectory of traffic vehicles in a local traffic scenario using an adaptive hybrid speed prediction method based on traffic vehicle status information; constructing an interaction matrix based on time-based collision warning based on the trajectory prediction results; and performing a vehicle traffic risk assessment based on the interaction matrix based on time-based collision warning.

[0043] As a preferred embodiment, the adaptive hybrid vehicle speed prediction method includes: performing acceleration anomaly detection for each vehicle in a local traffic scene; detecting vehicles with sudden deceleration whose acceleration is less than a preset abnormal deceleration threshold; when a vehicle with sudden deceleration is detected, using a constant acceleration model to extrapolate the speed / position of the vehicle with sudden deceleration, and using an intelligent driver model to predict the following vehicles of the vehicle with sudden deceleration; when no vehicle with sudden deceleration is detected, using an intelligent driver model to predict the trajectory of all vehicles in the scene.

[0044] In this embodiment, a speed prediction strategy of "abnormal sudden deceleration triggering + object-specific modeling" (adaptive hybrid vehicle speed prediction method) is adopted in local traffic scenarios: when a vehicle with sudden deceleration is detected in the scenario, a constant acceleration model (CA) is used to perform short-term extrapolation (speed / position extrapolation) for the vehicle; an intelligent driver model (IDM) is used to predict the following vehicles affected behind it; if no abnormal sudden deceleration vehicle is detected, IDM prediction is used for all vehicles in the scenario.

[0045] For each vehicle in the scene Based on the most recent time window Filtered acceleration sequence Perform an anomaly detection. If the condition is met... , Then the vehicle Vehicles that decelerate rapidly are designated as such and form a set. .in This is the threshold for abnormal deceleration (which is a negative value). The detection time window is set as described above in this embodiment.

[0046] When there are vehicles decelerating rapidly, for each vehicle decelerating rapidly... The CA model is used to predict the step size. Internal velocity / position extrapolation (short time domain): in This refers to the state variables after preprocessing (filtering / padding).

[0047] For vehicles that decelerate rapidly Vehicles behind IDM is used for longitudinal acceleration prediction, and the CA prediction result of the rapidly decelerating vehicle is used as the input for the preceding vehicle: Wherein, the desired spacing function is in, For the desired speed, To achieve the desired headway, For minimum spacing, For maximum acceleration, To reduce speed for comfort, This is the acceleration index. It is derived from IDM. Then, update the state of the following vehicle in discrete form (based on the sampling period). (Step size) When there are no rapidly accelerating vehicles, the scene is considered to be in a normal car-following / free-flow state, where the longitudinal behavior of vehicles is mainly constrained by the vehicle in front and driven by the desired speed. In this case, IDM is uniformly used to predict all vehicles in the scene.

[0048] The motivation behind the CA–IDM adaptive hybrid strategy is that a rapidly decelerating vehicle is in a forced dynamic unsteady state, and its short-term longitudinal motion is closer to the kinematic extrapolation dominated by the current acceleration; while the longitudinal response of the vehicle behind it is a typical interactive car-following process, and IDM can explicitly characterize the influence of distance and relative speed on acceleration decisions. Therefore, this strategy can maintain robust prediction of the trajectory of the vehicle in front under abnormal conditions, and can also provide prediction results of the braking response of the vehicle behind that conform to traffic flow patterns, making it suitable for short-term risk assessment and early warning scenarios.

[0049] This invention is based on a cloud control platform for comprehensive perception and analysis of regional traffic conditions, providing more accurate and forward-looking early warnings of potential traffic risks to vehicles, thereby improving the comprehensiveness and timeliness of risk warnings.

[0050] As a preferred embodiment, an interaction matrix based on time-based collision warning is constructed based on the trajectory prediction results, including: calculating the collision risk of adjacent vehicle pairs in the longitudinal platoon at the prediction time based on the trajectory prediction results; traversing multiple vehicles longitudinally arranged in the lane based on the calculated collision risk value, and constructing an interaction matrix based on time-based collision warning by combining the discretized prediction time step.

[0051] As a preferred embodiment, the time-based collision warning interaction matrix includes an environmental vehicle collision risk assessment matrix and a host vehicle safety risk assessment matrix; the environmental vehicle collision risk assessment matrix is ​​used to assess the collision risk of different interactive vehicles during the prediction period, and the host vehicle safety risk assessment matrix is ​​used to assess the collision risk between the host vehicle and the interactive vehicles during the prediction period.

