Multi-mode collaborative danger early warning method and device

Abstract

The invention relates to a multi-mode collaborative danger early warning method and device, and the method comprises the steps: S1, obtaining vehicle-mounted environment sensing data through an environment sensing module carried on a vehicle, and obtaining dynamic response data generated by the interaction of the vehicle and a road surface through a tire and suspension sensing module carried on the vehicle; the vehicle-mounted communication module receives early warning information sent by the roadside fixed sensing module; the roadside fixed sensing module is arranged on a high-risk road section which is pre-calibrated according to historical geological disaster data or topographic features; the vehicle central processing module obtains a third danger coefficient from the early warning information based on the vehicle-mounted environment sensing data and the dynamic response data, and calculates a comprehensive danger coefficient; and comparing the comprehensive danger coefficient with a first preset threshold value, and generating and outputting a danger early warning signal of a corresponding grade according to a comparison result, so that advanced early warning can be provided for mountain valley driving, and the driving safety is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle safety technology, and in particular to a multimodal collaborative hazard warning method and device. Background Technology

[0002] In mountainous and canyon areas, roads are often built along mountainsides or cliff edges, resulting in complex road conditions and extremely high potential risks. These road sections are frequently threatened by sudden geological disasters such as landslides, rockfalls, and roadbed collapses, especially during the rainy season or periods of geological activity, when the danger increases dramatically. Traditional driving safety relies mainly on the driver's real-time observation and experience-based judgment. However, in the aforementioned complex, changeable, and dangerous environments, human drivers have limitations in reaction time and judgment. Often, by the time obvious dangers (such as mudslides flowing onto the road or falling rocks) are observed, the best opportunity for avoidance has been lost, leading to serious traffic accidents.

[0003] To improve driving safety, existing technologies are mainly explored from two independent directions:

[0004] Single-vehicle intelligent perception and early warning: Modern vehicles are equipped with advanced driver assistance systems, such as collision warning and lane departure warning based on cameras and millimeter-wave radar. Some studies attempt to use onboard sensors to monitor the environment, such as visually identifying falling rocks or road anomalies. However, single-vehicle perception has inherent limitations: First, its perception range is limited, making it difficult to detect potential hazards such as slow deformation of side mountains or blind spots in curves ahead, or small cracks; second, relying solely on external environmental perception cannot effectively determine changes in the road surface's load-bearing capacity (such as roadbed softening or cavities); finally, severe weather (such as heavy rain or dense fog) significantly weakens the effectiveness of optical sensors.

[0005] Fixed roadside monitoring systems: In some high-risk road sections, management departments deploy monitoring equipment, such as surveillance cameras and geological sensors, to monitor mountain stability and road conditions. These systems typically operate independently, with warnings disseminated via broadcasts, signs, or independent monitoring centers. This makes real-time, accurate information exchange and collaborative decision-making with moving vehicles difficult. Vehicles cannot integrate these warnings as an immediate, quantifiable risk input into their own decision-making systems, resulting in insufficient timeliness and specificity of the warnings.

[0006] Therefore, there is an urgent need for a new multimodal collaborative hazard warning method to achieve advanced and accurate assessment of comprehensive driving risks in mountainous and canyon road sections. Summary of the Invention

[0007] (a) Technical problems to be solved

[0008] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a multimodal collaborative hazard early warning method and device, which solves the problems of insufficient comprehensive risk assessment capability and delayed early warning information in the prior art for complex environments such as mountainous canyons.

[0009] (II) Technical Solution

[0010] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0011] In a first aspect, embodiments of the present invention provide a multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas, comprising:

[0012] Step S1: Obtain vehicle-mounted environmental perception data through the environmental perception module mounted on the vehicle, and obtain dynamic response data generated by the interaction between the vehicle and the road surface through the tire and suspension perception module mounted on the vehicle; the environmental perception module operates based on simultaneous localization and mapping technology.

[0013] Step S2: Receive early warning information from roadside fixed sensing modules via vehicle-mounted communication modules; roadside fixed sensing modules are deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features.

[0014] Step S3, the vehicle central processing module performs the following operations:

[0015] Step S31: Based on the vehicle-mounted environmental perception data, generate a first hazard coefficient X1, where X1 represents the deformation hazard level of the terrain environment surrounding the vehicle.

[0016] Step S32: Based on the dynamic response data generated by the interaction between the vehicle and the road surface, a second hazard factor X2 is generated, which represents the vehicle state hazard level caused by road surface anomalies.

[0017] Step S32: Obtain the third hazard factor X3 from the early warning information. X3 is the environmental hazard level independently assessed by the roadside fixed sensing module based on the monitoring data of the high-risk road section it is stationed on.

[0018] Step S33: Calculate the comprehensive risk factor Y according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b; where W1, W2 and W3 represent the corresponding weighting coefficients, and b is the bias term;

[0019] S4 compares the comprehensive risk coefficient Y with the first preset threshold, and generates and outputs a corresponding level of risk warning signal based on the comparison result.

[0020] In one possible embodiment, generating a first hazard factor X1 based on vehicle-mounted environmental perception data includes: using the vehicle-mounted environmental perception data to construct or update a current map of the vehicle's surrounding environment in real time; extracting current environmental features from the current map and comparing the current environmental features with corresponding environmental features in a pre-stored reference map to detect deformation of mountain slopes, road shoulders, or canyon edges, and calculating their deformation amount or deformation rate; wherein the pre-stored reference map represents the state of the vehicle's surrounding environment at a previous time; inputting the deformation amount or deformation rate into a preset mapping function to convert it into a first hazard factor X1 representing the deformation hazard level; wherein the larger the deformation amount or deformation rate, the higher the mapped first hazard factor X1.

[0021] In one possible embodiment, generating a second hazard factor X2 based on dynamic response data generated by the interaction between the vehicle and the road surface includes: according to the mathematical expression X2= ×T+ ×S+ ×D+ ×E is used to calculate the second hazard factor X2; where T represents the tire condition hazard level judged based on tire pressure data and wheel speed data; S represents the road impact and deformation hazard level judged based on suspension displacement data; D represents the vehicle attitude instability hazard level judged based on vehicle body inertial measurement data; and E represents the road adhesion condition hazard level inferred based on at least one of T, S, and D and combined with onboard environmental perception data. , , and All are preset weighting coefficients, and satisfy the following conditions: + + + =1.

[0022] In one possible embodiment, the third hazard factor X3 is calculated by the roadside fixed sensing module based on the following mathematical expression and included in the warning information:

[0023] ;

[0024] Wherein, G represents the basic geological hazard level determined based on geological and topographic data of high-risk road sections; R represents the hydrometeorological hazard level determined based on real-time meteorological and soil moisture data. H represents the real-time surface deformation hazard determined based on synchronous positioning and mapping monitoring data from roadside fixed sensing modules; H represents the historical risk factor determined based on historical geological disaster data. , , and All of these are preset weighting coefficients.

[0025] In one possible embodiment, W1, W2, and W3 are dynamically adjusted according to the following strategy: in response to determining that a vehicle has entered a high-risk road segment covered by the roadside fixed sensing module, the weight of W3 is increased; in response to the first hazard factor X1 exceeding a second preset threshold, the weight of W1 is increased; in response to the second hazard factor X2 continuously exceeding a third preset threshold while the first hazard factor X1 does not exceed the second preset threshold, the weight of W2 is increased.

[0026] Secondly, embodiments of the present invention provide a multimodal collaborative hazard warning device for driving environments in mountainous and canyon areas, comprising:

[0027] The environmental perception module, mounted on the vehicle, is used to acquire onboard environmental perception data; the environmental perception module operates based on simultaneous localization and mapping technology.

[0028] The tire and suspension sensing module, mounted on the vehicle, is used to acquire dynamic response data generated by the interaction between the vehicle and the road surface;

[0029] The vehicle-mounted communication module is used to receive early warning information sent from the roadside fixed sensing module; the roadside fixed sensing module is deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features;

[0030] The vehicle central processing module is used to generate a first hazard coefficient X1 based on onboard environmental perception data, where X1 represents the deformation hazard level of the terrain environment surrounding the vehicle; generate a second hazard coefficient X2 based on dynamic response data generated by the interaction between the vehicle and the road surface, where X2 represents the vehicle state hazard level caused by road surface anomalies; obtain a third hazard coefficient X3 from the warning information, where X3 is the environmental hazard level independently assessed by the roadside fixed perception module based on monitoring data of the high-risk road sections it is stationed on; and calculate the comprehensive hazard coefficient Y according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b, where W1, W2, and W3 represent the corresponding weight coefficients, and b is a bias term.

