Method for predicting risk of collision between electric bicycle and vehicle based on multi-modal behavior of electric bicycle

Abstract

The invention belongs to the field of road vehicle control systems, and relates to a method for predicting the risk of collision between an electric bicycle and a vehicle based on the multi-modal behavior of the electric bicycle, which utilizes physical modeling and takes the influence of the braking distance and the transverse and longitudinal stability of the electric bicycle on the vehicle as a basic risk; automobile risks caused by the environment are further considered, and multi-dimensional environment risk assessment modeling is carried out by integrating four dimensions of dynamic distance-speed, adjacent vehicle types, traffic flow density and weather conditions (precipitation, temperature, wind speed and visibility); the risk of electric bicycle operation behaviors (acceleration change and steering angle and frequency) to the automobile is combined; the complex relation between the environment change and the whole vehicle risk is captured in real time by using the environment sensitivity coefficient, and finally a set of comprehensive, accurate and real-time multi-dimensional comprehensive risk assessment method is provided for the vehicle for the multi-mode riding behavior of the electric bicycle, so that the collision risk in the complex traffic flow environment is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of road vehicle control systems and relates to the prediction of upcoming collision risks, and in particular to a method for predicting the collision risk of an electric bicycle with a vehicle based on its multimodal behavior. Background Art

[0002] China is currently the country with the largest number of motor vehicles in the world. By 2024, the national motor vehicle ownership will reach 450 million, and it will continue to grow at a high rate, ranking first in the world in both total volume and increment. At the same time, China also has the world's largest electric bicycle market. As one of the most popular means of transportation in contemporary times, cars provide an efficient and flexible way for personal travel, greatly improving the convenience of life and the efficiency of time utilization. Electric bicycles are a green, environmentally friendly, people-oriented, and highly autonomous national industry. Both are showing a positive development trend in terms of market size, industrial chain development, technological progress, policy support, and market prospects.

[0003] With the increase in the number of motor vehicles and the development of electric bicycles, the negative impact they have on urban road traffic has become increasingly prominent, and motor vehicle accidents caused by electric bicycles have increased year by year. At present, the main measure taken to solve the problem of motor vehicle-non-motor vehicle conflicts is to install a blind spot video monitoring system on the car side. However, due to the flexible and changeable characteristics of electric bicycles, even if the blind spot monitoring system feeds back the traffic operation status in the blind spot to the car driver, it is difficult for the driver to take timely action, and the accident rate cannot be effectively reduced.

[0004] In order to predict the risk of machine-non-machine conflict in advance, scholars have designed a machine-non-machine conflict risk assessment system. However, the following problems are common in the existing machine-non-machine conflict risk assessment systems: (1) Most of them rely on traffic simulation models (VISSIM, AIMSUN, etc.) to simulate multiple levels such as traffic flow, traffic lights, non-motor vehicles and pedestrian behavior. The factors they consider are broad but not precise, and the operating characteristics of electric bicycles (such as lateral and longitudinal stability, acceleration changes) are rarely considered, making it difficult to make a targeted and refined risk assessment of the automobile risks caused by electric bicycles. In addition, the existing conflict models between motor vehicles and motor vehicles, and between motor vehicles and pedestrians mostly start from the analysis of motor vehicle driving behavior (such as lane changes, sudden braking) and pedestrian behavior analysis (such as walking speed, forward direction), without specifically considering the operating characteristics of electric bicycles, and cannot be applied to risk assessment for electric bicycles. In addition, the existing risk assessment systems for electric bicycles are even rarer, resulting in low accuracy in the risk identification of electric bicycles by cars in complex traffic scenarios.

[0005] (2) Focusing only on a single factor. For example, TTC (Time to Collision) calculates the time when a collision may occur based on the relative distance and relative speed of two objects, thereby determining the magnitude of the collision risk. Since the selected indicators are too single and cannot handle complex nonlinear behaviors, it is difficult to make a more accurate risk assessment for cars with electric bicycles that have high flexibility in complex scenarios.

[0006] (3) Most existing studies assume that road conditions are good and ignore the impact of dynamic changes in many factors such as traffic flow density, adjacent vehicle types, and weather conditions on the risk assessment system. Some studies even use the road geometry captured by map point cloud information as the environmental condition, resulting in a narrow scope of application of the risk assessment system, weak generalization ability, and low accuracy of the assessment results.

[0007] Chinese patent CN112581756A discloses "a driving risk assessment method based on mixed traffic", which uses a social force model to uniformly express the force relationship between motor vehicles, non-motor vehicles and road environment, and establishes a risk model for interactive collisions between motor vehicles and non-motor vehicles, thereby realizing dynamic driving risk assessment in mixed traffic scenarios. However, this method simply abstracts the individual behavior results of any motor vehicle or non-motor vehicle in the traffic scene into a combination of different forces, ignoring the differences in microscopic traffic individual types and the potential impact of changes in behavioral characteristics (such as the lateral and longitudinal stability, steering angle and frequency of non-motor vehicles) on driving risks. At the same time, the mechanical model used adopts simple linear superposition for different risk scenarios, which makes it difficult to respond in a timely manner to sudden traffic changes, and it is difficult to fully meet the needs of complex traffic environments in practical applications.

[0008] Chinese patent CN112015842B discloses a “Risk Assessment Method and System for Autonomous Driving Vehicles Based on Bicycle Trajectory Prediction”, which combines the cyclist’s motion characteristics and map point cloud information, and outputs the cyclist’s possible location through intention inference and trajectory prediction. However, when performing trajectory prediction, this method relies on a 2D static map that can only reflect the geometric structure of the road to reflect environmental factors, and ignores dynamically changing environmental conditions (such as weather conditions and traffic flow density), resulting in poor adaptability and robustness of the model in complex environments, insufficient generalization ability, and thus reducing the accuracy of the prediction results. At the same time, its information data needs to be positioned on a digital map before collection. Over-reliance on precise map information has resulted in relatively limited application scenarios and cumbersome data acquisition conditions.

[0009] Therefore, it is necessary to design a method to solve the traffic conflict problem between cars and electric bicycles, improve the traffic experience of travelers, and avoid the risk of road traffic accidents caused by it. Summary of the invention

[0010] In view of the above-mentioned technical problems and defects, the purpose of the present invention is to provide a method for predicting the risk of collision between electric bicycles and vehicles based on the multimodal behavior of electric bicycles. The method collects dynamic information of non-motor vehicles in real time, combines external environmental factors, and uses multi-dimensional modeling and intelligent algorithms to dynamically evaluate potential risks in a mixed traffic environment. This method can not only accurately predict the collision risk caused by electric bicycles and capture high-risk behaviors in a timely manner, but also adjust vehicle control strategies through environmental sensitivity, reduce the occurrence of traffic accidents, and improve traffic safety levels and driving experience.

