System and method for an intelligent vehicle to respond to a dangerous traffic participant

Abstract

This invention relates to a system for intelligent driving vehicles to respond to dangerous traffic participants. The system includes: a perception and acquisition module for collecting image information, dynamic information, position coordinates, and driving trajectories of surrounding vehicles; preprocessing the collected data to form a standardized traffic environment information data packet; and transmitting it to a hazard identification and marking module; a hazard identification and marking module for extracting behavioral characteristics of surrounding vehicles, identifying dangerous driving behaviors, and marking them as dangerous traffic participants; a decision-making and planning module for classifying hazard levels, dynamically setting corresponding warning distance thresholds and safe distance thresholds, generating hazard avoidance trajectories, and generating vehicle control commands; and a vehicle control module for converting these into execution signals, controlling the vehicle to perform hazard avoidance operations, and simultaneously collecting the actual driving status of the vehicle and feeding it back to the decision-making and planning module.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a system and method for intelligent driving vehicles to deal with dangerous traffic participants. Background Technology

[0002] Currently, the environmental perception and control of intelligent driving vehicles mainly rely on sensors such as cameras, millimeter-wave radar, and lidar. These sensors identify surrounding obstacles and predict their next trajectory to achieve conventional driving controls such as following, lane changing, and braking. However, existing technologies have significant shortcomings: in the face of sudden emergencies, the system often can only perform emergency braking to mitigate collision damage, lacking the ability to continuously identify, track, and preventively avoid potentially high-risk dangerous driving behaviors. In real-world driving scenarios, there are numerous dangerous road users with a high probability of accidents. Typical dangerous behaviors include, but are not limited to: multiple extreme lane changes within a short period, forced lane changes without turn signals, significant deceleration in a short time, frequent emergency braking, prolonged crossing or straddling of lane lines, and violations of traffic rules such as running red lights, driving against traffic, and illegal turns. Traditional intelligent driving systems only treat such vehicles as transient obstacles for single-use processing, without establishing a dedicated database of dangerous road users and a tiered avoidance strategy. This prevents the early identification, continuous tracking, and proactive avoidance of dangerous road users, resulting in insufficient safety redundancy and difficulty in minimizing the probability of collisions. Summary of the Invention

[0003] In view of the above problems, the present invention provides a system and method for intelligent driving vehicles to deal with dangerous traffic participants, so as to solve the technical problems of existing technologies that can only provide passive protection, cannot achieve early avoidance, have a narrow scope of application, and are prone to excessive or untimely avoidance.

[0004] This invention provides a system for intelligent driving vehicles to respond to dangerous traffic participants. The system includes: a perception and acquisition module, used to collect image information, dynamic information, position coordinates, and driving trajectories of surrounding vehicles; preprocessing the collected data to form a standardized traffic environment information data packet, which is then transmitted to a hazard identification and marking module; a hazard identification and marking module, connected to the perception and acquisition module, used to extract behavioral characteristics of surrounding vehicles based on the traffic environment information data packet; analyzing these behavioral characteristics through rule models and time-series trajectories to identify dangerous driving behaviors and mark the corresponding vehicles as dangerous traffic participants; a decision-making and planning module, connected to the hazard identification and marking module, used to classify the hazard level based on the dangerous behavior type, frequency, distance from the vehicle, and relative speed of the dangerous traffic participants; dynamically setting corresponding warning distance thresholds and safe distance thresholds for different hazard levels; when the dangerous traffic participant's distance is less than the warning distance threshold, selecting an avoidance strategy based on the current traffic environment, generating an avoidance trajectory, and generating vehicle control commands based on the avoidance trajectory; and a vehicle control module, connected to the decision-making and planning module, used to convert the avoidance trajectory and control commands into execution signals, control the vehicle to perform avoidance operations, and simultaneously collect the vehicle's actual driving status and feed it back to the decision-making and planning module for timely adjustment of control commands.

[0005] Furthermore, the system also includes a storage module, which is connected to the perception and acquisition module, the hazard identification and marking module, the decision planning module, and the vehicle control module, respectively, for storing various types of data during system operation, providing support for subsequent optimization and traceability.

[0006] Furthermore, the perception and acquisition module includes: a camera, mounted on the vehicle, covering a field of view of 180° in front, 60° behind, and 90° on each side, used to acquire image information of the shape of surrounding vehicles, turn signal status, and lane line position; a millimeter-wave radar, mounted on the vehicle, emitting electromagnetic waves at a frequency of not less than 10Hz and receiving reflected signals, used to detect the dynamic information of the distance, speed, and angle of surrounding vehicles within a 200-meter range; a lidar, mounted on the vehicle, emitting millions of laser pulses per second to construct a 360° high-precision point cloud map, with a detection range of 150 meters, used to accurately obtain the position coordinates and driving trajectories of surrounding vehicles; and a data preprocessing unit, connected to the camera, millimeter-wave radar, and lidar respectively, used to perform noise reduction, synchronization, and fusion processing on the acquired raw data, remove invalid data and measurement errors, output standardized traffic environment information data packets, and transmit them to the hazard identification and marking module.