[0031] The warning signal generation module is used to compare the comprehensive risk coefficient Y with the first preset threshold, and generate and output the corresponding level of risk warning signal based on the comparison result.

[0032] In one possible embodiment, the vehicle central processing module is specifically configured to: utilize onboard environmental perception data to construct or update a current map of the vehicle's surrounding environment in real time; extract current environmental features from the current map and compare the current environmental features with corresponding environmental features in a pre-stored reference map to detect deformation of mountain slopes, road shoulders, or canyon edges, and calculate their deformation amount or deformation rate; wherein the pre-stored reference map represents the state of the vehicle's surrounding environment at a previous moment; input the deformation amount or deformation rate into a preset mapping function to convert it into a first hazard coefficient X1 representing the deformation hazard level; wherein the larger the deformation amount or deformation rate, the higher the mapped first hazard coefficient X1.

[0033] In one possible embodiment, the vehicle central processing module is specifically used for: calculating the mathematical expression X2= ×T+ ×S+ ×D+ ×E is used to calculate the second hazard factor X2; where T represents the tire condition hazard level judged based on tire pressure data and wheel speed data; S represents the road impact and deformation hazard level judged based on suspension displacement data; D represents the vehicle attitude instability hazard level judged based on vehicle body inertial measurement data; and E represents the road adhesion condition hazard level inferred based on at least one of T, S, and D and combined with onboard environmental perception data. , , and All are preset weighting coefficients, and satisfy the following conditions: + + + =1.

[0034] In one possible embodiment, the third hazard factor X3 is calculated by the roadside fixed sensing module based on the following mathematical expression and included in the warning information:

[0035] ;

[0036] Wherein, G represents the basic geological hazard level determined based on geological and topographic data of high-risk road sections; R represents the hydrometeorological hazard level determined based on real-time meteorological and soil moisture data. H represents the real-time surface deformation hazard determined based on synchronous positioning and mapping monitoring data from roadside fixed sensing modules; H represents the historical risk factor determined based on historical geological disaster data. , , and All of these are preset weighting coefficients.

[0037] In one possible embodiment, the multimodal collaborative hazard warning device further includes: a weight dynamic adjustment module, configured to: increase the weight of W3 in response to determining that a vehicle has entered a high-risk road section covered by the roadside fixed sensing module; increase the weight of W1 in response to the first hazard coefficient X1 exceeding a second preset threshold; and increase the weight of W2 in response to the second hazard coefficient X2 continuously exceeding a third preset threshold while the first hazard coefficient X1 has not exceeded the second preset threshold.

[0038] Thirdly, embodiments of this application provide a storage medium for computer-readable storage, which stores one or more programs that can be executed by one or more processors to implement the multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas according to some embodiments of the first aspect of this application.

[0039] Fourthly, embodiments of this application provide an electronic device, which includes a memory, a processor, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas according to some embodiments of the first aspect of this application.

[0040] (III) Beneficial Effects

[0041] The beneficial effects of this invention are:

[0042] This invention proposes a multimodal collaborative hazard warning method and device for driving environments in mountainous and canyon areas. It acquires onboard environmental perception data through an environmental perception module mounted on the vehicle, and acquires dynamic response data generated by the interaction between the vehicle and the road surface through tire and suspension perception modules mounted on the vehicle. The environmental perception module operates based on synchronous positioning and mapping technology, and receives warning information from roadside fixed perception modules via an onboard communication module. The roadside fixed perception modules are deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features. The vehicle's central processing module performs the following operations: based on the onboard environmental perception data, it generates a first hazard coefficient X1, where X1 represents the deformation hazard level of the terrain environment surrounding the vehicle; based on... The dynamic response data generated by the interaction between the vehicle and the road surface generates a second hazard coefficient X2, which represents the hazard level of the vehicle's condition caused by road surface anomalies. A third hazard coefficient X3 is obtained from the warning information. X3 is the environmental hazard level independently assessed by the roadside fixed sensing module based on the monitoring data of the high-risk road sections it is stationed on. The comprehensive hazard coefficient Y is calculated according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b, where W1, W2, and W3 represent the corresponding weight coefficients, and b is a bias term. The comprehensive hazard coefficient Y is compared with a first preset threshold, and a corresponding level of hazard warning signal is generated and output based on the comparison result. This provides advanced warning for driving in mountainous canyons and significantly improves driving safety.

[0043] To make the above-mentioned objects, features and advantages to be achieved by the embodiments of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments of this disclosure will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this disclosure and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 The flowchart shown is a multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas provided by an embodiment of this application;

[0046] Figure 2 The diagram shows a structural block diagram of a multimodal collaborative hazard warning device for driving environments in mountainous and canyon areas, provided by an embodiment of this application. Detailed Implementation

[0047] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] To address the shortcomings of existing technologies in comprehensive risk assessment of complex environments such as mountainous areas and canyons, and the lag in early warning information, this invention proposes a multimodal collaborative hazard warning method and device for driving environments in mountainous and canyon areas. The method utilizes an environmental perception module mounted on the vehicle to acquire onboard environmental perception data, and a tire and suspension perception module mounted on the vehicle to acquire dynamic response data generated by the interaction between the vehicle and the road surface. The environmental perception module operates based on synchronous positioning and mapping technology. Furthermore, it receives early warning information from roadside fixed perception modules via an onboard communication module. The roadside fixed perception modules are deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features. The vehicle's central processing module performs the following operations: based on the onboard environmental perception data, it generates a first hazard coefficient X1. X1 represents the deformation hazard level of the surrounding terrain environment of the vehicle; based on the dynamic response data generated by the interaction between the vehicle and the road surface, a second hazard coefficient X2 is generated, which represents the vehicle state hazard level caused by road surface anomalies; a third hazard coefficient X3 is obtained from the warning information, which is the environmental hazard level independently assessed by the roadside fixed sensing module based on the monitoring data of the high-risk road section it is stationed on; the comprehensive hazard coefficient Y is calculated according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b; where W1, W2 and W3 represent the corresponding weight coefficients, and b is the bias term; and the comprehensive hazard coefficient Y is compared with a first preset threshold, and a corresponding level of hazard warning signal is generated and output according to the comparison result, thereby providing advanced warning for driving in mountainous canyons and significantly improving driving safety.

[0049] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0050] Please see Figure 1 , Figure 1A flowchart of a multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas, provided by an embodiment of this application, is shown. It should be understood that this multimodal collaborative hazard warning method can be executed by a multimodal collaborative hazard warning device for driving environments in mountainous and canyon areas, and the specific components of the multimodal collaborative hazard warning device can be configured according to actual needs; this embodiment is not limited thereto. For example, the multimodal collaborative hazard warning device can be a computer or a server, etc. Specifically, the multimodal collaborative hazard warning device includes:

[0051] Step S110 involves acquiring vehicle-mounted environmental perception data through an environmental perception module mounted on the vehicle, and acquiring dynamic response data generated by the interaction between the vehicle and the road surface through a tire and suspension perception module mounted on the vehicle. The environmental perception module operates based on simultaneous localization and mapping (SLAM) technology.

[0052] It should be understood that the devices included in the environmental sensing module can be configured according to actual needs, and the embodiments of this application are not limited thereto.

[0053] For example, this environmental perception module is an integrated sensing system based on Simultaneous Localization and Mapping (SLAM) technology. Its core function is to acquire high-precision spatial data (i.e., "vehicle-mounted environmental perception data") in real time to describe the terrain, topography, and road geometry around the vehicle during its movement. It typically includes one or more combinations of optical cameras, LiDAR, and millimeter-wave radar. The optical camera acquires high-resolution two-dimensional image information, providing rich texture and color features. Under good lighting conditions, the visual SLAM algorithm can achieve localization and dense mapping based on image sequences. The LiDAR directly acquires high-precision three-dimensional point cloud data of the surrounding environment by emitting laser beams and receiving reflected signals. It has strong geometric perception capabilities, is unaffected by lighting conditions, and is a core sensor for building accurate three-dimensional environmental maps. The millimeter-wave radar effectively detects the distance, speed, and orientation of objects by emitting millimeter waves and analyzing the echoes. It has stable detection performance in adverse weather conditions such as rain, fog, and dust, and can effectively supplement LiDAR and cameras, improving system robustness.