[0011] To achieve the above object, the present invention adopts the following technical solution: A method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of the electric bicycle, the method comprising the following steps: Step 1. Risk the vehicle Breaking down the risks of electric bikes to cars , Environmental risks to automobiles and the risks of electric bicycle operation to cars ; Step 2. Evaluate the risk of electric bikes’ inherent attributes to cars , Environmental risks to automobiles and the risks of electric bicycle operation to cars ; Among them, the inherent properties of electric bicycles pose risks to cars. Refers to the risks that electric bicycles bring to cars during driving due to their inherent properties, including the risks caused by the braking distance of electric bicycles , the risks that electric bicycle stability poses to cars ; ； in, is the proportionality coefficient, is the power index, Corrected reaction time for the rider, is the speed of the electric bicycle, is the acceleration of the e-bike; ; in, is the proportionality coefficient, is the power index, It is the lateral stability index of electric bicycles. It is the longitudinal stability index of electric bicycle. , is the weight coefficient, , , is the nonlinear index; Environmental risks to cars It refers to the risk caused to the car by changes in external environmental factors during the driving process, including the risk caused by dynamic distance-speed , the risk posed to the car by the type of adjacent vehicles , the risks posed by traffic density to cars , Risks of weather conditions to cars ; When evaluating the risk posed to the car by dynamic distance-speed, the relative speed and distance between the car and the first car behind it are comprehensively considered; when evaluating the risk posed to the car by adjacent vehicle types, the risks posed to the car by all vehicle types within the detection range are calculated, and the highest risk is taken for the risk prediction of the whole vehicle; when evaluating the risk posed to the car by traffic flow density, the risk function posed to the car by traffic flow density is constructed in the form of Gaussian distribution; when evaluating the risk posed to the car by weather conditions, the risks posed to the car by precipitation, temperature, wind speed and visibility are comprehensively considered, that is, the risk posed to the car by weather conditions includes the risk posed to the car by precipitation , Risks of temperature to cars , the risks of wind speed to cars and visibility risks to cars ; Risks of E-bike Operation to Vehicles Refers to the risk to the driving safety of the car caused by the driving behavior of the rider during the driving process, including the risk caused by the frequency of acceleration changes of electric bicycles to the car , Risks of electric bicycle steering angle and frequency to cars ; ; in, , , Represents the weights of different acceleration change frequency bands, is the corrected acceleration rate of the electric bicycle, ， is the acceleration change rate of the electric bicycle, is the speed of the electric bicycle, To control the adjustment coefficient of the speed amplification effect on the acceleration change frequency; ; in, is the steering angle of the electric bicycle, To adjust the steering angle The constant coefficient for the risk impact, is the speed multiplication factor, which controls the exponential effect of speed on the steering angle risk, is the distance between the car and the electric bicycle, The coefficient representing the influence of steering frequency on risk, represents time, i.e. the time window for observing the steering frequency, Unit time Number of steering operations within It represents the adjustment coefficient of the speed on the amplification effect of the acceleration change frequency. is the interaction risk term coefficient between steering angle and frequency; Step 3. Predict vehicle risk based on the capital asset pricing model and introduce the environmental sensitivity coefficient , assess the sensitivity of environmental changes to vehicle risks; in addition, add and The interaction term between is used to describe the superimposed effect of electric bicycle operation behavior on vehicle risk under different environmental conditions; the high-order effect is added to reflect the nonlinear trend of environmental risk impact; Vehicle Risk The prediction model is: ; in, is the interaction risk coefficient, with a value range of [0.1, 1.0], which reflects and The synergistic effects between is the power of the higher-order effect, which determines the speed and magnitude of the impact of environmental risk changes on the overall risk. value ; is the high-order environmental sensitivity coefficient, and its value range is [0.05, 0.5].

[0012] As a preferred embodiment of the present invention, the inherent properties of electric bicycles pose risks to automobiles. The evaluation model is: ; in, , , is the nonlinear index, The value range is [0.5, 2.0], The value range of is [0.5, 2.0]; The value range is [0.5, 3.0].

[0013] As a preferred embodiment of the present invention, the automobile risk caused by the environment The evaluation model is: ; in, , , , Represents the weight coefficient of each risk factor, The value range is [0.25, 0.5], The value range is [0.05, 0.2], The value range is [0.15, 0.3], The value range is [0.2, 0.4]; and finally satisfies ; represents the power of each risk's contribution to the overall risk, The value range is [1.5, 3.0], The value range is [1.0, 2.0], The value range is [1.2, 2.5], Value range: [1.5, 3.0]; for The maximum value in .

[0014] As a preferred embodiment of the present invention, the risk of electric bicycle operation to the car The evaluation model is: ; in, , is the weight coefficient, The value range is [0.4, 0.7], Value range: [0.3, 0.6]; , is the nonlinear index, The value range is [1.0, 2.0]; Value range: [1.5, 3.0]; The value range is [1.0, 2.0].

[0015] As a preferred embodiment of the present invention, the risk of dynamic distance-speed to the car The evaluation model is: ; in, is the relative speed risk coefficient, is the critical value of relative velocity, , , To adjust the parameters, is the relative speed between the car and the first car behind it, is the distance between the car and the first car behind it, is the safe distance between the car and the first car behind it, As a minimum value, take 10 -6 ~10 -9 .

[0016] As a preferred embodiment of the present invention, the risk caused by the adjacent vehicle type to the car The evaluation model is: ; in, , , is the weight coefficient, is the size of the vehicle model, For the quality of the car model, is the blind area size.

[0017] As a preferred embodiment of the present invention, the risk of traffic flow density to cars The evaluation model is: ; in, represents the traffic flow density, It refers to the point with the highest impact on vehicle driving risk under medium traffic density. is the risk sensitivity coefficient of the low-density area, is the risk sensitivity coefficient of high-density areas, is the risk change rate in low-density areas, is the risk change rate in high-density areas, It is an adjustment parameter with a value range of [0.1, 0.5], which is used to control the overall risk level under low traffic flow density.

[0018] As a preferred embodiment of the present invention, the risk of weather conditions on the car The evaluation model is: ; in, , , , Indicates the weight coefficient of each factor, reflecting the contribution of different weather factors to the total risk; The value range is [0.2, 0.4], The value range is [0.1, 0.3], The value range is [0.1, 0.3], The value range is [0.2, 0.4]; and finally satisfies ; It represents the power of each risk's contribution to the overall risk, reflecting the nonlinear impact of each factor. The value range is [1.0, 2.0], The value range is [1.5, 3.0], The value range is [1.0, 2.5], The value range is [0.5, 1.5]; among them, for The maximum value in .

[0019] As a preferred embodiment of the present invention, the risk of precipitation on the car The evaluation model is: ; in, Indicates the minimum risk under dry conditions, which is a minimum value of 10 -6 ~10 -9 ; The coefficient of risk increase due to the decrease in friction factor, is the amount of precipitation, is the friction coefficient of dry road surface, is the friction coefficient under moderate precipitation, is the minimum friction coefficient under heavy precipitation, , are the critical values ​​of small, medium and large precipitation. is the exponential decrease coefficient in the initial precipitation stage, indicating the decrease in the initial friction coefficient. is the logarithmic decrease coefficient in the medium precipitation stage, indicating the effect of increased precipitation on the friction coefficient; Temperature risks to cars The evaluation model is: ; in, Indicates temperature; Indicates whether there is ice or snow on the road. If yes, it is 1; if no, it is 0. It represents the minimum risk under suitable temperature, which is a minimum value, taking 10 -6 ~10 -9 ; The coefficient indicating the increase in risk due to temperature drop, indicating the degree of impact of temperature drop on risk, is the friction coefficient of dry road surface, The temperature at which the road begins to freeze. The critical temperature that controls the rate at which the friction coefficient decreases; Risks of wind speed to cars The evaluation model is: ; in, is the spatial interaction coefficient, which reflects the possibility of an electric bicycle occupying a car lane or colliding with a car after losing control, and depends on the lane width and crosswind intensity; is the traffic flow coefficient, which indicates the amplification effect of the out-of-control electric bicycle on the traffic flow composed of cars. Indicates wind speed and the angle of incidence The risk of electric bicycle rollover under the conditions of Visibility risks to cars The evaluation model is: ; in, Indicates the current visibility. It represents the basic risk value, that is, the minimum risk under good visibility, and is set to 10 -6 ~10 -9 ; represents the risk adjustment parameter, which is used to control the impact of reduced visibility on risk. Indicates critical visibility.

[0020] As a preferred embodiment of the present invention, the environmental sensitivity coefficient The initial value range is [0.05, 0.5]. Subsequent iterative training is carried out based on the existing data to finally obtain the optimal value under different traffic scenarios. The calculation formula is: ; in, Risks for the entire vehicle Automotive risks to the environment The covariance of Automobile risks caused by the environment fluctuations; The risks of cars to the environment The variance of represents the fluctuation range of environmental risk.

[0021] Advantages and beneficial effects of the present invention: (1) The present invention provides a method for predicting the risk of collision between an electric bicycle and a vehicle based on its multimodal behavior, which can effectively solve the problem that the traditional traffic risk assessment system has poor adaptability to complex mixed traffic scenes. The method comprehensively considers the impact of a series of complex factors on the driving of the vehicle, such as the lateral and longitudinal stability of the electric bicycle, acceleration changes, traffic flow density of the vehicle, and weather conditions, and can better handle the coupling effects and complex interdependencies between various factors; the method selects a variety of indicators that can adapt to nonlinear changes, and can make more accurate vehicle risk assessments for highly flexible electric bicycles and complex and changing environments.

[0022] (2) The present invention utilizes physical modeling, taking the braking distance of electric bicycles and the impact of their lateral and longitudinal stability on automobiles as the basic risk; further considering the automobile risks caused by the environment, the present invention integrates the dynamic distance-speed, adjacent vehicle types, traffic flow density, and weather conditions (precipitation, temperature, wind speed, and visibility) into four dimensions to conduct multi-dimensional environmental risk assessment modeling; the risk to the automobile caused by the operation behavior of electric bicycles (acceleration changes and steering angles and frequencies) is combined; the environmental sensitivity coefficient is used to capture the complex relationship between environmental changes and vehicle risks in real time, and finally, a comprehensive, accurate, and real-time multi-dimensional comprehensive risk assessment method is provided for the automobile based on the multi-modal riding behavior of electric bicycles. The method comprehensively considers various situations that the automobile may face during driving and has strong practicality; at the same time, the model performs nonlinear modeling for a variety of risk scenarios, thereby enhancing the generalization ability of the model, thereby effectively improving the driving safety of the vehicle and the road traffic efficiency.

