Disaster event vehicle warning and blocking system based on multi-modal spatio-temporal data collaboration

Abstract

The present application belongs to the technical field of road safety and emergency disposal of sudden events, and particularly relates to a disaster event vehicle warning and blocking system based on multi-modal spatio-temporal data cooperation, which comprises a disaster event information acquisition subsystem, a driving behavior feature acquisition subsystem, a risk assessment subsystem, a management and control level division subsystem, a dynamic risk field management and control subsystem, and a vehicle warning and blocking strategy output subsystem; the disaster event information acquisition subsystem and the driving behavior feature acquisition subsystem are respectively in communication connection with the risk assessment subsystem; the risk assessment subsystem is in communication connection with the management and control level division subsystem and the dynamic risk field management and control subsystem; the management and control level division subsystem is in communication connection with the dynamic risk field management and control subsystem; and the vehicle warning and blocking strategy output subsystem is in communication connection with the dynamic risk field management and control subsystem.

Description

Disaster Event Vehicle Warning and Interception System Based on Multimodal Spatiotemporal Data Collaboration Technical Field

[0001] This invention belongs to the field of road safety and emergency response technology, specifically relating to a vehicle warning and blocking system for disaster events based on multimodal spatiotemporal data collaboration. Background Technology

[0002] Due to the impact of torrential rains, floods, and mudslides, my country frequently experiences disasters such as landslides, subsidence, and collapses of roadbeds, slopes, and bridges. Existing highway infrastructure lacks a complete set of technologies for rapid and accurate perception, intelligent monitoring and early warning, and real-time response and dissemination, making it impossible to achieve immediate automatic early warning and interception, leading to mass vehicle falls. Therefore, there is an urgent need to develop an intelligent early warning, real-time dissemination, and automatic interception vehicle warning and interception strategy auxiliary decision-making system for severe highway and bridge damage events, improving vehicle hazard avoidance capabilities and reducing casualties and property losses.

[0003] The destruction of highways and bridges poses a serious threat to the safety and stability of transportation systems, causing not only traffic disruptions and property damage but also potential chain reactions leading to casualties and secondary accidents. Especially in complex traffic environments, driver behavior plays a decisive role in responding to such emergencies. Existing research largely focuses on individual drivers or specific scenarios, lacking systematic studies on the behavioral characteristics and emergency response mechanisms of intelligent connected vehicle drivers in traffic flow disruption situations. Furthermore, traditional emergency management strategies often overlook the heterogeneity and dynamism of driver behavior, resulting in ineffective responses. Significant differences exist in perception, decision-making, and execution among drivers of different ages and experience levels. With the increasing proportion of assisted driving vehicles and intelligent connected vehicles on the road, coordinating the behavior of different types of "drivers" and establishing efficient vehicle warning and interception strategies have become critical issues urgently needing to be addressed to reduce secondary accidents. Summary of the Invention

[0004] To address the problems of existing technologies, this invention provides a vehicle warning and interception system for disaster events based on multimodal spatiotemporal data collaboration. The system aims to utilize layered collaboration and dynamic optimization strategies. On one hand, it leverages the perception and decision-making characteristics of human drivers to construct a driver behavior classification model, quantitatively assessing the risk response capabilities of different behavioral patterns (conservative, risk-taking, and conformist) in emergency scenarios. On the other hand, it considers the back-end network collaborative control capabilities of intelligent connected vehicles, designing a multi-level collaborative framework including a warning layer, a guidance layer, and an interception layer to implement targeted emergency strategies. By integrating a risk field model and probabilistic risk assessment, a rear-end collision risk propagation mechanism is constructed, enabling dynamic adjustment of the interception range threshold. This achieves precise and real-time emergency management, reduces the incidence of secondary accidents, and provides scientific basis and technical support for emergency management of highway and bridge damage.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A disaster event vehicle warning and blocking system based on multimodal spatiotemporal data collaboration, the system includes: a disaster event information collection subsystem, a driving behavior feature collection subsystem, a risk assessment subsystem, a control level division subsystem, a dynamic risk field control subsystem, and a vehicle warning and blocking strategy output subsystem;

[0007] The disaster event information collection subsystem and the driving behavior characteristic collection subsystem are respectively connected to the risk assessment subsystem; the risk assessment subsystem is connected to the control level division subsystem and the dynamic risk field control subsystem; the control level division subsystem is connected to the dynamic risk field control subsystem; and the vehicle warning and blocking strategy output subsystem is connected to the dynamic risk field control subsystem.

[0008] The disaster event information collection subsystem is used to collect disaster event information, which includes map navigation information, mobile phone location information, weather condition information, highway ETC information, highway electromechanical facility information, and social alarm information.

[0009] The driving behavior feature acquisition subsystem is used to collect driving behavior feature information, which includes: human driver behavior features, assisted driving vehicle behavior features, and autonomous driving vehicle behavior features.

[0010] The risk assessment subsystem is used to assess the risks of vehicle fall, rear-end collision, traffic congestion, and overall risks following a highway disaster.

[0011] The control level classification subsystem is used to classify different control levels based on the assessment results output by the risk assessment subsystem;

[0012] The dynamic risk field control subsystem is used to define the dynamic risk field model and analyze the risk assessment results and control levels output by the risk assessment subsystem and the control level division subsystem through the dynamic risk field model to determine emergency response measures.

[0013] The vehicle warning and blocking strategy output subsystem is used to output blocking strategies, guidance strategies, and early warning strategies.

[0014] Preferably, the disaster event information collection subsystem includes a map navigation information collector, a mobile phone positioning information collector, a meteorological condition information collector, a highway ETC information collector, a highway electromechanical facility information collector, and a social alarm information collector;

[0015] Map navigation information collectors are used to collect basic map data, dynamic navigation data, route planning data, user behavior data, and auxiliary data.

[0016] Mobile phone location information collectors are used to collect spatiotemporal trajectory data, location source information, environmental parameters, and behavioral patterns;

[0017] Meteorological condition information collectors are used to collect surface meteorological elements, upper-air sounding data, and weather phenomena.

[0018] Highway ETC information collectors are used to collect vehicle passage data;

[0019] Highway electromechanical facility information collectors are used to collect equipment monitoring data, environmental monitoring information, traffic control information, and emergency data.

[0020] Social alarm information collectors are used to collect initial warning information about natural disasters.

[0021] Preferably, the driving behavior feature acquisition subsystem includes a human driver behavior feature acquisition device, an assisted driving vehicle behavior feature acquisition device, and an autonomous driving vehicle behavior feature acquisition device;

[0022] The human driver behavior data collector is used to collect driver physiological behavior and environmental data;

[0023] The driver assistance vehicle behavior feature collector is used to collect vehicle dynamic data, driver takeover data, sensor fusion data, and system status data.

[0024] The autonomous vehicle behavior feature collector is used to collect raw data from lidar, cameras, millimeter-wave radar, and ultrasonic sensors.

