Triangle warning board dynamic early warning distance adjusting system and method of AI road condition recognition

Abstract

The application relates to the technical field of vehicle safety warning, and discloses an AI road condition recognition triangular warning board dynamic early warning distance adjusting system and method, which comprises a main control system and a beacon device, the main control system comprises a benchmark establishing module, a path mapping module, a dynamic calculation module and a first wireless communication module; the beacon device comprises a second wireless communication module, a data acquisition module and a prompt module. The beacon device enables an operator to safely walk on an emergency lane and records a track through the data acquisition module, the main control system intelligently maps the safe track to a driving lane taking a real-time lane position of an accident vehicle as a benchmark through the path mapping module, generates a corresponding virtual driving path, obtains an accurate curve length through real-time integration of the virtual path through the dynamic calculation module, calculates a dynamic early warning distance, and prompts the operator to perform a placing operation through the prompt module when the two are matched, so that the safety guarantee for the operator is strengthened.

Description

Technical Field

[0001] This invention relates to the field of vehicle safety warning technology, and more specifically, to an AI-based road condition recognition system and method for adjusting the dynamic warning distance of a triangular warning sign. Background Technology

[0002] In the field of road traffic safety, correctly placing a warning triangle at a prescribed distance behind a vehicle after it breaks down or is involved in an accident within the driving lane is a crucial measure to warn oncoming vehicles and prevent secondary collisions. Currently, determining this distance mainly relies on the driver's visual estimation, which is subject to significant subjective errors and safety risks. To improve accuracy, the industry has introduced sensor-based technical solutions, but their evolution has revealed a series of profound and unresolved technical contradictions.

[0003] First, the initial automated solutions suffered from a fundamental flaw by ignoring road morphology. These solutions primarily involved two technical approaches: one was to identify warning triangles and estimate the straight-line distance using onboard vision sensors; the other was to use satellite positioning modules to calculate the straight-line distance between the vehicle and the warning triangle by obtaining their coordinates separately. Both approaches implicitly assumed an idealized view of the road as flat and straight. However, on actual road slopes, the pitch angle of the vision sensors introduces significant ranging errors; on curves, the complete obstruction of vision causes the vision system to malfunction, and the straight-line distance calculated by satellite positioning is far shorter than the actual braking path required by vehicles behind, ultimately outputting a safety distance that is severely insufficient in terms of dynamics, thus creating a new safety hazard.

[0004] To address the shortcomings of straight-line distance measurement, a driving path integration theory based on vehicle dynamics was proposed, which states that safe distances should be calculated cumulatively along the lane centerline. One intuitive approach to implementing this theory is to have personnel walk along the lane centerline with the equipment to conduct on-site measurements. However, this not only violates mandatory regulations requiring personnel to walk in safe areas such as emergency lanes, shoulders, or medians when placing warning triangles after an accident, but also exposes the operator to extreme danger from high-speed traffic.

[0005] However, if personnel are instructed to detour along safe areas such as emergency lanes, two problems arise. First, the detour itself adds extra distance to the calculation. Second, the direction and curvature of emergency lanes are not entirely consistent with those of driving lanes, especially at critical sections such as curves and ramps, where the path lengths differ significantly. If the system measures based on this walking path, the calculation results cannot accurately reflect the dangers of the driving lanes. If it measures based on the driving lanes, it cannot provide compliant guidance for personnel to safely reach the designated placement point. Therefore, this contradiction becomes a fundamental obstacle to the practical application of the technology. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution:

[0007] The first aspect is the AI ​​road condition recognition triangular warning sign dynamic warning distance adjustment system, which includes a main control system installed in the vehicle and a beacon device for operators to carry in their hands. The main control system includes: a reference establishment module, a path mapping module, a dynamic calculation module and a first wireless communication module.

[0008] The beacon device includes a second wireless communication module, a data acquisition module, and a notification module;

[0009] The baseline establishment module is used to respond to the warning activation command, obtain the real-time lane position of the vehicle, and establish a path calculation baseline based on it.

[0010] The data acquisition module is used to continuously collect safe walking path data, including displacement, heading and position information, when the operator moves along the safe area with the beacon device;

[0011] The path mapping module is used to pre-set or acquire AI road condition topology models in real time, and map the received safe walking path data to the center line of the driving lane determined based on the real-time lane position to generate a virtual driving path trajectory.

[0012] The dynamic calculation module is used to calculate the dynamic warning distance based on the real-time vehicle status and environmental parameters, and to obtain the cumulative path length by integrating the virtual driving path trajectory in real time. When the cumulative path length reaches the dynamic warning distance, an arrival command is generated.

[0013] The first wireless communication module establishes a connection with the second wireless communication module to transmit safe walking path data and arrival instructions;

[0014] The alert module is used to respond to arrival instructions and issue an alert signal to indicate that the current location of the beacon device corresponds to the recommended placement point of the warning triangle on the target lane.

[0015] Furthermore, the benchmark establishment module is specifically used to establish a two-dimensional path calculation coordinate system in the horizontal projection plane with the real-time lane position of the vehicle as the origin and the tangent direction of the road where the position is located as the benchmark direction, which serves as the path calculation benchmark.

[0016] Furthermore, the dynamic calculation module performs real-time integration of the virtual driving path trajectory to obtain the cumulative path length using the following methods:

[0017] Sequentially obtain adjacent trajectory points on the virtual driving path trajectory;

[0018] For each pair of adjacent trajectory points forming a differential line segment, based on the road curvature and slope information at the location of the line segment, its length in the actual three-dimensional road space is calculated as the corrected line segment length.

[0019] The cumulative path length is obtained by summing up the lengths of all corrected line segments from the path calculation reference point to the current position.

[0020] Furthermore, the AI ​​traffic topology model includes the geometric topological relationships of roads;

[0021] Geometric topology defines at least the spatial parallelism and width relationships between the safety zone and each driving lane, as well as between each driving lane and each other.

