Automobile risk potential field assessment method based on driving style coefficient

Abstract

The invention discloses an automobile risk potential field assessment method based on a driving style coefficient. On the basis of the risk potential field, the driving style coefficient obtained through driving behavior data and trajectory prediction are introduced, the risk is judged more accurately through the proper risk potential field, and the limitation and generalization of a traditional risk potential field method are overcome, especially the defects in the aspect of considering individual differences and behavior characteristics of drivers are overcome. According to the method, the driving style characteristics are accurately extracted by analyzing the existing driving data of the driver, and are integrated into the risk potential field model, so that the identification capability of the risk potential field on the driving safety is improved, the risk level of the driver under various driving conditions can be more comprehensively and accurately evaluated, and the driving safety of the vehicle is improved.

Description

Technical Field

[0001] The present invention relates to the field of driving risk assessment, and in particular to a vehicle risk potential field assessment method based on a driving style coefficient, which is used to assess driving risks and improve vehicle driving safety. Background Art

[0002] The risk assessment method of risk potential field has been widely used in the fields of traffic safety, autonomous driving and intelligent transportation systems. Potential field theory originates from physics and is mainly used to describe the distribution of forces in space and their impact on objects. In the field of risk assessment, potential field theory is used to describe the risk level at different locations. By constructing a risk potential field map, the risk distribution in different areas can be intuitively displayed.

[0003] Existing automobile driving safety technologies have been widely used in various intelligent assisted driving systems, such as automatic emergency braking systems (AEB), lane keeping assist systems (LKA), adaptive cruise control systems (ACC), etc. However, these systems mainly rely on objective factors of the external environment, such as traffic conditions, road conditions, weather conditions, etc., to judge the driving safety of the vehicle. Although these technologies can reduce traffic accidents caused by objective factors to a certain extent, there is a lack of effective evaluation methods for the driver's own behavior pattern, that is, the impact of driving style on driving safety. However, the current driving safety evaluation technology has not fully considered the driver's personalized driving behavior and style differences. The existing methods are more focused on monitoring and analyzing the external environment of the vehicle, and lack a systematic quantitative analysis of the driver's operating behavior. Summary of the invention

[0004] In order to solve the problems existing in the related art, the present invention provides a method for evaluating automobile risk potential field based on driving style coefficient. This method introduces the driving style coefficient and trajectory prediction obtained through driving behavior data on the basis of the risk potential field, and uses a suitable risk potential field to more accurately judge the risk, thereby overcoming the limitations and generalization of the traditional risk potential field method, especially in terms of considering the individual differences and behavioral characteristics of drivers. The traditional risk potential field method mainly relies on external environmental factors, while ignoring the impact of the driver's personalized behavior on risk. The present invention accurately extracts the driving style characteristics through the analysis of the driver's existing driving data, and integrates it into the risk potential field model, thereby improving the risk potential field's ability to identify driving safety, and can more comprehensively and accurately assess the driver's risk level under various driving conditions, thereby improving vehicle driving safety.

[0005] A method for evaluating automobile risk potential field based on driving style coefficients comprises the following steps:

[0006] Step 1: Establish velocity risk function and acceleration risk function.

[0007] Step 2: Collect the driver's driving behavior data, including acceleration, speed data and historical trajectory in different time windows, and calculate the risk integral value of different time windows through the established speed risk function and acceleration risk function. Record the vehicle driving information in different time windows and calculate the normalized driving style coefficient on the time window in chronological order.

[0008] Step 3: Predict the target vehicle and surrounding vehicles based on the hybrid neural network model.

[0009] Step 4: Generate a risk potential field with different driving vehicle emphases using the driving style coefficient to assess driving risk.

[0010] Step 5: Based on the predicted trajectory of the target vehicle and surrounding vehicles and their generation probability, calculate the target vehicle in the interactive vehicle potential field E V The expected risk potential field E.

[0011] Step 6: Couple the driving style coefficient, the predicted expected risk and the Gaussian vehicle risk potential field to obtain the final risk potential field to assess the risk of the vehicle trip.