[0025] Preferably, the risk assessment subsystem includes a vehicle fall risk assessment module, a rear-end collision risk assessment module, a traffic congestion risk assessment module, and a comprehensive risk assessment module; the risk assessment subsystem outputs vehicle fall risk assessment results, rear-end collision risk assessment results, traffic congestion risk assessment results, and comprehensive risk assessment results respectively based on disaster event information and driving behavior characteristic information.

[0026] Preferably, the control layer division subsystem includes an interception layer division module, a guidance layer division module, and an early warning layer division module.

[0027] Preferably, the dynamic risk field model defined by the dynamic risk field control subsystem is as follows:

[0028]

[0029] Kinetic energy calculation + road adhesion coefficient correction + V2X collaborative risk coefficient;

[0030] Among them, R total The total risk perceived by the driver, n is the number of risk items, and w i Let λ be the weight of the i-th risk term, λ be the kinetic energy correction coefficient, and R be the weight of the i-th risk term. i Let E be the risk value of the i-th risk item. kinetic For dynamic risk, θ i Let m be the original risk value of the i-th risk item, m be the vehicle mass, v be the vehicle speed, and V2X be the coordination between the vehicle and roads, pedestrians, and other vehicles using Internet of Things (IoT) technology.

[0031] Preferred, updated dynamic risk field:

[0032] R ped (t)=D·e -β·t ·e -k·d(t) ;

[0033] Among them, R ped The risk generated by the dynamic update of road blockage events is t, where t is time, D is the risk intensity coefficient of the dynamic risk field, β is the control risk, k is the attenuation coefficient, and d is the distance between the risk source and the target vehicle.

[0034] Preferably, the dynamic risk field management subsystem is a VLIW server equipped with a road disaster event risk field model.

[0035] Preferably, the vehicle warning and blocking strategy output subsystem includes an interception strategy output module, a guidance strategy output module, and a warning strategy output module.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] The vehicle warning and blocking system for disaster events based on multimodal spatiotemporal data collaboration provided by this invention includes: a disaster event information collection subsystem, a driving behavior feature collection subsystem, a risk assessment subsystem, a control level division subsystem, a dynamic risk field control subsystem, and a vehicle warning and blocking strategy output subsystem. The disaster event information collection subsystem and the driving behavior feature collection subsystem are respectively communicatively connected to the risk assessment subsystem. The disaster event information collection subsystem is used to collect multi-source spatiotemporal data such as map navigation information, mobile phone location information, meteorological condition information, highway ETC information, highway electromechanical facility information, and social alarm information. The driving behavior feature collection subsystem is used to collect driving behavior features such as human driver behavior features, assisted driving vehicle behavior features, and autonomous driving vehicle behavior features. The risk assessment subsystem... The system communicates with the control level division subsystem. The risk assessment subsystem is used to assess the risks of vehicle falls, rear-end collisions, traffic congestion, and overall risks after a highway disaster. The control level division subsystem is used to divide different control levels, including interception, guidance, and early warning, based on the assessment results output by the risk assessment subsystem. The control level division subsystem also communicates with the dynamic risk field control subsystem. The dynamic risk field control subsystem is used to define a dynamic risk field model and analyze the risk assessment results and control levels output by the risk assessment subsystem and the control level division subsystem through the dynamic risk field model to determine emergency response measures. The vehicle warning and blocking strategy output subsystem communicates with the dynamic risk field control subsystem and is used to output interception, guidance, and early warning strategies.

[0038] By applying the system of this invention, based on the behavioral characteristics of different drivers such as traditional vehicles and intelligent connected vehicles on the road, it outputs vehicle warning and blocking strategies for the early warning layer, guidance layer, and interception layer. This solves the current problems of insufficient layered collaboration and difficulty in dynamic optimization in highway bridge disaster events. Targeting the different risk characteristics of congestion risk, rear-end collision risk, and fall risk, it improves the efficiency of multi-level collaborative emergency response by integrating risk field models and probabilistic risk assessments. It can dynamically adjust the threshold of the interception range, achieve precise and real-time emergency management, reduce the incidence of secondary accidents, and provide a scientific basis and technical support for emergency management of highway bridge damage. Attached Figure Description

[0039] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 is a schematic diagram of the structure of the disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to an embodiment of the present invention;

[0041] Figure 2 is a schematic diagram of the handling process of the disaster event vehicle warning and blocking system based on multimodal spatiotemporal data collaboration according to an embodiment of the present invention. Detailed Implementation

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

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Example 1

[0045] This invention provides a disaster event vehicle warning and blocking system based on multimodal spatiotemporal data collaboration, including: a disaster event information collection subsystem, a driving behavior feature collection subsystem, a risk assessment subsystem, a control level division subsystem, a dynamic risk field control subsystem, and a vehicle warning and blocking strategy output subsystem;

[0046] The disaster event information collection subsystem and the driving behavior characteristic collection subsystem are respectively connected to the risk assessment subsystem; the risk assessment subsystem is connected to the control level division subsystem and the dynamic risk field control subsystem; the control level division subsystem is connected to the dynamic risk field control subsystem; and the vehicle warning and blocking strategy output subsystem is connected to the dynamic risk field control subsystem.

[0047] The disaster event information collection subsystem is used to collect disaster event information, which includes map navigation information, mobile phone location information, weather condition information, highway ETC information, highway electromechanical facility information, and social alarm information.

[0048] The driving behavior feature acquisition subsystem is used to collect driving behavior feature information, which includes: human driver behavior features, assisted driving vehicle behavior features, and autonomous driving vehicle behavior features.

[0049] The risk assessment subsystem is used to assess the risks of vehicle fall, rear-end collision, traffic congestion, and overall risks following a highway disaster.

[0050] The control level classification subsystem is used to classify different control levels based on the assessment results output by the risk assessment subsystem;

[0051] The dynamic risk field control subsystem is used to define the dynamic risk field model and analyze the risk assessment results and control levels output by the risk assessment subsystem and the control level division subsystem through the dynamic risk field model to determine emergency response measures.

[0052] The vehicle warning and blocking strategy output subsystem is used to output blocking strategies, guidance strategies, and early warning strategies.

[0053] Furthermore, the system of this invention achieves full-element traffic risk perception through a multimodal spatiotemporal data collaboration mechanism, and the specific implementation process is as follows:

[0054] In terms of multi-source data acquisition architecture, the disaster event information acquisition subsystem integrates multiple sensor arrays such as visual perception (deploying 4K high-definition cameras and infrared thermal imagers), environmental perception (configuring lidar and millimeter-wave radar), and communication perception (accessing 5G-V2X vehicle-road cooperative modules); the driving behavior feature acquisition subsystem adopts multi-modal data fusion, including human drivers (collecting eye movement trajectories and hand movements through DMS cameras), driver assistance systems (extracting longitudinal / lateral control data from ADAS systems), and autonomous driving units (acquiring planned paths and perception decision logs);

[0055] Regarding the spatiotemporal data collaborative processing flow, the process involves three main steps: First, spatiotemporal alignment, which synchronizes video frames, radar point clouds, and vehicle positioning data to centimeter-level accuracy based on BeiDou satellite positioning coordinates and high-precision maps. Second, feature extraction, which uses the YOLOv8 algorithm for multi-target detection and combines it with BiLSTM (Bi-directional Long Short-Term Memory) to extract temporal features of driving behavior. Third, dynamic fusion, which involves designing an adaptive weight allocation network to automatically adjust the weights of each modality of data according to weather conditions (e.g., increasing the weight of radar data to 0.7 on rainy days).