[0054] It should also be understood that the devices included in the tire and suspension sensing module can be configured according to actual needs, and the embodiments of this application are not limited thereto.

[0055] For example, a tire and suspension sensing module is deployed in the vehicle chassis and running gear, and it can monitor and acquire various physical signals (i.e., "dynamic response data") generated in real time when the vehicle chassis system interacts with the road surface. This data directly reflects the impact of the real-time road surface condition on vehicle dynamics. It typically includes one or more combinations of tire pressure monitoring sensors mounted on each tire, wheel speed sensors for measuring the angular velocity of each wheel, suspension displacement sensors for monitoring the suspension travel relative to the vehicle body, and an inertial measurement unit mounted on the vehicle body. The suspension displacement sensors measure the compression or extension travel of the suspension relative to the vehicle body, i.e., the relative vertical position of the wheel and the vehicle body; the inertial measurement unit measures the triaxial linear acceleration and triaxial angular velocity at the vehicle's center of gravity.

[0056] To facilitate understanding of step S110, it will be described below through specific embodiments.

[0057] Specifically, the environmental perception module acquires high-precision spatial data by continuously running the SLAM algorithm. Its workflow is as follows: First, it synchronously acquires raw data frames from various sensors, such as camera images or LiDAR point clouds, and extracts stable features for matching, such as image corner points or point cloud planes. Next, by matching the features of the current frame with historical features from the previous frame or local map, and using motion estimation algorithms, it calculates the vehicle's precise position and attitude changes in real time, completing localization. Simultaneously, the system integrates newly observed features and continuously optimizes a 3D spatial model describing the surrounding environment, achieving mapping. When the algorithm detects that the vehicle has returned to a historical area, it triggers closed-loop detection and global optimization to correct accumulated errors and ensure global consistency between the pose trajectory and the environmental map. Through this cyclical process, the module continuously outputs two core data streams: high-precision real-time vehicle pose, and a 3D map or feature set of the vehicle's surrounding environment containing rich geometric information at the current moment. These together constitute the "vehicle-mounted environmental perception data" required for subsequent analysis.

[0058] The aforementioned engineering steps, from feature extraction and matching to localization and map construction, are all based on the unified theoretical foundation of probabilistic state estimation. From a mathematical perspective, the core of SLAM technology can be expressed as a unified probabilistic inference problem. Essentially, at each time k, the system needs to simultaneously estimate the vehicle's own state X. k (Including position and attitude) and the surrounding environment map m. This joint estimation problem can be formalized as given all sensor observation data Z from the initial time to the current time. 0:k And (if available) all vehicle control inputs U 0:k Under the given conditions, calculate the joint posterior probability distribution P(X) of the vehicle state and the environment map. k,m |Z0:k U 0:k ,X0), where X0 is the initial state. The subscript 0:k represents a sequence from the initial time (time 0) to the current time (time k).

[0059] Solving for this posterior probability distribution depends on our understanding of the two fundamental models of the system: the motion model and the observation model.

[0060] The motion model describes how the vehicle's state evolves over time. It is typically modeled as a Markov process, i.e., the current state X. k Depends only on the state X of the previous time step k-1 and current control input (such as displacement and steering information derived from inertial measurement units and wheel speedometers), and independent of earlier historical states and observations. The model uses conditional probability P(X... k |X k-1 , )express;

[0061] The observation model describes the situation when the vehicle pose X k When map m is known, the sensor generates observations. The model characterizes the uncertainties of sensor measurement noise and the environment, using the conditional probability P(…). |X k ,m) represents.

[0062] Currently, most mainstream SLAM algorithms employ a two-step recursive filtering framework of prediction and correction to efficiently solve for the aforementioned posterior probabilities.

[0063] Prediction step (time update): Using the motion model, based on the state estimate from the previous moment and the current control input, predicts the prior state probability distribution of the vehicle at the current moment. Its mathematical expression reflects the recursive relationship of state prediction;

[0064] Calibration step (measurement update): When new sensor observation data is obtained. Subsequently, using an observation model, the predicted prior state is fused with the new observation information (based on Bayes' theorem) to calculate a more accurate posterior state probability distribution, and the environmental map m is updated simultaneously. Its mathematical expression describes the corrective effect of observation information on the state and map estimates.

[0065] Through this recursive process, the system can continuously output the optimal estimated vehicle pose sequence {X}. k The system provides a globally consistent environment map m. Ultimately, the optimized vehicle pose and environment map provided by the SLAM system form the high-precision spatiotemporal data foundation upon which the first hazard factor X1 depends.

[0066] Furthermore, when the tire and suspension sensing module includes one or more sensors such as tire pressure monitoring sensors, wheel speed sensors, suspension displacement sensors, and an inertial measurement unit (IMU), these sensors operate continuously during vehicle operation, directly capturing the physical signals generated by the interaction between the vehicle and the road surface. The tire pressure sensors provide real-time pressure values ​​for each tire, the wheel speed sensors output the rotational speed of each wheel, the suspension displacement sensors measure the real-time travel changes of each suspension element, and the IMU collects the vehicle's linear acceleration, angular velocity, and the resulting attitude information. All these pre-processed raw physical quantities are collectively referred to as "dynamic response data generated by the interaction between the vehicle and the road surface." They directly and sensitively reflect changes in vehicle state caused by road surface anomalies (such as bumps, collapses, or slipperiness), providing a basis for subsequent diagnosis of road conditions.

[0067] Step S120: Receive early warning information from roadside fixed sensing modules via the vehicle-mounted communication module. These roadside fixed sensing modules are deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features.

[0068] Specifically, the roadside fixed sensing module is pre-deployed in high-risk road sections that are pre-marked based on a comprehensive assessment of historical geological disaster data, geotechnical engineering survey reports, and topographic features (such as steep slopes, loose rock layers, and areas prone to water accumulation), such as areas prone to landslides, rockfalls, or unstable roadbeds.

[0069] Furthermore, the fixed roadside sensing module integrates sensors such as lidar and cameras, which continuously collect data on specific high-risk mountain slopes or roadbed sections under its monitoring at fixed installation angles and parameters. This raw point cloud and image data is fed into the module's built-in processing unit in real time. The processing unit first runs a SLAM algorithm to construct a high-precision, high-stability 3D baseline map of the fixed scene using sequential data, and continuously tracks and maintains the map. Based on this map, by accurately registering and comparing the real-time collected sensing data with historical baseline maps, the deformation and rate of change of the surface or slope within the monitoring area are calculated, achieving millimeter-level precision in detecting abnormal deformation. Simultaneously, the fixed roadside sensing module can also access local meteorological station data, geological sensor information, etc., forming a multi-dimensional environmental sensing input. This processed deformation data, environmental data, and pre-stored geological risk factors are input into a pre-defined risk assessment model (e.g., based on a linear weighted or machine learning model) for calculation. The core output of this model is a quantified third hazard coefficient X3, which directly and objectively characterizes the current environmental hazard level of the road section. Finally, the module's processing unit encapsulates this X3 coefficient value, along with key metadata such as the data timestamp, the module's unique geographic identifier, and suggested risk types (e.g., "landslide" or "rockfall"), into a lightweight, structured digital early warning message for broadcast by the wireless communication unit. The entire process is executed automatically and cyclically within the module, ensuring the real-time nature and professionalism of the early warning information.

[0070] Furthermore, during vehicle operation, the onboard communication module (such as a V2X onboard unit) continuously monitors the communication channels broadcast by roadside units. Once the vehicle enters the wireless communication coverage of a roadside fixed sensing module, it can receive real-time warning information from that module. This process enables real-time linking of the infrastructure's "static professional monitoring capabilities" with the vehicle's "dynamic mobile sensing needs," allowing the vehicle to obtain quantitative warning inputs several seconds to tens of seconds in advance regarding blind spot risks ahead or to the sides, providing a crucial external information source for subsequent onboard fusion decisions.

[0071] In step S130, the vehicle central processing module generates a comprehensive risk factor Y.