[0012] Furthermore, the specific operations of step 1 are as follows:

[0013] Speed ​​risk function R v The formula for (v(t)) is:

[0014]

[0015] Where a(t) is the vehicle speed at time t. The acceleration risk function formula is:

[0016]

[0017] Furthermore, the specific method of step 2 is as follows:

[0018] Based on the risk function established in step 1, within a given time window [tw, t], calculate the acceleration a(t) at each moment, and the formula is:

[0019] a(t)=F(t) / m

[0020] Where F(t) is the total external force that varies with time and m is the mass of the vehicle. Then the acceleration risk function R is applied a Calculate the acceleration risk value R at each moment a (a(t)).

[0021] For each moment in the time window [tw,t], perform the integral operation of the risk function:

[0022]

[0023] Where R v (a(t)) is the speed risk function, v(t) is the vehicle speed at time t, and a(t) is the vehicle acceleration at time t. This integral result comprehensively considers the impact of acceleration and speed on driving risk.

[0024] The calculated time window integration result R t (t) is normalized to eliminate the influence of different time window lengths on the evaluation results and obtain the overall driving style score:

[0025]

[0026] Among them, R max and R min are the maximum and minimum values ​​of the time window integration results, respectively, and N is the number of acquired time windows.

[0027] As the driver continues to drive, the generation of new time windows will affect the driving style score. After each new time window is generated, the formula for updating the driving style coefficient can be expressed as:

[0028]

[0029] Where N(t) is the number of valid time windows at time t, R max (t) and R min (t) is the maximum and minimum value in the current time window.

[0030] Furthermore, the specific method of step 3 is as follows:

[0031] Collect the historical state data of the driving vehicle, including position, speed, acceleration and steering angle. Before inputting into the hybrid neural network, preprocess the collected data, including denoising, normalization and filling missing values, to improve the performance of the model. Define the state of the vehicle, including normal driving, deceleration and lane change. Based on the existing state and historical data, establish a Markov chain model to describe the transition probability between vehicle states, that is, the state transition matrix.

[0032] Then the hybrid neural network model is established. Firstly, the historical trajectory, speed, acceleration data and state transition probability of the vehicle and surrounding vehicles are used as the input of the hybrid neural network.

[0033] Then, the convolutional neural network (CNN) extracts the spatial trajectory features of the moving vehicle and its surrounding vehicles; the long short-term memory network (LSTM) is used to process the time series data and extract the speed and acceleration features. The fully connected layer is used to integrate the features extracted from the CNN and LSTM, and a series of coordinate points of continuous paths are output. The probability of each possible path is output through the softmax layer to ensure that the sum of the probabilities of all paths is 1 and the cross entropy loss function is used to train the probability output part to ensure that the model can accurately predict the probability of each path. Finally, the similarity is used to measure the accuracy of the model output path.

[0034] The target vehicle is predicted through the hybrid neural network model, generating k different trajectories and the generation probability p of each trajectory. k At the same time, for the surrounding m vehicles, k different trajectories are generated and the generation probability p of each trajectory is mk , where m = 1, 2, ... M. Finally, the trajectory generation probabilities of the target vehicle and surrounding vehicles are normalized by softmax to ensure that the sum of the probabilities of all generated trajectories is 1, that is:

[0035] Σ k p k =1

[0036] Furthermore, the specific method of step 4 is as follows:

[0037] A Gaussian automobile risk potential field is established. Through the definition of the Gaussian potential field, the potential fields from different sources are uniformly modeled. The automobile is defined as the charge in the Gaussian field, and its charge q is defined as:

[0038] q=k*m

[0039] Where m is the mass of the car and k is the constant of proportionality.

[0040] The general Gaussian potential field is defined as follows:

[0041]

[0042] Where N is the total number of potential field sources, A i is the field strength coefficient of the ith potential source, d i is the distance between the current vehicle and the i-th potential field source, σ i is the influence range of the i-th potential field source.