[0056] Furthermore, spatiotemporal alignment includes:

[0057] 1. Coordinate transformation model (BeiDou → High-precision map):

[0058] [X map ,Y map Z map ] T =[ΔX,ΔY,ΔZ] T +(1+δ)·R(ε x ,ε y ,ε z )·[X BDS ,Y BDS Z BDS ] T ;

[0059] Among them, [Xmap ,Y map Z map [X] represents the coordinates in a high-precision map coordinate system, and T represents the transpose; BDS ,Y BDS Z BDS [ΔX, ΔY, ΔZ] represents the coordinates in the BeiDou coordinate system (BDS); [ΔX, ΔY, ΔZ] represents the translation parameters, indicating the offset between the origins of the two coordinate systems; δ is the scale factor, used to adjust the scale difference between the two coordinate systems; R(ε) x ,ε y ,ε z ) is a rotation matrix, consisting of three Euler angles ε. x ,ε y ,ε z Composition; ε x ,ε y ,ε z These represent the rotation angles around the X, Y, and Z axes, respectively.

[0060] 2. Time synchronization function:

[0061] t aligned =t base +((t sensor -t base start ) / (t base end -t base start ))·Δt;

[0062] Among them, t aligned The aligned timestamp; t base The starting time or reference time point of the baseline timeline; t sensor For the sensor's original timestamp; t base start t is the start time of the baseline timeline; base end Δt represents the end time of the baseline timeline; Δt is the time interval or period used for scaling.

[0063] Furthermore, feature extraction includes:

[0064] 1. YOLOv8 multi-object detection loss function:

[0065] L total =λ1·L cls +λ2·L box +λ3·L obj ;

[0066] Among them, L total L represents the total loss value used to train the object detection model. cls For classification loss; L boxThis refers to the Bounding Box Loss; L obj λ1, λ2, and λ3 are the weight coefficients of classification loss, localization loss, and confidence loss, respectively.

[0067] 2. BiLSTM temporal feature extraction:

[0068] Forward propagation:

[0069]

[0070] Backpropagation:

[0071]

[0072] Feature fusion:

[0073]

[0074] in, The hidden state of the feedforward LSTM (Long Short-Term Memory) network at time step t; This represents the hidden state of the forward LSTM at time step t-1; This represents the hidden state of the backward LSTM at time step t; Let x be the hidden state of the backward LSTM at time step t+1; t The input feature vector at time step t; The input x corresponds to the forward and backward propagation respectively. t The weight matrix; The weight matrices corresponding to the hidden state h in the forward and backward propagation, respectively; b f b b These correspond to the bias vectors for forward and backward propagation, respectively; σ represents the activation function (usually Sigmoid or Tanh); h t This is the final output feature vector of the BiLSTM at time step t, which is formed by concatenating the forward and backward hidden states.

[0075] Furthermore, dynamic fusion includes:

[0076] Adaptive weight allocation network:

[0077] w i =exp(f θ (c weather ) i ) / (Σ j exp(f θ(c weather ) j ));

[0078] Among them, w i f represents the calculated fusion weight of the i-th data modality; θ Let c represent a lightweight fully connected neural network with parameter θ; weather For external conditions (such as weather), the feature encoding vector is f. θ (c weather ) i Σ represents the raw score output by the network for the i-th data mode; j This represents summing over all modes j; the formula is essentially a softmax function, which applies to the neural network f. θ The output is converted into a probability distribution, and all fusion weights w i Satisfy Σ i w i =1, and 0≤w i ≤1; f θ Encode c based on the external conditions of the input. weather Adaptively assign importance weights to different modalities.

[0079] All sensor data achieves nanosecond-level time alignment via a time synchronization module (PTPv2 protocol). Spatial registration employs the ICP algorithm, with 50 iterations and a convergence threshold of 0.01m. Data preprocessing includes: denoising (wavelet thresholding), calibration (lens distortion correction), and format conversion (ROS message type standardization). This invention integrates trimodal driving behavior data (human eye movement trajectory / ADAS control data / autonomous driving path planning), overcoming the limitations of traditional systems that only collect vehicle dynamic data.

[0080] Furthermore, the disaster event information collection subsystem includes: map navigation information collector, mobile phone positioning information collector, meteorological condition information collector, highway ETC information collector, highway electromechanical facility information collector, and social alarm information collector;

[0081] The map navigation information collector is used to collect basic map data, dynamic navigation data, route planning data, user behavior data, and auxiliary data. Basic map data includes road networks (highways / national highways / provincial highways, etc.), natural geographic features (rivers / green spaces), and the names and locations of points of interest (POIs). Dynamic navigation data includes real-time traffic flow, road congestion, traffic accidents, and construction information. Route planning data includes optimal route calculation, estimated travel time, and fuel consumption estimation. User behavior data includes navigation search records, frequently visited locations, and preferred routes. Auxiliary data includes intersection views, lane guidance signs, and voice navigation commands.

[0082] Mobile phone location information collectors are used to collect spatiotemporal trajectory data, location source information, environmental parameters, and behavioral patterns. Spatiotemporal trajectory data includes latitude and longitude coordinates, movement speed, timestamp, and altitude. Location source information includes GPS signal strength, Wi-Fi access point MAC address, and base station cell ID. Environmental parameters include positioning accuracy (such as meter-level error range) and signal obstruction status. Behavioral patterns include dwell time, speed, acceleration, and direction.

[0083] Meteorological condition information collectors are used to collect surface meteorological elements, upper-air sounding data, and weather phenomena. Surface meteorological elements include temperature, humidity, air pressure, precipitation, wind speed / direction, visibility, sunshine hours, evaporation, and soil temperature and humidity. Upper-air sounding data includes geopotential height, wind direction, and wind speed at different altitudes (such as 500 hPa). Weather phenomena include thunderstorms, icing, fog, haze, and sandstorms.

[0084] Highway ETC information collectors are used to collect vehicle passage data, including vehicle type classification, entrance / exit toll stations, and passage time.

[0085] Highway electromechanical facility information collectors are used to collect equipment monitoring data, environmental monitoring information, traffic control information, emergency data, etc. Equipment monitoring data includes power supply system, communication system, and lighting facility information; environmental monitoring information includes CO concentration in tunnels, road surface ice thickness, and noise decibel values; traffic control information includes variable message sign display content, traffic light timing schemes, and vehicle detector traffic statistics; and emergency data includes emergency telephone call records, broadcast voice content, and emergency lane occupancy detection.