[0022] Furthermore, the path mapping module maps the safe walking path data to the centerline of the associated lane based on the real-time lane position using a translation mapping algorithm, according to the geometric topology, in order to generate a virtual driving path trajectory.

[0023] Furthermore, the translation mapping algorithm is as follows:

[0024] Based on the fixed lateral offset between the safe area defined in the geometric topology and the centerline of the determined driving lane, each coordinate point in the safe driving path data is translated by a fixed lateral offset in a direction perpendicular to the reference direction, thereby obtaining the corresponding point on the virtual driving path trajectory.

[0025] Furthermore, the data acquisition module specifically includes an odometer, an inertial measurement unit, a global positioning system receiver, and a magnetometer;

[0026] Odometers and inertial measurement units are used to measure the displacement and heading changes of the beacon device, GPS receivers are used to obtain the position coordinates of the beacon device, and magnetometers are used to provide heading angle correction references.

[0027] Furthermore, the dynamic calculation module calculates the vehicle status and environmental parameters on which the dynamic warning distance is based, specifically including the road slope obtained by the vehicle attitude sensor, the road surface adhesion coefficient obtained through the vehicle bus or wireless network, and the legal speed limit value of the driving lane determined based on the real-time lane position extracted from the AI ​​road condition topology model.

[0028] Furthermore, the prompting signal emitted by the prompting module is a combination of sound and light signals, including intermittent buzzing at a specific frequency and flashing of a bright LED.

[0029] Secondly, this invention discloses a method for adjusting the dynamic warning distance of a triangular warning sign based on AI road condition recognition, applied to the aforementioned AI road condition recognition system for adjusting the dynamic warning distance of a triangular warning sign. The method includes:

[0030] In response to the warning activation command, the system obtains the vehicle's real-time lane position through the main control system and calculates the baseline based on this path;

[0031] Data on the safe walking path of operators as they move along the safe zone is collected using beacon devices.

[0032] By using the AI ​​road condition topology model pre-set or acquired in real time within the main control system, the safe walking path data is mapped in real time to the center line of the driving lane determined based on the real-time lane position, thereby generating a virtual driving path trajectory.

[0033] The dynamic warning distance is calculated based on the real-time vehicle status and environmental parameters, and the cumulative path length is obtained by real-time integration of the virtual driving path trajectory.

[0034] When the cumulative path length reaches the dynamic warning distance, an arrival command is sent to the beacon device.

[0035] A warning signal is sent through the beacon device to indicate that the current location is the recommended placement point for the warning triangle on the designated driving lane.

[0036] Compared with related technologies, the present invention has the following beneficial effects:

[0037] This invention establishes a collaborative architecture between a main control system and a handheld beacon device, physically and logically separating safe path acquisition from hazardous path calculation. The beacon device allows operators to safely walk in the emergency lane and record their trajectory, while the main control system uses an AI-powered road topology model to intelligently map the safe trajectory onto the driving lane based on the real-time lane position of the accident vehicle, generating a corresponding virtual driving path. This mapping is the core of this invention, enabling all subsequent calculations to be based on real lane geometry. Subsequently, the system obtains the precise curve length by performing real-time integration on this virtual path and simultaneously integrates real-time vehicle and environmental parameters to calculate the dynamic warning distance. When the two match, the system prompts the operator to place the warning triangle. Throughout the entire process, the operator can accurately deploy the warning triangle without taking any risks, enhancing the operator's safety.

[0038] The translation mapping algorithm used in this invention is defined as a vertical translation transformation based on a fixed lateral offset, which eliminates the arbitrariness or cumulative error in the mapping process. This ensures that the final virtual driving path can strictly maintain geometric consistency with the lane centerline, regardless of whether the pedestrian walking path is a straight line or a complex curve. This guarantees that the cumulative path length calculated by the system can accurately reflect the actual trajectory length of the vehicle on complex roads such as winding mountain roads or interchange ramps.

[0039] In this invention, the AI ​​road condition topology model is concretized into a composite model that simultaneously includes static geometric topological relationships and dynamic road condition information. When the system performs path mapping and safety decisions, it can not only perform geometric calculations based on the physical layout of lanes, but also actively integrate dynamic risk elements such as temporary construction barriers, accident obstacles, or slippery road surfaces. This allows the final warning area to proactively avoid secondary risk points or provide extended protection for vehicles that change lanes to avoid dynamic obstacles. This elevates the targeting of warnings from general scenarios to personalized instantaneous scenarios, enhancing the system's robustness and protective effectiveness under complex and ever-changing real road conditions. Attached Figure Description

[0040] Figure 1 A flowchart illustrating the steps of the AI-based road condition recognition method for dynamically adjusting the warning distance of a triangular warning sign provided by this invention.

[0041] Figure 2 The data processing flowchart is shown in the AI ​​road condition recognition triangular warning sign dynamic warning distance adjustment system provided by this invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example 1

[0044] Please see Figure 1 As shown, this embodiment provides an AI-powered road condition recognition-based dynamic warning distance adjustment system for warning triangles. This system is activated when a vehicle malfunctions or is involved in an accident, and is designed to guide operators to safely and accurately place the warning triangle at a designated distance. The system includes a main control system (subsystem) integrated into the vehicle and a handheld beacon device for the operator.

[0045] Specifically, after an accident occurs, the main control system calculates and analyzes the dynamic warning distance setting and monitors the distance between the beacon device and the placement point in real time. When the operator arrives at the placement point, it sends a prompting command to the beacon device. The beacon device can be used as a triangular warning sign in this embodiment, mainly to respond to the prompting command of the main control system and send a prompting signal to indicate that the current position of the beacon device corresponds to the placement point.