[0052] Considering that collision risk assessment calculates TTC (Time-Based Collision Warning) values ​​based on the current motion state (such as relative speed and current distance), this transient calculation method ignores dynamic interactions and future evolution trends, making it difficult to meet the needs of proactive safety decision-making in complex scenarios. Therefore, this embodiment proposes a risk assessment framework based on forward trajectory prediction. By introducing a prediction time domain, the pairwise interaction risks of all environmental vehicles within a certain range in the lane where the driver vehicle is located are calculated, constructing a dynamic risk matrix to improve the assessment capability of future traffic environment risks. The interaction matrix based on time-based collision warning (Dynamic Risk Matrix of TTC Interaction) (TTCIM) is divided into a collision risk assessment matrix for environmental vehicles (Traffic Risk Assessment Matrix, TRAM) and a safety risk assessment matrix for the driver vehicle (Ego Vehicle Security Assessment Matrix, ESAM).

[0053] Set the prediction time domain as ,in For the current moment, To predict the duration. For any vehicle in the scene. Its predicted trajectory set is defined as , respectively representing the time of prediction The longitudinal position and velocity.

[0054] For adjacent vehicle pairs (following vehicles) in a longitudinal platoon With the car in front ), defining its time at the prediction time Collision risk This metric considers not only the current motion state but also the virtual overlap in the predicted trajectory: Note that in the formula, when When the value is 0, it indicates that the following vehicle has physically overlapped or crossed the preceding vehicle in the predicted trajectory, i.e., a virtual collision has occurred. In this case, TTC is assigned a value of 0, which represents an extremely high risk.

[0055] Based on the above calculation logic, the system analyzes the longitudinally arranged lanes... Vehicles (formed) The process iterates through adjacent vehicle pairs, combined with a discretized prediction time step. Construct a traffic vehicle collision risk assessment matrix (environmental vehicle collision risk assessment matrix). .

[0056] Each row of this matrix represents a pair of adjacent vehicles. and The matrix elements represent the interaction relationships, with each column representing a future prediction time step. This describes the risk level of a specific vehicle at a specific future time: By analyzing the matrix The decision-making system can capture potential chain collision risks in traffic flow from a global perspective, thereby achieving more forward-looking proactive safety control.

[0057] In calculation When dealing with matrices, the following calculation conditions are in place to ensure the stability of the matrix values.

[0058] Condition 1: If ( If the value is set to 1m, it is considered to be in a collision state, and the TTC value is set to 0. Condition 2: If ( If the value is set to 2m, it is considered that a vehicle has cut out of the lane in the vehicle pair, and the TTC value is set to inf. Furthermore, a safety threat matrix (safety risk assessment matrix of the main vehicle) is calculated for the vehicle and surrounding vehicles in the future prediction time domain. This matrix is ​​used to analyze the degree of safety threat to the main vehicle after the risk targets are identified, providing a basis for proactive safety decisions.

[0059] This invention combines the collaborative interaction between a cloud control platform and connected vehicles to achieve personalized and precise traffic risk warnings based on the real-time driving status and location information of specific connected vehicles.

[0060] As a preferred embodiment, determining the risk impact of corresponding risk points on the main vehicle includes: extracting feature values ​​based on the interaction matrix of time-based collision warning; and determining the deterministic and indeterminate risk points on the main vehicle based on the extracted feature values ​​and a multi-dimensional feature fusion traffic risk classification and determination mechanism.

[0061] As a preferred embodiment, the extracted feature values ​​include the minimum collision risk value for each pair of interactive vehicles during the prediction period, the time point at which the vehicle first falls into the dangerous time collision warning zone, and the proportion of time during the prediction period when the calculated collision risk value is lower than a preset risk threshold.

[0062] The time-based collision warning interaction matrix predicts and models collision scenarios for vehicles traveling in the forward direction. To facilitate the decision-making module's use of key information from the matrix and reduce numerical disturbances caused by prediction uncertainties, key features need to be extracted for robust decision-making in active safety. The following are three extracted feature indicators: Feature 1: Minimum collision risk value for each pair of interacting vehicles during the prediction period. : Within the prediction window, vehicles The minimum TTC value between these values ​​represents the most dangerous moment and the level of danger. Feature 2: The time point at which the vehicle first falls into the hazardous TTC (Time Collision Warning) zone. : If there are no moments during the entire prediction period that fall below the threshold, then it is defined as... ; Feature 3: The proportion of time during the prediction period when the TTC (calculated collision risk) is below a threshold. : In the formula, It represents the total number of sampling times within the prediction time window.