[0086] Social alarm information collectors are used to collect initial alarm information on major traffic accidents, highway collapses and interruptions, bridge and tunnel damage, earthquakes, floods and other natural disasters.

[0087] Furthermore, the driving behavior feature acquisition subsystem includes a human driver behavior feature acquisition device, an assisted driving vehicle behavior feature acquisition device, and an autonomous driving vehicle behavior feature acquisition device.

[0088] Human driver behavior data collectors mainly rely on in-vehicle cameras and sensors to collect driver physiological behavior and environmental data for scenarios such as fatigue driving warning and distraction detection. This includes driving operation data such as accelerator / brake pedal pressure, steering wheel turning speed, and gear shifting frequency, as well as environmental interaction data such as the percentage of time spent looking at the road ahead, rearview mirror / instrument panel scanning rate, voice command content, and emotional fluctuations (through voiceprint analysis).

[0089] The assisted driving vehicle behavior feature collector is used to collect vehicle dynamic data, driver takeover data, sensor fusion data, and system status data. It integrates multi-sensor data to monitor vehicle dynamic response and driver takeover behavior, and is applied to L2-L3 level assisted driving systems. Vehicle dynamic data includes motion state and control signal information. Motion state information includes longitudinal acceleration, yaw rate, and vehicle pitch angle. Control signal information includes AEB (Automatic Emergency Braking) trigger count and ACC (Adaptive Cruise Control) following distance. Driver takeover data includes takeover response and operational intervention information. Takeover response information includes the takeover delay time after system prompt and vehicle stability after takeover (e.g., frequency of emergency braking). Operational intervention information includes steering wheel correction angle and accelerator / brake coverage. Sensor fusion data includes environmental perception and positioning information. Environmental perception includes millimeter-wave radar target list (distance, speed, azimuth) and camera lane line recognition results. Positioning information includes BeiDou satellite navigation system fusion trajectory and map matching accuracy. System status data includes algorithm operation and fault record information. Algorithm operation information includes perception module CPU / GPU utilization and decision tree inference time. Fault record information includes sensor false alarm rate and system restart count.

[0090] The autonomous vehicle behavior feature collector is used to collect raw data from various sensors such as LiDAR, cameras, millimeter-wave radar, and ultrasonic sensors. This includes vehicle status data such as vehicle speed, acceleration, steering angle, and braking status, as well as environmental data such as road conditions, weather conditions, and traffic light status. The collected data is processed in real time to extract key features and identify and extract vehicle behavior features, such as driving trajectory, lane changing behavior, and obstacle avoidance strategies. Based on machine learning algorithms, the specific behavior patterns of the vehicle are identified.

[0091] Furthermore, the risk assessment subsystem includes modules for vehicle fall risk assessment, rear-end collision risk assessment, traffic congestion risk assessment, and comprehensive risk assessment. Following a highway bridge collapse, upstream vehicles within the core area of ​​the incident face the risk of falling, upstream vehicles outside the core area face the risk of rear-end collision due to sudden braking by vehicles ahead, and other upstream vehicles face the risk of traffic congestion due to the sudden interruption of the road. Based on disaster event information and driving behavior characteristics, the risk assessment subsystem uses a dynamic risk field model to output vehicle fall risk assessment results, rear-end collision risk assessment results, traffic congestion risk assessment results, and comprehensive risk assessment results, respectively.

[0092] Furthermore, the control level division subsystem divides the upstream area into control levels based on the risk assessment results output by the risk assessment subsystem, including an interception layer division module, a guidance layer division module, and an early warning layer division module.

[0093] Furthermore, the dynamic risk field management subsystem is used to define the dynamic risk field model and analyze the risk assessment results and management levels output by the risk assessment subsystem and the management level classification subsystem through the dynamic risk field model to determine emergency response measures. The dynamic risk field model treats highway disasters as core hazards, generating a risk field with each potential hazard in the environment (vehicles, pedestrians, obstacles, etc.). Driver behavior is considered a response to the "risk potential field." Drivers adjust their behavior (such as acceleration, deceleration, and lane changes) by perceiving the gradient of the comprehensive risk field. The dynamic risk field model proposes a quantitative risk assessment method, distinct from traditional static threshold judgment, specifically:

[0094]

[0095] Among them, R total The overall risk value is given by n, where n is the number of risk items, and w is the total risk value. i Let λ be the weight of the i-th risk term, λ be the kinetic energy correction coefficient, and R be the weight of the i-th risk term. i Let E be the risk value of the i-th risk item. kinetic This is a dynamic risk.

[0096] Weights of each risk item w i The multi-source data weighting algorithm based on Softmax is as follows:

[0097]

[0098] Where, θ i Let be the original risk value of the i-th risk item;

[0099] Dynamic Risk E kinetic The calculation formula is:

[0100] Kinetic energy calculation + road adhesion coefficient correction + V2X collaborative risk coefficient;

[0101] Where m is the vehicle mass, v is the vehicle speed, and V2X is the intelligentization of the entire road transportation system by using Internet of Things (IoT) technology to coordinate between vehicles and roads, pedestrians, and other vehicles.

[0102] A dynamic risk field model is used for risk assessment. This model combines the operating parameters of both vehicles and the risk of the preceding vehicle to establish a risk propagation mechanism for rear-end collisions. The core of risk field theory lies in quantifying the driver's perceived risk and transforming it into behavioral decisions. For example, factors such as the speed and position of surrounding vehicles, road curvature, and visibility can all affect the calculation of the risk field. Each factor can be considered a risk source, and its impact diminishes with distance. Drivers adjust their driving behavior based on the intensity of the overall risk field. Risk field theory simulates the driver's dynamic decision-making process in complex traffic scenarios by quantifying their risk perception of the surrounding environment.

[0103] The first step is to define the risk source and design the field function. In the risk field of highway disaster events, the risk caused by road blockage increases exponentially as the distance between the vehicle and the blockage event decreases, specifically as follows:

[0104] R rbe (d)=A·e k·d ;

[0105] In a vehicle risk field, the risk posed by surrounding vehicles decreases with distance, specifically as follows:

[0106] R veh (d)=A·e -m·d ;

[0107] Among them, R rbe R veh The risk posed to surrounding vehicles is denoted by d, where d is the distance between the risk source and the target vehicle, A is the risk intensity coefficient of the vehicle risk field, e is the exponent, and k and m are the attenuation coefficients.

[0108] In the risk field of highway boundaries, the risk field at the road edge or guardrail increases with the increase of lateral offset, specifically:

[0109]

[0110] Among them, R road Risks arising from road edges or guardrails, where y represents the lateral position of the vehicle, |yy edge | represents the location of the road boundary, B represents the risk intensity coefficient of the highway boundary risk field, and ∈ prevents the denominator from being zero.

[0111] The second step is to define other risks. In the static obstacle risk field, the risk field of a fixed obstacle can be set as a Gaussian distribution with a fixed range, specifically:

[0112]

[0113] Among them, R obsThe risk is posed by a fixed obstacle, (x,y) is the position of the target vehicle, C is the risk intensity coefficient of other risks, (x0,y0) is the center of the obstacle, and σ is the control influence range.