[0046] The main control system is typically integrated into the vehicle, connected to the vehicle's CAN bus, continuously powered, and in a low-power monitoring state. In this embodiment, the main control system includes a baseline establishment module, a path mapping module, a dynamic calculation module, and a first wireless communication module.

[0047] The beacon device is a standalone unit, typically stored in the trunk of a vehicle or in a designated location, and is in a powered-off or deep sleep state to conserve power. In this embodiment, the beacon device includes a second wireless communication module, a data acquisition module, and a notification module.

[0048] The following sections will provide a detailed description of each module of the main control system and beacon device.

[0049] The baseline establishment module is used to respond to the warning activation command, obtain the vehicle's real-time lane position, and establish a path calculation baseline based on it.

[0050] In this embodiment, when the vehicle stops due to a malfunction or accident, the operator can send a warning start command to the main control system via a dedicated "emergency warning" button on the in-vehicle infotainment system touchscreen, or by pressing an independent physical emergency button inside the vehicle (which can also be triggered by an airbag deployment signal or a vehicle fault code). Subsequently, the reference establishment module in the main control system is activated, and its primary task is to acquire and determine the real-time lane position.

[0051] In this embodiment, the real-time lane position specifically refers to the precise lane assignment information of a vehicle on the cross section of the road. It is not a simple latitude and longitude coordinate, but a lane number identifier with semantics. For example, on a three-lane highway from west to east, the lanes from left to right (from the driver's perspective) may be defined by the model as lane 1 (the leftmost fast lane), lane 2 (the middle lane), lane 3 (the rightmost slow lane), and emergency lane.

[0052] Specifically, since high-precision lane-level positioning services are already mature, real-time lane location can be output in real time through the vehicle's built-in GNSS receiver (such as a receiver that supports BeiDou-3 and GPS-L5 bands) and combined with a high-precision map. Therefore, in this embodiment, it is regarded as a basic service that can be directly invoked.

[0053] In some other embodiments, high-precision positioning services can be omitted, relying primarily on visual perception combined with ordinary GPS positioning. By visually determining the vehicle's position relative to lane lines, and then matching it with a rough road segment using ordinary GPS, combined with lane number information from a navigation map (non-high-precision), the real-time lane position can be inferred with a relatively high probability, although the accuracy and reliability will be lower than the preferred solution.

[0054] After determining the semantic real-time lane position, the benchmark establishment module needs to establish a unified mathematical benchmark for all subsequent geometric calculations, namely the path calculation benchmark, which is also a two-dimensional plane coordinate system.

[0055] For example, the reference establishment module is specifically used to establish a two-dimensional path calculation coordinate system in the horizontal projection plane with the real-time lane position of the vehicle as the origin and the tangent direction of the road where the position is located as the reference direction, as the path calculation reference.

[0056] Here we will first introduce the definition of this coordinate system:

[0057] Origin point O: Taken as the centerline of the lane corresponding to the vehicle's real-time lane position, which is the point closest to the vehicle's centroid projection. This point is calculated using high-precision maps or perception results and serves as the starting point (0-meter point) for all path length integrals.

[0058] Reference direction (positive X-axis direction): Take the tangent direction of the road centerline passing through the origin O, pointing to the original direction of vehicle travel (i.e. the opposite direction of oncoming traffic). This direction can be directly obtained from the lane centerline geometry data of that point in the high-precision map.

[0059] Positive Y-axis direction: In the horizontal plane, perpendicular to the X-axis and pointing to the left in the direction of vehicle movement (following the conventions of a right-handed coordinate system or a conventional mathematical plane coordinate system).

[0060] Plane: This coordinate system is a horizontal projection plane, ignoring the impact of road vertical slope variations on planar distance calculations. The effect of slope will be considered separately as a dynamic factor in subsequent dynamic warning distance calculations. This aligns with the conventional principle of simplifying vehicle kinematics modeling and effectively balances computational complexity and accuracy.

[0061] Specifically, assuming that the coordinates (E, N) of the origin O in the UTM coordinate system are determined by high-precision map matching, the azimuth angle of the lane centerline at that point is obtained from the map, such as the angle of clockwise rotation from due north.

[0062] Therefore, the origin of the established two-dimensional path calculation coordinate system is (0, 0). The coordinates (x, y) of any point P in space in this coordinate system can be obtained through a two-dimensional coordinate transformation: first, transform the UTM coordinates (E) of point P... p N p Subtract the UTM coordinates of the origin O, and then perform a rotation transformation with a rotation angle of θ.

[0063] Although the principles of transformation are described herein, those skilled in the art are well aware of how to program such basic coordinate system transformations, so no specific transformation matrix formula needs to be given in the embodiments.

[0064] Thus, in this embodiment, the three-dimensional spatial problem is reduced to a two-dimensional planar problem, focusing on the core factors that determine the safe distance and treating the length of the planar driving path as an independent factor, thereby significantly reducing unnecessary computation.

[0065] After the baseline establishment module successfully creates the path calculation baseline, the system will enter the core human-computer interaction and data acquisition stage. The goal of this step is to accurately and continuously record the complete trajectory of the operator holding a beacon device, starting from a safe starting point next to the vehicle, and walking along the emergency lane or shoulder or other safe areas towards oncoming traffic.

[0066] The beacon device includes a second wireless communication module, a data acquisition module, and a prompting module. The data acquisition module is used to continuously collect safe walking path data, including displacement, heading, and location information, when the operator carries the beacon device and moves along a safe area.

[0067] The beacon device can be built into the warning triangle or it can be the warning triangle itself. When collecting safe walking path data, the operator can remove the beacon device from the vehicle. This device is usually designed as an ergonomic cylindrical or handheld terminal, integrating the necessary sensors, processor, battery, and alert module.

[0068] Then the operator can press the power button on the beacon device to power it on. After the device is started, its internal second wireless communication module (which can be Bluetooth Low Energy or a proprietary radio frequency module) will automatically attempt to establish a connection with the main control system (first wireless communication module) in the vehicle.

