Expressway ice and snow covered pavement traffic safety risk grade prediction method

Abstract

The invention discloses a highway ice and snow covered pavement traffic safety risk grade prediction method. The method comprises the following steps: firstly, collecting vehicle operation data, meteorological data and road information data; then, dynamic response generated in the braking or accelerating process of the vehicle is analyzed, key parameters such as the speed, the acceleration and the braking force of the vehicle are combined, and the numerical value of the road surface friction coefficient is estimated; afterwards, environmental factors such as ice and snow, temperature and snowfall intensity are comprehensively considered, the road adhesion coefficient obtained through estimation is corrected, a numerical value better fitting the actual road condition is obtained, and road adhesion coefficient data finally used for risk assessment are calculated; and then, in combination with the corrected road adhesion coefficient under the ice and snow weather condition, carrying out detailed analysis on the adhesion coefficients of different areas by applying a risk assessment function. According to the method, low-cost, high-frequency and wide-coverage attachment coefficient estimation is realized, and the comprehensive risk condition of the ice and snow covered pavement can be reflected more comprehensively and stereoscopically.

Description

A method for predicting traffic safety risk levels on icy and snowy highway surfaces Technical Field

[0001] This invention belongs to the field of road traffic safety technology and relates to a method for predicting the traffic safety risk level of highways on icy and snowy roads. Background Technology

[0002] Icy and snowy roads on highways have a significant impact on road safety. Analysis shows that 10% of road accidents occur in adverse weather conditions primarily caused by ice and snow. When vehicles travel on icy and snowy roads, differences in vehicle type and speed, coupled with a significantly reduced coefficient of friction, lead to decreased friction between the vehicle and the icy surface. When vehicles are turning or need to slow down in special circumstances, the expected deceleration time deviates from the actual deceleration time, increasing braking distance and potentially causing driver stress, easily leading to traffic accidents. Therefore, how to effectively utilize weather, road, and vehicle information to assess road safety risk levels and issue traffic safety warnings has become a key research direction for preventing traffic accidents on icy and snowy roads.

[0003] Currently, the main methods for early warning of traffic safety risks on icy and snowy roads are:

[0004] (1) Risk prediction method based on meteorological conditions: This method relies on meteorological observation data or weather forecast information provided by meteorological stations or road meteorological monitoring stations to obtain road snowfall and snow accumulation conditions in the future time period, and further analyzes the impact of snowfall, snow accumulation or icing on the traffic capacity of highways. This method lacks the analysis of the real-time road adhesion coefficient of highways, and can only indirectly reflect road safety risks, and cannot accurately reflect the real-time safety of highway traffic.

[0005] (2) Risk prediction method based on road sensors: This method uses sensors and other equipment installed on roads or vehicles to collect information on the road adhesion coefficient in real time, thereby predicting the safety risks of highways. This method can accurately reflect the real-time situation of the road adhesion coefficient, but its deployment cost is high and the coverage of the sensors is limited, making it difficult to achieve large-scale popularization and application.

[0006] (3) Risk prediction method based on traffic operation status: This method indirectly infers the safety risk level of highways by collecting data on the characteristics of vehicle groups, such as speed, acceleration, vehicle spacing, and accident rate. This method can collect large-scale data using existing on-board equipment, but its prediction is lagging and is usually only predicted after the danger has already begun to occur.

[0007] In summary, existing methods for predicting road safety risks in icy and snowy weather still have some shortcomings. First, current methods extract information about vehicles traveling on highways in icy and snowy weather that is too simplistic, thus only indirectly reflecting the road risk level and failing to provide a direct analysis of road risk conditions. Second, methods for determining road surface adhesion using road sensors require expensive sensor equipment, limiting their widespread application across most highway sections. Finally, current highway risk warning methods for icy and snowy weather lack specific risk level classifications for highways. This prevents the development of targeted control measures and accident prevention measures tailored to different levels of road risk severity. Summary of the Invention

[0008] In view of the shortcomings and deficiencies of existing technologies, the purpose of this invention is to provide a method for predicting the traffic safety risk level of highway icy and snowy road surfaces. This method first collects vehicle operation data, meteorological data, and road information data through onboard equipment, road weather stations (or weather warning departments), and vehicle navigation systems, respectively, and uploads the onboard data to a meteorological data platform. Next, by analyzing the dynamic response of the vehicle during braking or acceleration, and combining key parameters such as vehicle speed, acceleration, and braking force, the road surface friction coefficient is estimated. Subsequently, considering environmental factors such as ice and snow, temperature, and snowfall intensity, the estimated road adhesion coefficient is corrected to obtain a value more closely reflecting actual road conditions, thereby calculating the final road adhesion coefficient data used for risk assessment. Next, combining the corrected road adhesion coefficient under icy and snowy weather conditions, a risk assessment function is applied to conduct a detailed analysis of the adhesion coefficient in different areas, and the corresponding risk confidence level is calculated accordingly. Finally, based on the risk level assessment results, road risks are divided into four levels, and targeted control measures are proposed according to the characteristics of each level to ensure road safety and smooth traffic flow.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A method for predicting traffic safety risk levels on icy and snowy highway surfaces, comprising the following steps:

[0011] Step S1. Snow and Ice Weather Data Collection:

[0012] Acquire vehicle operation data, meteorological data, and road information data under icy and snowy weather conditions;

[0013] Step S2. Estimation of adhesion coefficient:

[0014] Based on the vehicle operation data, meteorological data, and road information data collected in step S1, the dynamic response of the vehicle during braking or acceleration is analyzed. Combined with vehicle speed, acceleration, and braking force, the longitudinal friction utilization and lateral friction utilization are calculated. A single-peak morphology function is used to simulate the longitudinal friction model of the tire, and the peak value and uncertainty of the longitudinal friction coefficient are estimated online using EKF. Then, the minimum feasible limit value under the combination of longitudinal and lateral friction is calculated through friction ellipse constraints. Finally, the longitudinal estimation results, lateral estimation results, and friction ellipse constraints are combined, and the minimum value of the three is taken as the "maximum usable road adhesion coefficient".