[0063] This invention can directly provide real-time safety decision support for connected vehicles, effectively connecting risk prediction with vehicle safety decision-making, and meeting the immediacy and targeted needs of connected vehicles for safety decisions in dynamic traffic environments.

[0064] As a preferred embodiment, the traffic risk classification and determination mechanism based on multi-dimensional feature fusion includes: if there is an abnormal target ahead of the main vehicle's driving path and the speed of the abnormal target is less than a preset static determination threshold, the abnormal target is determined as a deterministic risk target; or, if there is an abnormal target ahead of the main vehicle's driving path and the speed of the abnormal target is not less than a preset static determination threshold, and the extracted feature value does not meet the preset feature threshold, the abnormal target is determined as a non-deterministic risk target.

[0065] To achieve hierarchical decision-making and precise intervention in active safety systems, this embodiment proposes a traffic risk classification and determination mechanism based on multi-dimensional feature fusion. This embodiment divides traffic risks into two dimensions: deterministic risks and uncertain risks, and provides the determination criteria and state transition conditions for each.

[0066] When an abnormal target (such as a stationary vehicle, road closure, or obstacle in the lane) exists ahead of the vehicle's path, and continuing along the current lane would inevitably result in a collision threat, this scenario is defined as a deterministic risk. For the case of a "stationary / quasi-stationary target," a speed threshold-based judgment criterion is used: , in, This is the threshold for determining whether an object is stationary. Therefore, an abnormal target is considered stationary if and only if it is located in front of the main vehicle and meets the following conditions. At that time, it was determined to be a target with certain risk.

[0067] Uncertainty risk refers to the risk that occurs at the current moment. There is no definite risk ahead of the main vehicle, but by predicting the behavior of traffic participants ahead, it is possible to infer future moments. This refers to a type of risk with a high probability of vehicle-to-vehicle collisions or strong interference risks (such as sudden deceleration of the vehicle in front). This embodiment identifies and escalates this type of risk using the feature quantities of the traffic risk assessment matrix.

[0068] For high-risk dynamic interaction scenarios, the perception system continuously... Environmental vehicle accelerations within a time window are sampled to construct acceleration sequences. If a target is detected to be continuously and abnormally decelerating within the time window, and the collision risk index exceeds a preset characteristic threshold, the scenario is upgraded from a non-deterministic risk to a deterministic risk. The joint determination conditions for risk upgrade are as follows: In the formula, The abnormal deceleration threshold, The minimum collision risk threshold, This is the accident probability threshold. This is the acceleration sampling time window. This embodiment sets... .

[0069] Please refer to Figure 2 , Figure 2 This is a schematic diagram illustrating the principle of a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0070] As a preferred embodiment, it also includes: switching the primary decision objective when a risk point shift is triggered.

[0071] If abnormal deceleration has been detected (acceleration condition is met), but not simultaneously met... and If the threshold condition is met, the scenario is classified as an uncertain risk. In real traffic flow, certain and uncertain risks may alternate. The significance of uncertain risks lies in exposing potential conflicts in advance, giving the system time margin, thereby optimizing intervention timing and comfort performance under safety constraints, and achieving the proactive safety control goal of "improving decision-making space with time margin". This refers to the target switching between the primary vehicle and the deterministic risk P1 when the existence of an uncertain risk P2 is detected and the risk transfer condition is triggered.

[0072] This invention combines a cloud control platform to collaboratively manage multiple connected vehicles, and coordinates traffic risks and safety decisions of vehicles in the region from a global perspective. In complex traffic scenarios with multiple vehicle interactions, the decisions of a single vehicle are matched with the overall traffic flow demand.

[0073] 102: Based on the determination of risk impact and the threshold of safety impact on the main vehicle, make predictive proactive safety decisions.

[0074] As a preferred embodiment, predictive proactive safety decisions are made based on the determination of risk impact and the safety impact threshold on the main vehicle, including: constructing a risk candidate set based on the determination of risk impact; selecting a primary risk target based on the risk candidate set; making predictive proactive safety decisions based on the primary risk target and a warning and braking triggering strategy; and using the safety impact threshold on the main vehicle in the warning and braking triggering strategy.