[0072] Specifically, based on vehicle-mounted environmental perception data, a first hazard coefficient X1 is generated, which represents the deformation hazard level of the terrain environment surrounding the vehicle; based on the dynamic response data generated by the interaction between the vehicle and the road surface, a second hazard coefficient X2 is generated, which represents the vehicle state hazard level caused by road surface anomalies; a third hazard coefficient X3 is obtained from the warning information, which is the environmental hazard level independently assessed by the roadside fixed perception module based on the monitoring data of the high-risk road section it is stationed on; the comprehensive hazard coefficient Y is calculated according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b; where W1, W2 and W3 represent the corresponding weight coefficients, and b is a bias term.

[0073] The process of generating a first hazard factor X1 based on vehicle-mounted environmental perception data includes: using vehicle-mounted environmental perception data to construct or update a current map of the vehicle's surrounding environment in real time; extracting current environmental features from the current map and comparing the current environmental features with corresponding environmental features in a pre-stored reference map to detect deformation of mountain slopes, road shoulders, or canyon edges, and calculating their deformation amount or deformation rate; wherein the pre-stored reference map represents the state of the vehicle's surrounding environment at a previous moment; inputting the deformation amount or deformation rate into a preset mapping function to convert it into a first hazard factor X1 representing the deformation hazard level; wherein the larger the deformation amount or deformation rate, the higher the first hazard factor X1 obtained by mapping.

[0074] For example, the vehicle's central processing module first uses real-time acquired onboard environmental perception data (such as LiDAR point cloud sequences and visual images) to run a simultaneous localization and mapping (SMR) algorithm. This process not only determines the vehicle's precise pose but, more importantly, dynamically constructs and updates a current map reflecting the fine geometric structure of the surrounding mountain slopes, road edges, and canyon terrain. This map typically exists in the form of a 3D point cloud or a set of feature points.

[0075] Next, the system extracts key environmental features from the current map. These features may be sets of points, lines, and surfaces with significant geometric characteristics, or specific object surfaces identified through semantic segmentation (such as an exposed rock face or a section of concrete shoulder). The purpose of extraction is to transform continuous map data into a series of computable and comparable geometric or semantic identifiers. Subsequently, the extracted current environmental features are precisely compared with corresponding environmental features in a pre-stored reference map. This pre-stored reference map represents the state of the same geographical area at a previous point in time (such as when the vehicle last passed through, or a baseline map provided by the cloud). Through algorithms such as point cloud registration and feature matching, the vehicle's central processing module can establish a one-to-one correspondence between current and historical features, and based on this, calculate the differences in location, shape, or distribution of the corresponding features, thereby detecting the deformation of targets such as mountain slopes, road shoulders, or canyon edges, and accurately calculating their deformation (such as displacement distance and area change) and deformation rate (deformation per unit time).

[0076] Finally, the calculated deformation and deformation rate are input into a predefined mapping function. This mapping function defines the conversion relationship from physical deformation parameters to risk level values. Its design follows a core principle: the larger the input deformation or deformation rate, the higher the mapped output first hazard factor X1 value. For example, this function can be a linear proportional function or a piecewise function with different sensitivities, aiming to distinguish between "slow settlement at the centimeter level" and "instantaneous landslide at the meter level" as different levels of risk values. Through this series of steps, the raw environmental geometric data, which is difficult to assess directly, is transformed into a unified, quantifiable first hazard factor X1, which directly and objectively characterizes the immediate hazard level of the terrain environment surrounding the vehicle due to deformation.

[0077] Furthermore, based on the dynamic response data generated by the interaction between the vehicle and the road surface, a second hazard factor X2 is generated, including: according to the mathematical expression X2= ×T+ ×S+ ×D+ ×E calculates the second hazard factor X2; T represents the tire condition hazard level judged based on tire pressure and wheel speed data; S represents the road impact and deformation hazard level judged based on suspension displacement data; D represents the vehicle attitude instability hazard level judged based on vehicle body inertial measurement data; E represents the road adhesion condition hazard level inferred based on at least one of T, S, and D and combined with onboard environmental perception data. , , and All are preset weighting coefficients, and satisfy the following conditions: + + + =1.

[0078] It should also be understood that the process of obtaining the T value, S value, D value and E value can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0079] Optionally, the process of obtaining the T value includes: the generation of the tire condition risk level T is completed collaboratively by the tire and suspension sensing module and the vehicle central processing module, and it is a continuous calculation process based on real-time sensor signals for feature extraction, pattern recognition, and risk quantification. The tire and suspension sensing module synchronously collects pressure monitoring sensor signals from each tire and wheel speed sensor signals from each wheel at a millisecond frequency, and sends these raw data to the vehicle central processing module. For the received tire pressure data, the algorithm built into the vehicle central processing module first calculates its rate of change over time. If it detects that the pressure of any tire decreases at a rate exceeding a preset "rapid leakage threshold" in a very short time (e.g., less than 0.5 seconds), it immediately determines that a high-risk puncture event has occurred. At the same time, the algorithm also monitors slow trend deviations in pressure and significant pressure differences with tires on the same axle or side, which may indicate chronic leaks or abnormal temperatures. For wheel speed data, the vehicle central processing module combines the vehicle reference speed (usually provided by the non-drive wheel speed or inertial measurement unit) to calculate the slip ratio of each tire in real time. When an unexpected surge in positive slip ratio of the drive wheels is detected (indicating wheel spin), or an abnormal loss of synchronicity in negative slip ratio of all wheels during braking (potentially suggesting a sudden change in unilateral road surface adhesion), it is marked as a grip anomaly event. Subsequently, the vehicle's central processing module inputs the identified anomaly patterns (including the rate and magnitude of pressure transients, and the magnitude and duration of slip ratios) into a preset risk mapping function. This risk mapping function maps physical anomalies of different characteristics and degrees to a standardized, quantified risk value. For example, a tire blowout might be directly mapped to a high-risk value close to the upper limit, while a minor, brief slippage would be mapped to a lower value. Finally, the vehicle's central processing module comprehensively considers the independent risk status of all tires, typically taking the maximum of the four tire risk values ​​as the overall tire condition hazard level T output, ensuring that any serious failure of a single tire is fully reflected in the calculation of the second hazard factor X2.

[0080] Optionally, the process of obtaining the S value includes: the suspension displacement sensor in the tire and suspension sensing module continuously collects real-time travel data of each wheel suspension relative to the vehicle body, and sends it to the vehicle central processing module. The algorithm built into the central processing module first preprocesses the raw displacement signal, including filtering to eliminate noise, and determines the suspension travel when the vehicle is traveling at a constant speed on a smooth road surface as a reference benchmark for the "static equilibrium position".

[0081] Subsequently, the algorithm performs in-depth feature extraction on the signal from both the time and frequency domains:

[0082] Temporal shock detection: The algorithm calculates in real-time the instantaneous deviation of the suspension travel relative to a reference point and its rate of change (speed). When a large positive compression (i.e., rapid wheel drop) occurs in the suspension travel of a single wheel within a very short period of time, accompanied by a peak in the rate of change, this is identified as a transient shock event, typically corresponding to driving over a deep pothole, the edge of a road collapse, or a large obstacle. The algorithm quantifies the severity of the event based on the magnitude of the shock, the peak speed, and the duration.

[0083] Frequency domain vibration analysis: The algorithm simultaneously performs (e.g., Fast Fourier Transform) or applies a set of bandpass filters on the suspension displacement signal to analyze its vibration spectrum. Sustained, high-frequency vibration components with concentrated energy (e.g., components much higher than the vehicle's normal suspension resonance frequency) are identified, typically corresponding to persistent high-frequency bumps caused by the vehicle traversing gravel roads, severely damaged paved surfaces, etc. The algorithm quantifies the intensity of this persistent road unevenness by calculating the signal energy or amplitude within a specific high-frequency band.

[0084] Next, the vehicle's central processing module inputs the extracted time-domain impact features and frequency-domain vibration features into a pre-defined pattern recognition and risk assessment model. This model can distinguish between different types of road surface anomalies. For example, it maps high-intensity transient impacts to extremely high single-instance risk values ​​(suggesting potential collapses or deep potholes), while mapping medium-to-high level high-frequency vibration energy to sustained medium-risk values ​​(suggesting poor road surface quality). Comprehensive analysis of multiple wheel signals can also help determine whether the anomaly is global (similar vibrations occur on all wheels) or local (severe impact occurs on only a single wheel), the latter potentially indicating more dangerous unilateral road surface instability.