[0015] Step S3. Environmental Modification:

[0016] Based on the road adhesion coefficient obtained in step S2, the road adhesion coefficient is corrected by comprehensively considering the effects of ice and snow, temperature and snowfall intensity, and finally a corrected value that is more in line with the actual road conditions is obtained. The risk prediction confidence is calculated based on the uncertainty of the corrected adhesion coefficient.

[0017] Step S4. Risk Level Prediction:

[0018] Based on the corrected road adhesion coefficient under icy and snowy weather, and combined with the risk assessment function, the calculated adhesion coefficient is divided into regions, and the risk confidence level is assessed.

[0019] Step S5. Implement road control measures based on the risk level:

[0020] Based on the scope of risk level assessment, the risk level is divided into four levels, and control measures are adopted accordingly.

[0021] As a preferred embodiment of the present invention, the vehicle operation data includes longitudinal acceleration, lateral acceleration, wheel rotation speed, braking pressure, and vehicle speed; the meteorological data includes temperature T, snow thickness D, and snowfall S; and the road information data includes road longitudinal slope angle and cross slope angle.

[0022] As a preferred embodiment of the present invention, the longitudinal friction utilization rate is:

[0023] Lateral friction utilization:

[0024] Where, μ x,util Longitudinal friction utilization; μ y,util : Utilization of lateral friction; Effective longitudinal and lateral accelerations after removing the gravitational component; m: vehicle mass; F drag Resistance; F roll Rolling resistance; θ g :Road longitudinal slope angle.

[0025] As a preferred embodiment of the present invention, the expression for the maximum road adhesion coefficient that can be utilized is:

[0026] μ avail (t)=min(μ x (t),μ y (t),μ ellipse (t),μ phys_cap );

[0027] Where: μ phys_cap ∈[0.9,1.1] is set according to tire / vehicle type; μ x (t) is the available longitudinal friction coefficient, μ x (t)=min(μ x,peak (t),μ x,util (t)+δ),δ∈[0.02,0.05],μ y (t) is the available transverse friction coefficient, μ y (t)=min(μ y,peak (t),μ y,util (t)+δ), μ x,peak (t) is the longitudinal peak friction coefficient, μ y,peak (t) represents the transverse peak friction coefficient, μ. y,peak (t) is equal to the longitudinal peak friction coefficient μ x,peak (t), μ ellipse This represents the minimum feasible limit value under the combined longitudinal and transverse friction calculated using the friction ellipse constraint.

[0028] As a preferred embodiment of the present invention, the minimum feasible limit value μ under the combination of longitudinal and transverse friction is calculated by means of the friction ellipse constraint. ellipse The expression is:

[0029]

[0030] in, ∑F z =mgcosθ g , n≈2.

[0031] As a preferred embodiment of the present invention, before correcting the road adhesion coefficient, the snow thickness, temperature, and snowfall intensity are first normalized, as expressed by:

[0032]

[0033] in: These represent the normalized snow depth, snowfall intensity, and effective temperature drop, respectively. ref S ref ΔT refThese are the corresponding reference values, usually D. ref Take 10-20mm, S ref Take 1-2 mm / h, ΔT ref Take 10-15℃, D is the snow thickness; S is the snowfall intensity; ΔT is the effective temperature drop;

[0034] The three processed terms are then substituted into the linear correction:

[0035]

[0036] Where: α is taken as 0.0015-0.0040mm -1 β is taken as 0.004-0.012°C. -1 γ is taken as 0.003-0.010h / mm, μ is the adhesion coefficient before environmental correction, and μ′ is the adhesion coefficient after environmental correction.

[0037] Then perform physical boundary clipping:

[0038] μ′←min(μ cap ,max(0,μ′)),μ cap ∈[0.9,1.1]

[0039] Where, μ cap : This represents the dry high-friction limit.

[0040] As a preferred embodiment of the present invention, the expression for the uncertainty of the corrected adhesion coefficient is:

[0041]

[0042] Where Var[μ′] is the uncertainty of the corrected adhesion coefficient. To correct for the uncertainty of the pre-adhesion coefficient, Var[D], Var[ΔT], and Var[S] are the measurement variances of snow thickness, effective temperature drop, and snowfall intensity, respectively.

[0043] The confidence level for risk prediction is c = exp(-Var[μ′] / τ), where τ is the scale.

[0044] As a preferred embodiment of the present invention, the expression of the risk assessment function is:

[0045] R = 100 × exp(-kμ′)

[0046] Where k>0, the risk is controlled to accelerate as the adhesion coefficient decreases; four levels are divided according to μ′, and when the R value exceeds the set value, the risk level corresponding to the same μ′ value is increased; when assessing the risk confidence level, standard control measures are selected to be implemented or strengthened or higher-level measures are taken based on the strength of the confidence level.

