A vehicle control method and system based on traffic accident prediction analysis

Abstract

The application relates to a vehicle control method and system based on traffic accident prediction analysis, which comprises the following steps: obtaining self-state information, environment perception information and traffic auxiliary information from external information sources of a vehicle; obtaining dynamic targets on a driving path of the vehicle according to the environment perception information and the traffic auxiliary information, and performing motion trend analysis on the dynamic targets to obtain a predicted motion state space and a probability distribution of each dynamic target; then, combining the self-state information of the vehicle to evaluate a collision risk index under multiple preset conflict scenarios; generating multiple candidate control schemes according to the collision risk index, and calculating comprehensive evaluation values of the candidate control schemes according to a preset optimization objective function; selecting a candidate control scheme with the minimum comprehensive evaluation value as a target control scheme, and outputting the target control scheme to an execution mechanism of the vehicle to control the vehicle to perform corresponding motion operations. The application realizes early prediction and active safety control of the vehicle on multiple-accident-scene such as ghost probe.

Description

Technical Field

[0001] This invention relates to the field of vehicle control technology, and in particular to a vehicle control method and system based on traffic accident prediction analysis. Background Technology

[0002] With the rapid development of the automotive industry, road traffic accidents have become one of the leading causes of personal injury and property damage worldwide. Statistics show that the vast majority of traffic accidents are closely related to driver perception delays, misjudgments, or improper operation. To reduce the accident rate, advanced driver assistance systems (ADAS) and autonomous driving technologies have emerged, one of their core objectives being to identify potential collision risks in advance and take corresponding control measures through perception and analysis of the surrounding traffic environment.

[0003] Existing vehicle active safety control technologies typically rely on onboard sensors to detect and track obstacles and dynamic targets around the vehicle in real time. The system then uses the target state information collected by the sensors at the current moment, combined with the vehicle's motion state, to determine whether a collision risk exists through a pre-set collision warning algorithm. When a risk is determined, an alarm is triggered or pre-set braking, steering, or other control actions are executed.

[0004] However, existing technologies have several shortcomings in practical applications. First, due to the limited sensing range of vehicle sensors and their susceptibility to obstructions, when a dynamic target is in the sensor's blind spot, its state information cannot be obtained. This can lead to insufficient reaction time when the target suddenly appears from the obstructed area, making it difficult to effectively avoid a collision, especially in complex traffic scenarios such as urban intersections and curves. Second, existing technologies typically assess collision risk based solely on the target's current state, lacking predictive analysis of the target's future movement trends. This makes it difficult to handle sudden changes in the target's motion state, resulting in delayed or misjudged risk assessments. Furthermore, existing control schemes are relatively simplistic, often pre-setting limited control strategies such as braking deceleration or emergency stopping. They fail to adequately consider the balance between safety and traffic efficiency in different conflict scenarios, potentially leading to overly conservative control actions in non-emergency situations, impacting traffic efficiency and driving comfort.

[0005] Therefore, those skilled in the art urgently need a vehicle control technology solution that can overcome the above-mentioned defects in order to improve the vehicle's ability to predict potential traffic accidents and its active safety control capabilities in complex traffic scenarios. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a vehicle control method and system based on traffic accident prediction analysis, which solves the technical problems in the prior art, such as the inability to perceive occluded targets due to sensor blind spots, the lack of prediction analysis of the movement trend of dynamic targets, and the single control strategy that fails to take into account both safety and traffic efficiency.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0010] In a first aspect, embodiments of the present invention provide a vehicle control method based on traffic accident prediction analysis, comprising:

[0011] Acquire vehicle status information, environmental perception information, and traffic assistance information from external information sources;

[0012] Based on environmental perception information and traffic assistance information, dynamic targets on the vehicle's driving path are obtained, and motion trend analysis is performed on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period.

[0013] Based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target, the collision risk indicators under various preset conflict scenarios are evaluated.

[0014] Multiple candidate control schemes are generated based on collision risk indicators, and the comprehensive evaluation value of each candidate control scheme is calculated based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints.

[0015] The candidate control scheme with the smallest comprehensive evaluation value is selected as the target control scheme and output to the vehicle's actuator to control the vehicle to perform the corresponding motion operation.

[0016] Optionally, acquiring the vehicle's own status information, environmental perception information, and traffic assistance information from external information sources includes:

[0017] Acquire the vehicle's own status information, including the vehicle's current position, heading angle, speed, acceleration, and braking system status parameters;

[0018] The vehicle collects environmental perception information around it through environmental perception sensors. The environmental perception information includes the position, speed, and direction of motion of dynamic targets on the vehicle's driving path, as well as the relative distance and relative speed between the dynamic targets and the vehicle.

[0019] Traffic assistance information is received from external sources via V2X communication. This traffic assistance information includes target information and traffic light status information broadcast by roadside units in obstructed areas, as well as perception information and avoidance event information shared by surrounding vehicles.

[0020] Optionally, obtaining dynamic targets along the vehicle's path based on environmental perception information and traffic assistance information includes:

[0021] The vehicle's environmental perception sensors collect visible targets along the driving path, perform target identification and state estimation on the visible targets, and obtain the visible dynamic targets along the vehicle's driving path.

[0022] Based on the target information of the obstructed area broadcast by the roadside unit in the traffic assistance information and / or the avoidance event information sent by the remote vehicle, combined with the relative positional relationship between the vehicle and the obstructed area and the vehicle speed, the blind spot risk coefficient of the vehicle passing through the obstructed area is obtained.