[0085] Finally, the model integrates all analysis results and outputs a quantified road impact and deformation risk level, S. The value of S directly reflects the algorithm's judgment on the sudden defects or overall severity of the current road surface structure, providing a key input for assessing the risk of vehicle loss of control due to road surface anomalies.

[0086] Optionally, the process of obtaining the D value includes: the inertial measurement unit in the tire and suspension sensing module continuously collects highly dynamic three-axis angular velocity and linear acceleration data at the vehicle's center of gravity and sends it to the vehicle's central processing module. This central processing module first preprocesses the raw data, fusing wheel speed information using algorithms such as Kalman filtering to accurately estimate vehicle speed and other states, and filters out noise. Subsequently, its built-in algorithm enters the core analysis stage: based on real-time steering wheel angle, vehicle speed, and other signals, it calculates an expected and safe range of vehicle attitude changes as a dynamic benchmark using a built-in vehicle dynamics model to distinguish between normal driver operation and dangerous responses triggered by abnormal road conditions. The algorithm continuously compares the angular velocities measured by the inertial measurement unit (especially yaw and roll angular velocities) with this dynamic benchmark, focusing on identifying unexpected and abrupt deviations. For example, a sudden increase in yaw rate without corresponding steering input (indicating a risk of sideslip), or a sudden change in roll rate during smooth driving (indicating unilateral road collapse or instantaneous suspension failure). For each identified abnormal symptom, a risk value can be quantified based on its deviation amplitude, rate of change (angular acceleration), and duration using a preset risk mapping function. The principle is that the larger the deviation, the more rapid the change, and the longer the duration, the higher the risk value. Finally, the vehicle's central processing module integrates the abnormal risk values ​​from all dimensions to generate a unified, quantified vehicle attitude instability risk level D. This value directly represents the immediate risk level of the vehicle tending to lose dynamic control due to road anomalies.

[0087] Optionally, the process of obtaining the E value is as follows: The vehicle central processing module first receives real-time input from different paths: the tire condition hazard T, road impact hazard S, and vehicle posture hazard D directly derived from the chassis sensors, as well as processed visual information from the environmental perception module (e.g., semantic segmentation results of the road surface material ahead, which can identify asphalt, water accumulation, reflective ice surface, or loose gravel).

[0088] Subsequently, the module's built-in inference algorithm begins to work. It doesn't simply add up the input values, but performs pattern matching and logical reasoning based on a pre-defined causal association rule base. For example, if the T value increases significantly (tire slippage detected), while the S value remains low (no severe bumps or large potholes detected), and visual information identifies continuous reflective areas or white coverings on the road surface, the system will strongly infer that it is currently encountering ice or compacted snow, resulting in an extremely low global adhesion coefficient. This pattern will trigger a significant increase in the E value. As another example, when the module simultaneously identifies a sustained moderate increase in tire condition risk T (indicating a continuous slippage trend not caused by rapid acceleration), and road impact risk S exhibits typical mid-to-high frequency vibration characteristics (indicating the presence of dense, small irregularities on the road surface), the chassis comprehensive indicators strongly suggest a special road condition combining low adhesion and continuous bumps. At this moment, the visual semantic information provided by the environmental perception module (such as the "slippery road surface" and "particulate matter present" labels) provides crucial evidence for this inference. The vehicle's central processing module jointly evaluates the numerical combination patterns of T and S values ​​with visual evidence. When the three highly match, the system determines that the current surface is a wet, slippery gravel road, triggering a moderately increased E value corresponding to this comprehensive risk pattern. For example, when the module detects a sudden increase in the D value with dynamic characteristics exhibiting a unilateral yaw trend, and the abnormal T value only appears in isolation on one side of the wheel (e.g., the left wheel slip rate spikes while the right is normal), the system initially locates the risk originating from one side of the vehicle. At this point, visual evidence provided by the environmental perception module, such as "local puddles" consistent with the trajectory of the abnormal wheel, becomes crucial in confirming the external cause. The core task of the vehicle's central processing module is to perform spatiotemporal correlation judgment: confirming whether the abnormal timing of D and T precisely matches the moment the vehicle passes through the low-adhesion area reported visually. If the match is valid, the system infers that the unilateral dynamic anomaly was caused by passing through a local low-adhesion area and generates a high-weighted, transient E value component, specifically characterizing the instantaneous high risk of this local adhesion mutation.

[0089] In addition, the algorithm also considers independent judgments based on environmental perception data. For example, even if T, S, and D do not show obvious abnormalities, but the camera clearly identifies a large area of ​​water accumulation or the start of rainfall ahead, the system will appropriately increase the base value of E based on prior knowledge as a risk warning.

[0090] Finally, the vehicle's central processing module integrates all the above reasoning results (i.e., the various E-value components calculated based on different rule triggers) (e.g., taking the maximum value or weighted fusion) and outputs a unified, quantified road adhesion condition hazard level E. This E-value represents the system's comprehensive judgment on the implicit risk of "whether the current road surface can provide sufficient grip," and it complements other hazard levels (T, S, D) that directly reflect physical anomalies, together forming a panoramic risk assessment of the driving environment.

[0091] The rules described above are merely examples. The vehicle's central processing module can implement the inference logic in various ways, such as a preset rule engine, decision tree, or machine learning model.

[0092] In addition, the third hazard factor X3 is calculated by the roadside fixed sensing module based on the following mathematical expression and included in the warning information:

[0093] ;

[0094] Wherein, G represents the basic geological hazard level determined based on geological and topographic data of high-risk road sections; R represents the hydrometeorological hazard level determined based on real-time meteorological and soil moisture data. H represents the real-time surface deformation hazard determined based on synchronous positioning and mapping monitoring data from roadside fixed sensing modules; H represents the historical risk factor determined based on historical geological disaster data. , , and All of these are preset weighting coefficients.

[0095] It should also be understood that G value, R value, The process of obtaining the H value can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0096] Optionally, the process of obtaining the G value includes: The process begins with the roadside processing module accessing and parsing its internal storage or cloud-synchronized high-precision geographic information system database. This module extracts the underlying geological and topographic vector data of the road segment corresponding to the monitoring point from the database. Key parameters may include: the geological and soil type and structure of the slope (e.g., determining it to be loose residual slope soil, strongly weathered fractured rock, or other easily unstable materials), the slope angle, slope height, and morphology (e.g., extracting near-vertical steep slopes), and key topographic factors such as topographic curvature and catchment area. Subsequently, the evaluation algorithm built into the roadside processing module can independently quantify the risk of each factor based on a set of preset engineering geological experience rules and weight tables. For example, the algorithm assigns a high-risk base value to "loose soil," and if "slope greater than 50 degrees" is also added, a risk multiplication calculation is performed. If unfavorable structural surface information such as "bedding slope" exists, additional risk bonuses are triggered. Finally, the algorithm uses either a linear weighted model or a lightweight machine learning classifier to fuse the scores of all factors into a normalized value between 0 and 1, namely the basic geological hazard level G. This G value is a relatively stable background risk indicator, representing the potential likelihood level of geological disasters occurring in this road section due to its inherent geological and topographical conditions under the absence of external disturbances such as rainfall and vibration, providing a crucial baseline reference for subsequent dynamic risk assessment.

[0097] Optionally, the process of obtaining the R value includes: This process begins with the roadside processing module synchronously acquiring and fusing multi-source real-time environmental data. This module continuously receives millisecond-level rainfall intensity, cumulative rainfall, and temperature data measured by a miniature weather station integrated near the monitoring point via an industrial bus or wireless network; simultaneously, it receives soil moisture and stress state change data from soil moisture and pore water pressure sensors buried at key depths of the slope (such as near potential slip surfaces). Furthermore, this module can also access regional meteorological radar data or short-term forecasts to obtain dynamic information on rain clouds over a wider area above the monitoring point. After obtaining the raw data, the module's built-in hydrogeological coupling analysis algorithm begins its core calculations. This algorithm first performs time-history analysis on the rainfall data, calculating key indicators such as "previous effective rainfall" and "current rainfall intensity-duration," and then couples these with the measured moisture content change curve of the slope soil in real time. This algorithm, based on the unsaturated soil seepage theory and slope stability model in geotechnical mechanics, dynamically simulates and calculates the physical process by which rainwater infiltration leads to an increase in the weight of slope soil, loss of matrix suction, and consequently, a decrease in shear strength and an increase in sliding moment. Through real-time iterative calculations, the algorithm outputs an index characterizing the immediate stability state of the slope, such as the reciprocal of the real-time safety factor Fs or the probability of instability. Finally, the calculation results of this physical model are mapped to a hydrometeorological hazard R between 0 and 1 through a preset normalization function. The R value is a highly dynamic index that accumulates rapidly with continuous rainfall, intuitively quantifying the immediate "catalytic" risk of current meteorological conditions triggering slope instability, debris flows, and other disasters.