[0047] The present invention also provides an electronic device, comprising: one or more processors and a memory; wherein the memory is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described method for predicting traffic safety risk levels on icy and snowy roads of highways.

[0048] The present invention also provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described method for predicting traffic safety risk levels on icy and snowy highway surfaces.

[0049] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0050] (1) Multi-source data fusion enhances the comprehensiveness of risk perception: This invention integrates data from vehicle-mounted equipment, road weather stations, and vehicle navigation systems to achieve deep fusion of vehicle operating status, meteorological environment, and road structure information. Compared with prediction methods based on a single data source, this solution can more comprehensively and three-dimensionally reflect the overall risk status of icy and snowy roads, avoiding omissions or false alarms in early warnings due to incomplete information.

[0051] (2) Online estimation of adhesion coefficient based on vehicle dynamics, balancing accuracy, real-time performance, and low cost: Traditional methods rely on expensive road surface sensors or indirect inference based on post-hoc statistics. The former has high deployment costs and limited coverage, while the latter has predictive lag. This invention innovatively utilizes real-time data such as the vehicle's longitudinal acceleration, braking pressure, and speed, combined with a vehicle dynamics model, to calculate the road surface adhesion coefficient online. This achieves low-cost, high-frequency, and wide-coverage adhesion coefficient estimation, providing a feasible path for real-time monitoring of large-scale road networks.

[0052] (3) Introducing an environmental correction mechanism to enhance the model's adaptability and reliability in icy and snowy weather. This invention not only estimates the basic adhesion coefficient, but also integrates environmental factors such as temperature, snow thickness, and snowfall intensity to dynamically correct the adhesion coefficient. This mechanism significantly improves the model's interpretability and prediction accuracy in complex and variable icy and snowy environments, making risk assessment more consistent with actual road conditions and avoiding assessment bias caused by environmental changes.

[0053] (4) The "friction utilization degree" and "friction ellipse" constraints are proposed to improve the physical rationality and safety of the estimation results: By calculating the longitudinal and lateral friction utilization degrees and introducing the friction ellipse theory for coupling constraints, the system can more scientifically determine the stability boundary of the vehicle under the current road conditions. This method not only improves the physical consistency of the adhesion coefficient estimation, but also provides a more reliable theoretical basis for subsequent risk level classification and enhances the safety margin of the early warning system.

[0054] (5) Establishing a clear risk level classification system to support hierarchical control and precise decision-making: Addressing the issue of ambiguous risk level classification in previous methods, this invention constructs a four-level risk system based on a modified adhesion coefficient, and provides quantitative thresholds and confidence assessments for each level. This provides clear and operable decision support for traffic management departments to implement differentiated and precise control measures, improving emergency management efficiency.

[0055] (6) Achieving a leap from "static early warning" to "dynamic prediction" and supporting real-time road network risk situation awareness: This invention, through cloud computing and real-time data stream processing, can realize dynamic risk assessment and visualization of the entire road segment and continuous time series, and supports the generation and evolution prediction of road network-level risk heat maps. This changes the limitation of traditional methods that can only provide point-like or fragmented early warnings, and helps to achieve a global grasp and advanced deployment of road network safety situation.

[0056] (7) High system integration, easy to integrate and deploy with existing intelligent transportation systems: The method described in this invention can be encapsulated as a standardized software module or cloud service, and can be connected with existing traffic management platforms, vehicle terminals or navigation apps through APIs. It does not require large-scale modification of existing infrastructure, is flexible in deployment and highly scalable, and is conducive to rapid promotion and application and the formation of collaborative early warning capabilities.

[0057] (8) It is both forward-looking and practical, and supports the optimization of autonomous driving and assisted driving systems: The real-time adhesion coefficient and risk level information provided by this invention can directly serve the path planning and control decision of high-level autonomous driving vehicles, and can also provide richer road information input for assisted driving systems (such as ESC and ABS), thereby improving the driving safety and adaptability of intelligent vehicles in icy and snowy weather. Attached Figure Description

[0058] Figure 1 is a flowchart of the method for predicting traffic safety risk levels on icy and snowy roads provided by the present invention. Detailed Implementation

[0059] To enable those skilled in the art to better understand the technical solutions and advantages of the present invention, the present application will be described in detail below with reference to the accompanying drawings, but this is not intended to limit the scope of protection of the present invention.

[0060] As shown in Figure 1, this embodiment provides a method for predicting the traffic safety risk level of icy and snowy road surfaces on highways. The method includes the following steps:

[0061] Step S1. Collection of highway data

[0062] Vehicle operation data, meteorological data, and road information data are acquired using in-vehicle equipment, road weather stations, and vehicle navigation systems, respectively. The in-vehicle data is then uploaded to a meteorological data platform. Next, all data is transmitted to the cloud for processing, and a fusion method is applied to perform comprehensive data analysis.

[0063] Specifically, in this embodiment, the vehicle operation data acquisition method is as follows: A vehicle-mounted acceleration sensor (IMU) is used to measure longitudinal and lateral acceleration in real time. Wheel speed sensors (WSS) are used to compare the speed differences between the wheels, and combined with the vehicle dynamics model, to calculate the slip ratio, thereby assisting in the estimation of the road adhesion coefficient. Braking pressure p is collected through pressure sensors in the braking system. b The vehicle speed v is obtained using a vehicle speed sensor.