[0023] When the blind spot risk coefficient exceeds the preset risk threshold, the moving target located in the obstructed area reported by the roadside unit or remote vehicle is marked as an invisible dynamic target, and the position and motion state of the invisible dynamic target in the traffic assistance information are converted into equivalent state information in the vehicle coordinate system.

[0024] Optionally, motion trend analysis is performed on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in future time periods, including:

[0025] Obtain the motion characteristics of visible dynamic targets, including historical position sequences, velocity sequences, and heading angle sequences;

[0026] The motion features are input into the pre-trained trajectory prediction model, which outputs multiple candidate trajectories for each visible dynamic target in the future time period and uses the probability density function to obtain the probability value corresponding to each candidate trajectory.

[0027] Based on the equivalent state information of the invisible dynamic target and the location information of the occluded area, multiple possible entry points for the invisible dynamic target to enter the vehicle's driving path from the occluded area and the entry probability corresponding to each entry point are obtained.

[0028] Starting from a possible entry point, and combining the historical movement trend of the invisible dynamic target with traffic rule constraints, multiple equivalent candidate trajectories are generated and an equivalent probability value is assigned to each equivalent candidate trajectory. The equivalent probability value is then corrected based on the entry probability to obtain the optimal equivalent probability value.

[0029] The predicted motion state space is constructed using multiple candidate trajectories and / or multiple equivalent candidate trajectories for each dynamic target, and the probability distribution is constructed using the corresponding probability values ​​and / or the optimal equivalent probability values.

[0030] Optionally, based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target, collision risk indicators under various preset conflict scenarios are evaluated, including:

[0031] Based on the vehicle's own status information, obtain the braking safety distance under the current vehicle operating conditions;

[0032] Based on the predicted positions of each candidate trajectory in the motion state space at each time point and the position of the vehicle at each corresponding time point, the predicted distance between the vehicle and the dynamic target at each future time point is obtained.

[0033] The minimum collision time corresponding to each candidate trajectory is extracted based on the predicted distance. The minimum collision time is the time value corresponding to the first time when the predicted distance is less than the braking safety distance.

[0034] The weight coefficients for dynamic targets are determined based on their types, with pedestrians and two-wheeled vehicles having higher weight coefficients than four-wheeled motor vehicles.

[0035] Collision risk values ​​for various preset conflict scenarios are calculated based on minimum collision time and weighting coefficients. The collision risk value is negatively correlated with the minimum collision time and positively correlated with the weighting coefficients.

[0036] By iterating through the probability values ​​of each candidate trajectory, the collision risk values ​​of each candidate trajectory under different conflict scenarios are corrected to obtain the collision risk index of vehicles and dynamic targets under various conflict scenarios.

[0037] Optionally, multiple candidate control schemes are generated based on collision risk indicators, and the comprehensive evaluation value of each candidate control scheme is calculated based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints, including:

[0038] The collision risk level of the road segment through which the vehicle passes is determined based on the collision risk index, and multiple candidate control schemes are generated according to the collision risk level. The candidate control schemes include any one or a combination of maintaining the current motion state, braking deceleration, acceleration, steering control, and braking to a stop.

[0039] The acceleration control sequence and / or steering angle control sequence corresponding to the candidate control scheme are mapped to the predicted time period of the dynamic target according to a preset time window in order to obtain the collision probability between the vehicle and the dynamic target at each time point after the candidate control scheme is executed.

[0040] With the optimization objectives of minimizing collision probability and maximizing traffic efficiency, the comprehensive evaluation value of each candidate control scheme is solved.

[0041] Optionally, the candidate control scheme with the smallest comprehensive evaluation value is selected as the target control scheme and output to the vehicle's actuators to control the vehicle to perform corresponding motion operations, including:

[0042] Compare the comprehensive evaluation values ​​of each candidate control scheme, and select the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme;

[0043] Obtain the acceleration control sequence and / or steering angle control sequence corresponding to the target control scheme, and parse the acceleration control sequence into brake pedal opening command and / or drive pedal opening command, and parse the steering angle control sequence into steering wheel angle command;

[0044] The parsed control commands are encapsulated into control signals according to the communication protocol of the actuators, and output to at least one of the vehicle's braking actuators, drive actuators, and steering actuators via the controller local area network bus;

[0045] During the execution of the target control scheme, the real-time motion status of each dynamic target is continuously monitored. If the actual motion trajectory of any dynamic target deviates from the corresponding candidate trajectory by more than a preset deviation threshold, the current execution is interrupted and the collision risk with the dynamic target is reassessed.

[0046] Secondly, embodiments of the present invention provide a vehicle control system based on traffic accident prediction analysis, comprising:

[0047] The information acquisition module is used to acquire the vehicle's own status information, environmental perception information, and traffic assistance information from external information sources;

[0048] The dynamic target analysis module is used to obtain dynamic targets on the vehicle's driving path based on environmental perception information and traffic assistance information, and to perform motion trend analysis on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period.

[0049] The collision risk assessment module is used to assess collision risk indicators under various preset conflict scenarios based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target.

[0050] The control scheme generation module is used to generate multiple candidate control schemes based on collision risk indicators, and calculate the comprehensive evaluation value of each candidate control scheme based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints.

[0051] The control scheme execution module is used to select the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme and output it to the vehicle's actuators to control the vehicle to perform the corresponding motion operations.

[0052] Thirdly, embodiments of the present invention provide a vehicle controller, comprising:

[0053] At least one processor;

[0054] and memory that is communicatively connected to at least one processor;

[0055] The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to execute the vehicle control method based on traffic accident prediction analysis described above.