[0007] This invention also provides a method for intelligent driving vehicles to deal with dangerous traffic participants. The method includes: Step 1, acquiring real-time image information, dynamic information, location coordinates, and driving trajectories of surrounding vehicles, preprocessing the collected data to form a standardized traffic environment information data package; Step 2, based on the traffic environment information data package, identifying dangerous driving behaviors based on rule models and time-series trajectory analysis, and marking the corresponding vehicles as dangerous traffic participants; Step 3, classifying the danger level according to the dangerous behavior type, frequency, distance from the vehicle, and relative speed of the dangerous traffic participants, and dynamically setting corresponding warning distance thresholds; Step 4, when the dangerous traffic participant is less than the warning distance threshold, selecting a risk avoidance strategy based on the current traffic environment, generating a risk avoidance trajectory, and generating vehicle control commands based on the risk avoidance trajectory; Step 5, converting the risk avoidance trajectory and control commands into execution signals, controlling the vehicle to perform risk avoidance operations, and simultaneously collecting the actual driving status of the vehicle and feeding it back to the decision planning module for timely adjustment of control commands.

[0008] Furthermore, step 2 includes: step 21, receiving traffic environment information data packets output by the perception and acquisition module, and extracting behavioral characteristics of surrounding vehicles, including lane-changing frequency, braking status, and duration of crossing the line; step 22, identifying whether surrounding vehicles exhibit at least one dangerous driving behavior through rule model and time-series trajectory analysis; step 23, if dangerous driving behavior exists, marking the vehicle as a dangerous traffic participant, assigning it a unique identification ID, establishing and dynamically updating a list of dangerous targets, and synchronizing the ID, behavioral characteristics, danger level, and location information of the dangerous targets to the decision planning module and storage module.

[0009] Furthermore, the dangerous driving behaviors include: multiple extreme lane changes within a unit of time, forced lane changes without turn signals; significant deceleration and frequent emergency braking in a short period of time; long-term driving across lane lines or straddling lane lines; running red lights, driving against traffic, making illegal turns, and not driving in the designated lane.

[0010] Furthermore, step 22 includes: step 221, establishing a rule model based on the temporal characteristics, threshold characteristics, and violation characteristics of dangerous driving behavior; step 222, making a preliminary judgment on whether the vehicle has dangerous behavior through the judgment rules in the rule model; step 223, identifying the frequency of dangerous behavior of the vehicle by continuously tracking, statistically analyzing, comparing, and predicting the vehicle's position, speed, heading, steering state, braking state, and lane position over multiple consecutive frames; and step 224, using the rule model to determine whether the vehicle has dangerous driving behavior based on the frequency of dangerous behavior.

[0011] Further, step 221 includes: Step 2211, defining a dangerous behavior type library, the types including multiple extreme lane changes, lane changes without turn signals, frequent emergency braking, prolonged lane crossing, running red lights, driving against traffic, and illegal turns; Step 2212, setting quantitative judgment conditions for each dangerous behavior, the quantitative judgment conditions including lane change frequency, braking intensity, lane crossing duration, speed change, steering angular velocity, and turn signal activation duration; Step 2213, using a fixed-length time window to construct a temporal behavior window and continuously statistically analyze vehicle behavior; Step 2214, establishing a graded judgment logic, the grade specifically divided into single dangerous behavior, multiple dangerous behavior, and continuous dangerous behavior, with the level increasing sequentially; Step 2215, adding legality elements to determine whether there is illegal driving, the legality elements including maps, traffic lights, guide lanes, and prohibitory markings; Step 2216, setting the model output as: dangerous traffic participant ID + danger level + continuous tracking marker.

[0012] Furthermore, the judgment rules in the rule model include: multiple extreme lane change behavior rules, where if there are no less than 2 lane changes within a unit time window and the steering angular velocity is no less than 15° / s, it is judged as multiple extreme lane changes; forced lane change behavior rules without turn signals, where if a lane change is made without using a turn signal or the turn signal is on for less than 3 seconds, it is judged as a forced lane change without turn signals; and short-term large deceleration / frequent emergency braking rules, where the deceleration duration is no more than 2 seconds, the speed drop is no less than 30 km / h, the number of emergency brakings within 1 minute is no less than 2, and the braking acceleration is no greater than -5 m / s². 2 If a vehicle crosses the lane line for an extended period of at least 5 seconds and its center is permanently on the lane line, it is considered a long-term violation of lane line regulations. If a vehicle runs a red light, drives against traffic, or makes an illegal turn, it is considered a serious violation of dangerous driving regulations.

[0013] Further, step 223 includes: step 2231, continuously collecting trajectory data of the target vehicle at a frequency of 10Hz, the trajectory data including position, speed, heading angle, steering angular velocity, brake light status, turn signal status, and lane position; step 2232, using a first-in-first-out time-series queue to store the most recent N consecutive trajectories; step 2233, performing behavior slicing on the continuous trajectory, identifying lane-changing behavior segments, braking behavior segments, line-crossing behavior segments, and steering behavior segments based on the collected trajectory data; step 2234, extracting the behavior features of each behavior segment, including duration, speed change, steering angular velocity, braking intensity, turn signal on duration, line-crossing duration, and number of lane changes; step 2235, statistically analyzing the behavior features and matching them with judgment rules to obtain the judgment result of dangerous driving behavior; step 2236, if any judgment rule is met, marking the driver as a dangerous traffic participant and assigning a unique ID; if there is no dangerous behavior in N consecutive frames, the marking is lifted.