[0064] Meteorological data is collected by using weather data such as temperature (T), snow depth (D), and snowfall (S) obtained from road weather stations. This data is used to assist in calculating the road adhesion coefficient, thereby enhancing its reliability and accuracy. Meteorological data is provided by the roadside RWIS and third-party meteorological APIs, and accessed through a standard protocol.

[0065] The road information data collection method is as follows: information such as road type and road slope is obtained by using the road basic database and the navigation system of the experimental vehicle.

[0066] In this embodiment, vehicle data is collected via the CAN bus, uploaded to the vehicle gateway, and then transmitted to the cloud via a 4G / 5G network. Before data fusion, a timestamp synchronization mechanism is used to unify the data to second-level UTC time. Then, Kalman filtering and Bayesian updates are used to fuse data from different sources to ensure accuracy and consistency.

[0067] Step S2. Estimation of adhesion coefficient

[0068] Using the vehicle operation data, meteorological data, and road information data collected above, the longitudinal available adhesion coefficient, lateral available adhesion coefficient, combined available adhesion coefficient, and estimated variance / confidence level are calculated based on the dynamic relationship during vehicle braking or acceleration, and used to estimate the road friction coefficient (road adhesion coefficient).

[0069] Step S2.1. Sensor time alignment and preprocessing

[0070] Vehicle sensor data (such as acceleration, wheel speed, vehicle speed, and brake pressure) are unified onto a single time axis. Since the sampling frequencies and timestamps of different sensors are not entirely consistent, interpolation is needed to align the data. The signal is then filtered to remove high-frequency noise, making the calculations more stable.

[0071] Step S2.2. Calculation of transverse / longitudinal friction utilization

[0072] Slope compensation:

[0073] When a vehicle is traveling uphill or on a road with a lateral incline, the acceleration measured by the sensors includes a component of gravity. Therefore, it is necessary to remove the gravity component from the original acceleration based on the road slope (longitudinal slope angle) and cross slope angle to obtain the "effective acceleration" truly caused by the tire-road interaction.

[0074] Given the longitudinal slope angle θ of the road g With cross slope angle φ b Remove the gravitational component from the longitudinal acceleration of the IMU:

[0075]

[0076] Among them, a x IMU longitudinal acceleration, a y IMU lateral acceleration, Effective acceleration after removing the gravitational component;

[0077] Slip ratio calculation:

[0078] The tire slip ratio is calculated by comparing the actual rotational speed of the wheel (angular velocity multiplied by the tire radius) with the overall vehicle speed. The slip ratio describes the degree of wheel slippage and is a key parameter for estimating the longitudinal friction force of the tire.

[0079] For each wheel i∈{fl,fr,rl,rr}:

[0080]

[0081] Among them, κ i : Longitudinal slip ratio, R: Effective tire radius, ω i : Wheel angular velocity, v: Vehicle speed;

[0082] Dynamic load calculation:

[0083] Because a vehicle experiences a load transfer between the front and rear axles during acceleration or braking, the actual vertical load on the front and rear wheels changes. The dynamic load distribution between the front and rear axles can be estimated using vehicle mass, geometric parameters (wheelbase, center of gravity height), and longitudinal acceleration, and then redistributed to each wheel.

[0084] Let the wheelbase L = a + b, and the height of the center of gravity be h:

[0085]

[0086] Single-wheel approximation: F z,i Let θ be the load on the i-th wheel, and a and b be the distances from the center of mass to the front and rear axles, respectively. g The longitudinal slope angle of the road.

[0087] Drag and rolling resistance estimation:

[0088] In addition to the friction between the tires and the ground, vehicles are also affected by air resistance and rolling resistance. By using the vehicle's drag coefficient, air density, speed, and rolling resistance coefficient, the effects of these forces can be calculated and compensated for, avoiding interference with the estimation of the friction coefficient.

[0089] F roll =C r mg cosθ g ·sgn(v).

[0090] Among them, F drag Resistance, F roll Rolling resistance, ρ: air density, C d A: Product of aerodynamic drag area, m: Vehicle mass, C r Rolling resistance coefficient.

[0091] Friction utilization rate calculation:

[0092] Using Newton's second law, the vehicle's actual acceleration is multiplied by its mass to obtain the net force. Combined with drag and rolling resistance, the frictional force provided by the tires can be calculated. Normalizing this frictional force to the vehicle's total weight yields the "currently utilized coefficient of friction." This value reflects how much of the vehicle's frictional potential is currently being used.

[0093] Longitudinal force balance:

[0094]

[0095] Among them, F x,i Let be the longitudinal force of the i-th wheel;

[0096] Define longitudinal friction utilization:

[0097]

[0098] Where, μ x,util Longitudinal friction utilization represents the "current proportion of longitudinal friction used," which approaches the upper limit of availability under forced dynamic / strong drive; the peak value is underestimated under light load excitation.

[0099] In addition to longitudinal friction, lateral friction also needs to be considered. Ignoring crosswinds and the first-order approximation of aerodynamic side forces, the resultant lateral force... Define the degree of lateral friction utilization:

[0100]

[0101] Where, μ y,util : Transverse friction utilization.

[0102] Step S2.3. Calculation of the maximum available road adhesion coefficient (maximum road friction coefficient)

[0103] Tire model fitting:

[0104] The relationship between tire friction and slip ratio is nonlinear. Commonly used models include the "magic formula" or a simplified unimodal model. This method uses a simplified model, employing several parameters to describe the variation of friction with slip ratio. These parameters include the maximum coefficient of friction and the curve shape factor.