[0098] Optionally, The data acquisition process includes: the roadside processing module's LiDAR and other sensors continuously scan the specific high-risk slopes or roadbed sections they are "guarding" at a fixed frequency, generating ordered time-series point cloud data. The dedicated SLAM processing unit built into the roadside processing module constructs a millimeter-precision 3D reference map of the fixed scene using initial or periodically calibrated data. This map serves as the geometric reference system for judging all subsequent deformations. In each data processing cycle, the system executes rigorous point cloud registration and change detection algorithms: First, through iterative nearest-point algorithms, the latest acquired point cloud frame is precisely aligned with the reference map to eliminate errors caused by minor sensor vibrations. Then, in the registered unified coordinate system, the 3D coordinate differences between the current point cloud and corresponding points on the historical reference map are compared point-by-point or region-by-region. By statistically analyzing these differences (such as calculating the root mean square error of the overall point cloud and the deformation vector field of a specific area of ​​interest), the algorithm can quantitatively calculate the deformation of the monitored area in the vertical and horizontal directions, and further calculate the deformation rate based on the timestamps of multiple consecutive frames of data. Finally, the system inputs the two key physical quantities, deformation and deformation rate, into a pre-defined risk mapping function. This function is designed to be sensitive to both the "amplitude" and "velocity" of deformation; for example, a slow but cumulative settlement and a sudden instantaneous displacement may both be mapped to a higher risk. Value. Through this continuous process, the module outputs the real-time surface deformation hazard level. It directly and objectively quantifies the physical displacement of mountains or roadbeds and its urgency, and is the most critical and direct quantitative indicator for revealing that geological disasters have entered the pre-disaster deformation stage.

[0099] Optionally, the process of obtaining the H value includes: the process begins with the construction and access of a structured historical disaster database built in the cloud or regional center. This database archives in detail, in chronological order, all key metadata of past landslides, rockfalls, collapses and other events that have occurred on the target road section, including the precise spatiotemporal coordinates of each event, the type and scale of the disaster, and the associated meteorological and seasonal background. When calculating the H-value, the roadside processing module accesses and extracts complete event sequence data for the road segment via the communication network, primarily following two analytical paths for quantification: First, frequency-based statistical modeling, where the module calculates the annual average disaster incidence rate for the road segment and can perform seasonal decomposition or weighting by event size to derive a static risk baseline value reflecting historical activity. Second, a more advanced spatiotemporal prediction-based machine learning model, where the module or cloud platform uses the timestamps and location information of historical events to train models such as spatiotemporal Poisson processes or survival analysis models. These models not only learn the average density of disaster occurrences but also capture their temporal clustering (such as aftershock effects) and spatial migration patterns. Furthermore, by combining historical meteorological sequences, they output a priori risk value with probabilistic predictive significance that dynamically adjusts with the current date and previous weather conditions. Ultimately, regardless of the path used to calculate the original index, it is normalized into a historical risk factor H between 0 and 1. The H value, as a correction term integrating empirical wisdom, is injected into the calculation formula of X3. Its core function is to endow the system with "historical insight": even if the current real-time deformation monitoring ( The lack of immediate threat from hydrological and meteorological data (R) and the relatively high H value will keep the overall risk rating (X3) of the road section at a cautiously elevated baseline level, thereby ensuring continuous key monitoring and risk prevention for high-risk road sections with "poor records" and enhancing the depth and continuity of the entire early warning system assessment.

[0100] In addition, W1, W2, and W3 can be dynamically adjusted according to the following strategies:

[0101] In response to the determination that a vehicle has entered a high-risk road segment covered by a roadside fixed sensing module, the weight of W3 is increased. For example, when a vehicle determines, through matching its onboard high-precision positioning module with electronic map data, that it has entered a pre-marked high-risk road segment covered by a roadside fixed sensing module, the system immediately increases the weight of the third hazard factor X3, W3. The principle is that within this geographical area, the warning information (X3) provided by the roadside unit has the highest relevance and authority because it originates from long-term, professional, fixed-point monitoring of that road segment. Increasing W3 essentially means that, in the short period of time a vehicle passes through this most dangerous area, the system places greater trust and reliance on the professional judgment from the infrastructure when making decisions.

[0102] In response to a first hazard factor X1 exceeding a second preset threshold, the weight of W1 is increased. For example, when the vehicle's central processing module continuously monitors the first hazard factor X1 (environmental deformation hazard level), if the value of X1 exceeds a preset second threshold (which corresponds to a deformation level considered to require high vigilance), the system will significantly increase its weight W1. This is mainly because once the vehicle's own sensors clearly detect significant and dangerous deformation of the surrounding mountains or roadbed, this indicates that environmental risk has become the primary concern, and the system must immediately shift its assessment focus to this direct external threat. Increasing W1 ensures that the imminent terrain change risk detected in real time by SLAM is most fully reflected in the comprehensive assessment; the specific value of the second preset threshold can be set according to actual needs.

[0103] In response to a situation where the second hazard factor X2 continuously exceeds a third preset threshold, while the first hazard factor X1 does not exceed the second preset threshold, the weight of W2 is increased. For example, when the system detects that the second hazard factor X2 (the vehicle's condition hazard level caused by road surface anomalies) continuously exceeds the third preset threshold (indicating that the vehicle is continuously experiencing severe bumps, instability, or tire abnormalities), while the first hazard factor X1 does not exceed its second threshold (meaning no corresponding significant external environmental deformation is detected), the system will increase the weight of X2, W2. This situation usually reveals the existence of hidden road risks, such as a roadbed that has been hollowed out but the surface has not yet collapsed, or encountering deep potholes or black ice that are not visually obvious. In this case, the "personal experience" (X2) of the vehicle chassis reflects the real danger better than the external environmental perception (X1). Increasing W2 allows the system to issue timely warnings based on the vehicle's own dynamic anomalies even when there is no external perception warning; the specific value of the third preset threshold can also be set according to actual needs.

[0104] Step S140: Compare the comprehensive risk coefficient Y with the first preset threshold, and generate and output a corresponding level of hazard warning signal based on the comparison result. The specific value of the first preset threshold can be set according to actual needs. Its setting can be determined based on a large amount of historical data statistics, vehicle model parameter calibration, or specific road safety standards, and can be dynamically fine-tuned according to different road segment attributes or weather patterns.

[0105] It should be understood that the specific process of comparing the comprehensive risk coefficient Y with the first preset threshold and generating and outputting the corresponding level of risk warning signal based on the comparison result can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0106] Optionally, the vehicle's central processing module pre-stores a tiered warning threshold table. This table defines at least two (e.g., three) incremental thresholds to classify different levels of danger. For example, it can be set as follows:

[0107] The first-level threshold (green / traffic threshold): When Y ≤ the first preset threshold T1, it indicates that the comprehensive risk is low and within the safe range;

[0108] The second-level threshold (yellow / warning threshold): When T1 < Y ≤ the fourth preset threshold T4, it indicates that the comprehensive risk has increased and the driver needs to be vigilant. Among them, the specific value of the fourth preset threshold T4 can be set according to actual needs, and its setting can be determined based on a large amount of historical data statistics, vehicle type parameter calibration, or specific road safety standards, and can be dynamically fine-tuned according to different road section attributes or weather patterns;

[0109] The third-level threshold (red / danger threshold): When Y > T4, it indicates that the comprehensive risk is high and there is an imminent danger.