[0114] The third step is to calculate the overall risk field. The total risk field perceived by the driver is the superposition of all risk sources, specifically:

[0115]

[0116] Among them, R total Total risk perceived by the driver.

[0117] The fourth step is driving decision modeling, where the driver adjusts their behavior based on the risk field gradient, with the goal of minimizing risk exposure.

[0118] In terms of acceleration decision-making, the speed is adjusted according to the gradient of the risk field ahead, and an improved intelligent driving model is adopted:

[0119]

[0120] Where a is acceleration, α is risk sensitivity coefficient, and a IDM This represents the risk sensitivity coefficient of the risk field ahead. This represents the gradient of the risk field.

[0121] Regarding lane-changing decisions, if the current lane risk field intensity exceeds the threshold R... th Triggering lane change intention:

[0122] Lane change condition: R total >R th And the target lane has a lower risk;

[0123] Fifth, update the dynamic risk field, taking into account the impact of real-time changes in the risk of road blockage events, such as the increasing area of ​​road collapses and the continuous expansion of the incident zone. The risk field needs to be updated in real time.

[0124]

[0125] Among them, R ped The risk generated by the dynamic update of road blockage events is represented by t, where t is time, D is the risk intensity coefficient of the dynamic risk field, and β is the control risk, which increases with the blockage time, such as the increase in the length of the upstream vehicle queue and the expansion of the road collapse area.

[0126] Furthermore, the dynamic risk field management subsystem is a VLIW server equipped with a road disaster event risk field model.

[0127] Furthermore, the vehicle warning and blocking strategy output subsystem includes an interception strategy output module, a guidance strategy output module, and a warning strategy output module, which are used to output interception strategies, guidance strategies, and warning strategies.

[0128] In summary, the disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration provided by this invention includes: a disaster event information collection subsystem, a driving behavior feature collection subsystem, a risk assessment subsystem, a control level division subsystem, a dynamic risk field control subsystem, and a vehicle warning and interception strategy output subsystem. The disaster event information collection subsystem and the driving behavior feature collection subsystem are respectively communicatively connected to the risk assessment subsystem. The disaster event information collection subsystem is used to collect multi-source spatiotemporal data such as map navigation information, mobile phone location information, meteorological condition information, highway ETC information, highway electromechanical facility information, and social alarm information. The driving behavior feature collection subsystem is used to collect driving behavior features such as human driver behavior features, assisted driving vehicle behavior features, and autonomous driving vehicle behavior features. The risk assessment subsystem... The subsystem communicates with the control level division subsystem. The risk assessment subsystem is used to assess the risks of vehicle falls, rear-end collisions, traffic congestion, and overall risks after a highway disaster. The control level division subsystem is used to divide different control levels, including interception, guidance, and early warning layers, based on the assessment results output by the risk assessment subsystem. The control level division subsystem also communicates with the dynamic risk field control subsystem. The dynamic risk field control subsystem is used to define a dynamic risk field model and analyze the risk assessment results and control levels output by the risk assessment subsystem and the control level division subsystem to determine emergency response measures. The vehicle warning and blocking strategy output subsystem communicates with the dynamic risk field control subsystem and is used to output interception, guidance, and early warning strategies.

[0129] By applying the system of this invention, based on the behavioral characteristics of different drivers such as traditional vehicles and intelligent connected vehicles on the road, it outputs vehicle warning and blocking strategies for the early warning layer, guidance layer, and interception layer. This solves the current problems of insufficient layered collaboration and difficulty in dynamic optimization in highway bridge disaster events. Targeting the different risk characteristics of congestion risk, rear-end collision risk, and fall risk, it improves the efficiency of multi-level collaborative emergency response by integrating risk field models and probabilistic risk assessments. It can dynamically adjust the threshold of the interception range, achieve precise and real-time emergency management, reduce the incidence of secondary accidents, and provide a scientific basis and technical support for emergency management of highway bridge damage.

[0130] Example 2

[0131] As shown in Figure 1, the present invention provides a disaster event vehicle warning and blocking system based on multimodal spatiotemporal data collaboration, including: a disaster event information collection subsystem 101, a driving behavior feature collection subsystem 102, a risk assessment subsystem 103, a control level division subsystem 104, a dynamic risk field control subsystem 105, and a vehicle warning and blocking strategy output subsystem 106.

[0132] Among them, the disaster event information collection subsystem 101 and the driving behavior characteristic collection subsystem 102 are respectively connected to the risk assessment subsystem 103. The risk assessment subsystem 103 is connected to the control level division subsystem 104 and the dynamic risk field control subsystem 105. The control level division subsystem 104 is connected to the dynamic risk field control subsystem 105. The vehicle warning and blocking strategy output subsystem 105 is connected to the dynamic risk field control subsystem 106.

[0133] Specifically, the disaster event information collection subsystem 101 is used to collect multi-source spatiotemporal data such as map navigation information, mobile phone positioning information, meteorological condition information, highway ETC information, highway electromechanical facility information, and social alarm information; the driving behavior characteristic collection subsystem 102 is used to collect driving behavior characteristics such as human driver behavior characteristics, assisted driving vehicle behavior characteristics, and autonomous driving vehicle behavior characteristics; the risk assessment subsystem 103 is used to assess the risk of vehicle falling, rear-end collision, traffic congestion, and comprehensive risks after a highway disaster event; and the control level division subsystem 104 is used to divide different control levels based on the assessment results output by the risk assessment subsystem, including the interception layer, guidance layer, and early warning layer.

[0134] Subsequently, the dynamic risk field control subsystem 105 is used to define the dynamic risk field model, and analyzes the risk assessment results and control levels output by the risk assessment subsystem 103 and the control level division subsystem 104 through the dynamic risk field model to determine emergency response measures; the vehicle warning and blocking strategy output subsystem 106 outputs the interception strategy, guidance strategy and early warning strategy.

[0135] By applying the system of this invention, based on the behavioral characteristics of different drivers such as traditional vehicles and intelligent connected vehicles on the road, it outputs vehicle warning and blocking strategies for the early warning layer, guidance layer, and interception layer. This solves the current problems of insufficient layered collaboration and difficulty in dynamic optimization in highway bridge disaster events. Targeting the different risk characteristics of congestion risk, rear-end collision risk, and fall risk, it improves the efficiency of multi-level collaborative emergency response by integrating risk field models and probabilistic risk assessments. It can dynamically adjust the threshold of the interception range, achieve precise and real-time emergency management, reduce the incidence of secondary accidents, and provide a scientific basis and technical support for emergency management of highway bridge damage.

[0136] Furthermore, the disaster event information collection subsystem 101 provided by the present invention includes: a map navigation information collector 1011, a mobile phone positioning information collector 1012, a meteorological condition information collector 1013, a highway ETC information collector 1014, a highway electromechanical facility information collector 1015, and a social alarm information collector 1016.