[0014] This invention provides a system and method for intelligent driving vehicles to deal with dangerous traffic participants. It mainly addresses the shortcomings of existing technologies, such as the lack of ability to identify dangerous driving behaviors, mark and continuously track dangerous traffic participants, the inability to achieve proactive preventive avoidance, insufficient safety redundancy, the single avoidance strategy that cannot be adjusted according to the degree of danger, the inability to balance safety and traffic efficiency, the tendency to cause secondary accidents or untimely avoidance, and the inability to dynamically adjust the avoidance trajectory according to changes in the behavior of dangerous traffic participants and the driving status of the vehicle, resulting in poor adaptability and difficulty in meeting the needs of intelligent driving in multiple scenarios. Attached Figure Description

[0015] Figure 1 A flowchart of a method for intelligent driving vehicles to deal with dangerous traffic participants provided by the present invention; Figure 2 A system structure diagram of an intelligent driving vehicle responding to dangerous traffic participants is provided by the present invention; Figure 3 This is a flowchart of a method for identifying dangerous driving behavior and marking dangerous traffic participants provided by the present invention; Figure 4 This is a flowchart of a method for identifying whether surrounding vehicles are engaging in dangerous driving behavior, provided by the present invention. Figure 5 This is a flowchart of a method for establishing a rule model provided by the present invention; Figure 6 This is a flowchart of a time-series trajectory analysis algorithm method provided by the present invention. Detailed Implementation

[0016] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0017] Device Example: This invention provides a system for intelligent driving vehicles to respond to dangerous traffic participants, comprising: a perception and acquisition module, a hazard identification and marking module, a decision-making and planning module, a vehicle control module, and a storage module. It is adaptable to L2 to L4 level intelligent driving vehicles. Each module has a clear division of labor and works collaboratively. The modules are connected via a data transmission link to achieve information interaction and closed-loop control. Figure 2 As shown.

[0018] The perception and acquisition module is used to collect image information, dynamic information, location coordinates and driving trajectories of surrounding vehicles. The collected data is preprocessed to form a standardized traffic environment information data package, which is then transmitted to the hazard identification and marking module. The perception and acquisition module, consisting of an onboard camera, millimeter-wave radar, lidar, and a data preprocessing unit, is the system's "perception terminal." It includes: a camera mounted on the vehicle to cover a 180° field of view in front, 60° behind, and 90° to each side, used to collect image information of surrounding vehicles' shapes, turn signal status, and lane line positions; a millimeter-wave radar mounted on the vehicle that emits electromagnetic waves at a frequency of at least 10Hz and receives reflected signals, used to detect the distance, speed, and angle of surrounding vehicles within a 200-meter range; a lidar mounted on the vehicle that emits millions of laser pulses per second to construct a 360° high-precision point cloud map with a detection range of 150 meters, used to accurately obtain the position coordinates and driving trajectories of surrounding vehicles; and a data preprocessing unit connected to the camera, millimeter-wave radar, and lidar, used to denoise, synchronize, and fuse the collected raw data, removing invalid data and measurement errors, and outputting standardized traffic environment information data packets for transmission to the hazard identification and marking module. In this module, the following three types of signals are transmitted: Signal 1: Static information (position, shape), dynamic information (speed, heading, turning status), and behavioral time-series data of surrounding vehicles collected by the intelligent camera group, millimeter-wave radar, and lidar; Signal 2: Standardized traffic environment information data packets output by the perception and acquisition module after denoising, synchronizing, and fusing the raw data; Signal 3: Signal by which the perception and acquisition module transmits the standardized data packets to the hazard identification and marking module.

[0019] The hazard identification and marking module, connected to the perception and acquisition module, is used to extract the behavioral characteristics of surrounding vehicles based on traffic environment information data packets. These behavioral characteristics are then analyzed using rule models and time-series trajectories to identify dangerous driving behaviors and mark the corresponding vehicles as dangerous traffic participants. Dangerous driving behaviors include: multiple extreme lane changes within a unit of time (1-5 minutes, adjustable according to the actual scenario) (steering angular velocity ≥15° / s during lane changes); forced lane changes without turn signals (no corresponding turn signal activated before lane change or turn signal activation time <3s); significant deceleration within a short period (≤2s) (speed decrease ≥30km / h); and frequent emergency braking (≥2 emergency braking times within 1 minute, emergency braking judgment standard is braking acceleration ≤-5m / s²). 2 This includes behaviors such as long-term crossing of lane lines, driving on the lane lines (endurance of crossing the lane lines ≥ 5 seconds); running red lights, driving against traffic, illegal turns (such as turning at intersections where turning is prohibited), and not driving in the designated lane. The following signals are transmitted in this module: Signal 4: The signal by which the hazard identification and marking module transmits the hazard target list, hazard level, and other data to the decision planning module and storage module; Signal 10: The de-marking signal issued by the hazard identification and marking module after hazard avoidance is completed or the hazard target disappears, synchronously updating the hazard target list.