[0069] After a successful connection, the main control system sends the origin coordinates and reference direction of the newly established path calculation benchmark to the beacon device. Simultaneously, the main control system sends a "start acquisition" command to the beacon device, and the beacon device's data acquisition module immediately enters high-frequency data acquisition mode.

[0070] The data acquisition module is a multi-sensor fusion system. For example, the data acquisition module specifically includes an odometer, an inertial measurement unit, a global positioning system receiver, and a magnetometer. The odometer and inertial measurement unit are used to measure the displacement and heading changes of the beacon device, the global positioning system receiver is used to obtain the position coordinates of the beacon device, and the magnetometer is used to provide a heading angle correction reference.

[0071] Specifically, an odometer is typically a microelectromechanical system (MEMS) device that estimates linear displacement by detecting the movement of its internal microstructures. In this embodiment, it is responsible for providing a high-frequency displacement increment signal that has accumulated errors, thereby reflecting the displacement changes of the beacon device in real time.

[0072] The inertial measurement unit can be a MEMS chipset that integrates a three-axis accelerometer and a three-axis gyroscope. The accelerometer is responsible for measuring the proportion of the carrier (i.e., the beacon device) in three directions, while the gyroscope is responsible for measuring the angular velocity of the carrier around the three axes. Finally, after inertial navigation calculation, it can provide the attitude (pitch, roll, heading) changes and velocity information of the beacon device.

[0073] The Global Positioning System receiver can be a civilian-grade GNSS chip that supports GPS, GLONASS, or BeiDou. It is responsible for outputting the antenna position and latitude and longitude coordinates of the beacon device. In open environments, the absolute accuracy is controlled at approximately 2-5 meters, and it provides a UTC timestamp.

[0074] A magnetometer can be a three-axis magnetic sensor, mainly used to measure the components of the Earth's magnetic field vector in the carrier coordinate system, and to assist in correcting the heading angle, especially during long-term travel and when gyroscope errors accumulate.

[0075] Specific methods for generating safe walking path data that includes displacement, heading, and position information include:

[0076] First, determine the heading angle (by combining the gyroscope and magnetometer). The heading angle can be obtained by integrating the gyroscope in the IMU, but there is a drift error that increases over time. The heading angle provided by the magnetometer has no cumulative error, but it is easily affected by surrounding ferromagnetic materials (such as guardrails and vehicles).

[0077] Therefore, a fusion method can be adopted, using the angular velocity ω (around the vertical axis) output by the gyroscope as the main prediction input for the filter. At the same time, when it is determined that the current environmental magnetic field strength and tilt angle conform to the characteristics of the geomagnetic field (which can be determined by a preset threshold range, such as the total magnetic field strength being between 45-55 μT and the tilt angle differing from the local geomagnetic tilt angle by less than 5 degrees), the heading angle calculated by the magnetometer is used as the observation value input to the filter to correct the integral error of the gyroscope.

[0078] The preset threshold settings, namely the threshold ranges for geomagnetic field strength and tilt angle, are derived from the calculation values ​​of the device's approximate current geographical location (provided by GNSS) by the International Geomagnetic Reference Field Model, with a certain interference tolerance (usually ±5μT). These thresholds are not fixed values, but can be adaptively set according to the location.

[0079] Next, position and displacement are determined (by fusing GNSS, odometry, and inertial navigation). GNSS provides absolute position but has a low update rate (typically 1-10Hz) and is ineffective when obstructed; odometry and inertial navigation provide high-frequency relative displacement but errors accumulate.

[0080] The method for fusing these signals is as follows: when the GNSS signal is valid and the Positioning Accuracy Factor (PDOP) is less than a threshold Th (e.g., Th=4, this threshold is set based on experience, PDOP<4 usually indicates good satellite geometry and high positioning accuracy), the filter will use the latitude and longitude coordinates of the GNSS (coordinates after transformation to the path calculation reference coordinate system) as the high-weighted observation value.

[0081] Finally, the real-time output shows the two-dimensional coordinate sequence (x, y) of the beacon device in the path calculation reference coordinate system. i y i ), where i represents the sampling time. This sequence {(x i y i This data represents a continuous and smooth safe walking path, which can then be sent back to the main control system in real time via a wireless communication link.

[0082] Through the above detailed description, those skilled in the art can clearly understand how to construct and operate a beacon device to obtain a precise, coordinate-system-unified, safe walking path for subsequent intelligent mapping.

[0083] When the path mapping module of the main control system receives the safe walking path data (i.e., the two-dimensional coordinate sequence) from the beacon device in real time, its core task is to intelligently convert this safe walking path into a dangerous driving path for vehicles, that is, to generate a virtual driving path trajectory. This conversion is the key to realizing the core idea of ​​this invention.

[0084] For example, the path mapping module can pre-set or acquire an AI road condition topology model in real time, and map the received safe walking path data onto the center line of the driving lane determined based on the real-time lane position to generate a virtual driving path trajectory.

[0085] Furthermore, the AI ​​traffic topology model includes the geometric topology of the road; the geometric topology defines at least the spatial parallelism and width relationships between the safety zone and each driving lane, as well as between each driving lane and each other.

[0086] Specifically, the path mapping module first calls a preset or real-time acquired AI traffic topology model. In this embodiment, the model is a structured database containing two main categories of information:

[0087] The first major category of information is the road geometric topology (static core), which not only precisely defines the spatial attributes of all lanes in the current road segment—for each lane, the model stores the geometric coordinate sequence of its centerline—but more importantly, it also defines the topological relationships between different lane elements, including:

[0088] Parallelism and width relationship: For example, the center line of the emergency lane is parallel to the center line of the adjacent driving lane 1, and the lateral distance between them is constant; the distance between the center lines of driving lane 1 and driving lane 2.