[0110] After the vehicle's central processing module obtains the real-time calculated Y value, it immediately compares it with the above thresholds to determine the current danger level. Subsequently, the module generates a structured danger warning signal according to the determined level and outputs it to the driver in multiple modalities through the vehicle's human-machine interface:

[0111] Visual warning: On the dashboard, head-up display or center console screen, display color icons (such as red, yellow, green) corresponding to the level and brief text prompts (such as "Dangerous road conditions, please drive carefully", "High risk ahead, it is recommended to stop"), and may be supplemented with a flashing effect to enhance attention;

[0112] Auditory warning:发出不同频率、节奏和音调的警告音。例如,黄色预警可能伴随间歇性提示音,而红色预警则触发持续、急促的蜂鸣声;

[0113] Tactile warning: Generate different intensities of tactile feedback through the vibration module of the steering wheel or seat to physically remind the driver.

[0114] In addition, the warning signal can also be linked with other vehicle control systems. For example, in the red danger level, the system can automatically pre-tighten the seat belt, adjust the air conditioner air volume to reduce noise interference, or provide higher-priority intervention preparation for the adaptive cruise and emergency braking systems.

[0115] It should be noted here that the algorithms, models, etc. involved in the above content of this application can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0116] Note: There is an error in the original text for item . The correct English translation for "发出不同频率、节奏和音调的警告音。例如,黄色预警可能伴随间歇性提示音,而红色预警则触发持续、急促的蜂鸣声;" should be "发出 warning sounds with different frequencies, rhythms, and tones. For example, a yellow warning may be accompanied by intermittent beeps, while a red warning triggers continuous, rapid beeping;" However, I have translated it based on the content you provided as accurately as possible while following the rules.In summary, by utilizing the aforementioned technical solutions, this application's embodiments construct a three-dimensional monitoring network that coordinates "vehicle-mounted mobile sensing" and "roadside fixed sensing." The system simultaneously possesses the vehicle's own real-time, high-precision detection capability of its surrounding environment, as well as the professional, fixed-point monitoring capability of roadside facilities for blind spots on curves and high-risk slopes. This completely overcomes the limitations of single-vehicle sensing range and the difficulty in detecting slow deformations to the side and in the distance, extending the risk warning time lead from "second-level reaction" to "tens of seconds-level warning," providing drivers with valuable time for hazard avoidance decisions.

[0117] Furthermore, this application combines the perception of external environmental deformation (through SLAM technology) with the perception of road surface load-bearing status (through vehicle chassis dynamics response). This enables the system to not only provide early warnings of visible risks such as landslides and rockfalls, but also to sensitively diagnose less visually apparent, hidden, and potentially fatal threats such as roadbed collapses, cavities beneath the road surface, and black ice, achieving a comprehensive risk assessment of the driving environment from the surface to the core.

[0118] Furthermore, by fusing heterogeneous data from multiple sources such as vision, lidar, millimeter-wave radar, and chassis sensors, this application enables the system to complement each other under different weather (rain, fog, night) and lighting conditions. The introduction of roadside information provides an independent data verification source, ensuring that even when the performance of a single sensor declines or fails, the system can still make reliable judgments based on other information sources, achieving stable all-weather early warning.

[0119] Furthermore, this application can dynamically adjust the weights (i.e., W1, W2, and W3) of different risk inputs in the final decision based on the vehicle's location and the real-time confidence level of each risk source (such as whether the onboard sensors detect significant deformation or whether the vehicle itself senses any abnormality). This makes the warning output no longer a simple information aggregation, but an intelligent judgment with contextual understanding capabilities, greatly reducing the false alarm and false negative rates and making the warning information more authoritative and actionable.

[0120] It should be understood that the above-described multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas is merely exemplary. Those skilled in the art can make various modifications based on the above method, and the modified solutions also fall within the protection scope of this application.

[0121] Please see Figure 2 , Figure 2This diagram illustrates a structural block diagram of a multimodal collaborative hazard warning device 200 for driving environments in mountainous and canyon areas, according to an embodiment of this application. It should be understood that the multimodal collaborative hazard warning device 200 is capable of performing the steps described in the above method embodiments. The specific functions of the multimodal collaborative hazard warning device 200 can be found in the description above; detailed descriptions are omitted here to avoid repetition. The multimodal collaborative hazard warning device 200 includes at least one software function module that can be stored in a memory or embedded in the operating system (OS) of the multimodal collaborative hazard warning device 200 in the form of software or firmware. Specifically, the multimodal collaborative hazard warning device 200 includes:

[0122] The environmental perception module 210 is mounted on the vehicle and is used to acquire onboard environmental perception data; the environmental perception module operates based on simultaneous localization and mapping technology.

[0123] The tire and suspension sensing module 220, mounted on the vehicle, is used to acquire dynamic response data generated by the interaction between the vehicle and the road surface;

[0124] The vehicle-mounted communication module 230 is used to receive early warning information sent from the roadside fixed sensing module; the roadside fixed sensing module is deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features;

[0125] The vehicle central processing module 240 is used to generate a first hazard coefficient X1 based on onboard environmental perception data, where X1 represents the deformation hazard level of the terrain environment surrounding the vehicle; generate a second hazard coefficient X2 based on dynamic response data generated by the interaction between the vehicle and the road surface, where X2 represents the vehicle state hazard level caused by road surface anomalies; obtain a third hazard coefficient X3 from the warning information, where X3 is the environmental hazard level independently assessed by the roadside fixed perception module based on monitoring data of the high-risk road sections it is stationed on; and calculate the comprehensive hazard coefficient Y according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b; where W1, W2 and W3 represent the corresponding weight coefficients, and b is a bias term;

[0126] The warning signal generation module 250 is used to compare the comprehensive risk coefficient Y with the first preset threshold, and generate and output the corresponding level of risk warning signal based on the comparison result.

[0127] In one possible embodiment, the vehicle central processing module 240 is specifically configured to: construct or update a current map of the vehicle's surrounding environment in real time using onboard environmental perception data; extract current environmental features from the current map and compare the current environmental features with corresponding environmental features in a pre-stored reference map to detect deformation of mountain slopes, road shoulders, or canyon edges, and calculate their deformation amount or deformation rate; wherein the pre-stored reference map represents the state of the vehicle's surrounding environment at a previous moment; input the deformation amount or deformation rate into a preset mapping function to convert it into a first hazard coefficient X1 representing the deformation hazard level; wherein the larger the deformation amount or deformation rate, the higher the first hazard coefficient X1 obtained by mapping.

[0128] In one possible embodiment, the vehicle central processing module 240 is specifically configured to: calculate based on the mathematical expression X2= ×T+ ×S+ ×D+ ×E is used to calculate the second hazard factor X2; where T represents the tire condition hazard level judged based on tire pressure data and wheel speed data; S represents the road impact and deformation hazard level judged based on suspension displacement data; D represents the vehicle attitude instability hazard level judged based on vehicle body inertial measurement data; and E represents the road adhesion condition hazard level inferred based on at least one of T, S, and D and combined with onboard environmental perception data. , , and All are preset weighting coefficients, and satisfy the following conditions: + + + =1.

[0129] In one possible embodiment, the third hazard factor X3 is calculated by the roadside fixed sensing module based on the following mathematical expression and included in the warning information:

[0130] ;

[0131] Wherein, G represents the basic geological hazard level determined based on geological and topographic data of high-risk road sections; R represents the hydrometeorological hazard level determined based on real-time meteorological and soil moisture data. H represents the real-time surface deformation hazard determined based on synchronous positioning and mapping monitoring data from roadside fixed sensing modules; H represents the historical risk factor determined based on historical geological disaster data. , , and All of these are preset weighting coefficients.

[0132] In one possible embodiment, the multimodal collaborative hazard warning device 200 further includes: a weight dynamic adjustment module (not shown), used to: increase the weight of W3 in response to determining that a vehicle enters a high-risk road section covered by the roadside fixed sensing module; increase the weight of W1 in response to the first hazard coefficient X1 exceeding a second preset threshold; and increase the weight of W2 in response to the second hazard coefficient X2 continuously exceeding a third preset threshold while the first hazard coefficient X1 does not exceed the second preset threshold.

[0133] Since the apparatus described in the above embodiments of the present invention is an apparatus used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and variations of the apparatus based on the methods described in the above embodiments of the present invention, and therefore will not be described again here. All apparatuses used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0134] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0136] It should be noted that any reference numerals placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims that enumerate several means, several of these means may be embodied by the same hardware. The use of the terms first, second, third, etc., is merely for convenience of expression and does not indicate any order. These terms can be understood as part of the component names.