[0020] The decision-making and planning module, connected to the hazard identification and marking module, is used to classify hazard levels based on the type, frequency, distance from the vehicle, and relative speed of hazardous traffic participants. For different hazard levels, it dynamically sets corresponding warning distance thresholds and safe distance thresholds. When a hazardous traffic participant is within the warning distance threshold, it selects an avoidance strategy based on the current traffic environment, generates an avoidance trajectory, and generates vehicle control commands based on the avoidance trajectory. The following signals are transmitted in this module: Signal 5: The decision-making and planning module receives signals of the vehicle's real-time driving status (speed, position, heading) and road environment information (number of lanes, speed limit); Signal 6: The decision-making and planning module generates avoidance strategies (automatic avoidance and departure / rapid departure) and time-series trajectory data signals based on hazard target information and environmental information; Signal 7: The decision-making and planning module transmits the avoidance trajectory and control commands to the vehicle control module.

[0021] The vehicle control module, connected to the decision-making and planning module, converts the avoidance trajectory and control commands into execution signals to control the vehicle to perform avoidance maneuvers. Simultaneously, it collects the vehicle's actual driving status and feeds it back to the decision-making and planning module for timely adjustments to control commands. The execution signals include acceleration, braking, and steering angle signals, used to control the vehicle to precisely execute avoidance maneuvers such as acceleration, deceleration, and steering, ensuring accurate execution of the avoidance trajectory. The vehicle's actual driving status includes speed, position, and heading, which is fed back to the decision-making and planning module to achieve closed-loop control, allowing for timely adjustments to control commands and avoiding deviations. This module transmits the following signals: Signal 8: The signal from which the vehicle control module converts control commands into execution signals and transmits them to the vehicle's power system, braking system, and steering system; Signal 9: The signal from which the vehicle control module collects the vehicle's actual driving status and feeds it back to the decision-making and planning module for closed-loop adjustment. The storage module, connected to the sensing and acquisition module, hazard identification and marking module, decision-making and planning module, and vehicle control module, stores various types of data during system operation, providing support for subsequent optimization and traceability. The storage module stores the unique IDs, behavioral characteristic data, hazard level records, marking times, and trajectory data of hazardous traffic participants; it also stores the system's preset hazardous behavior rules, grading standards, and algorithm parameters. The storage period can be set according to needs (1-7 days), automatically overwriting expired data. The stored data can be used for subsequent hazardous behavior model optimization, accident tracing, and analysis.

[0022] This invention provides a system and method for intelligent driving vehicles to deal with dangerous traffic participants. It mainly addresses the shortcomings of existing technologies, such as the lack of ability to identify dangerous driving behaviors, mark and continuously track dangerous traffic participants, the inability to achieve proactive preventive avoidance, insufficient safety redundancy, the single avoidance strategy that cannot be adjusted according to the degree of danger, the inability to balance safety and traffic efficiency, the tendency to cause secondary accidents or untimely avoidance, and the inability to dynamically adjust the avoidance trajectory according to changes in the behavior of dangerous traffic participants and the driving status of the vehicle, resulting in poor adaptability and difficulty in meeting the needs of intelligent driving in multiple scenarios.

[0023] Method Example 2: This invention provides a method for intelligent driving vehicles to respond to dangerous road users, such as... Figure 1 As shown, the method includes: Step 1: Acquire real-time image information, dynamic information, location coordinates and driving trajectories of surrounding vehicles; preprocess the collected data to form a standardized traffic environment information data package. This step uses a multi-sensor fusion system to acquire static and dynamic information and behavioral time-series data of surrounding traffic participants in real time, with a collection frequency of no less than 10Hz.

[0024] Step 2: Based on the traffic environment information data package, and using rule models and time-series trajectory analysis, identify dangerous driving behaviors and mark the corresponding vehicles as dangerous traffic participants. like Figure 3 As shown, step 2 includes: Step 21: Receive the traffic environment information data packet output by the perception and acquisition module, and extract the behavioral characteristics of surrounding vehicles, including lane change frequency, braking status, and duration of crossing the line. Step 22: Through rule modeling and time-series trajectory analysis, identify whether there is at least one dangerous driving behavior in the surrounding vehicles; like Figure 4 As shown, step 22 includes: Step 221: Establish a rule model based on the temporal characteristics, threshold characteristics, and violation characteristics of dangerous driving behavior; The rule-based model is a multi-layered judgment model built upon the temporal, threshold, and violation characteristics of dangerous driving behavior. It is used to accurately identify "dangerous traffic participants" from among normal vehicles. For example... Figure 5 As shown, 221 includes: Step 2211: Define a library of dangerous behavior types, including multiple extreme lane changes, lane changes without turn signals, frequent emergency braking, prolonged crossing of lane lines, running red lights, driving against traffic, and illegal turns; Step 2212: Set quantitative judgment conditions for each dangerous behavior. The quantitative judgment conditions include lane change frequency, braking intensity, line crossing duration, speed change, steering angular velocity, and turn signal activation duration. Step 2213: Using a fixed-length time window, construct a time-series behavior window to continuously statistically analyze vehicle behavior; The fixed-length time window is configurable, for example, 1–5 minutes.