[0056] Fourthly, embodiments of the present invention provide a vehicle, comprising:

[0057] Vehicle body;

[0058] The vehicle controller described above;

[0059] The actuators, including braking actuators, drive actuators and steering actuators, are connected to the vehicle controller and are used to receive control commands output by the vehicle controller in order to control the vehicle body to perform corresponding motion operations.

[0060] (III) Beneficial Effects

[0061] The beneficial effects of this invention are as follows: The vehicle control method based on traffic accident prediction analysis, employing a multi-source information fusion perception strategy, can effectively acquire complete traffic environment information, including invisible dynamic targets within sensor blind spots, overcoming the limitations of single-sensor perception range and occlusion problems. Simultaneously, by analyzing the motion trends of dynamic targets and constructing a predictive motion state space with probability distribution, it achieves a quantitative description of the uncertainty of future behavior of traffic participants, providing a reliable data foundation for early collision risk prediction. Furthermore, by using an optimization objective function including safety and traffic efficiency constraints to comprehensively evaluate multiple candidate control schemes, it can adaptively select the optimal control strategy according to different risk levels, achieving synergistic optimization of safety and traffic efficiency. Compared to existing technologies, this invention significantly improves the vehicle's ability to predict potential traffic accidents and its active safety control capabilities in complex traffic scenarios, effectively reducing the collision accident rate while simultaneously considering traffic efficiency and driving comfort. Attached Figure Description

[0062] Figure 1 A flowchart illustrating a vehicle control method based on traffic accident prediction analysis, provided as an embodiment of the present invention;

[0063] Figure 2 This is a schematic flowchart of a traffic information acquisition method according to an embodiment of the present invention;

[0064] Figure 3 This is a schematic flowchart of a method for extracting dynamic targets on a driving path according to an embodiment of the present invention;

[0065] Figure 4 This is a schematic flowchart of a method for dynamic target operation trend analysis provided in an embodiment of the present invention;

[0066] Figure 5 This is a schematic diagram of a collision risk assessment method provided in an embodiment of the present invention;

[0067] Figure 6 This is a schematic flowchart of a control scheme evaluation method provided in an embodiment of the present invention;

[0068] Figure 7 This is a schematic flowchart illustrating the method for executing a control scheme according to an embodiment of the present invention. Detailed Implementation

[0069] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0070] refer to Figures 1 to 7 As shown in the figure, the vehicle control method based on traffic accident prediction analysis proposed in this embodiment of the invention mainly uses an on-board controller as the execution subject. The method includes: acquiring the vehicle's own state information, environmental perception information, and traffic assistance information from external information sources; acquiring dynamic targets on the vehicle's driving path based on the environmental perception information and traffic assistance information, and performing motion trend analysis on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period; evaluating collision risk indicators under various preset conflict scenarios based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target; generating multiple candidate control schemes based on the collision risk indicators, and calculating the comprehensive evaluation value of each candidate control scheme based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints; selecting the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme, and outputting it to the vehicle's actuator to control the vehicle to perform corresponding motion operations.

[0071] This embodiment, by employing a multi-source information fusion perception strategy, can effectively acquire complete traffic environment information, including invisible dynamic targets within sensor blind spots, overcoming the limitations of single-sensor perception range and occlusion problems. Simultaneously, by performing motion trend analysis on dynamic targets and constructing a predictive motion state space with probability distribution, it achieves a quantitative description of the uncertainty of future behavior of traffic participants, providing a reliable data foundation for early collision risk prediction. Furthermore, by using an optimization objective function that includes safety and traffic efficiency constraints to comprehensively evaluate multiple candidate control schemes, it can adaptively select the optimal control strategy according to different risk levels, achieving synergistic optimization of safety and traffic efficiency. Compared to existing technologies, this invention significantly improves the vehicle's ability to predict potential traffic accidents and its active safety control capabilities in complex traffic scenarios, effectively reducing the collision accident rate while simultaneously considering traffic efficiency and driving comfort.

[0072] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0073] Specifically, refer to Figure 1 As shown, the vehicle control method based on traffic accident prediction analysis proposed in this embodiment may include the following steps S100 to S500:

[0074] S100: Acquire vehicle's own status information, environmental perception information, and traffic assistance information from external information sources.

[0075] In this embodiment, reference Figure 2 As shown, step S100 may include the following sub-steps S110 to S130:

[0076] S110. Obtain the vehicle's own status information, including the vehicle's current position, heading angle, speed, acceleration, and braking system status parameters.

[0077] Furthermore, the vehicle's own status information is continuously updated according to a preset sampling period and cached in the vehicle controller's memory for subsequent steps. Among these, the status parameters of the braking system include at least the brake disc temperature, brake master cylinder pressure, and brake fluid temperature.

[0078] S120. Collect environmental perception information around the vehicle through the vehicle-mounted environmental perception module. The environmental perception information includes the position, speed, and direction of movement of dynamic targets on the vehicle's driving path, as well as the relative distance and relative speed between the dynamic targets and the vehicle.

[0079] S130. Receive traffic assistance information from external information sources via V2X communication. The traffic assistance information includes target information and traffic light status information of the obstructed area broadcast by the roadside unit, as well as perception information and avoidance event information shared by surrounding vehicles.

[0080] Furthermore, V2X communication includes, but is not limited to, V2V (vehicle-to-vehicle communication), V2I (vehicle-to-infrastructure communication), and V2P (vehicle-to-pedestrian communication). Among these, obstructed area target information refers to information about dynamic targets located in areas that may obstruct the vehicle's sensor line of sight, detected by the roadside unit through its own sensors.

[0081] S200: Based on environmental perception information and traffic assistance information, obtain dynamic targets on the vehicle's driving path, and perform motion trend analysis on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period.