[0075] To achieve stable and interpretable proactive safety decisions in traffic scenarios with multiple objectives and concurrent risks, this embodiment addresses this at each discrete control time. The system performs risk stratification and primary target selection on the set of environmental targets, and then constructs the minimum safe braking requirements for the vehicle based on its future motion predictions after determining the primary targets. This process consists of two parts: "risk target selection" and "demand deceleration calculation".

[0076] Please refer to Figure 3 , Figure 3 This is one of the schematic diagrams illustrating the specific process of a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0077] Please refer to Figure 4 , Figure 4 This is the second schematic diagram of the specific process of a vehicle traffic risk prediction and safety decision-making method provided by the present invention.

[0078] In a preferred embodiment, the risk candidate set includes a deterministic risk candidate set and a non-deterministic risk candidate set; selecting a primary risk objective based on the risk candidate set includes: if the deterministic risk candidate set is not empty, selecting the objective that minimizes urgency in the deterministic risk candidate set as the primary risk objective; or, if the deterministic risk candidate set is empty and the non-deterministic risk candidate set is not empty, selecting the objective that minimizes urgency in the non-deterministic risk candidate set as the primary risk objective.

[0079] Set time The set of observable targets in the scene is For each target The risk assessment module provides a risk level identifier. and urgency measurement This embodiment divides the risk level into two layers: Certainty risk layer: This indicates that, under the current observation and judgment conditions, the conflict between the target and the vehicle is already deterministic (e.g., an unavoidable collision trend or the mandatory intervention triggering criterion has been met), and should be the priority source for braking intervention.

[0080] Uncertainty risk layer: This indicates that the conflict is inferred from future trajectory predictions and involves a certain degree of uncertainty. In principle, it is used for early warning or as a backup source for certain risks.

[0081] Based on this, risk candidate sets are constructed respectively: .

[0082] When the deterministic risk candidate set is not empty, this embodiment adopts the "determinism priority" principle, that is, active braking is only implemented for the main target in the deterministic risk layer; when the deterministic risk candidate set is empty, the main target is then selected from the predictive risk layer for early warning decision-making. The selection of the main target is achieved by minimizing urgency. like ,but like and ,but .

[0083] This strategy ensures that when multiple objectives exist simultaneously, the system's proactive safety output always points to the objective with the "most certain risk nature and the most imminent conflict," thereby avoiding missing the opportunity for proactive safety intervention due to the failure to detect the primary objective in advance.

[0084] This invention combines a cloud control platform to make real-time safety decisions for multiple connected vehicles in a region in a global collaborative manner. Based on the real-time traffic situation, it dynamically generates optimal and collaborative driving strategies for different vehicles, thereby achieving optimal allocation of traffic resources and maximizing vehicle driving safety when dealing with complex and ever-changing traffic conditions.

[0085] As a preferred embodiment, the warning and braking triggering strategy includes: using time-based collision warning threshold criteria and distance-based threshold criteria to perform collaborative safety determination, and combining braking triggering logic to determine whether to perform parking braking.

[0086] The overall framework of the PSDM strategy is as follows: First, the strategy obtains the risk target ID and its risk type from the traffic risk assessment module and determines whether a risk target exists. If no risk target exists, the system remains inactive. If a risk target exists, further branch decisions are made based on the risk type: When the risk type is deterministic, the system enters the core decision layer, using rule-based TTC threshold criteria and distance domain threshold criteria for collaborative safety judgment, and combining this with braking trigger logic to determine whether to execute parking brake; when the risk type is non-deterministic, the system directly outputs warning information. To suppress frequent control mode switching caused by perception jitter, the core decision layer introduces a mode switching strategy with hysteresis characteristics. After the decision is made, if braking is triggered, the system calculates the recommended deceleration and target speed and executes braking control; if braking is not triggered, the system remains in the warning state.

[0087] The core decision-making process is shown in the diagram. First, the stop-triggering logic is determined: if the current vehicle speed... And the previous frame was already in braking mode ( If the system determines that the risk persists, it will maintain braking until the vehicle stops. If the above conditions are not met, the system will enter a multi-level risk assessment phase: Braking determination: Prioritize determining whether the time to collision (TTC) is less than the braking threshold. or longitudinal relative distance Is it less than the distance threshold? If any of the conditions is met, the braking mode will be triggered.