[0089] Lane connection relationship: This usually describes the connection logic of lanes at ramps, diverging points, and merging points.

[0090] It should be noted that this basic data comes from commercial or self-developed high-precision maps and is loaded into the main control system during system initialization or periodic updates.

[0091] The second major category of information is dynamic road condition information (AI-enhanced), which is information that can be sensed and fused in real time through vehicle sensors (such as cameras and radar) or vehicle-to-everything (V2X) networks. For example:

[0092] Temporary roadblock locations: Cones and construction areas identified and located by the vehicle's forward-facing camera, and their locations are marked in the topology model.

[0093] Slippery road surface areas: Locations of low adhesion events reported by the vehicle's ESP system, or areas of icing or water accumulation warnings broadcast by the roadside unit.

[0094] Abnormal traffic flow: The origin of abnormal slow traffic or congestion inferred based on V2V information.

[0095] It should be noted that this data is generated in real time by the vehicle's environmental perception AI algorithm, or received from the roadside facility (RSU) via the C-V2X module and dynamically updated into the AI ​​road condition topology model.

[0096] The next step is to determine the mapping target, which is the conversion from the real-time lane position to the associated driving lane. The path mapping module will determine the set of associated driving lanes that need to be warned based on the real-time lane position provided by the baseline module (here, we assume it is the second lane) and the AI ​​traffic topology model.

[0097] Basic rule: By default, the associated driving lanes must first include the real-time lane location (lane 2) where the accident vehicle was located, because this is the most direct source of danger.

[0098] AI extension rules: Dynamic information and topological relationships in the model are used for intelligent extension: If the accident vehicle partially intrudes into an adjacent lane (which can be determined by visual perception), the adjacent lane (lane 1 or 3) is also added to the association set.

[0099] If the topology indicates that the current location is a highway ramp entrance and the accident vehicle is stopped near the guide line, then the relevant lanes on the main road and the ramp may be associated.

[0100] If dynamic information indicates that a temporary roadblock ahead is forcing traffic to change lanes, the associated set may extend upstream to the affected lanes.

[0101] This decision-making process can be implemented by a series of rule-based logics or a lightweight neural network classifier, the output of which is a list of explicit target lane IDs.

[0102] Then, the translation mapping algorithm is executed. For example, the path mapping module maps the safe walking path data to the center line of the associated driving lane based on the real-time lane position using the translation mapping algorithm according to the geometric topology, so as to generate a virtual driving path trajectory.

[0103] For each coordinate point P in the safe walking path data i (x i ,y i The path mapping module executes a translation mapping algorithm to map the path onto the centerline of each target lane, thereby generating one or more parallel virtual trajectories.

[0104] Furthermore, the translation mapping algorithm is as follows: based on the fixed lateral offset between the safe area defined in the geometric topology and the centerline of the determined driving lane, each coordinate point in the safe driving path data is translated by a fixed lateral offset in a direction perpendicular to the reference direction, thereby obtaining the corresponding point on the virtual driving path trajectory.

[0105] Specifically, the first step is to calculate the fixed lateral offset d, which can be determined by finding the lateral distance between the centerline of the safe zone (the emergency lane or shoulder for operators) and the centerline of the target lane based on the topological relationship. This is a constant that is explicitly defined in the model.

[0106] Assuming the operator is walking in the emergency lane, their path data point P i This corresponds to the projection point on the center line of the emergency lane. The target lane is driving lane 2. According to the topology model, the fixed lateral offset d from the center line of the emergency lane to the center line of driving lane 2 is 7.25 meters (assuming the emergency lane is 3 meters wide, driving lane 2 is 3.75 meters wide, and there is a 0.5-meter shoulder in the middle).

[0107] Next, a translation transformation is performed. Since the direction of translation is crucial, it is not a simple translation along the Y-axis, but a translation along the reference direction (X-axis) that is perpendicular to the path calculation reference, i.e., a translation along the Y-axis.

[0108] Therefore, for point P i (x i ,y i ), its mapping point P' on target lane 2 i (x' i ,y'i The computational logic of ) can be described as follows:

[0109] x' i =x i (The X-coordinate remains unchanged because it is along the parallel direction);

[0110] y' i =y i +d (Add a fixed offset to the Y coordinate; if the target lane is to the left of the walking path, the offset is negative).

[0111] Perform this operation on all points along the safe path to obtain a new set of points {(x' i ,y' i This constitutes the virtual driving path trajectory on the center line of lane 2. If there are multiple associated target lanes, the corresponding 'd' values ​​can be used for translation to generate multiple parallel trajectories.

[0112] The path conversion using a translation mapping algorithm with a fixed offset is not only computationally efficient (with extremely low algorithm complexity, involving only addition operations) and capable of processing high-frequency path point data in real time, but also accurate. As long as the geometric relationship of the high-precision map is accurate, the mapped virtual trajectory can perfectly match the shape of the actual lane (whether it is a straight line, a curve, or a transition curve), perfectly simulating the driving lane path parallel to the safe walking path. This intuitively solves the problem of inconsistency between the distance traveled by a person and the path traveled by a vehicle in the background technology.

[0113] After generating the virtual driving path trajectory, the dynamic calculation module is responsible for completing two core calculations. The dynamic calculation module is used to calculate the dynamic warning distance based on the real-time acquired vehicle status and environmental parameters, and to obtain the cumulative path length by integrating the virtual driving path trajectory in real time. When the cumulative path length reaches the dynamic warning distance, an arrival command is generated.

[0114] The dynamic warning distance is calculated based on the real-time vehicle status and environmental parameters. In this embodiment, a calculation method based on a kinematic model is used, taking into account the braking requirements under the worst-case scenario.

[0115] For example, the vehicle state and environmental parameters on which the dynamic calculation module calculates the dynamic warning distance include road slope obtained by the vehicle attitude sensor (IMU), road surface adhesion coefficient obtained through the vehicle bus or wireless network, and legal speed limit value of the driving lane determined based on the real-time lane position extracted from the AI ​​road condition topology model.