[0023] (3) The present invention is useful in assessing the automobile risks caused by the environment When using independent power and differentiated weights, it is possible to better consider the impact of different types of environmental risks on The nonlinear influence of The benchmark power of the proposed method ensures the accuracy of risk assessment while balancing the computational efficiency and sensitivity. It adopts adjustable parameters and can be flexibly adjusted according to the complex environmental changes in different traffic scenarios, which has strong adaptability. Compared with complex methods such as traditional neural networks, this method can be optimized with a small amount of data, and has the significant advantage of being lightweight, especially when applied to embedded systems, and can meet the needs of accurately assessing the automobile risks caused by the environment in a short time.

[0024] (4) The method provided by the present invention has the advantages of real-time, dynamic and multi-dimensional comprehensive analysis. It can capture high-risk behaviors in a timely manner and provide a basis for providing early warning or automatic control solutions to drivers, thereby effectively reducing the probability of traffic accidents and improving driving safety and traffic efficiency in mixed traffic scenarios. It provides an innovative solution for intelligent traffic management in complex road environments.

[0025] (5) When evaluating the risk of traffic density, an environmental factor, to automobiles, the present invention adopts a method combining piecewise Gaussian function with dynamic parameter adjustment, which breaks through the limitations of traditional linear models and can accurately characterize the nonlinear risk characteristics of low, medium and high density areas. By introducing independent parameters ( , and , ) respectively regulate the risk sensitivity coefficient and change rate of different density areas, and combine the adjustment parameters Calibrate the potential risks in low-density areas, solving the problem of underestimation of risks in low-traffic and high-speed scenarios and large prediction deviations in medium-density peak areas in traditional models; the model ensures The smooth transition of parameters significantly improves the physical rationality and calculation stability of the risk curve; at the same time, the parameter fitting mechanism based on actual accident data (such as , The model’s adaptability to real traffic scenarios is enhanced, and a high-precision and high-robust dynamic risk assessment tool is provided for autonomous driving decision-making and traffic control, which can effectively reduce the risk of collision in complex traffic flow environments.

[0026] (6) The present invention decomposes the risk of the whole vehicle into the risk of the inherent attributes of the electric bicycle to the vehicle. , Environmental risks to automobiles and the risks of electric bicycle operation to cars Three independent modules support independent optimization and dynamic adjustment of each module parameter, ensuring the rapid adaptation of the model in different scenarios and significantly improving the dynamic response capability and robustness of risk prediction. For example, in the case of sudden strong side wind conditions, the interaction coefficient between electric bicycles and cars can be dynamically improved (such as ), and calibrate the nonlinear index of the steering angle and frequency of the electric bicycle based on real-time sensor data (such as ), to achieve refined adaptation of risk prediction under extreme weather conditions, thereby enhancing the model's anti-interference ability to complex environmental disturbances and real-time warning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] By referring to the following description in conjunction with the accompanying drawings, and with a more comprehensive understanding of the present invention, other objects and results of the present invention will become more apparent and easy to understand. In the accompanying drawings: Figure 1 A flow chart for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of the electric bicycle provided by the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the technical solution and advantages of the present invention, the present application is described in detail below in conjunction with the accompanying drawings, but it is not intended to limit the protection scope of the present invention.

[0029] This embodiment provides a method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of the electric bicycle. The method starts from the perspective of the vehicle and analyzes the impact of the multimodal behavior of the electric bicycle and environmental factors on its driving safety. It combines various sensor devices (such as radar, camera) and vehicle networking technology (5G-V2X) installed on the vehicle, uses physical modeling, and takes the braking distance of the electric bicycle and the impact of the lateral and longitudinal stability on the vehicle as the basic risk. It further considers the vehicle risk caused by the environment, and comprehensively considers the dynamic distance-speed, adjacent vehicle type, traffic flow density, weather conditions (precipitation, temperature, wind speed, visibility The multi-dimensional environmental risk assessment model is built based on four dimensions (acceleration, steering angle and frequency); the risk of electric bicycle operation behavior (acceleration change, steering angle and frequency) to the car is combined; the environmental sensitivity coefficient is used to capture the complex relationship between environmental changes and vehicle risk in real time, and finally a comprehensive, accurate and real-time multi-dimensional comprehensive risk assessment method is provided for the car based on the multi-modal riding behavior of electric bicycles; this method considers in detail the various situations that the car may face during driving and has strong practicality; at the same time, the model performs nonlinear modeling for a variety of risk scenarios, which enhances the generalization ability of the model, thereby effectively improving the driving safety of the vehicle and the road traffic efficiency.

[0030] like Figure 1 As shown, this embodiment provides a method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of the electric bicycle, comprising the following steps: Step 1. Risk the vehicle Breaking down the risks of electric bikes to cars , Environmental risks to automobiles and the risks of electric bicycle operation to cars ; Step 2. Evaluate the risk of electric bikes’ inherent attributes to cars , Environmental risks to automobiles and the risks of electric bicycle operation to cars ; Among them, the inherent properties of electric bicycles pose risks to cars. It refers to the risk that the inherent properties of electric bicycles bring to the car during driving. It is a risk factor that is not affected by the external environment, including the risk caused by the braking distance of electric bicycles to the car. , the risks that electric bicycle stability poses to cars ; Environmental risks to cars It refers to the risk caused to the car by changes in external environmental factors during the driving process, including the risk caused by dynamic distance-speed , the risk posed to the car by the type of adjacent vehicles , the risks posed by traffic density to cars , Risks of weather conditions to cars ; The risk of weather conditions on the car includes the risk of precipitation on the car , Risks of temperature to cars , the risks of wind speed to cars and visibility risks to cars ; Risks of E-bike Operation to Vehicles Refers to the risk to the driving safety of the car caused by the driving behavior of the rider during the driving process, including the risk caused by the frequency of acceleration changes of electric bicycles to the car , Risks of electric bicycle steering angle and frequency to cars Among them, the risks posed by the steering angle and frequency of electric bicycles to cars Including the risks posed by the steering angle of electric bicycles to cars , Risks of electric bicycle steering frequency to cars ; Step 3. Predict vehicle risk based on the capital asset pricing model (CAPM) and introduce the environmental sensitivity coefficient , assess the sensitivity of environmental changes to vehicle risks; in addition, considering the mutual influence between vehicle environmental risks and electric bicycle behavior characteristics, add and The interaction term between is used to describe the superimposed effect of electric bicycle operation behavior on vehicle risk under different environmental conditions; and in order to more accurately capture Risks to the entire vehicle More detailed impacts, adding higher-order effects, are used to reflect the nonlinear trend of environmental risk impacts, which is applicable when the growth of risks under adverse conditions is much greater than linear expectations; Vehicle Risk The prediction model is:

[0031] in ， is the interaction risk coefficient, with a value range of [0.1, 1.0], which reflects and This interaction term can capture the amplification effect of electric bicycle operation behavior on car risk in complex driving scenarios, especially in high-risk environments such as bad weather or complex traffic; It is the power of the higher-order effect, which determines the speed and magnitude of the impact of environmental risk changes on the overall risk. Generally speaking, value ; Under extreme environmental conditions, the risk is amplified in a high-order form, and the value can be greater than 2; It is a high-order environmental sensitivity coefficient with a value range of [0.05, 0.5]. It is used to measure the amplification effect of larger environmental risk changes on the overall risk. This coefficient determines the increase in nonlinear risk. For example, under extreme weather conditions, the risk grows faster.

[0032] Furthermore, in this embodiment, the risk of the electric bicycle braking distance to the car The evaluation method is as follows: The braking distance refers to the distance required for an electric bicycle to stop completely from the time the brake pedal is applied, which directly affects the driving risk of the vehicle. It can be calculated by the kinematic formula:

[0033] in: is the final speed of the e-bike (when it comes to a complete stop ); is the speed of the electric bicycle, which can be obtained by Doppler speed measurement and visual displacement tracking using the radar installed on the car; is the acceleration of the electric bicycle (negative value during braking, i.e. deceleration), which can be calculated based on the acquired speed change rate; is the braking distance of an electric bicycle.