[0105] Using a single-peak shape function:

[0106] θ={μ max ,C1,C2}.

[0107] Longitudinal force of a single wheel: F x,i =μ x (κ i ;θ)F z,i ; where μ x (κ;θ): a simplified model of the tire's longitudinal friction; C1, C2 are the tire model's shape parameters, μ max The core parameter of this model is the longitudinal peak friction coefficient under the current road surface conditions.

[0108] This embodiment estimates the maximum coefficient of friction online:

[0109] state: Transfer: x t+1 =x t +w t Among them, w t This is process noise.

[0110] Observation (matching longitudinal resultant force):

[0111] Where, η t To observe noise.

[0112] EKF Standard Update:

[0113] predict:

[0114] P t|t-1 =P t-1 +Q

[0115] in: P represents the true state value and the estimated state value at time t-1; t|t-1 P t-1 Let be the estimation error covariance matrix of the corresponding true and estimated values; Q is the process noise.

[0116] The linearized measurement model h(x) yields Jacobi

[0117] Gain and Correction:

[0118] Kalman benefit calculation:

[0119]

[0120] Where: K t P is the Kalman gain; t|t-1 The prior estimate is the error covariance matrix; R is the Jacobian matrix of the observation matrix; R is the observation noise covariance matrix.

[0121] When estimating the road surface adhesion coefficient, K t This determines the vehicle dynamics data measured at the current moment (such as the strong braking acceleration z). t Update tire model parameters Its influence.

[0122] Status update (correction):

[0123]

[0124] in: Update the state; z represents the state prior prediction value from the prediction step; t These are the data actually observed at time t; This is the information or residual. It is the driving force behind the entire update process and represents the difference between actual observations and model predictions.

[0125] Covariance update:

[0126] P t =(IK t H t )P t|t-1

[0127] Among them, P t K is the covariance; I is the identity matrix; t H t This is the information gain matrix.

[0128] Vertical peak output and uncertainty:

[0129]

[0130] This step directly "extracts" from the EKF the estimated value of the longitudinal adhesion coefficient and its confidence index, which are of most interest to this invention.

[0131] Triggering conditions:

[0132] (e.g., 0.5-0.8 m / s) 2 ), or p b Ascend / ABS = 1; otherwise, only propagate.

[0133] EKF observation updates require a sufficiently significant "excitation" signal. If the vehicle is traveling at a constant speed and smoothly, the observation data contains almost no information about the friction limit, making updates meaningless and potentially misleading due to noise.

[0134] When the absolute value of the vehicle's effective longitudinal acceleration exceeds a threshold (e.g., 0.5-0.8 m / s²), 2 This indicates that the vehicle is significantly accelerating or braking. This dynamic change contains a wealth of information about tire-road interaction.

[0135] p b Rising: Braking pressure is increasing, indicating that the driver is actively braking.

[0136] ABS=1: The anti-lock braking system is activated. This is a direct indication that the wheels are about to lock up and the friction is approaching its limit. It is an estimate of μ. max The golden moment.

[0137] Otherwise, only propagation: If none of the above conditions are met, the EKF observation update step is not performed, and only the prediction step is performed. This is equivalent to keeping the previous estimate unchanged, while allowing the uncertainty P to increase slightly due to process noise Q (meaning that confidence in the old estimate decreases over time).

[0138] To ensure physical consistency, vertical availability consistency is achieved through integration:

[0139] μ x (t)=min(μ x,peak (t),μ x,util (t)+δ),δ∈[0.02,0.05]

[0140] This embodiment adds a small safety margin δ to the current friction utilization rate, forming a "reasonable lower limit" slightly higher than the current usage level; min(μ x,peak (t),μ x,util (t)+δ): Take the smaller of the EKF estimate and the reasonable lower bound based on current use as the final output longitudinal available friction coefficient μ. x (t).

[0141] This rule acts like a pair of "physical scissors," preventing the EKF from outputting an overly high and unsafe estimate when data stimulus is insufficient. It ensures that the final output μ... x (t) is based on model learning and always consistent with the real-time dynamics of the vehicle, resulting in conservative and safe outcomes.

[0142] Friction Ellipse Constraint:

[0143] The longitudinal and lateral frictions of a vehicle are not independent but follow a "friction ellipse" relationship. That is, if the longitudinal friction approaches its limit, the usable lateral friction will decrease. Using the friction ellipse constraint, the minimum feasible limit value under the combination of longitudinal and lateral friction can be calculated.

[0144] Using a hyperellipse n≈2 to constrain the resultant forces in both the longitudinal and transverse directions:

[0145]

[0146] in: ∑F z =mg cosθ g

[0147] Minimum feasible peak value consistent with current operating conditions:

[0148]

[0149] Overall estimate:

[0150] By combining the longitudinal estimation results, the lateral estimation results, and the friction ellipse constraint, the minimum value of the three is taken as the "usable adhesion coefficient." This ensures safety while avoiding overestimation.