[0137] The map navigation information collector 1011 is used to collect basic map data, dynamic navigation data, route planning data, user behavior data, and auxiliary data. Basic map data includes road networks (highways / national highways / provincial highways, etc.), natural geographic features (rivers / green spaces), and the names and locations of points of interest (POIs). Dynamic navigation data includes real-time traffic flow, road congestion, traffic accidents, and construction information. Route planning data includes optimal route calculation, estimated travel time, and fuel consumption estimation. User behavior data includes navigation search records, frequently visited locations, and preferred routes. Auxiliary data includes intersection real-view images, lane guidance signs, and voice navigation commands.

[0138] The mobile phone location information collector 1012 is used to collect spatiotemporal trajectory data, location source information, environmental parameters, behavior patterns, etc. The spatiotemporal trajectory data includes latitude and longitude coordinates, movement speed, timestamp, altitude, etc. The location source information includes GPS signal strength, Wi-Fi access point MAC address, base station cell ID, etc. The environmental parameters include positioning accuracy (such as meter-level error range), signal obstruction status, etc. The behavior patterns include dwell time, speed, acceleration, direction, etc.

[0139] The meteorological condition information collector 1013 is used to collect ground meteorological elements, upper-air sounding data, and weather phenomena. Ground meteorological elements include temperature, humidity, air pressure, precipitation, wind speed / direction, visibility, sunshine hours, evaporation, and soil temperature and humidity. Upper-air sounding data includes geopotential height, wind direction, and wind speed at different altitudes (such as 500 hPa). Weather phenomena include thunderstorms, icing, fog, haze, and sandstorms.

[0140] The 1014 highway ETC information collector is used to collect vehicle passage data, including vehicle type classification, entrance / exit toll stations, and passage time.

[0141] The Highway Electromechanical Facilities Information Collector 1015 is used to collect equipment monitoring data, environmental monitoring information, traffic control information, emergency data, etc. Equipment monitoring data includes power supply system, communication system, and lighting facility information; environmental monitoring information includes CO concentration in tunnels, road surface ice thickness, and noise decibel value; traffic control information includes variable message sign display content, traffic light timing scheme, and vehicle detector traffic statistics; and emergency data includes emergency telephone call records, broadcast voice content, and emergency lane occupancy detection.

[0142] The Social Alarm Information Collector 1016 is used to collect initial alarms for major traffic accidents, highway collapses and interruptions, bridge and tunnel damage, earthquakes, floods and other natural disasters.

[0143] For example, the map navigation information collector 1011 is a high-precision GNSS (Global Navigation Satellite System) module (such as the Jisibao MG8 series), which supports Beidou / GPS / GLONASS multi-mode positioning with centimeter-level accuracy and has a built-in high-sensitivity antenna, making it suitable for complex terrain.

[0144] The 1012 mobile phone location information collector is a portable detection device (such as ZXTY-DW900A), which uses 4G / 5G signal simulation + IMSI code capture, has a positioning accuracy of 0.1 meters, and supports concurrent positioning of 100+ targets.

[0145] The Xiang Condition Information Acquisition Unit 1013 is a multi-parameter weather station (such as the JZ-PH type), integrating 16 sensors including temperature, humidity, air pressure, wind speed, rainfall, and radiation. It supports RS485 / GPRS / 4G transmission and has a storage capacity of 12M (capable of storing one year's worth of data).

[0146] The highway ETC information collector 1014 is an integrated ETC cabinet (such as the Conexant C2000), which includes an integrated device of RSU (Roadside Unit) antenna controller, lane controller, UPS power supply (Uninterruptible Power Supply) and environmental monitoring module.

[0147] The 1015 highway electromechanical facility information collector is a lightning protection monitoring terminal (such as SPD in ZVD), achieving IP20 protection and an operating range of -40℃ to 85℃. It monitors lightning peak value, grounding resistance, temperature and humidity. The intelligent sensor network consists of current / voltage sensors and LoRa (Long Range Radio) wireless transmission modules, which monitor power supply anomalies and provide early warnings of faults in real time.

[0148] The 1016 social alarm information collector is a telephone alarm host (such as a multi-channel acquisition and output system), with 20 alarm input channels, supports RS485 networking, and has a built-in AT24C02 memory chip.

[0149] The map navigation information collector 1011, mobile phone location information collector 1012, meteorological condition information collector 1013, highway ETC information collector 1014, highway electromechanical facility information collector 1015, and social alarm information collector 1016 provided in this embodiment of the invention comprehensively collect data on traffic flow, infrastructure, road environment, and other elements after a highway bridge disaster, providing reliable and accurate data support for safety risk assessment.

[0150] Furthermore, the driving behavior feature acquisition subsystem 102 provided by the present invention includes a human driver behavior feature acquisition device 1021, an assisted driving vehicle behavior feature acquisition device 1022, and an autonomous driving vehicle behavior feature acquisition device 1023. The human driver behavior feature acquisition device 1021 mainly relies on vehicle-mounted cameras and sensors to collect driver physiological behavior and environmental data for scenarios such as fatigue driving warning and distraction monitoring. This includes driving operation data such as accelerator / brake pedal pressure, steering wheel turning angle speed, and gear shifting frequency, as well as environmental interaction data such as the percentage of time spent looking at the road ahead, rearview mirror / instrument panel scanning rate, voice command content, and emotional fluctuations (through voiceprint analysis).

[0151] The 1022 driver assistance vehicle behavior feature collector is used to collect vehicle dynamic data, driver takeover data, sensor fusion data, system status data, etc. It integrates multi-sensor data to monitor vehicle dynamic response and driver takeover behavior, and is applied to L2-L3 level driver assistance systems. Vehicle dynamic data includes motion state and control signal information. Motion state information includes longitudinal acceleration, yaw rate, and vehicle pitch angle. Control signal information includes AEB (Automatic Emergency Braking) trigger count and ACC (Adaptive Cruise Control) following distance. Driver takeover data includes takeover response and operation intervention information. Takeover response information includes takeover delay time after system prompt and vehicle stability after takeover (such as emergency braking frequency). Operation intervention information includes steering wheel correction angle and accelerator / brake coverage force. Sensor fusion data includes environmental perception and positioning information. Environmental perception includes millimeter-wave radar target list (distance, speed, azimuth) and camera lane line recognition results. Positioning information includes Beidou satellite navigation system fusion trajectory and map matching accuracy. System status data includes algorithm operation and fault record information. Algorithm operation information includes perception module CPU / GPU utilization and decision tree inference time. Fault record information includes sensor false alarm rate and system restart count.

[0152] The autonomous vehicle behavior feature collector 1023 is used to collect raw data from various sensors such as LiDAR, cameras, millimeter-wave radar, and ultrasonic sensors. This includes vehicle state data such as vehicle speed, acceleration, steering angle, and braking status, as well as environmental data such as road conditions, weather conditions, and traffic light status. It is applied to L3 and above intelligent connected vehicles, processes the collected data in real time, extracts key features, and identifies and extracts vehicle behavior features such as driving trajectory, lane changing behavior, and obstacle avoidance strategies. Based on machine learning algorithms, it identifies specific behavior patterns of the vehicle.