[0034] Will Substitute into the formula to get the braking distance of the electric bicycle for:

[0035] Considering that in actual driving, the rider's reaction time will significantly affect the braking distance. Before the rider reacts, the electric bicycle will continue to travel at the initial speed for a distance. , which can be expressed as:

[0036] in, is the rider's reaction time, which can be 0.3-1.0 seconds.

[0037] Considering that the traffic flow ahead of the e-bike may trigger unexpected events through unstable speed changes, it will increase the attention needs of the rider, thus affecting his reaction time. In order to reflect the impact of the randomness of the environment on the rider’s reaction time, a modified reaction time can be defined :

[0038] in: It indicates the sensitivity coefficient of the rider's increased attention when the electric bicycle approaches a car, which can be set to 0.5~2.0 according to the actual situation; It represents the rate of change of the distance between the electric bicycle and the car, that is, the ratio of the change in distance between the electric bicycle and the car to the change in time.

[0039] Combined with the rider's reaction distance, the corrected total braking distance It can be expressed as:

[0040] in ， is the distance traveled within the corrected reaction time; A positive nonlinear relationship between automobile risk and electric bicycle braking distance is established, and the braking distance is linked to the risk. The expression is: ; in, The risk of electric bicycle braking distance to the car is considered. The impact caused is the proportionality coefficient, is the power index.

[0041] In this embodiment, historical accident data is used to optimize the parameters in the model through nonlinear regression method. and , The value range is [0.001, 0.1], The value range is [0.5, 2.0], with mean square error is the objective function, which measures the error between the model prediction and the observed data:

[0042] in, is the number of observation samples, It is The actual risk value of the observed sample is It is calculated according to the model The risk of a sample.

[0043] When performing nonlinear estimation, we first need to give and The initial value of can be determined based on a simple linear regression method. The initial value can be 0.01. The initial value can be 1.0, and then the gradient descent method is used to update the parameters by finding the partial derivative of the objective function:

[0044]

[0045] in, For the updated value; Before update value; For the updated value; Before update value; is the learning rate considering the impact of braking distance on vehicle risk, controlling the step size of parameter update, and its value range is [0.001, 0.1]. represent MSE 1 right The partial derivative of represent MSE 1 right The partial derivative of .

[0046] Furthermore, in this embodiment, the risk of the electric bicycle's stability to the car The evaluation method is as follows: Stability reflects the ability of an electric bicycle to maintain balance and control under various operating conditions. An electric bicycle with poor stability is prone to losing control during steering, acceleration, braking, etc., thus posing a higher risk to the vehicle. In this embodiment, the stability of the electric bicycle in the horizontal and vertical directions is mainly considered.

[0047] (1) Lateral stability: The lateral stability of electric bicycles is mainly reflected in the risk of rollover, which is mainly affected by the following key physical factors: Center of gravity position (center of mass height) :The model can be determined by the visual recognition system installed on the car, matching the default value in the vehicle design specifications and estimating the rider's weight. The value range is [0.8, 1.2] meters, and a conservative value can be taken. The height of the center of gravity of an electric bicycle has an important influence on its lateral rollover tendency when turning; the higher the center of gravity, the easier it is to roll over when turning at high speed.

[0048] Wheelbase : Refers to the horizontal distance between the center of the front wheel and the center of the rear wheel of an electric bicycle. The model can be determined by the visual recognition system installed on the car, and the default value in the vehicle design specification can be matched. The value range is [1.0, 1.5] meters, and a conservative value can be taken. The wheelbase has an important impact on its driving stability; when driving at high speed, the longer the wheelbase, the higher the stability of the electric bicycle.

[0049] quality : Refers to the total mass of the electric bicycle and the rider. It can be estimated by the visual recognition system installed on the car, with a range of [80, 150] kg, and a conservative value can be taken.

[0050] Travel speed The current speed of the electric bicycle can be obtained by Doppler speed measurement and visual displacement tracking by the radar installed on the car. Especially when turning, the higher the speed, the greater the centrifugal force and lateral acceleration the electric bicycle is subjected to.

[0051] Turning Radius : The curvature radius of the path of the electric bicycle when turning can be achieved by fitting the path curvature through the point cloud using the radar sensor installed on the car, and the value range is [2, 5] meters. The smaller the turning radius, the greater the centrifugal force of the electric bicycle when turning, and the worse the stability.

[0052] Lateral stability index of electric bicycle Indicates the risk of lateral rollover, mainly caused by centrifugal force and gravity When the centrifugal force exceeds the stabilizing torque generated by gravity, the electric bicycle may roll over:

[0053] in, It is half of the wheelbase of the electric bicycle, that is, the distance from the center of mass of the electric bicycle to the steering axis; is the total weight of the electric bicycle and the rider; is the centrifugal force between the electric bicycle and the rider, Represents the acceleration due to gravity, and its value is 9.8N / kg.

[0054] when When the stabilizing torque generated by gravity is greater than the centrifugal force, the electric bicycle can maintain stability; when When the centrifugal force exceeds the stabilizing torque generated by gravity, the electric bicycle is at risk of tipping over.

[0055] (2) Longitudinal stability: The longitudinal stability of an electric bicycle mainly refers to its stability during acceleration and deceleration, especially the load transfer and the change of the grip of the front and rear wheels. Since electric bicycles are light in weight, have a short wheelbase, and a relatively high center of gravity, the grip distribution of the front and rear wheels is prone to change significantly during dynamic driving, especially during rapid acceleration and braking, which in turn affects the longitudinal stability, thereby affecting the driving safety of the vehicle.

[0056] Dynamic load transfer occurs during acceleration or deceleration of an e-bike, which is central to longitudinal stability. The vehicle's center of mass height and wheelbase determine the amount of load transfer between the front and rear wheels. Load transfer causes a change in the normal force between the front and rear wheels, which in turn affects the grip distribution.

[0057] When an electric bicycle accelerates (decelerates), the center of gravity moves backward (forward), the load on the front wheel (rear wheel) decreases, and the load on the rear wheel (front wheel) increases. The load transfer amount is defined as and can be calculated by the following formula:

[0058] in, is the acceleration component of the electric bicycle in the direction of travel; During acceleration, the load is transferred from the front wheel to the rear wheel, and the normal force of the front wheel is reduced, resulting in insufficient grip of the front wheel, affecting stability. The greater the acceleration, the greater the load transfer, and the more obvious the decrease in the front wheel's grip. During emergency braking, the load on the rear wheel may be reduced to near zero, causing the rear wheel to lose grip. Therefore, the load transfer during braking has a more significant impact on the longitudinal stability of the electric bicycle.

[0059] Load transfer directly affects the positive pressure of the front and rear tires on the ground, thus determining the vehicle's grip. The coefficient of friction between the front and rear wheels and the ground is defined as , the front wheel base (i.e. the horizontal distance from the front wheel axle to the center of mass) is , the rear wheelbase (i.e. the horizontal distance from the rear wheel axle to the center of mass) is , the normal force on the front wheel is , the normal force on the rear wheel is , the front wheel grip is , the rear wheel grip is ,but:

[0060]

[0061]

[0062]

[0063] Taking into account the grip of the front and rear wheels, the longitudinal stability can be defined as the relative proportion of the difference in grip between the front and rear wheels. The following formula can be established to construct a risk model and define the longitudinal stability index of electric bicycles: Characterize the risk.

[0064]

[0065] in, It is an adjustment coefficient used to reflect the impact of load transfer on risk, and its value range is [0.1, 1.0].

[0066] The lateral stability index of electric bicycles and longitudinal stability index Comprehensive considerations to establish the stability of electric bicycles In this embodiment, a nonlinear combination method is used to integrate these two risk factors, and a nonlinear index of longitudinal and lateral stability risks on the stability of the electric bicycle is introduced. , a nonlinear index used to adjust the overall combination of the two risk terms , and adjust the weight coefficients of lateral stability and longitudinal stability to the stability of the electric bicycle , , as follows:

[0067] Among them, the parameters The value range is [0.5, 2.0], The value range is [0.5, 2.0], The value range is [0.5, 3.0], The value range of is [0, 1], , the value range is [0, 1].

[0068] Stability of electric bikes Risks to cars As the stability of electric bicycles is greater, the risk to cars The lower it is, the negative correlation nonlinear relationship between automobile risk and electric bicycle stability is established, linking stability with risk, and the expression is: ; in, The risk that the stability of electric bicycles brings to cars, that is, considering the risk of electric bicycle stability to cars The impact caused is the proportionality coefficient, is the power index, The bigger, The smaller it is, the more stable the e-bike will be.