[0151] Combining multiple source upper bounds, we take the uniform minimum upper bound:

[0152] μ avauil (t)=min(μ x (t),μ y (t),μ ellipse (t),μ phys_cap )

[0153] Where: μ phys_cap ∈[0.9,1.1] is set according to tire / vehicle type; μ y (t) represents the available lateral friction coefficient (available adhesion coefficient). Considering the isotropic assumption of road surface adhesion characteristics, in this embodiment, the peak lateral friction coefficient μ is set to... y,peak (t) is equal to the longitudinal peak friction coefficient μ x,peak (t), μ y (t)=min(μ y,peak (t),μ y,util (t)+δ), μ avail (t) represents the maximum available road adhesion coefficient.

[0154] Step S2.4. Uncertainty and Confidence Analysis

[0155] Uncertainty and confidence level:

[0156] Due to sensor errors and variations in operating conditions, each estimation result carries uncertainty. By propagating the covariance of the filter, the variance of the friction estimate can be given, and a "confidence index" can be further defined. This not only provides a numerical value but also reflects its reliability.

[0157] Longitudinal peak uncertainty: From EKF covariance;

[0158] Propagated to μ via function g(.) avail Approximate variance:

[0159]

[0160] Define confidence level τ is the scale hyperparameter;

[0161] Step S2.5. Output of correlation coefficient

[0162] Finally, the maximum available road adhesion coefficient, variance uncertainty (variance), and confidence level (reliability level) are calculated and output to the environmental correction platform.

[0163]

[0164] Step S3. Environmental Modification

[0165] The goal of the environmental correction module is to take into account the effects of ice, snow, temperature, and snowfall intensity on the existing adhesion coefficient to obtain a correction value that more closely reflects actual road conditions. Based on the maximum road adhesion coefficient μ mentioned above... avail By collecting environmental data, the maximum road adhesion coefficient is corrected and optimized to obtain further road adhesion coefficient data, which is then used for road risk level prediction and management in icy and snowy weather.

[0166] Step S3.1. Weather data correction

[0167] To avoid physics-inappropriate "bonus points," the three input terms are processed to be non-negative and truncated:

[0168] Temperature correction:

[0169] Temperature correction refers to the algorithmic process in this invention that dynamically adjusts the estimated road surface adhesion coefficient based on ambient temperature data. Its core objective is to quantify the negative impact of low temperatures on road surface adhesion, making the estimated adhesion coefficient more consistent with the actual physical laws under icy and snowy conditions.

[0170] If the temperature is above zero degrees Celsius, the road surface will not become slippery due to the temperature drop, so no correction is made. If the temperature is below zero degrees Celsius, the coefficient of friction will decrease according to the extent of the temperature drop.

[0171] ΔT = max(0, T0 - T)

[0172] Where: ΔT is the effective temperature drop, representing the degree to which the temperature is below the freezing point; T0 is the reference temperature threshold, usually set to 0℃. Temperature correction is activated when the ambient temperature is below this value; T: current ambient temperature (unit: ℃), obtained from a road weather station or vehicle external temperature sensor.

[0173] Corrections for snow depth and snowfall intensity:

[0174] Snowfall and snow accumulation are non-negative quantities:

[0175] The thicker the snow, the less friction there is between the vehicle and the ground. For every 1 millimeter increase in snow thickness, the coefficient of adhesion decreases.

[0176] D←max(D,0), S←max(S,0)

[0177] Where: D is the snow thickness; S is the snowfall intensity.

[0178] Step S3.2. Integrating and correcting the road surface adhesion coefficient

[0179] Scale normalization:

[0180] Scale normalization, in this invention, refers to the technical process of converting environmental variables (snow thickness, temperature change, snowfall intensity) with different physical dimensions and numerical ranges to a unified numerical range. Its core purpose is to eliminate dimensional differences, enabling different environmental variables to be compared and linearly combined at the same scale, thereby achieving a scientific correction of the road surface adhesion coefficient.

[0181]

[0182] in: These represent the normalized snow depth, snowfall intensity, and effective temperature drop, respectively. ref S ref ΔT ref These are the corresponding reference values, usually D. ref Take 10-20mm, S ref Take 1-2 mm / h, ΔT ref Set the temperature to 10–15℃;

[0183] Substitute the processed three terms into the linear correction:

[0184]

[0185] Where: α is taken as 0.0015-0.0040mm -1 β is taken as 0.004-0.012°C. -1 γ is taken as 0.003-0.010h / mm, μ is the adhesion coefficient before environmental correction, and μ′ is the adhesion coefficient after environmental correction.

[0186] Then perform physical boundary clipping to prevent out-of-bounds errors:

[0187] μ′←min(μ cap ,max(0,μ′)),μ cap ∈[0.9,1.1]

[0188] μ cap : This is the dry high friction limit (set according to tire / road type), typically 0.9 to 1.1.

[0189] Uncertainty and Quality Control:

[0190] If the uncertainty of the overall available adhesion coefficient given by the upstream is (i.e., the output of step S2.4), the environmental quantities have measurement variances Var[D], Var[ΔT], and Var[S], and they are independent of each other, then:

[0191]

[0192] Based on this, the risk prediction confidence level is defined as c = exp(-Var[μ′] / τ) (τ is the scale).

[0193] Based on the above calculations, the output {μ′,c} is provided for use by the risk prediction module.

[0194] Step S4. Risk Level Prediction

[0195] Based on the corrected road adhesion coefficient under icy and snowy weather, the calculated adhesion coefficient is divided into regions using a risk assessment function, and a risk confidence assessment is conducted.