[0025] Step 2214: Establish a graded judgment logic, wherein the graded judgment is specifically divided into single dangerous behavior, multiple dangerous behavior, and continuous dangerous behavior, and the grade increases in sequence. Establish a tiered judgment logic: single dangerous behavior → multiple dangerous behaviors → continuous dangerous behavior → risk level increase.

[0026] Step 2215: Add legality elements to determine whether there is a traffic violation. The legality elements include maps, traffic lights, guide lanes, and prohibition markings. Step 2216: Set the model output as: Hazardous Traffic Participant ID + Hazard Level + Continuous Tracking Marker.

[0027] Step 222: Initially determine whether the vehicle has dangerous behavior using the decision rules in the rule model; The judgment rules in the rule model include: multiple extreme lane change behavior rule, when the number of lane changes is not less than 2 within a unit time window and the steering angular velocity is not less than 15° / s, it is judged as multiple extreme lane changes; forced lane change behavior without turn signal rule, when changing lanes without using the turn signal or when the turn signal is on for less than 3 seconds, it is judged as forced lane change without turn signal; short-term large deceleration / frequent emergency braking rule, when the deceleration time is not greater than 2 seconds, the speed drop is not less than 30km / h, the number of emergency brakings within 1 minute is not less than 2, and the braking acceleration is not greater than -5m / s². 2 If a vehicle crosses the lane line for an extended period of at least 5 seconds and its center is permanently on the lane line, it is considered a long-term violation of lane line regulations. If a vehicle runs a red light, drives against traffic, or makes an illegal turn, it is considered a serious violation of dangerous driving regulations.

[0028] Step 223: By continuously tracking, statistically analyzing, comparing, and predicting the vehicle's position, speed, heading, steering status, braking status, and lane position over multiple consecutive frames, the frequency of dangerous behaviors of the vehicle is identified. Temporal trajectory analysis algorithms continuously track, statistically analyze, compare, and predict the position, speed, heading, steering state, braking state, and lane position of a target vehicle across multiple frames to identify "repeated, repetitive, continuous, and malicious" dangerous driving behaviors. For example... Figure 6 As shown, step 223 includes: Step 2231: Continuously collect trajectory data of the target vehicle at a frequency of 10Hz. The trajectory data includes position, speed, heading angle, steering angular velocity, brake light status, turn signal status, and lane position. At a frequency of 10Hz, continuously collect the target vehicle's position (x, y), velocity v, heading angle, yaw rate, brake light status, turn signal status, and lane position. Step 2232: Use a first-in-first-out time queue to store the most recent N consecutive trajectories; A first-in-first-out (FIFO) time queue is used to store the most recent N consecutive trajectories (N=60~300, corresponding to the time window length).

[0029] Step 2233: Perform behavior slicing on the continuous trajectory and identify lane change behavior segments, braking behavior segments, line crossing behavior segments, and steering behavior segments based on the collected trajectory data; Behavioral slicing is performed on continuous trajectories to identify: 1. Lane change determination: lane number change: lane(t+T)−lane(t)≥1 and steering angular velocity: yaw_rate≥15° / s; 2. Frequent braking determination: Δv=v(t)−v(t+Δt)≥30km / h, Δt≤2s and braking acceleration: abrake≤−5m / s² 2 3. Long-term lane crossing judgment: The vehicle center continuously fluctuates on both sides of the lane line: tpress≥5s; 4. Lane change without turn signal judgment: The turn signal status is always OFF during the lane change, or the on time is <3s: tturn<3s.

[0030] Step 2234: Extract the behavioral features of each behavior segment, including duration, speed change, steering angular velocity, braking intensity, turn signal duration, line crossing duration, and number of lane changes. Step 2235: Statistically analyze the behavioral characteristics and match them with the judgment rules to obtain the judgment result of dangerous driving behavior; Step 2236: If any of the judgment rules are met, mark the participant as a dangerous traffic participant and assign a unique ID. If there is no dangerous behavior for N consecutive frames, remove the mark.

[0031] This step assigns a unique identification ID (e.g., "Dangerous_Vehicle_XXX") to each hazardous traffic participant, establishes and dynamically updates a list of hazardous targets, and synchronizes data such as the ID, behavioral characteristics, hazard level, and location information of the hazardous targets to the decision-making and planning module and the storage module. When hazard avoidance is completed (the distance between the vehicle and the hazardous traffic participant exceeds the warning distance threshold, and the hazardous traffic participant does not engage in any further hazardous behavior), the system controls the vehicle to resume normal driving. When the hazardous traffic participant moves out of the vehicle's sensor detection range, or does not engage in any hazardous behavior for 30 consecutive seconds, the system automatically removes it from the list of hazardous targets and demarks it. In other words, meeting any one of these criteria: marking as a hazardous traffic participant and assigning a unique ID; meeting multiple criteria: hazard level upgraded; N consecutive frames without hazardous behavior: downgraded and demarked.

[0032] Step 224: The rule model determines whether a vehicle is engaging in dangerous driving behavior based on the frequency of such behavior.

[0033] Step 23: If dangerous driving behavior exists, mark the vehicle as a dangerous traffic participant, assign it a unique identification ID, establish and dynamically update the list of dangerous targets, and synchronize the ID, behavioral characteristics, danger level, and location information of the dangerous targets to the decision planning module and the storage module.