[0069] In order to better fit the above risk model, in this embodiment, historical accident data is used to optimize the parameters in the model through nonlinear regression method. and , The value range is [0.001, 0.1], The value range is [0.5, 2.0], with mean square error is the objective function, which measures the error between the model prediction and the observed data:

[0070] in, is the number of observation samples, It is The actual risk value of the observed sample is It is calculated according to the model The risk of a sample.

[0071] When performing nonlinear estimation, we first need to give and The initial value of can be determined based on a simple linear regression method. The initial value can be 0.01. The initial value can be 1.0, and then the gradient descent method is used to update the parameters by finding the partial derivative of the objective function:

[0072]

[0073] in, For the updated value; Before update value; For the updated value; Before update value; is the learning rate considering the impact of electric bicycle stability on car risk, controlling the step size of parameter update, and its value range is [0.001, 0.1]. Represents MSE2 The partial derivative of Represents MSE2 The partial derivative of .

[0074] Furthermore, in this embodiment, the risk of electric bicycle braking distance to the car is taken into consideration. The risks posed to cars by e-bike stability They may affect each other, and when the risks of both increase, a greater combined effect may occur, so a nonlinear relationship between the two should be established:

[0075] , , It is a nonlinear index, which is adjusted by optimization algorithm based on historical data or experimental data. ; The value range is [0.5, 2.0], The value range is [0.5, 2.0], The value range of is [0.5, 3.0]. To minimize the error between the model prediction value and the actual value, the risk model can be accurately fitted. Under extreme driving conditions, it can be appropriately reduced. , enhances the effect of nonlinear combination, reflecting the situation that the two risk terms may jointly amplify the risk of the car in an emergency situation.

[0076] Furthermore, in this embodiment, the risk of the dynamic distance-speed to the vehicle The evaluation method is as follows: When a car interacts with other vehicles in the driving environment, the driving risk of the car is not only related to the distance between adjacent vehicles, but also to the distance between adjacent vehicles. It is also related to the relative speed of the two The closer the distance and the higher the relative speed, the greater the driving risk to the car. Conversely, when the relative speed is low, even if the distance is close, the risk is relatively low.

[0077] First, the safe distance between the car and other vehicles is required. Determine, defined as the minimum distance at which the first car behind the car can stop safely:

[0078] in, is the current speed of the first car behind; The driver's reaction time can be 0.3-1.0 seconds; is the deceleration of the first vehicle behind; Based on the distance between the car and the first car behind and safe distance By comparing the distance control risk ;if Less than , the car may be in a high-risk state; when Greater than When , the risk is reduced; the specific quantitative formula is as follows:

[0079] in, , , To adjust the parameters, used to adjust the nonlinearity of risk, Controls the risk growth rate when the distance between two vehicles is less than the safe distance. It is usually related to the reaction time and braking ability of the following vehicle, and the value range is [0.1, 2.0]; Controls the risk attenuation rate when the distance between the two vehicles is greater than the safe distance, reflecting the risk reduction when the distance between the two vehicles exceeds the safe range, and the value range is [0.2, 1.5]; The risk attenuation degree when the control distance exceeds the safe distance can be adjusted to reflect different scenarios outside the safe distance, and the value range is [0.05, 0.5]. The determination of the three parameters can be based on the existing data for fitting, or obtained through simulation experiments.

[0080] Secondly, the relative speed risk factor is introduced To describe the additional effect of relative speed on risk:

[0081] in, is the maximum relative speed, which can be 50km / h for general urban roads; Represents the impact of relative speed on risk. As speed increases, the risk rises significantly.

[0082] Finally, we get the dynamic distance-speed risk to the car. :

[0083] in, is the critical value of relative speed, which can be determined based on factors such as traffic rules, driving behavior, vehicle characteristics and data analysis. For general urban roads, it can be set at between; is a minimum value, which can be 10 -6 ~10 -9 , used to indicate a relatively near-zero risk.

[0084] Furthermore, in this embodiment, the risk posed to the car by the adjacent vehicle type The evaluation method is as follows: Under complex environmental conditions, the types of adjacent vehicles in the same lane also have an impact on the driving risk of the car. In particular, when large vehicles (such as trucks and buses) appear in the same lane as cars, it will increase the attention burden of the car driver and amplify the risk of accidents; there is also the possibility of causing a chain accident. The type of adjacent vehicles is limited to vehicles adjacent to the car in the same lane. Different types of vehicles pose different threats to cars, and thus pose different risks to cars. Generally speaking, the larger the vehicle model, the heavier the mass, and the larger the blind spot of the vehicle, the higher the potential risk. Based on these characteristics, different risk weights are assigned to different types of vehicles to reflect the collision threats they pose to cars.

[0085] The detection range can be limited to 50 to 150 meters in front of the car and 20 to 50 meters behind the car in the same lane. The specific values ​​can be adjusted according to the camera's recognition range and risk calculation accuracy requirements.

[0086] For the calculation method of vehicle model risk weight, assume Represents the vehicle type. Four vehicle types are considered in the model, and each type corresponds to a risk weight: Non-motor vehicles: such as bicycles, tricycles, electric bicycles, etc.; Small car: usually refers to the car with engine displacement of about 1.0-1.3, such as sedans, electric cars, etc.; Mid-sized car: usually refers to the car with engine displacement of about 1.3-3.0, such as vans, SUVs, etc.; Large vehicles: usually refers to vehicles with engine displacement of 3.0 or above, such as vans, trucks, buses, etc.

[0087] Each variable defined by physical characteristics can be integrated into a risk weight model by linear combination. The model takes size, weight, and blind area as independent risk factors and uses weight coefficients , , To control the impact of each factor, the formula of the risk weight model (the risk of adjacent vehicle types to the car) is as follows:

[0088] in: The size of the vehicle is measured in cubic meters (m 3 ) as a unit, the external dimensions can be estimated by camera, or accurately measured by millimeter-wave radar; is the mass of the vehicle model, in tons (t). The vehicle appearance or type can be identified by a camera to simply estimate the mass of the vehicle; The blind spot size is in square meters (m²), which indicates the area that the driver cannot directly observe. The camera can be used to identify which areas are blocked based on the vehicle type. , , is the weight coefficient ( ), The value range is [0.2, 0.5], The value range is [0.3, 0.6], The value range is [0.1, 0.3], which is used to reflect the influence of each physical characteristic on the overall risk weight. The determination of the parameters requires analyzing the performance of different vehicle models in actual traffic accidents, fitting the relationship between these three physical characteristics and the actual collision risk, and then estimating the contribution weight of each characteristic to the overall risk.

[0089] The risks posed to the car by all vehicle types within the calculated detection range are taken as the dominant risk source and used to predict the risk of the entire vehicle.

[0090] Furthermore, in this embodiment, the risk posed by the traffic flow density to the car The evaluation method is as follows: Traffic flow density is a parameter that describes the road traffic status at a certain moment. It indicates the number of vehicles passing through a certain traffic area per unit time, which directly poses a risk to car driving.

[0091] When the traffic density is very low, the distance between the car and the surrounding vehicles is large, there is less interaction between vehicles, and the car has enough space to run, so the risk of collision is low. However, if the traffic density is too low, the car is likely to encounter other vehicles traveling at high speeds. In this case, the relative speed is high, which may increase the risk for both parties.

[0092] As traffic density increases, the headroom decreases, the frequency of interaction between cars and other vehicles increases, and the complexity of driving operations increases, the risk gradually increases. Usually, when traffic density is medium, it is a time when cars are at greater risk, because at this time, all vehicles are traveling at a higher speed, but the distance between vehicles is not enough to provide sufficient reaction time.

[0093] In high traffic density, the distance between vehicles becomes smaller, but the speed of vehicles usually decreases accordingly. In this case, although the driving space between vehicles is compressed, the risk may not increase significantly, because the vehicles are traveling very slowly and the consequences of a collision may not be so serious. Therefore, in congestion, the risk of accidents may not necessarily be higher despite the high frequency of interactions.