[0153] For example, the human driver behavior feature collector 1021 is a DMS camera (non-invasive), a DAMS system that complies with the national standard GB / T41797-2022, supports behavior monitoring such as closed eyes, abnormal head posture, and making and receiving phone calls, with a day / night detection rate of ≥95%, a response delay of <1.5 seconds, and supports optical + acoustic + tactile multimodal prompts.

[0154] The 1022 driver assistance vehicle behavior feature collector consists of a 77GHz millimeter-wave radar and surround-view cameras. The Minshi STONKAM 77GHz radar (costing approximately 500 RMB per unit) has a detection range of >200m and supports AEB and ACC functions. It also features four 1080P surround-view cameras (covering a 360° field of view) and a Mobileye EyeQ4 chip to achieve lane line detection and traffic sign recognition.

[0155] The 1023 autonomous vehicle behavior feature collector is a solid-state LiDAR+4D radar, Luminar Hydra (solid-state, angular resolution 0.05°×0.05°), supports a detection distance of 250m, meets automotive-grade reliability requirements, and Arbe Phoenix (4 times higher resolution), supports multi-target tracking in complex scenarios.

[0156] The embodiments of the present invention do not limit the form of the data collector, which can be determined by those skilled in the art according to actual needs.

[0157] Furthermore, the dynamic risk field control subsystem 105 is a VLIW (Very Long Instruction Word) server equipped with a road environmental element interference model. It breaks down VLIW instructions into sub-instructions of multiple functional units (FUs), supporting parallel operations such as arithmetic, logic, and memory access, providing high-bandwidth read / write support, and reducing data conflicts. The VLIW server includes multiple data access ports for connecting to the risk assessment subsystem 103 and the control level division subsystem 104, respectively, to analyze risk assessment results and control levels, and determine emergency response measures.

[0158] Furthermore, the VLIW server analyzes emergency response measures based on the following dynamic risk field model:

[0159]

[0160] Among them, R total The overall risk value is given by n, where n is the number of risk items, and w is the total risk value. i Let λ be the weight of the i-th risk term, λ be the kinetic energy correction coefficient, and R be the weight of the i-th risk term. i Let E be the risk value of the i-th risk item. kinetic This is a dynamic risk.

[0161] Weights of each risk item w i The multi-source data weighting algorithm based on Softmax is as follows:

[0162]

[0163] Where, θ i Let be the original risk value of the i-th risk item;

[0164] Dynamic Risk E kinetic The calculation formula is:

[0165] Kinetic energy calculation + road adhesion coefficient correction + V2X collaborative risk coefficient;

[0166] Where m is the vehicle mass, v is the vehicle speed, and V2X is the intelligentization of the entire road transportation system by using Internet of Things (IoT) technology to coordinate between vehicles and roads, pedestrians, and other vehicles.

[0167] A dynamic risk field model is used for risk assessment. This model combines the operating parameters of both vehicles and the risk of the preceding vehicle to establish a risk propagation mechanism for rear-end collisions. The core of risk field theory lies in quantifying the driver's perceived risk and transforming it into behavioral decisions. For example, factors such as the speed and position of surrounding vehicles, road curvature, and visibility can all affect the calculation of the risk field. Each factor can be considered a risk source, and its impact diminishes with distance. Drivers adjust their driving behavior based on the intensity of the overall risk field. Risk field theory simulates the driver's dynamic decision-making process in complex traffic scenarios by quantifying their risk perception of the surrounding environment.

[0168] The first step is to define the risk source and design the field function. In the risk field of highway disaster events, the risk caused by road blockage increases exponentially as the distance between the vehicle and the blockage event decreases, specifically as follows:

[0169] R rbe (d)=A·e k·d ;

[0170] In a vehicle risk field, the risk posed by surrounding vehicles decreases with distance, specifically as follows:

[0171] r veh (d)=A·e -m·d ;

[0172] Among them, R rbe R veh The risk posed to surrounding vehicles is denoted by d, where d is the distance between the risk source and the target vehicle, A is the risk intensity coefficient of the vehicle risk field, e is the exponent, and k and m are the attenuation coefficients.

[0173] In the risk field of highway boundaries, the risk field at the road edge or guardrail increases with the increase of lateral offset, specifically:

[0174]

[0175] Among them, R road Risks arising from road edges or guardrails, where y represents the lateral position of the vehicle, |yy edge | represents the location of the road boundary, B represents the risk intensity coefficient of the highway boundary risk field, and ∈ prevents the denominator from being zero.

[0176] The second step is to define other risks. In the static obstacle risk field, the risk field of a fixed obstacle can be set as a Gaussian distribution with a fixed range, specifically:

[0177]

[0178] Among them, R obs The risk is posed by a fixed obstacle, (x,y) is the position of the target vehicle, C is the risk intensity coefficient of other risks, (x0,y0) is the center of the obstacle, and σ is the control influence range.

[0179] The third step is to calculate the overall risk field. The total risk field perceived by the driver is the superposition of all risk sources, specifically:

[0180]

[0181] Among them, R total Total risk perceived by the driver.

[0182] The fourth step is driving decision modeling, where the driver adjusts their behavior based on the risk field gradient, with the goal of minimizing risk exposure.

[0183] In terms of acceleration decision-making, the speed is adjusted according to the gradient of the risk field ahead, and an improved intelligent driving model is adopted:

[0184]

[0185] Where a is acceleration, α is risk sensitivity coefficient, and a IDM This represents the risk sensitivity coefficient of the risk field ahead. This represents the gradient of the risk field.

[0186] Regarding lane-changing decisions, if the current lane risk field intensity exceeds the threshold R... th Triggering lane change intention:

[0187] Lane change condition: R total >R th And the target lane has a lower risk;

[0188] Fifth, update the dynamic risk field, taking into account the impact of real-time changes in the risk of road blockage events, such as the increasing area of ​​road collapses and the continuous expansion of the incident zone. The risk field needs to be updated in real time.

[0189] R ped (t)=D·e -β·t ·e -k·d(t) ;

[0190] Among them, R ped The risk generated by the dynamic update of road blockage events is represented by t, where t is time, D is the risk intensity coefficient of the dynamic risk field, and β is the control risk, which increases with the blockage time, such as the increase in the length of the upstream vehicle queue and the expansion of the road collapse area.

[0191] Furthermore, the vehicle warning and blocking strategy output subsystem 106 includes an interception strategy output module 1061, a guidance strategy output module 1062, and a warning strategy output module 1063, which are used to output interception strategies, guidance strategies, and warning strategies.