[0034] Step 3: Classify the danger level according to the dangerous behavior type, frequency, distance from the vehicle, and relative speed of dangerous traffic participants, and dynamically set the corresponding warning distance threshold; This step receives hazard target information transmitted by the hazard identification and marking module, and combines it with the vehicle's real-time driving status and road environment information; it is the system's "decision-making core." It dynamically updates the hazard level of hazardous traffic participants, classifying them into three levels—Level 1 (High Risk), Level 2 (Medium Risk), and Level 3 (Low Risk)—based on the type and frequency of dangerous behavior, distance from the vehicle, and relative speed. For each hazard level, corresponding warning distance thresholds and safe distance thresholds are dynamically set to ensure the accuracy and timeliness of the classification: Level 1 (High Risk): The hazardous traffic participant is less than 50m from the vehicle and is engaging in dangerous behaviors that are highly likely to cause a collision, such as malicious lane changes (e.g., continuous lane changes across multiple lanes), or emergency braking; the relative speed difference is ≥20km / h; the warning distance threshold is set to 50m. The safe distance threshold is set at 30m; Level 2 (Medium Risk): Dangerous traffic participants are 50-100m away from the vehicle, exhibiting dangerous behaviors such as continuous lane crossing or multiple violations (e.g., two illegal lane changes within one minute), with a relative speed difference of 10-20km / h; the warning distance threshold is set at 100m, and the safe distance threshold is set at 20m; Level 3 (Low Risk): Dangerous traffic participants are >100m away from the vehicle, exhibiting only occasional dangerous behaviors (e.g., a single illegal lane crossing <5s), with a relative speed difference <10km / h; the warning distance threshold is set at 150m, and the safe distance threshold is set at 15m.

[0035] Step 4: When the dangerous traffic participant is less than the warning distance threshold, select a risk avoidance strategy based on the current traffic environment, generate a risk avoidance trajectory, and generate vehicle control instructions based on the risk avoidance trajectory. When the vehicle approaches a dangerous road user within warning distance, based on the current traffic environment (number of lanes, distribution of surrounding vehicles, road speed limits), the vehicle will choose either an automatic avoidance and departure strategy or a rapid departure strategy: 1. Automatic avoidance and departure: Without violating traffic regulations or affecting surrounding vehicles, the vehicle will adjust its lateral trajectory to maintain a lateral distance greater than the safe distance threshold (≥1.5m) from the dangerous road user; simultaneously, it will smoothly reduce its speed (deceleration ≥-2m / s²). 2 1. Avoid sudden braking; maintain a safe following distance and avoid driving alongside or closely following dangerous road users. 2. Quickly move away: Within the legal speed limit and provided there are no other obstacles or dangerous vehicles in adjacent lanes, control your vehicle to accelerate smoothly (acceleration ≤ 2m / s²). 2The system can quickly move away from the danger zone where dangerous traffic participants are located; or plan a lane-changing trajectory in advance to change lanes away from dangerous traffic participants, thereby breaking away from parallel sections and avoiding close contact; 3. Auxiliary constraints: The vehicle is prohibited from driving to the side of dangerous traffic participants for an extended period (≥3s), and is prohibited from following dangerous traffic participants at close range (less than the safe distance threshold). If it is not possible to immediately move away, the vehicle will be controlled to decelerate to a safe speed, continuously monitor the dynamics of dangerous traffic participants, and wait for an opportunity to avoid danger. The system combines the vehicle's dynamic parameters (such as steering system response speed, braking system performance), road and lane constraints (such as lane width, lane line type), and traffic regulations to generate a smooth and safe avoidance trajectory. The trajectory planning cycle is ≤0.1s, ensuring the real-time performance and stability of the trajectory; the trajectory planning results are then converted into specific vehicle control commands.

[0036] Step 5: Convert the avoidance trajectory and control commands into execution signals to control the vehicle to perform avoidance operations. At the same time, collect the actual driving status of the vehicle and feed it back to the decision planning module so as to adjust the control commands in a timely manner.

[0037] This invention provides a system and method for intelligent driving vehicles to deal with dangerous traffic participants. It mainly addresses the shortcomings of existing technologies, such as the lack of ability to identify dangerous driving behaviors, mark and continuously track dangerous traffic participants, the inability to achieve proactive preventive avoidance, insufficient safety redundancy, the single avoidance strategy that cannot be adjusted according to the degree of danger, the inability to balance safety and traffic efficiency, the tendency to cause secondary accidents or untimely avoidance, and the inability to dynamically adjust the avoidance trajectory according to changes in the behavior of dangerous traffic participants and the driving status of the vehicle, resulting in poor adaptability and difficulty in meeting the needs of intelligent driving in multiple scenarios.