[0094] In order to capture the nonlinear relationship described above, a peak function is used to describe the relationship between traffic flow density and risk. Assume that the risk is The maximum value is reached at , and it shows a normal distribution trend as the traffic flow changes. The risk function caused by traffic flow density to cars can be constructed in the form of Gaussian distribution:

[0095] in: It represents the traffic flow density, that is, the number of vehicles per unit length on the road at a certain moment, in units of ; Indicates the risk situation of traffic flow density to cars; It refers to the point with the highest impact on vehicle driving risk under medium traffic density, with a value range of [20, 40] veh / km. At this time, the traffic flow density reaches a high value, but it is not completely congested. The speeds of vehicles are fast and the interactions are frequent, which makes collisions easy to occur. It can be obtained by analyzing the accident peak based on medium traffic flow density, and finding the traffic flow density value with the highest accident rate as ; is the risk sensitivity coefficient of low-density areas, ranging from [0.5, 1.5], is the risk sensitivity coefficient of the high-density area, with a value range of [0.2, 1.0]. According to the continuity condition, Department and equal; is the risk change rate in low-density areas, with a range of [5, 10] veh / km, is the risk change rate in high-density areas, with a value range of [10, 20] veh / km. and The specific value can be determined by observing a large number of accident rates under low and high traffic flow density and then fitting; It is an adjustment parameter with a value range of [0.1, 0.5], which is used to regulate the overall risk level under low traffic flow density to reflect the potential risks existing under low traffic density in reality.

[0096] Furthermore, in this embodiment, the risk of the weather condition to the car The evaluation method is as follows: In order to more comprehensively evaluate the risk conditions posed to automobiles under different weather conditions, this embodiment mainly considers the risk impact of variable weather conditions on automobiles from four aspects: precipitation, temperature, wind speed and visibility.

[0097] In terms of precipitation, the main consideration is the impact of the ground friction coefficient on the risk. In the small precipitation stage, the oil and dust on the road surface mix with a small amount of rainwater to form a very slippery surface, causing the friction coefficient to drop sharply; in the medium precipitation stage, the oil on the road surface is gradually washed away by the rain, a water film is formed, and the friction coefficient continues to drop, but more slowly than in the initial stage; in the heavy precipitation stage, water may accumulate on the road surface, at which time the friction coefficient hardly changes and tends to a stable low value. Based on this, a model for the change of friction coefficient with precipitation is established:

[0098] in, is the amount of precipitation, in millimeters; is the road friction coefficient corresponding to different precipitation levels; is the friction coefficient of dry road surface, ranging from [0.7 to 0.9]; is the friction coefficient under moderate precipitation, with a value range of [0.4, 0.6]; is the minimum friction coefficient under heavy precipitation, with a value range of [0.15, 0.35]; , are the critical values ​​of small, medium and large precipitation. The value range is [5, 15] mm / 24h. Value range: [20, 30] mm / 24h; is the exponential decrease coefficient of the initial precipitation stage, indicating the magnitude of the decrease in the initial friction coefficient, with a value range of [0.1, 0.5] mm -1 ; is the logarithmic decrease coefficient in the medium precipitation stage, indicating the effect of increased precipitation on the friction coefficient, with a value range of [0.05, 0.2] mm -1 .

[0099] Once the friction factor changes with precipitation, we can construct a driving risk model through the relationship between the friction factor and the risk of automobile driving; the lower the friction factor, the higher the risk, and the risk is inversely proportional to the friction factor:

[0100] in, Indicates precipitation Impact on automobile driving risks; Indicates risk, which indicates the minimum risk under dry conditions. It is a minimum value and can be 10 -6 ~10 -9 ; The coefficient of risk increase caused by the decrease of friction factor indicates the influence of the decrease of friction factor on risk, and the value range is [0.01, 0.5]; In summary, the piecewise function relationship between precipitation and automobile risk is obtained: ; In terms of temperature, the main consideration is the impact of icy and snowy roads on cars under low temperature conditions.

[0101] Under sub-zero temperature conditions, if there is snowfall, the road surface may freeze, causing a sharp drop in the friction coefficient, and the risk of car driving increases:

[0102] in: Indicates temperature; Indicates whether there is ice or snow on the road. If yes, it is 1; if no, it is 0. It can be obtained through camera recognition. Indicates the risk of driving a car in low-temperature, icy and snowy conditions (the risk that temperature poses to the car); Indicates risk, which is the minimum risk under suitable temperature. It is a minimum value, which can be 10 -6 ~10 -9 ; The coefficient indicating the increase in risk due to temperature drop, indicating the degree of impact of temperature drop on risk, with a value range of [1.0, 5.0]; is the friction coefficient of dry road surface, ranging from [0.7 to 0.9]; Indicates the temperature at which the road surface begins to freeze, with a value range of [-2, 0] degrees Celsius; The critical temperature that controls the rate at which the friction coefficient decreases is in the range of [2, 10] degrees Celsius. The lower the temperature, the faster the friction coefficient decreases.

[0103] In terms of wind speed, crosswinds can easily cause small vehicles such as electric bicycles to become unstable: when the crosswind speed is strong, the lateral moment of the electric bicycle will increase, and there is a risk of rollover or lane deviation, which will affect the car. First, the risk of crosswinds to electric bicycles is modeled:

[0104] in, Indicates wind speed; Indicates wind speed and the angle of incidence The risk of electric bicycle rollover under the conditions of is the risk value, which indicates the lowest risk under windless conditions and can be 10 -6 ~10 -9 ; The crosswind risk coefficient has a value range of [0.5, 2.0], reflecting the impact of wind on the rollover risk; is the lateral drag coefficient, with a value range of [0.3, 1.2], which determines the force exerted by the side wind on the electric bicycle; The windward area of ​​the side of the electric bicycle affects the lateral force generated by the wind on the electric bicycle; is the air density, usually taking the standard value of 1.225 kg / m³; is the height of the vehicle's center of mass. The higher the center of mass, the greater the risk of rollover. represents the sine value of the crosswind incident angle, The closer it is to 90°, that is, the closer the wind is to being perpendicular to the direction of travel of the electric bicycle, the greater the lateral force of the side wind and the higher the risk of rollover.

[0105] For cars, since crosswinds can cause electric bicycles to lose control (roll over or deviate from the lane), thus affecting the driving safety of the car, the risk model transmitted to the car is constructed as follows:

[0106] in, is the spatial interaction coefficient, with a value range of [1.0, 2.5], reflecting the possibility of an electric bicycle occupying a car lane or colliding with a car after losing control, which depends on the lane width and crosswind intensity; is the traffic flow coefficient, with a value range of [1.0, 3.0], which represents the amplifying effect of an out-of-control electric bicycle on the traffic flow composed of cars.

[0107] In terms of visibility, an exponential decay model is constructed to characterize the nonlinear relationship between automobile risk and visibility:

[0108] in, Indicates the current visibility in meters; Indicates the risk to the car caused by visibility; It represents the basic risk value, that is, the minimum risk under good visibility, which can be 10 -6 ~10 -9 ; It represents the risk adjustment parameter, which is used to control the impact of reduced visibility on risk. The value range is [0.5, 5.0]. The specific value can be determined by analyzing and fitting historical data, reflecting the rate of increase of risk when visibility decreases. The larger the value, the more significant the impact of visibility on risk. Represents critical visibility, with a value range of [50, 200] meters. When visibility drops below this critical value, the risk will increase significantly and safe driving will not be possible. The specific value can be estimated based on historical visibility data and traffic accident data.

[0109] When visibility When is very high, such as on sunny days or during the day, the risk function approaches the risk , indicating a low risk level; when visibility When it drops to a medium level, the risk begins to increase rapidly, and the exponential term causes the risk to rise faster and faster; when visibility Continue to descend close to When the driver's reaction time and foresight are greatly reduced, the risk will increase significantly, and it is not recommended to continue driving at this time.

[0110] The precipitation, temperature, wind speed and visibility are combined, considering that the four factors may affect each other. In order to better handle the interaction and nonlinear relationship between weather factors, in this embodiment, a weighted power combination method is used to construct a car driving risk model under variable weather conditions:

[0111] in, Indicates the risk that weather factors pose to cars; , , , Indicates the weight coefficient of each factor, reflecting the contribution of different weather factors to the total risk; The value range is [0.2, 0.4], The value range is [0.1, 0.3], The value range is [0.1, 0.3], The value range is [0.2, 0.4]; and finally satisfies ; It represents the power of each risk's contribution to the overall risk, reflecting the nonlinear impact of each factor. The value range is [1.0, 2.0], The value range is [1.5, 3.0], The value range is [1.0, 2.5], The value range is [0.5, 1.5]; among them, for The maximum value in .