[0082] In this embodiment, dynamic targets are acquired in two categories: visible dynamic targets and invisible dynamic targets. Visible dynamic targets are directly detected by the vehicle's sensors, while invisible dynamic targets are indirectly acquired through V2X information sources. The unified acquisition and fusion of these two types of targets provides a complete target data foundation for subsequent motion trend analysis and collision risk assessment. Specifically, refer to... Figure 3 As shown, the specific steps for obtaining dynamic targets along the vehicle's driving path based on environmental perception information and traffic assistance information may include the following sub-steps S211 to S213:

[0083] S211. The vehicle collects visible targets along the driving path using its environmental perception sensors, performs target identification and state estimation on the visible targets, and obtains the visible dynamic targets along the vehicle's driving path.

[0084] For example, environmental perception sensors include a front-facing camera, a forward-facing millimeter-wave radar, and a lidar. A multi-sensor fusion perception algorithm is used to perform target detection, classification, and tracking on the raw data collected by the sensors. Specifically, camera data can be used to identify the categories and pixel positions of pedestrians, two-wheeled vehicles, three-wheeled vehicles, and four-wheeled motor vehicles through a deep learning target detection network; millimeter-wave radar data obtains the relative speed and relative distance of targets through the Doppler effect; and lidar data obtains the 3D contours and precise positions of targets through a point cloud clustering algorithm. After spatiotemporally unifying the above multi-source perception results, Kalman filtering is used for target state estimation and tracking to obtain the attribute information of each visible dynamic target. The attribute information includes: target type label, geometric dimensions, position information, current speed, heading angle, and historical trajectory sequence. Finally, the identified visible dynamic targets are added to a dynamic target list, and a unique identifier label is assigned to each target.

[0085] S212. Based on the target information of the obstructed area broadcast by the roadside unit in the traffic auxiliary information and / or the avoidance event information sent by the remote vehicle, combined with the relative positional relationship between the vehicle and the obstructed area and the vehicle's driving speed, obtain the blind spot risk coefficient of the vehicle passing through the obstructed area.

[0086] Furthermore, the formula for calculating the blind spot risk coefficient η is as follows:

[0087] ;

[0088] In the formula, η0 is the basic risk coefficient of the obstructed area, which is determined by historical experience data of avoidance events generated in the obstructed area. For example, when the roadside unit reports that no target activity is taking place in the obstructed area, η0=0; when the obstructed area generates more avoidance events than expected within a set time period, η0=1; when the obstructed area generates more avoidance events than expected within a set time period, η0=0.6; α and β are adjustment coefficients, t0 is the estimated time for the vehicle to arrive at the entrance of the obstructed area; and γ is the obstructed area type factor.

[0089] S213. When the blind spot risk coefficient exceeds the preset risk threshold, the moving target reported by the roadside unit or the remote vehicle that is located in the obstruction area is marked as an invisible dynamic target, and the position and motion state of the invisible dynamic target in the traffic assistance information are converted into equivalent state information in the vehicle coordinate system.

[0090] Furthermore, the risk threshold η τ Determined based on system calibration, for example, taking η. τ =0.6. When the real-time calculated blind zone risk coefficient η > η τ At that time, the invisible dynamic target acquisition process is triggered.

[0091] In this embodiment, for visible and / or invisible dynamic targets in the dynamic target list, their future motion trends need to be analyzed and predicted. Visible dynamic targets undergo multimodal trajectory prediction based on their historical observation trajectories using a pre-trained trajectory prediction model. Invisible dynamic targets, lacking direct historical trajectories observed by the vehicle's sensors, require entry probability modeling based on their equivalent state information and the exit position of the occluded area to generate equivalent candidate trajectories. Through this differentiated yet unified prediction framework, the predicted motion state space and probability distribution of all dynamic targets in the future time period are finally obtained. (Reference) Figure 4 As shown, motion trend analysis is performed on dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in future time periods. The specific steps may include the following sub-steps S221 to S225:

[0092] S221. Obtain the motion characteristics of visible dynamic targets, including historical position sequence, velocity sequence and heading angle sequence.

[0093] Furthermore, for each visible dynamic target in the dynamic target list, historical data of the target at the most recent M sampling times are extracted from the vehicle controller's cache to form an initial motion feature vector: historical position sequence P=[(x1,y1),(x2,y2),…,(x M ,y M The velocity sequence V = [v1, v2, ..., v] M ] and the heading angle sequence θ=[θ1,vθ2,…,θ M Subsequently, higher-order statistical features of the above sequences are extracted as supplementary features, including the mean and standard deviation of velocity, the mean rate of change of heading angle, and the dynamic target type label. Finally, the initial motion feature vector and the supplementary features are normalized to form a fixed-length final feature vector F, which is used as input to the trajectory prediction model.

[0094] S222. Input the motion features into the pre-trained trajectory prediction model. The trajectory prediction model outputs multiple candidate trajectories for each visible dynamic target in the future time period, and uses the probability density function to obtain the probability value corresponding to each candidate trajectory.

[0095] For example, the trajectory prediction model employs a Long Short-Term Memory (LSTM) network, which is trained offline using a large amount of real-world traffic scenario data. The network structure includes an encoder-decoder architecture, where the encoder encodes historical motion sequences and the decoder generates predicted trajectories for the next N time points. During training, multiple loss functions are used, including mean squared error of trajectory position, trajectory smoothness loss, and diversity loss, to ensure the model can output multiple reasonable candidate trajectories.