[0038] In summary, this invention provides a system and method for intelligent driving vehicles to deal with dangerous traffic participants. This technical solution breaks through the passive protection mode of existing technologies that rely solely on emergency braking, constructing a closed-loop active risk avoidance system of "identification-marking-classification-planning-execution." This system enables early identification, continuous tracking, and preventative avoidance of dangerous traffic participants, significantly reducing the probability of collisions and improving the driving safety of intelligent driving vehicles. By employing multi-source sensor fusion acquisition and time-series trajectory analysis, it can accurately identify various dangerous driving behaviors. Combined with dynamic hazard classification and distance threshold settings, it makes risk avoidance decisions more targeted and flexible, adapting to the needs of different traffic scenarios and varying levels of danger, such as highways, urban roads, and expressways. It offers two core hazard avoidance strategies: automatic avoidance and rapid departure, balancing driving safety and traffic efficiency. This avoids secondary accidents caused by single emergency braking, while smooth acceleration, deceleration, and steering control enhance the comfort and experience for passengers. The system architecture is simple with clearly defined modules, making it compatible with L2 to L4 level autonomous driving vehicles, demonstrating strong versatility. Furthermore, the hazardous behavior data stored in the storage module can be used to optimize subsequent hazardous behavior judgment models, further improving the system's recognition accuracy and hazard avoidance performance. The dynamic marking and continuous tracking functions of the hazard identification and marking module can quickly respond to changes in the behavior of hazardous traffic participants. When a hazardous vehicle re-enters the perception range, its hazardous attributes can be quickly identified, improving hazard avoidance response efficiency.

[0039] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A system for intelligent driving vehicles to respond to dangerous road users, characterized in that, The system includes: The perception and acquisition module is used to collect image information, dynamic information, location coordinates and driving trajectory of surrounding vehicles. The collected data is preprocessed to form a standardized traffic environment information data package, which is then transmitted to the hazard identification and marking module. The hazard identification and marking module, connected to the perception and acquisition module, is used to extract the behavioral characteristics of surrounding vehicles based on traffic environment information data packets. The behavioral characteristics are then analyzed using rule models and time-series trajectories to identify dangerous driving behaviors and mark the corresponding vehicles as dangerous traffic participants. The decision planning module, connected to the hazard identification and marking module, is used to classify hazard levels based on the type, frequency, distance from the vehicle, and relative speed of dangerous traffic participants. For different hazard levels, corresponding warning distance thresholds and safe distance thresholds are dynamically set. When a dangerous traffic participant is less than the warning distance threshold, a risk avoidance strategy is selected based on the current traffic environment, a risk avoidance trajectory is generated, and vehicle control commands are generated based on the risk avoidance trajectory. The vehicle control module, connected to the decision-making and planning module, is used to convert the avoidance trajectory and control commands into execution signals, control the vehicle to perform avoidance operations, and collect the actual driving status of the vehicle and feed it back to the decision-making and planning module so that the control commands can be adjusted in a timely manner.

2. The system for intelligent driving vehicles to respond to dangerous traffic participants according to claim 1, characterized in that, The system also includes: The storage module is connected to the sensing and acquisition module, the hazard identification and marking module, the decision planning module, and the vehicle control module, respectively. It is used to store various types of data during the operation of the system, providing support for subsequent optimization and traceability.

3. The system for intelligent driving vehicles to respond to dangerous traffic participants according to claim 1, characterized in that, The sensing and acquisition module includes: The camera is installed on the vehicle to cover a field of view of 180° in front of the vehicle, 60° behind the vehicle, and 90° on each side, and is used to collect image information of the shape of surrounding vehicles, turn signal status, and lane line position. Millimeter-wave radar, installed on a vehicle, emits electromagnetic waves at a frequency of not less than 10Hz and receives reflected signals to detect dynamic information such as distance, speed, and angle of surrounding vehicles within a 200-meter range. LiDAR, installed on vehicles, emits millions of laser pulses per second to build a 360° high-precision point cloud map with a detection range of 150 meters, used to accurately obtain the position coordinates and driving trajectory of surrounding vehicles; The data preprocessing unit is connected to the camera, millimeter-wave radar, and lidar, respectively. It is used to denoise, synchronize, and fuse the collected raw data, remove invalid data and measurement errors, and output standardized traffic environment information data packets, which are then transmitted to the hazard identification and marking module.

4. A method for using a system for responding to dangerous traffic participants with an intelligent driving vehicle as described in any one of claims 1-3, characterized in that, The method includes: Step 1: Acquire real-time image information, dynamic information, location coordinates and driving trajectories of surrounding vehicles; preprocess the collected data to form a standardized traffic environment information data package. Step 2: Based on the traffic environment information data package, and using rule models and time-series trajectory analysis, identify dangerous driving behaviors and mark the corresponding vehicles as dangerous traffic participants. Step 3: Classify the danger level according to the dangerous behavior type, frequency, distance from the vehicle, and relative speed of dangerous traffic participants, and dynamically set the corresponding warning distance threshold; Step 4: When the dangerous traffic participant is less than the warning distance threshold, select a risk avoidance strategy based on the current traffic environment, generate a risk avoidance trajectory, and generate vehicle control instructions based on the risk avoidance trajectory. Step 5: Convert the avoidance trajectory and control commands into execution signals to control the vehicle to perform avoidance operations. At the same time, collect the actual driving status of the vehicle and feed it back to the decision planning module so as to adjust the control commands in a timely manner.