[0088] Warning determination: If the braking conditions are not met, the system will compare the following in sequence. and If the corresponding threshold is met, an early warning mode will be triggered.

[0089] Risk-free determination: If none of the conditions are met, the output is no action (None).

[0090] All judgment results are ultimately input into the mode switching logic.

[0091] The system switches between three states—"no action," "warning," and "braking"—following the logic defined in the table. To ensure control stability and prevent frequent state oscillations, the switching strategy introduces a 1-second time delay when exiting the warning (condition). ), and employs a buffer-based decision-making mechanism (condition) when exiting braking mode. This hierarchical approach with buffering mechanisms ensures both immediate response to urgent threats and a smooth transition during risk mitigation, as shown in Table 1.

[0092] Table 1 State Transition Table In the main target When the risk level is within the deterministic risk layer and the conditions for triggering / holding automatic emergency braking are met, the real-time deceleration is further calculated based on the target's future stopping point.

[0093] Assume the target is in the time domain The predicted longitudinal velocity and relative position within are respectively Define the "stop threshold". If an earliest time exists... Make , Then Defined as the predicted stopping point distance and vehicle speed of the target; if no stopping occurs within the prediction domain, the end position of the prediction domain is used as the approximate stopping point and vehicle speed: To ensure that the safety distance constraint is still met at the end of the parking area, the minimum safety margin is set as follows: (e.g., minimum safe following distance), then the effective distance that the vehicle can use to complete parking is , Let the current longitudinal velocity of the vehicle be... .like ( (If the value is a small positive number to avoid odd values), then the required deceleration is constructed based on the uniform deceleration stopping relationship: If demand slows down Not exceeding the driver's custom deceleration The system then uses a driver-defined deceleration. As a control variable; if demand decreases... Exceeding the driver's custom deceleration Then a redesigned deceleration will be adopted. .

[0094] This invention uses a cubic polynomial curve as the speed reference for driver following in the HMI display. The problem model is constructed as follows: Velocity function: Constraints: By constructing equations based on the problem, the unknown parameters of the velocity function can be solved. This system of equations has a unique solution, thus yielding a uniquely determined velocity function. .

[0095] Among the deceleration curves corresponding to different initial accelerations, there are two types: one that is more consistent with the deceleration curve of a typical driver, where the acceleration increases first and then decreases; and another that is more consistent with the deceleration curve of a sudden change in risk and a closer situation, where the initial deceleration is large and then gradually decreases, similar to the actual situation when a driver faces an emergency, where there is a rapid deceleration at first, and as the risk decreases and the driver's tension eases, the deceleration decreases and finally the vehicle comes to a stop.

[0096] This invention considers the mutual influence between multiple connected vehicles within a region. Based on the global traffic risk information provided by the cloud control platform, it dynamically generates collaborative and optimal driving strategies for different vehicles. In complex traffic scenarios, it achieves coordination between local decision-making and overall traffic flow, improving traffic efficiency and vehicle driving safety. The decision-making logic integrates the global collaborative decision-making advantages of the cloud control platform, enhancing the driving safety of connected vehicles and the regional traffic efficiency.

[0097] The vehicle traffic risk prediction and safety decision-making system provided by the present invention is described below. The vehicle traffic risk prediction and safety decision-making system described below can be referred to in correspondence with the vehicle traffic risk prediction and safety decision-making system method described above.

[0098] Please refer to Figure 5 , Figure 5 This is a schematic diagram of the structure of a vehicle traffic risk prediction and safety decision-making system provided by the present invention.

[0099] The present invention also provides a vehicle traffic risk prediction and safety decision-making system applied to a cloud control platform; the system includes: a prediction module 501, used to assess vehicle traffic risks based on traffic vehicle status information using trajectory prediction methods, and determine the risk impact of corresponding risk points on the main vehicle; and a decision-making module 502, used to make predictive proactive safety decisions based on the determination results of risk impact and the safety impact threshold on the main vehicle.