[0196] Step S4.1. Risk Classification

[0197] Define a continuous risk assessment function:

[0198] R = 100 × exp(-kμ′)

[0199] Where k>0, the degree to which risk decreases with the decrease of the adhesion coefficient is controlled.

[0200] μ'≥0.5: Safety level (sufficient vehicle grip);

[0201] 0.5>μ'≥0.3: Medium risk (risk of slippage exists, caution is advised);

[0202] 0.3>μ'≥0.2: High risk (significantly increased vehicle braking distance, dangerous cornering);

[0203] μ'<0.2: Extremely high risk (severe slippage or even loss of control);

[0204] When the R value is high (e.g., R>60), the threshold is automatically lowered, thus increasing the risk level corresponding to the same μ′ value.

[0205] Step S4.2. Risk Confidence Assessment

[0206] The confidence level is included when outputting risk results.

[0207] Control intensity grading based on confidence level:

[0208] When c≥0.7, the strength grade is the standard strength, and the control principle is to implement the standard measures for this grade;

[0209] When 0.4≤c<0.7, the strength grade is reinforced, and the control principle is to strengthen it by 20-50% on the basis of standard measures;

[0210] When c < 0.4, the intensity level is conservative, and the control principle is to take the next higher level of measures or implement the highest intensity.

[0211] Step S5. Implement road control measures based on the risk level.

[0212] Step S5.1. Risk Classification Basis

[0213] Based on the risk level assessment scope above and in accordance with highway severe weather control measures, the risk levels are divided into four levels:

[0214] Level I (Safe / Green): μ'≥0.5;

[0215] With a high road surface adhesion coefficient, vehicles can maintain good grip even in icy and snowy weather.

[0216] Level II (Attention / Yellow): 0.5 > μ' ≥ 0.3;

[0217] When the coefficient of adhesion drops to a moderate level, the vehicle's braking distance becomes longer, and the probability of skidding increases.

[0218] Level III (High Risk / Orange): 0.3 > μ' ≥ 0.2;

[0219] With a low coefficient of friction, vehicles are prone to losing control, especially on dangerous sections of road such as curves and slopes.

[0220] Level IV (Extremely High Risk / Red): μ'<0.2;

[0221] When the risk is extremely high, continuing to pass may lead to a mass accident, and the safety of personnel must be the top priority.

[0222] Step S5.2. Tiered Control Measures

[0223] Level I (Safe / Green):

[0224] Normal traffic is permitted, but warning signs should be set up to indicate "road surface may be slippery," and the frequency of maintenance and inspections should be increased appropriately to ensure a rapid response should any issues arise.

[0225] Level II (Attention / Yellow):

[0226] Speed ​​limits will be appropriately reduced, overtaking and sudden lane changes will be prohibited, and the frequency of snow removal and salting operations will be increased to prevent rapid deterioration of road surfaces. Time-based controls will be implemented for hazardous materials vehicles and large trucks to reduce high-risk factors.

[0227] Level III (High Risk / Orange):

[0228] Significant speed limits will be imposed, and the use of certain lanes will be restricted, such as closing the fast lane. Dangerous vehicles will be restricted or rerouted, and trucks will be required to install snow chains when necessary. Maintenance departments will operate continuously, prioritizing the clearing of bridges, interchanges, and downhill sections. Traffic police and road administration will work together to rigorously inspect vehicle conditions.

[0229] Level IV (Extremely High Risk / Red):

[0230] Implement road closures or diversion measures to suspend vehicle access to high-risk road sections. Organize emergency snow and ice removal operations and deploy rescue vehicles on standby. Issue early warning information to the public through multiple channels to guide vehicles to detour or park nearby.

[0231] The present invention also provides an electronic device, comprising: one or more processors and a memory; wherein the memory is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the prediction method described above.

[0232] The present invention also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the prediction method described above.

[0233] Those skilled in the art will understand that all or part of the functions of the various methods / modules in the above embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, which may include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the above functions can be implemented by executing the program with a computer. For example, the program can be stored in the memory of a device, and when the program in the memory is executed by the processor, all or part of the above functions can be implemented.

[0234] In addition, when all or part of the functions in the above embodiments are implemented by computer programs, the programs can also be stored in storage media such as servers, other computers, disks, optical discs, flash drives, or portable hard drives. They can be downloaded or copied to the memory of the local device, or the system of the local device can be updated. When the program in the memory is executed by the processor, all or part of the functions in the above embodiments can be implemented.

[0235] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection described in the claims.

Claims

1. A method for predicting traffic safety risk levels on icy and snowy highway surfaces, characterized in that, The method includes the following steps: Step S1. Snow and ice weather data collection: Acquire vehicle operation data, meteorological data, and road information data under snow and ice weather conditions; Step S2. Adhesion coefficient estimation: Based on the vehicle operation data, meteorological data, and road information data collected in Step S1, analyze the dynamic response of the vehicle during braking or acceleration, combine vehicle speed, acceleration, and braking force to calculate the longitudinal friction utilization and lateral friction utilization, and use a single-peak morphology function to simulate the tire longitudinal friction model, and use EKF to estimate the peak value and uncertainty of the longitudinal friction coefficient online; Based on the longitudinal and lateral friction utilization and the peak values ​​of the longitudinal and lateral friction coefficients, determine the available longitudinal and lateral friction coefficients, and then calculate the minimum feasible limit value under the combination of longitudinal and lateral friction through friction ellipse constraints; Finally, the longitudinal friction coefficient is... The estimation results, lateral estimation results, and friction ellipse constraints are combined, and the minimum value of the three is taken as the "maximum usable road adhesion coefficient"; Step S3. Environmental correction: Based on the road adhesion coefficient obtained in step S2, the road adhesion coefficient is corrected by comprehensively considering the influence of ice and snow, temperature, and snowfall intensity, and finally a corrected value that is more in line with the actual road conditions is obtained. The risk prediction confidence is calculated based on the uncertainty of the corrected adhesion coefficient; Step S4. Risk level prediction: According to the corrected road adhesion coefficient under ice and snow weather, combined with the risk assessment function, the calculated adhesion coefficient is divided into regions, and the risk confidence is assessed; Step S5. Road control measures according to risk level: For the risk level assessment range, the risk level is divided into four levels, and control measures are adopted.