[0116] Specifically, dynamic warning distance It can be understood as consisting of three parts ( ):

[0117] (Perception Distance): This refers to the distance required for a driver of a vehicle approaching from behind to recognize the warning triangle. This value can be set as an empirical constant, typically 50 meters. In special weather conditions such as rain, snow, or heavy fog, this value is also related to weather visibility. Add an additional 50 meters or more.

[0118] (Reaction distance): This refers to the distance the vehicle travels within the driver's reaction time. , where v is the legal speed limit for the road segment (extracted from the AI ​​road condition topology model), and t is the average driver reaction time, which is usually 1.5 seconds.

[0119] (Braking distance): The distance a vehicle needs to decelerate from speed v to 0. ,in To effectively reduce speed.

[0120] For the sake of full explanation, the effective deceleration is given here. Detailed example:

[0121] ;

[0122] in, It is the acceleration due to gravity; The road surface adhesion coefficient can be obtained as a conservative percentage of the historical maximum value from the ESP system via the vehicle bus or as a road condition recommendation value provided by weather services via network access. The road slope angle can be directly measured by the vehicle's attitude sensor; it is positive for uphill and negative for downhill. Negative, leading to Decrease The increase is in accordance with the laws of physics.

[0123] Because this model is a classic simplification of vehicle kinematics, with clear physical meaning and moderate computational load, it can capture the key variables affecting braking distance—namely speed, coefficient of friction, and gradient—and dynamically adjust according to real-time environmental conditions, making it more scientific and safer than fixed-distance regulations.

[0124] At the same time, the dynamic calculation module calculates the generated virtual driving path trajectory {(x' i ,y' iFor example, the dynamic calculation module performs real-time integration on the virtual driving path trajectory to obtain the cumulative path length. This includes: sequentially acquiring adjacent trajectory points on the virtual driving path trajectory; for each pair of adjacent trajectory points forming a differential line segment, calculating its length in the actual three-dimensional road space based on the road curvature and slope information at the location of the line segment, and using this as the corrected line segment length; and accumulating all the corrected line segment lengths from the path calculation reference point to the current position, with the sum being the cumulative path length.

[0125] Calculate the straight-line distance between these two points in the two-dimensional path calculation coordinate system. This distance is called the planar projection length ΔL1. For example, starting from the origin O(0,0) of the path calculation reference, calculate the distance for each newly added point P' on the trajectory. i to the previous point P' {i-1} The straight-line distance ΔL1.

[0126] ΔL1=sqrt((x' i -x' {i-1} ) 2 +(y' i -y' {i-1} ) 2 );

[0127] Where sqrt is the square root calculation, and ΔL1 represents the distance between the two points under the assumption of an ideal, perfectly flat and straight road.

[0128] It's important to note that the integral calculates the actual length of the curved path. On curves (high curvature) or slopes, simply accumulating two-dimensional straight segments will underestimate the actual three-dimensional path length.

[0129] Therefore, ΔL1 needs to be corrected to obtain the actual distance ΔL2. Since the slope will make the actual driving path longer than the horizontal projection length ΔL1, for slope correction: the system can query the road slope angle at the midpoint of the line segment through the IMU. The essence of this correction is to "stretch" the horizontal projection length to the corresponding slope length based on the slope angle. The steeper the slope, the greater the stretching ratio. The calculation formula can be expressed as follows: .

[0130] Because vehicles travel along an arc on a curve, their actual path (arc length) will be greater than the straight distance (chord length) ΔL1 connecting the two ends of the arc. Therefore, for horizontal curvature correction, the system can query the horizontal curvature radius R of the lane where the midpoint of the line segment is located. The essence of this correction is to "unfold" the chord length into an arc length that is closer to the real one based on the curvature radius R. The sharper the curve (the smaller the curvature radius), the more significant the increase in arc length relative to chord length. Its calculation formula can be expressed as ΔL2=R×Δα (Δα is the central angle corresponding to the differential line segment).

[0131] Those skilled in the art can choose whether and how to perform corrections based on accuracy requirements. This embodiment only demonstrates one implementation method for performing such corrections.

[0132] Finally, the system will continuously accumulate all calculated corrected actual distances ΔL2 from the path calculation reference point (corresponding to the location of the accident vehicle) to the current beacon device mapping point (if the road segment does not require correction, ΔL1 can be used directly as the corrected path length). The sum is the cumulative path length L. 总 .

[0133] Finally, the system continuously compares L to determine the warning point. 总 and :

[0134] When L 总 < At this time, the system waits and continues accumulating points.

[0135] When L 总 ≥ At that time, the dynamic calculation module immediately generates an arrival command, which also means that, from a kinematic point of view, placing a warning triangle at the current road position corresponding to the beacon device can provide a minimum safe braking space for vehicles traveling at the legal speed limit behind.

[0136] When the dynamic calculation module determines the cumulative path length L 总 Reaching or exceeding the dynamic warning distance When the signal is received, the prompting module will respond to the arrival command and issue a prompting signal to indicate that the current position of the beacon device corresponds to the recommended placement point of the triangular warning sign on the target driving lane. The purpose is to notify the operator in a clear and unambiguous manner to stop walking and place the sign.

[0137] Specifically, the arrival command generated by the dynamic calculation module includes at least the command type (such as the distance to be reached) and necessary status information. The arrival command can be sent to the beacon device held by the operator via a wireless link established between the first and second wireless communication modules. To ensure reliability, the command is transmitted using data packets with an acknowledgment mechanism. If no response is received from the beacon device within a preset time (usually set to 100 milliseconds), the main control system will retransmit the command.