[0100] Compared with the prior art, the present invention has the following beneficial effects: Comprehensiveness and Accuracy of Traffic Risk Prediction: This invention employs a traffic risk prediction method based on a Time-to-Cross (TTC) interaction matrix. By constructing a collision risk assessment matrix for environmental vehicles and a safety risk assessment matrix for the primary vehicle, it can not only identify current risk points but also predict and analyze future risk points, forming a complete set of driving risk states. This approach can comprehensively and accurately capture the collision risk relationships between vehicles. Compared to existing single or partial risk assessment methods, it can more meticulously reflect the risk situation of multiple vehicles in the traffic system, providing a more reliable basis for subsequent safety decisions.

[0101] The targeted and timely nature of safety decision-making and early warning: For both deterministic and uncertain risks, predictive proactive safety decision-making and early warning are conducted using distance-domain decision-making and predictive TTC decision-making methods, respectively. For deterministic risks, by calculating the relative distance between the primary vehicle and the at-risk vehicle, the required deceleration, and the braking distance based on user-defined deceleration boundary values, the risk level can be accurately determined, triggering the corresponding AEB (Automatic Emergency Braking) level or warning. For uncertain risks, by predicting future risk points and combining different thresholds, corresponding control commands are triggered. This categorized decision-making approach makes safety decision-making and early warning more targeted, enabling timely responses to different types of traffic risks, effectively improving the risk response capabilities of connected vehicles, ensuring driving safety, and also contributing to optimizing the overall operational efficiency of regional traffic.

[0102] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include a processor 601, a communications interface 602, a memory 603, and a communication bus 604. The processor 601, communications interface 602, and memory 603 communicate with each other via the communication bus 604. The processor 601 can call logical instructions in the memory 603 to execute a vehicle traffic risk prediction and safety decision-making method, applied to a cloud control platform. This method includes: assessing vehicle traffic risks using trajectory prediction methods based on traffic vehicle status information, and determining the risk impact of corresponding risk points on the main vehicle; and making predictive proactive safety decisions based on the determined risk impact and the safety impact threshold on the main vehicle.

[0103] Furthermore, the logical instructions in the aforementioned memory 603 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the vehicle traffic risk prediction and safety decision-making methods provided by the above methods and apply them to a cloud control platform. The method includes: assessing vehicle traffic risks using trajectory prediction methods based on traffic vehicle status information and determining the risk impact of corresponding risk points on the main vehicle; and making predictive proactive safety decisions based on the determination results of the risk impact and the safety impact threshold on the main vehicle.

[0105] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the vehicle traffic risk prediction and safety decision-making methods provided by the above methods, and is applied to a cloud control platform; the method includes: assessing vehicle traffic risks using trajectory prediction methods based on traffic vehicle status information, and determining the risk impact of corresponding risk points on the main vehicle; and making predictive proactive safety decisions based on the determination results of the risk impact and the safety impact threshold on the main vehicle.

[0106] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0107] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting vehicle traffic risks and making safety decisions, characterized in that, Applied to a cloud control platform; the method includes: Based on traffic vehicle status information, a trajectory prediction method is used to assess vehicle traffic risks and determine the risk impact of corresponding risk points on the main vehicle. Based on the determination of risk impact and the threshold of safety impact on the main vehicle, predictive proactive safety decisions are made.

2. The vehicle traffic risk prediction and safety decision-making method according to claim 1, characterized in that, Before conducting vehicle traffic risk assessment using trajectory prediction methods based on traffic vehicle status information, the process also includes: Based on the real-time location and speed of the main vehicle, the status information of all traffic participants within a preset distance range ahead of the current lane is obtained as the traffic vehicle status information to construct a local traffic scene. Based on the traffic vehicle status information, traffic trajectory preprocessing is performed to conduct risk assessment based on the traffic vehicle status information after traffic trajectory preprocessing; the preprocessing method is one or more combinations of temporal filtering, frame loss completion, and abnormal target merging.

3. The vehicle traffic risk prediction and safety decision-making method according to claim 1, characterized in that, The method of assessing vehicle traffic risk based on traffic vehicle status information and using trajectory prediction methods includes: Based on the traffic vehicle status information, an adaptive hybrid vehicle speed prediction method is used to predict the motion trajectory of traffic vehicles in a local traffic scene. Based on the trajectory prediction results, an interaction matrix based on time-based collision warning is constructed; Based on the time-based collision warning interaction matrix, a vehicle traffic risk assessment is performed.