2. The method for predicting traffic safety risk levels on icy and snowy highway surfaces according to claim 1, characterized in that, Vehicle operation data includes longitudinal acceleration, lateral acceleration, wheel rotation speed, braking pressure, and vehicle speed; meteorological data includes temperature T, snow depth D, and snowfall S; road information data includes road longitudinal slope angle and cross slope angle.

3. A method for predicting traffic safety risk levels on icy and snowy highway surfaces according to claim 1 or 2, characterized in that, Longitudinal friction utilization: Lateral friction utilization: Where, μ x,util Longitudinal friction utilization; μ y,util : Utilization of lateral friction; Effective longitudinal and lateral accelerations after removing the gravitational component; m: vehicle mass; F drag Resistance; F roll Rolling resistance; θ g :Road longitudinal slope angle.

4. The method for predicting traffic safety risk levels on icy and snowy highway surfaces according to claim 3, characterized in that, The expression for the maximum available road adhesion coefficient is: μ avail (t)=min(μ x (t),μ y (t),μ ellipse (t),μ phys_cap ); where: μ phys_cap ∈[0.9,1.1] is set according to tire / vehicle type; μ x (t) is the available longitudinal friction coefficient, μ x (t)=min(μ x,peak (t),μ x,util (t)+δ),δ∈[0.02,0.05],μ y (t) is the available transverse friction coefficient, μ y (t)=min(μ y,peak (t),μ y,util (t)+δ), μ x,peak (t) is the longitudinal peak friction coefficient, μ y,peak (t) represents the transverse peak friction coefficient, μ. y,peak (t) is equal to the longitudinal peak friction coefficient μ x,peak (t), μ ellipse This represents the minimum feasible limit value under the combined longitudinal and transverse friction calculated using the friction ellipse constraint.

5. The method for predicting traffic safety risk levels on icy and snowy highway surfaces according to claim 4, characterized in that, The minimum feasible limit value μ under the longitudinal and transverse friction combination calculated using the friction ellipse constraint. ellipse The expression is: in, ∑F z =mg cos θ g ,n≈2.

6. A method for predicting traffic safety risk levels on icy and snowy road surfaces on highways according to any one of claims 1 to 5, characterized in that, Before correcting for the road adhesion coefficient, the snow thickness, temperature, and snowfall intensity are first normalized to their scales, as expressed by: in: These represent the normalized snow depth, snowfall intensity, and effective temperature drop, respectively. ref S ref ΔT ref These are the corresponding reference values, D ref Take 10-20mm, S ref Take 1-2 mm / h, ΔT ref Take a temperature range of 10–15℃, where D is the snow depth, S is the snowfall intensity, and ΔT is the effective temperature drop. Then, substitute the processed three terms into the linear correction: Where: α is taken as 0.0015-0.0040mm -1 β is taken as 0.004-0.012℃ -1 γ is set to 0.003-0.010 h / mm, μ is the adhesion coefficient before environmental correction, and μ′ is the adhesion coefficient after environmental correction; then physical boundary trimming is performed: μ′←min(μ cap ,max(0,μ′)),μ cap ∈[0.9,1.1] where, μ cap : This represents the dry high-friction limit.

7. A method for predicting traffic safety risk levels on icy and snowy road surfaces on highways according to any one of claims 1 to 6, characterized in that, The expression for the uncertainty of the corrected adhesion coefficient is: Where Var[μ′] is the uncertainty of the corrected adhesion coefficient. To correct for the uncertainty of the pre-adhesion coefficient, Var[D], Var[ΔT], and Var[S] are the measurement variances of snow thickness, effective temperature drop, and snowfall intensity, respectively; the risk prediction confidence level is c = exp(-Var[μ′] / τ), where τ is the scale.

8. A method for predicting traffic safety risk levels on icy and snowy road surfaces on highways according to any one of claims 1 to 7, characterized in that, The risk assessment function is expressed as: R = 100 × exp(-kμ′), where k > 0, which controls the rate at which the risk decreases with the attachment coefficient; it is divided into four levels according to μ′, and when the R value exceeds the set value, the risk level corresponding to the same μ′ value is increased; when assessing the risk confidence level, the standard control measures are selected based on the strength of the confidence level, or the standard measures are strengthened or the next higher level of measures are taken.

9. An electronic device, comprising: One or more processors and a memory; wherein the memory is used to store one or more programs, characterized in that, when the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting traffic safety risk levels on icy and snowy roads according to any one of claims 1 to 8.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for predicting traffic safety risk levels on icy and snowy roads on highways as described in any one of claims 1 to 8.

Images