[0138] Upon receiving a valid prompt command, the beacon device's prompting module is activated and generates a strong indication signal. For example, the prompting signal emitted by the module may be a combination of sound and light, including intermittent beeping at a specific frequency and flashing of a bright LED.

[0139] The design principle of the combined audio-visual signal is to ensure that it can still be clearly perceived in the complex environment of highways (wind noise, tire noise, ambient light).

[0140] In this embodiment, the auditory signal can be an intermittent buzzing sound at a specific frequency emitted by the device's built-in speaker. For example, a buzzing frequency of 2.5 kHz (which is within the sensitive range of the human ear and has strong penetrating power) can be used, cycling in a rhythm of "0.3 seconds on, 0.2 seconds off". This intermittent mode is more attention-grabbing and less annoying than a continuous beep, and the volume can be adjusted by the operator as needed.

[0141] Visual signals can be achieved by the high-brightness light-emitting diodes integrated into the device housing flashing at a high frequency. For example, a high-lumen red LED can be used, flashing at a frequency of 5 Hz (5 flashes per second) synchronized with or interleaved with the rhythm of a sound. Red light has a universal "warning" semantic in traffic environments. Visual signals are particularly critical at night or in low-light environments such as tunnels.

[0142] In other embodiments, tactile feedback can also be added. For example, the beacon device can also have a built-in vibration motor that generates strong vibrations when prompted, or even form a triple sensing channel of sound, light and touch to ensure that nothing goes wrong.

[0143] Next comes the operator's turn. Upon hearing and seeing the warning signal from the beacon device (triangle warning sign), the operator should immediately stop walking. They should then observe the surrounding environment, confirming they are in a safe area such as the emergency lane or shoulder, and use the gaps in the signal to monitor traffic conditions behind them.

[0144] Once safety is confirmed, the operator needs to move the handheld warning triangle from the safe area to the lane directly opposite their current position. This "moving" action is crucial; it utilizes the system's intelligent mapping from the safe path to the lane path, eliminating the need for the operator to cross dangerous lanes for distance measurement, requiring only a safe lateral movement.

[0145] For example, an operator holding a beacon device walks in the emergency lane and hears the warning at point A on the edge of the emergency lane. Point A is mapped to point A' on the driving lane in the system. The operator simply needs to place the triangular marker from point A, perpendicularly across the curb, near the edge or center line of the driving lane corresponding to point A' (as required by regulations) to complete the precise placement.

[0146] After the triangular warning sign is placed, the operator presses a large, icon-embedded "Confirm Placement" physical button on the device. This action sends a placement confirmation message to the main control system (this function is a simple interactive function and will not be described in detail in this embodiment).

[0147] After receiving the confirmation information, the main control system performs the following actions: First, it records and archives the complete task data package (including accident time, location, real-time lane location, and dynamic warning distance). Final cumulative path length L 总 The data (such as virtual trajectories and beacon device travel paths) are stored in a non-volatile manner, and this data can be used for accident analysis or system optimization.

[0148] Then a "stop prompt" command is sent to the beacon device, which then shuts off the audible and visual signals.

[0149] Finally, the main control system can announce to the occupants via the vehicle's audio system or screen that "the warning triangle has been placed as required," providing closed-loop feedback.

[0150] To further expand advanced functionality, a monitoring function could be added, allowing the system to enter continuous monitoring mode after the placement process is complete. For example, if the vehicle is equipped with rear radar or a camera, it could monitor whether the placed warning triangle has been knocked over or moved, and issue an alarm to the occupants of the vehicle if this occurs.

[0151] This concludes the introduction to all modules of the AI-powered traffic condition recognition-based dynamic warning distance adjustment system for triangular warning signs. Additionally, for better application of this system, please refer to... Figure 2 As shown in the figure, this embodiment also provides a method for adjusting the dynamic warning distance of the triangular warning sign based on AI road condition recognition as follows:

[0152] Step 1: Respond to the warning activation command, obtain the vehicle's real-time lane position through the main control system, and calculate the baseline based on this path.

[0153] Step 2: Collect safe walking path data of operators as they move along the safe area using the beacon device.

[0154] Step 3: Using the AI ​​road condition topology model pre-set or acquired in real time within the main control system, the safe walking path data is mapped in real time to the center line of the driving lane determined based on the real-time lane position, generating a virtual driving path trajectory.

[0155] Step 4: Calculate the dynamic warning distance based on the real-time vehicle status and environmental parameters, and perform real-time integration on the virtual driving path trajectory to obtain the cumulative path length.

[0156] Step 5: When the cumulative path length reaches the dynamic warning distance, send an arrival command to the beacon device.

[0157] Step six: Send a warning signal through the beacon device to indicate that the current location is the recommended placement point for the warning triangle on the determined driving lane.

[0158] Since this method is applied to the dynamic warning distance adjustment system for triangular warning signs based on AI road condition recognition, it has the same beneficial effects.

[0159] In summary, by establishing a collaborative architecture between the main control system and the handheld beacon device, the safe path acquisition and dangerous path calculation are physically and logically separated. The beacon device allows operators to safely walk in the emergency lane and record their trajectory, while the main control system uses an AI-powered road topology model to intelligently map the safe trajectory onto the driving lane based on the real-time lane position of the accident vehicle, generating a corresponding virtual driving path. This mapping is the core of this invention, enabling all subsequent calculations to be based on real lane geometry. Subsequently, the system obtains the precise curve length by performing real-time integration on this virtual path and simultaneously integrates real-time vehicle and environmental parameters to calculate the dynamic warning distance. When the two match, the system prompts the operator to place the warning triangle. Throughout the entire process, the operator can accurately deploy the warning triangle without taking any risks, enhancing the operator's safety.