[0096] Furthermore, the probability value corresponding to each candidate trajectory is calculated from the trajectory confidence score output by the model using the softmax function:

[0097] ;

[0098] In the formula, τ k Let s be the k-th candidate trajectory of a visible dynamic target, F be the motion feature vector of the visible dynamic target, and s be the path of the target. k Let K be the confidence score of the k-th trajectory, and K be the total number of subsequent candidate trajectories.

[0099] S223. Based on the equivalent state information of the invisible dynamic target and the location information of the occluded area, obtain multiple possible entry points for the invisible dynamic target to enter the vehicle's driving path from the occluded area and the entry probability corresponding to each entry point.

[0100] S224. Starting from a possible entry point, and combining the historical movement trend of the invisible dynamic target and traffic rule constraints, generate multiple equivalent candidate trajectories and assign an equivalent probability value to each equivalent candidate trajectory. Then, adjust the equivalent probability value according to the entry probability to obtain the optimal equivalent probability value.

[0101] Furthermore, to reflect the historical movement trend of invisible dynamic targets, the consistency of the travel time from the target's current equivalent position within the occlusion area to the entry point is considered: if the travel time calculated from the distance from the equivalent position to the entry point and the equivalent velocity is within a reasonable range, the equivalent probability value is positively corrected; otherwise, a negative correction is performed. The correction formula is as follows:

[0102] ;

[0103] In the formula, The optimal equivalent probability value τ of the l-th candidate trajectory for an invisible dynamic target. l Let F' be the k-th candidate trajectory of the invisible dynamic target, and let F' be the motion feature vector of the invisible dynamic target. d is the initial equivalent probability value; λ is the correction coefficient, obtained by normalizing the entry probability, and is directly proportional to the entry probability. l v represents the relative distance between the invisible dynamic target and the entrance. l T represents the velocity of an invisible dynamic target. l The expected travel time for invisible dynamic targets.

[0104] S225. The predicted motion state space is formed by multiple candidate trajectories and / or multiple equivalent candidate trajectories of each dynamic target, and the probability distribution is formed by the corresponding probability values ​​and / or the optimal equivalent probability values.

[0105] S300 assesses collision risk indicators under various preset conflict scenarios based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target.

[0106] In this embodiment, the conflict scenarios include at least: forward rear-end collision, lateral cutting-in, intersection "ghost peek" collision, and oncoming collision. Different conflict scenarios correspond to different collision geometry models and risk calculation methods. The collision risk index is used to quantitatively characterize the probability and severity of a collision between a vehicle and a dynamic target in a future time period under a specific conflict scenario. The risk index comprehensively considers factors such as the predicted trajectory probability of the dynamic target, the change in the predicted distance between the vehicle and the target, the target type weight, and the vehicle's current braking capability, providing a quantitative basis for the subsequent generation and optimization of multi-strategy candidate control schemes. Specifically, refer to... Figure 5 As shown, step S300 may include the following sub-steps S310 to S360:

[0107] S310. Based on the vehicle's own status information, obtain the braking safety distance under the current vehicle operating conditions.

[0108] Furthermore, the braking safety distance D safe The calculation formula is:

[0109] ;

[0110] In the formula, v is the current speed of the vehicle; t delay The system response time includes sensing and processing delays and decision-making delays, typically ranging from 0.2s to 0.5s; g is the gravitational acceleration; μ is the road surface friction coefficient; f is the slope normalization value; d margin This is for a safety margin distance.

[0111] S320. Based on the predicted positions of each candidate trajectory in the motion state space at each time point and the position of the vehicle at each corresponding time point, obtain the predicted distance between the vehicle and the dynamic target at each future time point.

[0112] S330. Extract the minimum collision time corresponding to each candidate trajectory based on the predicted distance. The minimum collision time is the time value corresponding to the first time when the predicted distance is less than the braking safety distance.

[0113] S340. Determine the weight coefficients corresponding to dynamic targets based on the type of dynamic targets, wherein the weight coefficients for pedestrians and two-wheeled vehicles are higher than the weight coefficients for four-wheeled motor vehicles.

[0114] For example, the acquired dynamic target type labels are used to assign type weight coefficients to each dynamic target. The principle for assigning weight coefficients is as follows: for vulnerable road users, such as pedestrians and two-wheeled and three-wheeled vehicles, higher weights are assigned to reflect tilt protection; for four-wheeled motor vehicles, a baseline weight is assigned. Example assignments are as follows: pedestrians ω1=2.0; two-wheeled vehicles ω2=1.8; three-wheeled vehicles ω3=1.5; four-wheeled motor vehicles ω4=1.0.

[0115] S350. Calculate the collision risk value of various preset conflict scenarios based on the minimum collision time and weight coefficient. The collision risk value is negatively correlated with the minimum collision time and positively correlated with the weight coefficient.

[0116] Furthermore, firstly, the orientation and direction of motion of the dynamic target relative to the vehicle are obtained based on the candidate trajectory to determine the corresponding conflict scenario type. Then, the collision risk value R for the corresponding conflict scenario is obtained using the collision risk value calculation formula, which is:

[0117] ;

[0118] In the formula, ω is the weight coefficient of the dynamic target, and T min Let ε be the minimum collision time, and ε be a small positive number to prevent the denominator from being zero. Where, if a minimum collision time is not met, T... min Taking infinity, the collision risk value is approximately 0.

[0119] S360: Iterate through the probability values ​​of each candidate trajectory, correct the collision risk values ​​of each candidate trajectory under different conflict scenarios, and obtain the collision risk index of vehicles and dynamic targets under various conflict scenarios.