5. The method for an intelligent driving vehicle to deal with dangerous traffic participants according to claim 4, characterized in that, Step 2 includes: Step 21: Receive the traffic environment information data packet output by the perception and acquisition module, and extract the behavioral characteristics of surrounding vehicles, including lane change frequency, braking status, and duration of crossing the line. Step 22: Through rule modeling and time-series trajectory analysis, identify whether there is at least one dangerous driving behavior in the surrounding vehicles; Step 23: If dangerous driving behavior exists, mark the vehicle as a dangerous traffic participant, assign it a unique identification ID, establish and dynamically update the list of dangerous targets, and synchronize the ID, behavioral characteristics, danger level, and location information of the dangerous targets to the decision planning module and the storage module.

6. The method for an intelligent driving vehicle to deal with dangerous traffic participants according to claim 5, characterized in that, The dangerous driving behaviors include: multiple extreme lane changes within a unit of time, forced lane changes without turn signals; significant deceleration and frequent emergency braking in a short period of time; long-term driving across lane lines or straddling lane lines; running red lights, driving against traffic, illegal turns, and not driving in the designated lane.

7. The method for an intelligent driving vehicle to deal with dangerous traffic participants according to claim 5, characterized in that, Step 22 includes: Step 221: Establish a rule model based on the temporal characteristics, threshold characteristics, and violation characteristics of dangerous driving behavior; Step 222: Initially determine whether the vehicle has dangerous behavior using the decision rules in the rule model; Step 223: By continuously tracking, statistically analyzing, comparing, and predicting the vehicle's position, speed, heading, steering status, braking status, and lane position over multiple consecutive frames, the frequency of dangerous behaviors of the vehicle is identified. Step 224: The rule model determines whether a vehicle is engaging in dangerous driving behavior based on the frequency of such behavior.

8. The method for an intelligent driving vehicle to deal with dangerous traffic participants according to claim 7, characterized in that, The 221 includes: Step 2211: Define a library of dangerous behavior types, including multiple extreme lane changes, lane changes without turn signals, frequent emergency braking, prolonged crossing of lane lines, running red lights, driving against traffic, and illegal turns; Step 2212: Set quantitative judgment conditions for each dangerous behavior. The quantitative judgment conditions include lane change frequency, braking intensity, line crossing duration, speed change, steering angular velocity, and turn signal activation duration. Step 2213: Using a fixed-length time window, construct a time-series behavior window to continuously statistically analyze vehicle behavior; Step 2214: Establish a graded judgment logic, wherein the graded judgment is specifically divided into single dangerous behavior, multiple dangerous behavior, and continuous dangerous behavior, and the grade increases in sequence. Step 2215: Add legality elements to determine whether there is a traffic violation. The legality elements include maps, traffic lights, guide lanes, and prohibition markings. Step 2216: Set the model output as: Hazardous Traffic Participant ID + Hazard Level + Continuous Tracking Marker.

9. A method for an intelligent driving vehicle to deal with dangerous traffic participants according to claim 7, characterized in that, The decision rules in the rule model include: The rule for multiple extreme lane changes is that if there are no less than 2 lane changes within a unit time window and the turning angular velocity is no less than 15° / s, it is judged as multiple extreme lane changes. The rule for forced lane change without turn signal is that if a lane change is made without using a turn signal or the turn signal is on for less than 3 seconds, it is considered a forced lane change without turn signal. The rules for short-duration, significant deceleration / frequent emergency braking are as follows: the deceleration duration is no more than 2 seconds, the speed reduction is no less than 30 km / h, the number of emergency braking events is no less than 2 times within 1 minute, and the braking acceleration is no greater than -5 m / s². 2 If so, it is judged as frequent sudden braking and malicious speed reduction; The rule for driving across the lane line for an extended period of time is as follows: if the vehicle is driving over the lane line for a period of time for no less than 5 seconds and the center of the vehicle is on the lane line for an extended period of time, it will be judged as a violation of driving over the lane line for an extended period of time. Running a red light, driving against traffic, and making illegal turns are considered serious violations of driving rules. For example, if a vehicle crosses the stop line during a red light, drives in the wrong lane, turns at a no-turn intersection, or fails to follow the designated lane, it will be considered a serious violation of driving rules.

10. A method for an intelligent driving vehicle to deal with dangerous traffic participants according to claim 9, characterized in that, Step 223 includes: Step 2231: Continuously collect trajectory data of the target vehicle at a frequency of 10Hz. The trajectory data includes position, speed, heading angle, steering angular velocity, brake light status, turn signal status, and lane position. Step 2232: Use a first-in-first-out time queue to store the most recent N consecutive trajectories; Step 2233: Perform behavior slicing on the continuous trajectory and identify lane change behavior segments, braking behavior segments, line crossing behavior segments, and steering behavior segments based on the collected trajectory data; Step 2234: Extract the behavioral features of each behavior segment, including duration, speed change, steering angular velocity, braking intensity, turn signal duration, line crossing duration, and number of lane changes. Step 2235: Statistically analyze the behavioral characteristics and match them with the judgment rules to obtain the judgment result of dangerous driving behavior; Step 2236: If any of the judgment rules are met, mark the participant as a dangerous traffic participant and assign a unique ID. If there is no dangerous behavior for N consecutive frames, remove the mark.

Images