[0192] Specifically, the interception strategy output module 1061, guidance strategy output module 1062, and early warning strategy output module 1063 all utilize electronic sand table displays. LED or LCD screens integrate GIS maps, supporting gesture operation and combining VR / AR technology to achieve real-time traffic monitoring and AR navigation interfaces. A 270° circular sand table is constructed using a three-fold seamless splicing technology. The surface integrates a capacitive sensing layer to achieve millimeter-level touch precision. The underlying layer uses LiDAR scanning to acquire a 1:500 three-dimensional terrain model, overlaid with real-time traffic heatmaps and emergency resource vector markers. A dynamic rendering engine supports four-quadrant split-screen display, simultaneously presenting composite information such as current event impact range prediction and contingency plan simulation. The GIS holographic projection wall consists of 4×6 unit 110-inch micro-pitch LED modules, supporting 8K resolution point cloud map projection, displaying strategy schemes for interception sections (red), guidance sections (yellow), and early warning sections (blue). Edge computing nodes achieve millisecond-level GIS data refresh, and ray tracing technology is used to present a three-dimensional road network.

[0193] Example 3

[0194] As shown in Figure 2, this invention provides a method for warning and blocking vehicles involved in disaster events based on multimodal spatiotemporal data collaboration. It is implemented using the disaster event vehicle warning and blocking system based on multimodal spatiotemporal data collaboration described in the preceding embodiments. The method includes:

[0195] The disaster event information collection subsystem collects disaster event information, and the driving behavior characteristic collection subsystem collects driving behavior characteristic information.

[0196] The risk assessment subsystem outputs vehicle fall risk assessment results, rear-end collision risk assessment results, traffic congestion risk assessment results, and comprehensive risk assessment results based on disaster event information and driving behavior characteristics information.

[0197] The control level division subsystem divides the upstream area into control levels based on the risk assessment results output by the risk assessment subsystem;

[0198] The dynamic risk field control subsystem defines a dynamic risk field model, defines risk sources and design field functions, analyzes risk assessment results and control levels through the dynamic risk field model, calculates the comprehensive risk field, updates the dynamic risk field, and determines emergency response measures.

[0199] The vehicle warning and blocking strategy output subsystem outputs interception strategies, guidance strategies, and early warning strategies based on emergency response measures.

[0200] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration, characterized in that, The system includes: a disaster event information collection subsystem, a driving behavior characteristic collection subsystem, a risk assessment subsystem, a control level classification subsystem, a dynamic risk field control subsystem, and a vehicle warning and blocking strategy output subsystem. The disaster event information collection subsystem and the driving behavior characteristic collection subsystem are respectively communicatively connected to the risk assessment subsystem. The risk assessment subsystem is communicatively connected to the control level classification subsystem and the dynamic risk field control subsystem. The control level classification subsystem is communicatively connected to the dynamic risk field control subsystem. The vehicle warning and blocking strategy output subsystem is communicatively connected to the dynamic risk field control subsystem. The disaster event information collection subsystem is used to collect disaster event information, including map navigation information, mobile phone location information, weather condition information, highway ETC information, highway electromechanical facility information, and social alarm information. The driving behavior characteristic collection subsystem is used to collect driving behavior characteristic information, including: human driver behavior characteristics, assisted driving vehicle behavior characteristics, and other relevant information. The system includes: a vehicle behavior characteristics subsystem and an autonomous vehicle behavior characteristics subsystem; a risk assessment subsystem for assessing vehicle fall risk, rear-end collision risk, traffic congestion risk, and overall risk after a highway disaster; a control level classification subsystem for classifying different control levels based on the assessment results output by the risk assessment subsystem; a dynamic risk field control subsystem for defining a dynamic risk field model and analyzing the risk assessment results and control levels output by the risk assessment subsystem and the control level classification subsystem to determine emergency response measures; and a vehicle warning and blocking strategy output subsystem for outputting interception, guidance, and warning strategies. The dynamic risk field model treats highway disasters as core hazards, generating a risk field with each potential hazard in the environment, and considers driver behavior as a response to the "risk potential field." Drivers adjust their behavior by perceiving the gradient of the comprehensive risk field. The dynamic risk field model proposes a quantitative risk assessment method, distinct from traditional static threshold judgment, specifically: ； ； Among them, R total Let n represent the total risk perceived by the driver, and n be the number of risk items. For the first The weights of each risk item, where λ is the kinetic energy correction coefficient. For the first The risk value of each risk item As for dynamic risks, For the first The original risk value of each risk item, For vehicle quality, For vehicle speed, This technology utilizes the Internet of Things (IoT) to enable coordination between vehicles, roads, pedestrians, and other vehicles.

2. The disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to claim 1, characterized in that, The disaster event information collection subsystem includes a map navigation information collector, a mobile phone location information collector, a meteorological condition information collector, a highway ETC information collector, a highway electromechanical facility information collector, and a social alarm information collector. The map navigation information collector is used to collect basic map data, dynamic navigation data, route planning data, user behavior data, and auxiliary data. The mobile phone location information collector is used to collect spatiotemporal trajectory data, location source information, environmental parameters, and behavioral patterns. The meteorological condition information collector is used to collect ground meteorological elements, upper-air detection data, and weather phenomena. The highway ETC information collector is used to collect vehicle traffic data. The highway electromechanical facility information collector is used to collect equipment monitoring data, environmental monitoring information, traffic control information, and emergency data. The social alarm information collector is used to collect initial warning information about natural disasters.

3. The disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to claim 1, characterized in that, The driving behavior feature acquisition subsystem includes a human driver behavior feature acquisition device, an assisted driving vehicle behavior feature acquisition device, and an autonomous driving vehicle behavior feature acquisition device. The human driver behavior feature acquisition device is used to collect driver physiological behavior and environmental data. The assisted driving vehicle behavior feature acquisition device is used to collect vehicle dynamic data, driver takeover data, sensor fusion data, and system status data. The autonomous driving vehicle behavior feature acquisition device is used to collect raw data from lidar, cameras, millimeter-wave radar, and ultrasonic sensors.

4. The disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to claim 1, characterized in that, The risk assessment subsystem includes a vehicle fall risk assessment module, a rear-end collision risk assessment module, a traffic congestion risk assessment module, and a comprehensive risk assessment module. Based on disaster event information and driving behavior characteristics, the risk assessment subsystem outputs vehicle fall risk assessment results, rear-end collision risk assessment results, traffic congestion risk assessment results, and comprehensive risk assessment results.

5. The disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to claim 1, characterized in that, The control layer division subsystem includes an interception layer division module, a guidance layer division module, and an early warning layer division module.

6. The disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to claim 1, characterized in that, Update dynamic risk field: ;in, The risks arising from the dynamic updating of road blockage events, For time, This represents the risk intensity coefficient of a dynamic risk field. To control risks, The attenuation coefficient is... The distance between the risk source and the target vehicle.

7. The disaster event vehicle warning and blocking system based on multimodal spatiotemporal data collaboration according to claim 6, characterized in that, The dynamic risk field management subsystem is a VLIW server equipped with a road disaster event risk field model.

8. The disaster event vehicle warning and interception system based on multimodal spatiotemporal data collaboration according to claim 1, characterized in that, The vehicle warning and blocking strategy output subsystem includes an interception strategy output module, a guidance strategy output module, and a warning strategy output module.

Images