[0120] Furthermore, to eliminate the differences in risk value dimensions caused by varying target numbers across different scenarios and to highlight the urgency of high-risk scenarios, the collision risk value needs to be probability-weighted and normalized. First, for each conflict scenario, all candidate trajectories in that scenario are obtained, and the cumulative collision risk for that scenario is calculated: ,in, Let be the probability of the k-th candidate trajectory in the i-th scenario. The collision risk value of the k-th candidate trajectory in the i-th scenario is then determined. Subsequently, the collision risk index for each conflict scenario is normalized to map it to a preset risk level range [0,1]. For a specific scenario, the risk level can be set to the highest level, R(s). normalize =1, for example, in the scenario of a ghost peeking out, because this scenario is sudden and has serious consequences, it requires more sensitive triggering of warnings and braking.

[0121] S400: Generate multiple candidate control schemes based on collision risk indicators, and calculate the comprehensive evaluation value of each candidate control scheme based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints.

[0122] In this embodiment, the collision risk index reflects the risk level of the vehicle under various preset conflict scenarios in the current environment. Based on this, it is necessary to determine the overall risk level currently faced by the vehicle according to the risk index and generate multiple corresponding candidate control schemes, forming a scheme set covering different risk response strategies. Subsequently, by constructing an optimization objective function including safety constraints and traffic efficiency constraints, a comprehensive evaluation value is calculated for each candidate control scheme, thereby providing a quantitative basis for the subsequent selection of the optimal scheme. Specifically, refer to... Figure 6 As shown, step S400 may include the following sub-steps S410 to S430:

[0123] S410. Determine the collision risk level of the road segment through which the vehicle passes based on the collision risk index, and generate multiple candidate control schemes according to the collision risk level. The candidate control schemes include any one or combination of maintaining the current motion state, braking deceleration, acceleration, steering control, and braking to a stop.

[0124] Furthermore, the mapping table of collision risk levels is shown in Table 1.

[0125]

[0126] For example, when the risk level is low, two candidate solutions are generated: Solution 1 is to maintain the current motion state, and Solution 2 is to apply light braking and deceleration; when the risk level is medium, three candidate solutions are generated: Solution 1 is to apply moderate braking and deceleration, Solution 2 is a control combination of light braking and small-angle steering to avoid collision, and Solution 3 is to maintain the current motion state but enhance monitoring.

[0127] When the risk level is high, three candidate solutions are generated: Solution 1 is emergency braking, Solution 2 is a combination of moderate braking and steering to avoid collision, and Solution 3 is steering to avoid collision. When the risk level is extremely high, two candidate solutions are generated: Solution 1 is full braking to a stop, and Solution 2 is a combination of full braking and steering to avoid collision.

[0128] S420. Map the acceleration control sequence and / or steering angle control sequence corresponding to the candidate control scheme to the prediction period of the dynamic target according to a preset time window, so as to obtain the collision probability of the vehicle and the dynamic target at each time point after the candidate control scheme is executed.

[0129] S430. With the minimum collision probability and maximum traffic efficiency as optimization objectives, solve for the comprehensive evaluation value of each candidate control scheme.

[0130] Furthermore, the expression for optimizing the objective function is:

[0131] ;

[0132] In the formula, ω a ω b ω c C represents the weighting coefficient. a Let C be the collision probability. b The difference between the expected speed and the actual speed, i.e., the traffic efficiency, C c To normalize the value for driving comfort.

[0133] S500 selects the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme and outputs it to the vehicle's actuator to control the vehicle to perform the corresponding motion operation.

[0134] In this embodiment, for each candidate solution, the smaller the comprehensive evaluation value, the better the solution performs in terms of the overall balance between safety and traffic efficiency. Based on this, this embodiment selects the optimal solution from the candidate solution set, converts it into control commands recognizable by the vehicle's actuators, and outputs them. Simultaneously, it continuously monitors environmental changes throughout the execution process, and can promptly interrupt and re-determine when the actual behavior of the dynamic target deviates from the prediction. Therefore, this embodiment, through the above mechanism, achieves complete closed-loop control from perception, prediction, evaluation, decision-making to execution, ensuring the system's adaptability and robustness in complex traffic environments. Specifically, refer to... Figure 7 As shown, step S500 may include the following sub-steps S510 to S540:

[0135] S510. Compare the comprehensive evaluation values ​​of each candidate control scheme and select the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme.

[0136] S520: Obtain the acceleration control sequence and / or steering angle control sequence corresponding to the target control scheme, and parse the acceleration control sequence into brake pedal opening command and / or drive pedal opening command, and parse the steering angle control sequence into steering wheel angle command.

[0137] S530. Based on the parsed control command, the control signal is encapsulated into a control signal according to the communication protocol of the actuator, and output to at least one of the vehicle's braking actuator, drive actuator, and steering actuator through the controller local area network bus.

[0138] S540. During the execution of the target control scheme, the real-time motion status of each dynamic target is continuously monitored. If the actual motion trajectory of any dynamic target deviates from the corresponding candidate trajectory by more than a preset deviation threshold, the current execution is interrupted and the collision risk with the dynamic target is reassessed.

[0139] Furthermore, this embodiment also provides a vehicle control system based on traffic accident prediction analysis, including:

[0140] The information acquisition module is used to acquire the vehicle's own status information, environmental perception information, and traffic assistance information from external information sources.

[0141] The dynamic target analysis module is used to acquire dynamic targets on the vehicle's driving path based on environmental perception information and traffic assistance information, and to perform motion trend analysis on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period.

[0142] The collision risk assessment module is used to evaluate collision risk indicators under various preset conflict scenarios based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target.

[0143] The control scheme generation module is used to generate multiple candidate control schemes based on collision risk indicators, and calculate the comprehensive evaluation value of each candidate control scheme according to a preset optimization objective function that includes safety constraints and traffic efficiency constraints.