[0137] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0138] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the claims should be interpreted to include both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0139] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, then this invention should also include these modifications and variations.

Claims

1. A multimodal collaborative hazard warning method for driving environments in mountainous and canyon areas, characterized in that, include: Step S1: Obtain vehicle-mounted environmental perception data through the environmental perception module mounted on the vehicle, and obtain dynamic response data generated by the interaction between the vehicle and the road surface through the tire and suspension perception module mounted on the vehicle; the environmental perception module operates based on synchronous positioning and mapping technology. Step S2: Receive warning information sent from the roadside fixed sensing module through the vehicle communication module; The roadside fixed sensing modules are deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features; Step S3, the vehicle central processing module performs the following operations: Step S31: Based on the vehicle environment perception data, generate a first hazard coefficient X1, whereby X1 represents the deformation hazard level of the terrain environment surrounding the vehicle. Step S32: Based on the dynamic response data generated by the interaction between the vehicle and the road surface, a second hazard factor X2 is generated, wherein X2 represents the vehicle state hazard level caused by road surface anomalies. Step S32: Obtain the third hazard factor X3 from the warning information. X3 is the environmental hazard level independently assessed by the roadside fixed sensing module based on the monitoring data of the high-risk road section it is stationed on. Step S33: Calculate the comprehensive risk factor Y according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b; where W1, W2 and W3 represent the corresponding weighting coefficients, and b is the bias term; S4, compare the comprehensive risk coefficient Y with the first preset threshold, and generate and output a corresponding level of risk warning signal based on the comparison result.

2. The multimodal collaborative hazard early warning method according to claim 1, characterized in that, The step of generating a first risk factor X1 based on the vehicle environment perception data includes: Using the in-vehicle environmental perception data, a current map of the vehicle's surrounding environment is constructed or updated in real time; The current environmental features are extracted from the current map and compared with the corresponding environmental features in a pre-stored reference map to detect the deformation of the mountain slope, road shoulder or canyon edge, and calculate its deformation amount or deformation rate; wherein, the pre-stored reference map represents the state of the vehicle's surrounding environment at a previous moment. The deformation amount or deformation rate is input into a preset mapping function to convert it into a first hazard factor X1 that characterizes the deformation hazard level; wherein, the larger the deformation amount or deformation rate, the higher the first hazard factor X1 obtained by mapping.

3. The multimodal collaborative hazard early warning method according to claim 1, characterized in that, The generation of a second hazard factor X2 based on the dynamic response data generated by the interaction between the vehicle and the road surface includes: According to the mathematical expression X2= ×T+ ×S+ ×D+ The second risk factor X2 is calculated using ×E; where T represents the tire condition risk level determined based on tire pressure data and wheel speed data; S represents the road impact and deformation risk level determined based on suspension displacement data; D represents the vehicle attitude instability risk level determined based on vehicle body inertial measurement data; and E represents the road adhesion condition risk level inferred based on at least one of T, S, and D, combined with the vehicle environment perception data. , , and All are preset weighting coefficients, and satisfy the following conditions: + + + =1.

4. The multimodal collaborative hazard early warning method according to claim 1, characterized in that, The third hazard factor X3 is calculated by the roadside fixed sensing module based on the following mathematical expression and is included in the warning information: ； Wherein, G represents the basic geological hazard level determined based on the geological and topographical data of the high-risk road section; R represents the hydrometeorological hazard level determined based on real-time meteorological and soil moisture data; H represents the real-time surface deformation hazard determined based on the synchronous positioning and mapping monitoring data of the roadside fixed sensing module; H represents the historical risk factor determined based on historical geological disaster data. , , and All of these are preset weighting coefficients.

5. The multimodal collaborative hazard early warning method according to claim 1, characterized in that, W1, W2, and W3 are dynamically adjusted according to the following strategy: In response to determining that the vehicle has entered the high-risk road section covered by the roadside fixed sensing module, the weight of W3 is increased; In response to the first risk factor X1 exceeding a second preset threshold, the weight of W1 is increased; In response to the second risk factor X2 continuously exceeding the third preset threshold, while the first risk factor X1 does not exceed the second preset threshold, the weight of W2 is increased.

6. A multimodal collaborative hazard warning device for driving environments in mountainous and canyon areas, characterized in that, include: An environmental perception module, mounted on the vehicle, is used to acquire in-vehicle environmental perception data; The environmental perception module operates based on synchronous positioning and mapping technology; The tire and suspension sensing module, mounted on the vehicle, is used to acquire dynamic response data generated by the interaction between the vehicle and the road surface; The vehicle-mounted communication module is used to receive early warning information sent from the roadside fixed sensing module; The roadside fixed sensing modules are deployed in high-risk road sections pre-marked based on historical geological disaster data or terrain features; The vehicle central processing module is used to generate a first hazard coefficient X1 based on the vehicle environmental perception data, wherein X1 represents the deformation hazard level of the terrain environment surrounding the vehicle. Based on the dynamic response data generated by the interaction between the vehicle and the road surface, a second hazard coefficient X2 is generated, which represents the vehicle state hazard level caused by road surface anomalies; a third hazard coefficient X3 is obtained from the warning information, which is the environmental hazard level independently assessed by the roadside fixed sensing module based on the monitoring data of the high-risk road section it is stationed on; the comprehensive hazard coefficient Y is calculated according to the mathematical expression Y=W1×X1+W2×X2+W3×X3+b; where W1, W2 and W3 represent the corresponding weight coefficients, and b is a bias term; The warning signal generation module is used to compare the comprehensive risk coefficient Y with a first preset threshold, and generate and output a corresponding level of risk warning signal based on the comparison result.

7. The multimodal collaborative hazard warning device according to claim 6, characterized in that, The vehicle central processing module is specifically used for: constructing or updating a current map of the vehicle's surrounding environment in real time using the vehicle-mounted environmental perception data; extracting current environmental features from the current map and comparing the current environmental features with corresponding environmental features in a pre-stored reference map to detect deformation of mountain slopes, road shoulders, or canyon edges, and calculating their deformation amount or deformation rate; wherein the pre-stored reference map represents the state of the vehicle's surrounding environment at a previous moment; inputting the deformation amount or deformation rate into a preset mapping function to convert it into a first hazard coefficient X1 representing the deformation hazard level; wherein the larger the deformation amount or deformation rate, the higher the mapped first hazard coefficient X1.

8. The multimodal collaborative hazard warning device according to claim 6, characterized in that, The vehicle central processing module is specifically used for: calculating the mathematical expression X2= ×T+ ×S+ ×D+ The second risk factor X2 is calculated using ×E; where T represents the tire condition risk level determined based on tire pressure data and wheel speed data; S represents the road impact and deformation risk level determined based on suspension displacement data; D represents the vehicle attitude instability risk level determined based on vehicle body inertial measurement data; and E represents the road adhesion condition risk level inferred based on at least one of T, S, and D, combined with the vehicle environment perception data. , , and All are preset weighting coefficients, and satisfy the following conditions: + + + =1.

9. The multimodal collaborative hazard warning device according to claim 6, characterized in that, The third hazard factor X3 is calculated by the roadside fixed sensing module based on the following mathematical expression and is included in the warning information: ； Wherein, G represents the basic geological hazard level determined based on the geological and topographical data of the high-risk road section; R represents the hydrometeorological hazard level determined based on real-time meteorological and soil moisture data; H represents the real-time surface deformation hazard determined based on the synchronous positioning and mapping monitoring data of the roadside fixed sensing module; H represents the historical risk factor determined based on historical geological disaster data. , , and All of these are preset weighting coefficients.

10. The multimodal collaborative hazard warning device according to claim 6, characterized in that, The multimodal collaborative hazard warning device further includes a weight dynamic adjustment module, used to: increase the weight of W3 in response to determining that the vehicle enters the high-risk road section covered by the roadside fixed sensing module; increase the weight of W1 in response to the first hazard coefficient X1 exceeding a second preset threshold; and increase the weight of W2 in response to the second hazard coefficient X2 continuously exceeding a third preset threshold while the first hazard coefficient X1 does not exceed the second preset threshold.

Images