4. The vehicle traffic risk prediction and safety decision-making method according to claim 3, characterized in that, The adaptive hybrid vehicle speed prediction method includes: For each vehicle in a local traffic scenario, perform acceleration anomaly detection; the acceleration of a vehicle decelerating rapidly is less than a preset abnormal deceleration threshold. In the event of detecting the rapidly decelerating vehicle, a constant acceleration model is used to extrapolate the speed / position of the rapidly decelerating vehicle, and an intelligent driver model is used to predict the following vehicles of the rapidly decelerating vehicle. In the absence of a vehicle undergoing rapid deceleration, an intelligent driver model is used to uniformly predict the trajectories of vehicles within the scene.

5. The vehicle traffic risk prediction and safety decision-making method according to claim 3, characterized in that, The step of constructing an interaction matrix based on time-based collision warning according to the trajectory prediction results includes: Based on the trajectory prediction results, the collision risk of adjacent vehicle pairs in the longitudinal convoy at the predicted time is calculated; Based on the calculated collision risk value, multiple vehicles arranged longitudinally within the lane are traversed, and the time-based collision warning interaction matrix is ​​constructed by combining the discretized prediction time step.

6. The vehicle traffic risk prediction and safety decision-making method according to claim 3, characterized in that, The time-based collision warning interaction matrix includes an environmental vehicle collision risk assessment matrix and a host vehicle safety risk assessment matrix; the environmental vehicle collision risk assessment matrix is ​​used to assess the collision risk of different interactive vehicles during the prediction period, and the host vehicle safety risk assessment matrix is ​​used to assess the collision risk between the host vehicle and the interactive vehicles during the prediction period.

7. The vehicle traffic risk prediction and safety decision-making method according to claim 3, characterized in that, The determination of the risk impact of the corresponding risk points on the main vehicle includes: Feature value extraction is performed based on the interaction matrix of the time-based collision warning. Based on the extracted feature values, and using a multi-dimensional feature fusion traffic risk classification and determination mechanism, the deterministic and indeterminate risk points for the main vehicle are identified.

8. The vehicle traffic risk prediction and safety decision-making method according to claim 7, characterized in that, The extracted feature values ​​include the minimum collision risk value for each pair of interactive vehicles during the prediction period, the time point at which the vehicle first falls into the dangerous time collision warning zone, and the proportion of time during the prediction period when the calculated collision risk value is lower than the preset risk threshold.

9. The vehicle traffic risk prediction and safety decision-making method according to claim 7, characterized in that, The traffic risk classification and determination mechanism based on multi-dimensional feature fusion includes: If there is an abnormal target ahead of the main vehicle's driving path, and the speed of the abnormal target is less than the preset static judgment threshold, the abnormal target will be judged as a deterministic risk target. or, If there is an abnormal target ahead of the main vehicle's driving path, and the speed of the abnormal target is not less than the preset static judgment threshold, and the extracted feature value does not meet the preset feature threshold, the abnormal target will be judged as an uncertain risk target.

10. The vehicle traffic risk prediction and safety decision-making method according to any one of claims 1 to 9, characterized in that, The process of making predictive proactive safety decisions based on the determination of risk impact and the safety impact threshold on the main vehicle includes: Based on the determination of the risk impact, a risk candidate set is constructed; Based on the aforementioned risk candidate set, select the primary risk objective; Based on the main risk objective, predictive active safety decisions are made according to the warning and braking triggering strategy; the warning and braking triggering strategy uses the safety impact threshold of the main vehicle.

11. The vehicle traffic risk prediction and safety decision-making method according to claim 10, characterized in that, The risk candidate set includes a deterministic risk candidate set and a non-deterministic risk candidate set; The step of selecting the primary risk target based on the risk candidate set includes: If the set of deterministic risk candidates is not empty, the objective that minimizes the urgency of the deterministic risk candidate set is selected as the primary risk objective. or, If the set of deterministic risk candidates is empty and the set of nondeterministic risk candidates is not empty, the objective that minimizes urgency in the set of nondeterministic risk candidates is selected as the primary risk objective.

12. The vehicle traffic risk prediction and safety decision-making method according to claim 10, characterized in that, The warning and braking triggering strategy includes: A collaborative safety assessment is performed using time-based collision warning threshold criteria and distance-based threshold criteria, and the decision on whether to execute parking brake is determined in conjunction with the braking trigger logic.

13. The vehicle traffic risk prediction and safety decision-making method according to any one of claims 1 to 9, characterized in that, Also includes: In the event of a risk point shift, switch the primary decision objective.

Images