[0144] The control scheme execution module is used to select the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme and output it to the vehicle's actuators to control the vehicle to perform the corresponding motion operations.

[0145] Furthermore, this embodiment also provides a vehicle controller, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the vehicle control method based on traffic accident prediction analysis described above.

[0146] Finally, this embodiment also provides a vehicle, including: a vehicle body; a vehicle controller; and actuators, including a braking actuator, a driving actuator, and a steering actuator, which are respectively connected to the vehicle controller and used to receive control commands output by the vehicle controller to control the vehicle body to perform corresponding motion operations.

[0147] In summary, the vehicle control method and system based on traffic accident prediction and analysis provided by this invention firstly achieves comprehensive perception of both visible dynamic targets and invisible dynamic targets in obstructed areas by deeply fusing onboard environmental perception information and V2X traffic assistance information, effectively overcoming the detection limitations of traditional single-vehicle sensors in blind spot scenarios such as "ghost pedestrians" (sudden pedestrians appearing from behind obstacles). Secondly, through a multimodal trajectory prediction method based on deep learning and entry probability modeling, the uncertainty and probability distribution of the future motion state of each dynamic target are quantitatively expressed, significantly improving the accuracy and robustness of the prediction. Furthermore, by comprehensively considering the minimum collision time, target type weight, and braking safety distance in a multi-scenario collision risk assessment model, it achieves comprehensive detection of forward rear-end collisions, lateral cut-in collisions, and intersection ghost pedestrians. Risk quantification of various pre-set conflict scenarios, such as head-on and oncoming collisions, provides a reliable basis for accurate decision-making. Then, through a multi-objective optimization function based on safety and traffic efficiency constraints, multiple candidate control schemes are generated and comprehensively evaluated, including maintaining the current motion state, braking deceleration, acceleration, steering control, and combinations thereof. The scheme with the smallest comprehensive evaluation value is selected as the target control scheme, achieving a globally optimal balance between safety and traffic efficiency. Finally, by parsing the control scheme into control commands recognizable by the actuator and outputting them, and by continuously monitoring the deviation between the actual trajectory and the predicted trajectory of the dynamic target during execution and triggering interruption and replanning when the deviation exceeds a threshold, a complete closed-loop control link of perception, prediction, evaluation, decision-making, execution, monitoring, and replanning is formed. Therefore, this invention can effectively address the safety challenges caused by blind spots, behavioral uncertainties, and multi-objective composite risks in complex urban road and intersection scenarios. It achieves advance prediction and active braking in sudden "ghost peek" scenarios, multi-strategy game optimization decision-making when multiple risks are intertwined, and rapid adaptive adjustment when target behavior changes abruptly, comprehensively improving the vehicle's active safety protection capabilities, traffic efficiency, and ride comfort in complex traffic environments.

[0148] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the systems / devices based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems / devices used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0149] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0151] It should be noted that in the description of this invention, the word "a" or "an" preceding a component does not exclude the existence of multiple such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. The use of terms such as first, second, third, etc., is merely for convenience and does not indicate any order. These terms can be understood as part of the component names.

[0152] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0153] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning of the basic inventive concept, can make other changes and modifications to these embodiments.

[0154] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from the spirit and scope of the invention.

Claims

1. A vehicle control method based on traffic accident prediction analysis, characterized in that, include: Acquire vehicle status information, environmental perception information, and traffic assistance information from external information sources; Based on environmental perception information and traffic assistance information, dynamic targets on the vehicle's driving path are obtained, and motion trend analysis is performed on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period. Based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target, the collision risk indicators under various preset conflict scenarios are evaluated. Multiple candidate control schemes are generated based on collision risk indicators, and the comprehensive evaluation value of each candidate control scheme is calculated based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints. The candidate control scheme with the smallest comprehensive evaluation value is selected as the target control scheme and output to the vehicle's actuator to control the vehicle to perform the corresponding motion operation.

2. The method as described in claim 1, characterized in that, Acquiring vehicle status information, environmental perception information, and traffic assistance information from external sources includes: Acquire the vehicle's own status information, including the vehicle's current position, heading angle, speed, acceleration, and braking system status parameters; The vehicle collects environmental perception information around it through environmental perception sensors. The environmental perception information includes the position, speed, and direction of motion of dynamic targets on the vehicle's driving path, as well as the relative distance and relative speed between the dynamic targets and the vehicle. Traffic assistance information is received from external sources via V2X communication. This traffic assistance information includes target information and traffic light status information broadcast by roadside units in obstructed areas, as well as perception information and avoidance event information shared by surrounding vehicles.

3. The method as described in claim 1, characterized in that, Dynamic targets along the vehicle's path are obtained based on environmental perception information and traffic assistance information, including: The vehicle's environmental perception sensors collect visible targets along the driving path, perform target identification and state estimation on the visible targets, and obtain the visible dynamic targets along the vehicle's driving path. Based on the target information of the obstructed area broadcast by the roadside unit in the traffic assistance information and / or the avoidance event information sent by the remote vehicle, combined with the relative positional relationship between the vehicle and the obstructed area and the vehicle speed, the blind spot risk coefficient of the vehicle passing through the obstructed area is obtained. When the blind spot risk coefficient exceeds the preset risk threshold, the moving target located in the obstructed area reported by the roadside unit or remote vehicle is marked as an invisible dynamic target, and the position and motion state of the invisible dynamic target in the traffic assistance information are converted into equivalent state information in the vehicle coordinate system.