[0112] In this embodiment, after the above calculation, the risk caused by the dynamic distance-speed to the car is finally obtained. , the risk posed to the car by the type of adjacent vehicles , the risks posed by traffic density to cars , Risks of weather conditions to cars ; In order to better deal with the interaction and nonlinear relationship between various environmental factors, a weighted power combination method is used to construct a car driving risk model in a complex environment. The expression is:

[0113] in, , , , It represents the weight coefficient of each risk factor, reflecting the contribution of different environmental factors to the total risk. The value range is [0.25, 0.5], The value range is [0.05, 0.2], The value range is [0.15, 0.3], The value range is [0.2, 0.4]; and finally satisfies ; It represents the power of each risk's contribution to the overall risk, reflecting the nonlinear impact of each factor. The value range is [1.5, 3.0], The value range is [1.0, 2.0], The value range is [1.2, 2.5], Value range: [1.5, 3.0]; for The maximum value in .

[0114] Furthermore, in this embodiment, the risk of the electric bicycle operation behavior to the car The evaluation method is as follows: The risk of electric bicycle operation to the car refers to the safety hazard caused by the unstable driving behavior and poor operation of the rider during the driving of the electric bicycle. It mainly includes the impact of key operating parameters such as acceleration change, steering angle and frequency. When the rider frequently accelerates and decelerates or makes sharp turns, it will have a significant impact on the driving safety of the car. Therefore, the risk of electric bicycle operation to the car It mainly assesses the risks posed by cyclists’ driving behavior to vehicle driving safety.

[0115] In this embodiment, the risk assessment method of the acceleration change to the vehicle is as follows: Acceleration change is an important indicator to measure the stability of the operation behavior of electric bicycle riders. It represents the frequency and degree of acceleration or deceleration of the rider in a unit of time. Frequent and large acceleration changes usually mean that the rider's operation is more aggressive, which has a greater impact on the driving safety of the vehicle. By evaluating the acceleration change, potential driving risks can be effectively predicted.

[0116] In order to evaluate the impact of the frequency of changes in the acceleration of an electric bicycle on the car, the acceleration change value of the electric bicycle, the time window and the following distance are comprehensively considered. That is, the electric bicycle rider will bring higher operational risks when the acceleration changes are more drastic in a shorter time and the following distance is closer.

[0117]

[0118] in, is the rate of change of acceleration; Represents time, that is, the time window in which the operation occurs. The value range is [5, 30] seconds. If it is too short, noise will be introduced, and if it is too long, instantaneous risks may be ignored. Indicates at time The number of acceleration or deceleration operations; is the acceleration change for each operation; is the time interval between each operation; It is the following distance between the car and the electric bicycle during each operation.

[0119] Considering that the acceleration changes have different effects on the stability of electric bicycles at different speeds, and thus the risks to cars are also different, the speed of electric bicycles is introduced. , so that the frequency of acceleration change is related to the velocity:

[0120] in, is the corrected acceleration rate of change; This is the adjustment coefficient for controlling the speed's amplification effect on the acceleration change frequency, with a value range of [0.5, 1.5].

[0121] Using high-order polynomials to establish the relationship between acceleration change frequency and vehicle risk The relationship between:

[0122] in, , , Represents the weights of different acceleration change frequency bands, which can flexibly respond to the risk changes of acceleration change frequency in different frequency bands. The value range is [0.1, 0.5], The value range is [0.05, 0.3], The value range is [0.01, 0.2].

[0123] In this embodiment, the risk assessment method of the steering angle and frequency to the vehicle is as follows: The steering angle reflects the deflection of the handlebars of an electric bicycle during steering, while the steering frequency reflects the rider's control stability during vehicle driving. A larger steering angle usually means a sharp turn, which increases the driving risk of adjacent cars, especially at high speeds. Frequent steering operations may indicate that the driver frequently changes lanes or evades in a complex traffic environment, increasing the complexity and uncertainty of the operation.

[0124] First, consider the impact of the steering angle of electric bicycles on the risk of automobiles. Risks to cars It is related to the square of the absolute value of the steering angle and changes with the speed of the electric bicycle. It increases exponentially and decreases as the distance between the car and the electric bicycle increases. The expression is:

[0125] in, is the steering angle of the electric bicycle, in , which can be obtained by using the target detection and tracking technology through the vehicle-mounted camera; To adjust the steering angle The constant coefficient of the risk impact degree has a value range of [0.2, 0.8] and can be set by data fitting or experience; is the speed amplification factor, which controls the exponential effect of speed on steering angle risk. Its value range is [0.1, 0.5] and can be set by data fitting or experience; is the distance between the car and the electric bicycle, which can be obtained through the on-board radar.

[0126] Similarly, by defining turning frequency as the number of times a rider turns the handlebars per unit time, we can establish a correlation between the turning frequency of electric bicycles and the risk of automobiles. The relationship between:

[0127] in, represents time, i.e., the time window for observing the steering frequency, with a value range of [5, 30] seconds. If it is too short, noise will be introduced, and if it is too long, instantaneous risks may be ignored; The coefficient representing the influence of steering frequency on risk has a value range of [0.1, 0.4] and can be set by data fitting or experience; Unit time The number of steering operations within is the critical steering angle, ranging from [10°, 30°]; only The number of turns is counted, if , the corresponding steering can be ignored; The adjustment coefficient that represents the amplification effect of velocity on the frequency of acceleration change. The value range is [0.5, 1.5] and can be set by data fitting or experience. is the distance between the car and the electric bicycle, which can be obtained through the on-board radar.

[0128] Considering the risk interaction between steering angle and frequency, the nonlinear superposition of the two is used in the final automobile risk model to obtain the relationship between steering angle and frequency and automobile risk. The relationship between them is:

[0129] in, is the interaction risk term coefficient of steering angle and frequency, with a value range of [0.05, 0.2], reflecting the influence of the superposition effect of steering angle and steering frequency at a certain driving speed on the operational risk.

[0130] Finally, the risk of electric bicycle operation to cars :

[0131] in, 、 is the weight coefficient, which is used to adjust the mutual influence between different risk items. Value range: [0.4, 0.7]; Value range: [0.3, 0.6]; , is a nonlinear index used to adjust the nonlinear growth range of each risk item under different operating conditions. The value range is [1.0, 2.0]; Value range: [1.5, 3.0]; The value range is [1.0, 2.0].

[0132] In this embodiment, the environmental risk sensitivity coefficient is introduced , used to measure the automotive risks posed by the environment Risks to the entire vehicle It reflects the fluctuation of vehicle risk when external environmental conditions (dynamic distance-speed, adjacent vehicle type, traffic flow density, weather conditions) change. The contribution of environmental changes to the risk of the entire vehicle can be quantified, allowing the risk assessment model to more accurately capture the potential impact of the external environment on vehicle risks. The initial value ranges from [0.05, 0.5]. It can be determined by preliminarily evaluating the contribution of environmental risk to the risk of the entire vehicle based on actual conditions. Subsequently, it can be iterated and trained continuously based on existing data to eventually obtain the optimal value under different traffic scenarios.

[0133] Specifically, calculate the automotive risks caused by the environment The variance can reflect the volatility of environmental changes:

[0134] in, It is Environmental risk value at a time point; is the average value of the automobile risk caused by the environment; is the number of time series of data.

[0135] Calculate vehicle risk Automotive risks to the environment The covariance of can reflect the joint volatility of overall risk and environmental risk:

[0136] in, It is The risk value of the vehicle at a point in time; is the average value of the vehicle risk; from this, the environmental sensitivity coefficient can be obtained The calculation formula is:

[0137] Among them, the covariance Indicates the risk of the entire vehicle Automobile risks caused by the environment If the covariance value is high, it means that environmental changes significantly affect the overall risk; It indicates the fluctuation range of environmental risk. The larger the variance, the more drastic the change in environmental conditions.

[0138] 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-mentioned method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of the electric bicycle.

[0139] The present invention also provides a computer-readable medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of the electric bicycle is implemented as described above.

[0140] Those skilled in the art will appreciate that all or part of the functions of the various methods / modules in the above embodiments may 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 may be stored in a computer-readable storage medium, which may include: a read-only memory, a random access memory, a disk, an optical disk, a hard disk, etc. The program is executed by a computer to implement the above functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above functions can be implemented.