[0160] Furthermore, the translation mapping algorithm used in the path mapping is defined as a vertical translation transformation based on a fixed lateral offset, which eliminates the arbitrariness or cumulative error in the mapping process. This ensures that no matter whether the pedestrian walking path is a straight line or a complex curve, the final generated virtual driving path can strictly maintain geometric consistency with the lane centerline. This guarantees that on winding mountain roads or complex road alignments such as interchange ramps, the cumulative path length calculated by the system can accurately reflect the actual trajectory length of the vehicle.

[0161] Furthermore, the AI ​​road condition topology model is concretized into a composite model that simultaneously includes static geometric topological relationships and dynamic road condition information. When the system performs path mapping and safety decisions, it can not only perform geometric calculations based on the physical layout of lanes, but also actively integrate dynamic risk elements such as temporary construction barriers, accident obstacles, or slippery road surfaces. This allows the final warning area to proactively avoid secondary risk points or provide extended protection for vehicles that change lanes to avoid dynamic obstacles. This elevates the targeting of warnings from general scenarios to personalized instantaneous scenarios, enhancing the system's robustness and protective effectiveness under complex and ever-changing real road conditions.

[0162] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the aforementioned scope.

[0163] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-based road condition recognition-based dynamic warning distance adjustment system for triangular warning signs, comprising a main control system installed in the vehicle and a beacon device for handheld operation, characterized in that: The main control system includes: a baseline establishment module, a path mapping module, a dynamic calculation module, and a first wireless communication module; The beacon device includes a second wireless communication module, a data acquisition module, and a notification module; The baseline establishment module is used to respond to the warning activation command, obtain the real-time lane position of the vehicle, and establish a path calculation baseline based on it. The data acquisition module is used to continuously collect safe walking path data, including displacement, heading and position information, when the operator moves along the safe area with the beacon device; The path mapping module is used to pre-set or acquire AI road condition topology models in real time, and map the received safe walking path data to the center line of the driving lane determined based on the real-time lane position to generate a virtual driving path trajectory. The dynamic calculation module is used to calculate the dynamic warning distance based on the real-time vehicle status and environmental parameters, and to obtain the cumulative path length by integrating the virtual driving path trajectory in real time. When the cumulative path length reaches the dynamic warning distance, an arrival command is generated. The first wireless communication module establishes a connection with the second wireless communication module to transmit safe walking path data and arrival instructions; The alert module is used to respond to arrival instructions and issue an alert signal to indicate that the current location of the beacon device corresponds to the recommended placement point of the warning triangle on the target lane.

2. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 1, characterized in that, The baseline establishment module is specifically used to establish a two-dimensional path calculation coordinate system in the horizontal projection plane, with the real-time lane position of the vehicle as the origin and the tangent direction of the road where the real-time lane position is located as the baseline direction, as the baseline for path calculation.

3. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 1, characterized in that, The dynamic calculation module uses the following methods to perform real-time integration of the virtual driving path trajectory to obtain the cumulative path length: Sequentially obtain adjacent trajectory points on the virtual driving path trajectory; For each pair of adjacent trajectory points forming a differential line segment, based on the road curvature and slope information at the location of the differential line segment, its length in the actual three-dimensional road space is calculated as the corrected line segment length. The cumulative path length is obtained by summing up the lengths of all corrected line segments from the path calculation reference point to the current position.

4. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 1, characterized in that, AI traffic topology models include the geometric and topological relationships of roads; Geometric topology defines at least the spatial parallelism and width relationships between the safety zone and each driving lane, as well as between each driving lane and each other.

5. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 4, characterized in that, The path mapping module maps the safe walking path data to the centerline of the associated driving lane based on the real-time lane position using a translation mapping algorithm, according to the geometric topology, in order to generate a virtual driving path trajectory.

6. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 5, characterized in that, The translation mapping algorithm is as follows: Based on the fixed lateral offset between the safe area defined in the geometric topology and the centerline of the determined driving lane, each coordinate point in the safe driving path data is translated by a fixed lateral offset in a direction perpendicular to the reference direction, thereby obtaining the corresponding point on the virtual driving path trajectory.

7. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 1, characterized in that, The data acquisition module specifically includes an odometer, an inertial measurement unit, a GPS receiver, and a magnetometer; Odometers and inertial measurement units are used to measure the displacement and heading changes of the beacon device, GPS receivers are used to obtain the position coordinates of the beacon device, and magnetometers are used to provide heading angle correction references.

8. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 1, characterized in that, The dynamic calculation module calculates the vehicle status and environmental parameters on which the dynamic warning distance is based, including the road slope obtained by the vehicle attitude sensor, the road surface adhesion coefficient obtained by the vehicle bus or wireless network, and the legal speed limit value of the driving lane determined based on the real-time lane position extracted from the AI ​​road condition topology model.

9. The AI-based road condition recognition system for dynamic warning distance adjustment of triangular warning signs according to claim 1, characterized in that, The prompting module emits a combination of sound and light signals, including intermittent buzzing at a specific frequency and flashing of a bright LED.

10. A method for adjusting the dynamic warning distance of a triangular warning sign based on AI road condition recognition, applied to the AI ​​road condition recognition system for adjusting the dynamic warning distance of a triangular warning sign as described in any one of claims 1 to 9, characterized in that, The method includes: In response to the warning activation command, the system obtains the vehicle's real-time lane position through the main control system and uses this as a reference for path calculation. Data on the safe walking path of operators as they move along the safe zone is collected using beacon devices. By using the AI ​​road condition topology model pre-set or acquired in real time within the main control system, the safe walking path data is mapped in real time to the center line of the driving lane determined based on the real-time lane position, thereby generating a virtual driving path trajectory. The dynamic warning distance is calculated based on the real-time vehicle status and environmental parameters, and the cumulative path length is obtained by real-time integration of the virtual driving path trajectory. When the cumulative path length reaches the dynamic warning distance, an arrival command is sent to the beacon device. A warning signal is sent through the beacon device to indicate that the current location is the recommended placement point for the warning triangle on the designated driving lane.

Images