4. The method as described in claim 3, characterized in that, Motion trend analysis of dynamic targets yields the predicted motion state space and probability distribution of each target in future time periods, including: Obtain the motion characteristics of visible dynamic targets, including historical position sequences, velocity sequences, and heading angle sequences; The motion features are input into the pre-trained trajectory prediction model, which outputs multiple candidate trajectories for each visible dynamic target in the future time period and uses the probability density function to obtain the probability value corresponding to each candidate trajectory. Based on the equivalent state information of the invisible dynamic target and the location information of the occluded area, multiple possible entry points for the invisible dynamic target to enter the vehicle's driving path from the occluded area and the entry probability corresponding to each entry point are obtained. Starting from a possible entry point, and combining the historical movement trend of the invisible dynamic target with traffic rule constraints, multiple equivalent candidate trajectories are generated and an equivalent probability value is assigned to each equivalent candidate trajectory. The equivalent probability value is then corrected based on the entry probability to obtain the optimal equivalent probability value. The predicted motion state space is constructed using multiple candidate trajectories and / or multiple equivalent candidate trajectories for each dynamic target, and the probability distribution is constructed using the corresponding probability values ​​and / or the optimal equivalent probability values.

5. The method as described in claim 1, characterized in that, Based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target, collision risk indicators are evaluated under various preset conflict scenarios, including: Based on the vehicle's own status information, obtain the braking safety distance under the current vehicle operating conditions; Based on the predicted positions of each candidate trajectory in the motion state space at each time point and the position of the vehicle at each corresponding time point, the predicted distance between the vehicle and the dynamic target at each future time point is obtained. The minimum collision time corresponding to each candidate trajectory is extracted based on the predicted distance. The minimum collision time is the time value corresponding to the first time when the predicted distance is less than the braking safety distance. The weight coefficients for dynamic targets are determined based on their types, with pedestrians and two-wheeled vehicles having higher weight coefficients than four-wheeled motor vehicles. Collision risk values ​​for various preset conflict scenarios are calculated based on minimum collision time and weighting coefficients. The collision risk value is negatively correlated with the minimum collision time and positively correlated with the weighting coefficients. By iterating through the probability values ​​of each candidate trajectory, the collision risk values ​​of each candidate trajectory under different conflict scenarios are corrected to obtain the collision risk index of vehicles and dynamic targets under various conflict scenarios.

6. The method as described in claim 1, characterized in that, Multiple candidate control schemes are generated based on collision risk indicators, and the comprehensive evaluation value of each candidate control scheme is calculated according to a preset optimization objective function that includes safety constraints and traffic efficiency constraints. The collision risk level of the road segment through which the vehicle passes is determined based on the collision risk index, and multiple candidate control schemes are generated according to the collision risk level. The candidate control schemes include any one or a combination of maintaining the current motion state, braking deceleration, acceleration, steering control, and braking to a stop. The acceleration control sequence and / or steering angle control sequence corresponding to the candidate control scheme are mapped to the predicted time period of the dynamic target according to a preset time window in order to obtain the collision probability between the vehicle and the dynamic target at each time point after the candidate control scheme is executed. With the optimization objectives of minimizing collision probability and maximizing traffic efficiency, the comprehensive evaluation value of each candidate control scheme is solved.

7. The method as described in claim 1, characterized in that, The candidate control scheme with the lowest comprehensive evaluation value is selected as the target control scheme, and the output is sent to the vehicle's actuators to control the vehicle to perform corresponding motion operations, including: Compare the comprehensive evaluation values ​​of each candidate control scheme, and select the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme; Obtain the acceleration control sequence and / or steering angle control sequence corresponding to the target control scheme, and parse the acceleration control sequence into brake pedal opening command and / or drive pedal opening command, and parse the steering angle control sequence into steering wheel angle command; The parsed control commands are encapsulated into control signals according to the communication protocol of the actuators, and output to at least one of the vehicle's braking actuators, drive actuators, and steering actuators via the controller local area network bus; During the execution of the target control scheme, the real-time motion status of each dynamic target is continuously monitored. If the actual motion trajectory of any dynamic target deviates from the corresponding candidate trajectory by more than a preset deviation threshold, the current execution is interrupted and the collision risk between the vehicle and the dynamic target is reassessed.

8. A vehicle control system based on traffic accident prediction analysis, characterized in that, include: The information acquisition module is used to acquire the vehicle's own status information, environmental perception information, and traffic assistance information from external information sources; The dynamic target analysis module is used to obtain dynamic targets on the vehicle's driving path based on environmental perception information and traffic assistance information, and to perform motion trend analysis on the dynamic targets to obtain the predicted motion state space and probability distribution of each dynamic target in the future time period. The collision risk assessment module is used to assess collision risk indicators under various preset conflict scenarios based on the vehicle's own state information and the predicted motion state space and probability distribution of each dynamic target. The control scheme generation module is used to generate multiple candidate control schemes based on collision risk indicators, and calculate the comprehensive evaluation value of each candidate control scheme based on a preset optimization objective function that includes safety constraints and traffic efficiency constraints. The control scheme execution module is used to select the candidate control scheme with the smallest comprehensive evaluation value as the target control scheme and output it to the vehicle's actuators to control the vehicle to perform the corresponding motion operations.

9. A vehicle controller, characterized in that, include: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which are executed by at least one processor to enable the at least one processor to perform a vehicle control method based on traffic accident prediction analysis as described in any one of claims 1-7.

10. A vehicle, characterized in that, include: Vehicle body; The vehicle controller as described in claim 9; The actuators, including braking actuators, drive actuators and steering actuators, are connected to the vehicle controller and are used to receive control commands output by the vehicle controller in order to control the vehicle body to perform corresponding motion operations.

Images