[0141] In addition, when all or part of the functions in the above-mentioned embodiments are implemented by means of a computer program, the program can also be stored in a storage medium such as a server, another computer, a disk, an optical disk, a flash drive or a mobile hard disk, and saved to the memory of a local device by downloading or copying, or the system of the local device is updated. When the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0142] The above specific examples are used to illustrate the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the art of the present invention, according to the idea of ​​the present invention, some simple deductions, deformations or substitutions can be made. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for predicting the risk of collision between an electric bicycle and a vehicle based on its multimodal behavior, characterized in that: The method comprises the following steps: Step 1. Risk the vehicle Breaking down the risks of electric bikes to cars , Environmental risks to automobiles and the risks of electric bicycle operation to cars ; Step 2. Evaluate the risk of electric bikes’ inherent attributes to cars , Environmental risks to automobiles and the risks of electric bicycle operation to cars ; Among them, the inherent properties of electric bicycles pose risks to cars. Refers to the risks that electric bicycles bring to cars during driving due to their inherent properties, including the risks caused by the braking distance of electric bicycles , the risks that electric bicycle stability poses to cars ; ； in, is the proportionality coefficient, is the power index, Corrected reaction time for the rider, is the speed of the electric bicycle, is the acceleration of the e-bike; ； in, is the proportionality coefficient, is the power index, It is the lateral stability index of electric bicycles. It is the longitudinal stability index of electric bicycle. , is the weight coefficient, , , is the nonlinear index; Environmental risks to cars It refers to the risk caused to the car by changes in external environmental factors during the driving process, including the risk caused by dynamic distance-speed , the risk posed to the car by the type of adjacent vehicles , the risks posed by traffic density to cars , Risks of weather conditions to cars ; When evaluating the risk posed to the car by dynamic distance-speed, the relative speed and distance between the car and the first car behind it are comprehensively considered; when evaluating the risk posed to the car by adjacent vehicle types, the risk posed to the car by all vehicle types within the detection range is calculated, and the highest risk is taken for the risk prediction of the whole vehicle; when evaluating the risk posed to the car by traffic flow density, the risk function posed to the car by traffic flow density is constructed in the form of Gaussian distribution; when evaluating the risk posed to the car by weather conditions, the risk posed to the car by precipitation, temperature, wind speed and visibility are comprehensively considered, that is, the risk posed to the car by weather conditions includes the risk posed to the car by precipitation , Risks of temperature to cars , the risks of wind speed to cars and visibility risks to cars ; Risks of E-bike Operation to Vehicles Refers to the risk to the driving safety of the car caused by the driving behavior of the rider during the driving process, including the risk caused by the frequency of acceleration changes of electric bicycles to the car , Risks of electric bicycle steering angle and frequency to cars ; ； in, , , Represents the weights of different acceleration change frequency bands, is the corrected acceleration rate of the electric bicycle, ， is the acceleration change rate of the electric bicycle, is the speed of the electric bicycle, To control the adjustment coefficient of the speed amplification effect on the acceleration change frequency; ； in, is the steering angle of the electric bicycle, To adjust the steering angle The constant coefficient for the risk impact, is the speed multiplication factor, which controls the exponential effect of speed on the steering angle risk, is the distance between the car and the electric bicycle, The coefficient representing the influence of steering frequency on risk, represents time, i.e. the time window for observing the steering frequency, Unit time Number of steering operations within It represents the adjustment coefficient of the speed on the amplification effect of the acceleration change frequency. is the interaction risk term coefficient between steering angle and frequency; Step 3. Predict vehicle risk based on the capital asset pricing model and introduce the environmental sensitivity coefficient , assess the sensitivity of environmental changes to vehicle risks; in addition, add and The interaction term between is used to describe the superimposed effect of electric bicycle operation behavior on vehicle risk under different environmental conditions; the high-order effect is added to reflect the nonlinear trend of environmental risk impact; Vehicle Risk The prediction model is: ； in, is the interaction risk coefficient, with a value range of [0.1, 1.0], which reflects The synergistic effect between is the power of the higher-order effect, which determines the speed and magnitude of the impact of environmental risk changes on the overall risk. value ; is the high-order environmental sensitivity coefficient, and its value range is [0.05, 0.5].

2. A method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: The risks inherent in electric bicycles to cars The evaluation model is: ； in, , , is the nonlinear index, The value range is [0.5, 2.0], The value range of is [0.5, 2.0]; The value range is [0.5, 3.0].

3. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Environmental risks to cars The evaluation model is: ； in, , , , represents the weight coefficient of each risk factor, The value range is [0.25, 0.5], The value range is [0.05, 0.2], The value range is [0.15, 0.3], The value range is [0.2, 0.4]; and finally satisfies ; represents the power of each risk's contribution to the overall risk, The value range is [1.5, 3.0], The value range is [1.0, 2.0], The value range is [1.2, 2.5], Value range: [1.5, 3.0]; for The maximum value in .

4. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Risks of E-bike Operation to Vehicles The evaluation model is: ； in, , is the weight coefficient, The value range is [0.4, 0.7], Value range: [0.3, 0.6]; , is the nonlinear index, The value range is [1.0, 2.0]; Value range: [1.5, 3.0]; The value range is [1.0, 2.0].

5. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Dynamic distance-speed risk to cars The evaluation model is: ； in, is the relative speed risk coefficient, is the critical value of relative velocity, , , To adjust the parameters, is the relative speed between the car and the first car behind it, is the distance between the car and the first car behind it, is the safe distance between the car and the first car behind it, As a minimum value, take 10 -6 ~10 -9 .

6. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Risk to the car by adjacent vehicle types The evaluation model is: ； in, , , is the weight coefficient, is the size of the vehicle model, For the quality of the car model, is the blind area size.

7. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Risks of traffic density to cars The evaluation model is: ； in, represents the traffic flow density, It refers to the point with the highest impact on vehicle driving risk under medium traffic density. is the risk sensitivity coefficient of the low-density area, is the risk sensitivity coefficient of high-density areas, is the risk change rate in low-density areas, is the risk change rate in high-density areas, It is an adjustment parameter with a value range of [0.1, 0.5], which is used to control the overall risk level under low traffic flow density.

8. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Risks of weather conditions to cars The evaluation model is: ； in, , , , Indicates the weight coefficient of each factor, reflecting the contribution of different weather factors to the total risk; The value range is [0.2, 0.4], The value range is [0.1, 0.3], The value range is [0.1, 0.3], The value range is [0.2, 0.4]; and finally satisfies ; It represents the power of each risk's contribution to the overall risk, reflecting the nonlinear impact of each factor. The value range is [1.0, 2.0], The value range is [1.5, 3.0], The value range is [1.0, 2.5], The value range is [0.5, 1.5]; among them, for The maximum value in .

9. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Risks of precipitation to cars The evaluation model is: ； in, Indicates the minimum risk under dry conditions, which is a minimum value of 10 -6 ~10 -9 ; The coefficient of risk increase due to the decrease in friction factor, is the amount of precipitation, is the friction coefficient of dry road surface, is the friction coefficient under moderate precipitation, is the minimum friction coefficient under heavy precipitation, , are the critical values ​​of small, medium and large precipitation. is the exponential decrease coefficient in the initial precipitation stage, indicating the decrease in the initial friction coefficient. is the logarithmic decrease coefficient in the medium precipitation stage, indicating the effect of increased precipitation on the friction coefficient; Temperature risks to cars The evaluation model is: ； in, Indicates temperature; Indicates whether there is ice or snow on the road. If yes, it is 1; if no, it is 0. It represents the minimum risk under suitable temperature, which is a minimum value, taking 10 -6 ~10 -9 ; The coefficient indicating the increase in risk due to temperature drop, indicating the degree of impact of temperature drop on risk, is the friction coefficient of dry road surface, The temperature at which the road begins to freeze. The critical temperature that controls the rate at which the friction coefficient decreases; Risks of wind speed to cars The evaluation model is: ； in, is the spatial interaction coefficient, which reflects the possibility of an electric bicycle occupying a car lane or colliding with a car after losing control, and depends on the lane width and crosswind intensity; is the traffic flow coefficient, which indicates the amplification effect of the out-of-control electric bicycle on the traffic flow composed of cars. Indicates wind speed and the angle of incidence The risk of electric bicycle rollover under the conditions of Visibility risks to cars The evaluation model is: ； in, Indicates the current visibility. It represents the basic risk value, that is, the minimum risk under good visibility, and is set to 10 -6 ~10 -9 ; represents the risk adjustment parameter, which is used to control the impact of reduced visibility on risk. Indicates critical visibility.

10. The method for predicting the risk of collision between an electric bicycle and a vehicle based on the multimodal behavior of an electric bicycle according to claim 1, characterized in that: Environmental sensitivity coefficient The initial value range is [0.05, 0.5]. Subsequent iterative training is carried out based on the existing data to finally obtain the optimal value under different traffic scenarios. The calculation formula is: ； in, Risks for the entire vehicle Automotive risks to the environment The covariance of Automobile risks caused by the environment fluctuations; The risks of cars to the environment The variance of represents the fluctuation range of environmental risk.

Images