[0043] In order to prevent vehicles from changing lanes at will during driving, the lane marking potential field E is set L , which is defined as:

[0044]

[0045] Among them, ALi is the field strength coefficient of the i-th lane marking, d Li is the distance between the vehicle and the lane marking, σ L is the attenuation coefficient of the lane marking.

[0046] In order to ensure that the vehicle drives safely on the lane, the road boundary potential field E is set B , which is defined as:

[0047]

[0048] Among them, A Bj is the field strength coefficient at the jth road boundary, d Bj is the distance between the vehicle and the road boundary, σ B is the attenuation parameter of the road boundary.

[0049] To prevent collision with other vehicles, an interactive vehicle potential field E is set V , which is defined as:

[0050]

[0051] Among them, A Vk is the field strength coefficient of the kth interacting vehicle, d Vk is the distance between the vehicle and the kth interacting vehicle, σ V is the attenuation parameter of the interacting vehicle potential field.

[0052] To prevent danger caused by changes in road conditions, set up signal equipment potential field E S , which is defined as:

[0053]

[0054] Among them, A Sl is the field strength coefficient of the lth signal device, d Sl is the distance between the vehicle and the lth signal device, σ S is the attenuation parameter of the signal device, and L is the total number of signal devices.

[0055] Therefore, the total potential field of all potential sources is E T is defined as:

[0056]

[0057] Where N is the total number of potential field sources.

[0058] Furthermore, the specific method of step 5 is as follows:

[0059] The predicted trajectories of the target vehicle and surrounding vehicles obtained in step 3 and their generation probabilities are calculated to calculate the target vehicle in the interactive vehicle potential field E VThe expected risk value of the target vehicle is E. Suppose the risk value of the interaction between the target vehicle and the surrounding vehicles in the predicted trajectory after t seconds is E km ,have to:

[0060] E km =E V

[0061] The target vehicle is in the interactive vehicle potential field E V The expected risk potential field E can be expressed as:

[0062]

[0063] Among them, p k is the generation probability of the kth predicted trajectory of the target vehicle, p km is the probability of generating the kth predicted trajectory of the mth surrounding vehicle, E km is the risk value between the target vehicle and surrounding vehicles.

[0064] Furthermore, the specific method of step 6 is as follows:

[0065] The total potential field E is divided into the static risk potential field E static and expected risk potential field E, static risk potential field E static Defined as:

[0066] E static =E L +E B +E S

[0067] In order to make the risk potential field focus on different drivers with different driving styles, drivers with aggressive driving styles pay more attention to the interaction with other vehicles, while drivers with conservative driving styles pay more attention to the safety of lane keeping and road boundaries. style (t) is coupled with the risk potential field to obtain the final risk potential field E final , defined as:

[0068] E final =E static +R style (t)*E

[0069] Compared with the prior art, the advantages of the present invention are as follows:

[0070] 1) By analyzing the existing driving data of the driver, the driving style characteristics are accurately extracted, which overcomes the deficiency of the traditional method in lacking systematic quantitative analysis of driver behavior. This overcomes the problem of ignoring individual differences of drivers in the traditional method;

[0071] 2) The traditional risk potential field method is limited to external environmental factors when judging risks, and its generalization is poor. The present invention introduces a driving style coefficient into the risk potential field model, which can more accurately judge driving risks and improve the risk potential field's ability to identify driving safety;

[0072] 3) The deep learning model can predict vehicle trajectories and their generation probabilities in real time. By combining the driving style coefficient with the predicted trajectory and its potential field probability, the expected risk value is calculated, which can more comprehensively and accurately assess the driver's driving risks in various environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0074] Figure 1 It is a flow chart of the method of the present invention;

[0075] Figure 2 is the lane marking potential field E L Schematic diagram of

[0076] Figure 3 is the road boundary potential field E B Schematic diagram of

[0077] Figure 4 is the interactive vehicle potential field E V Schematic diagram of

[0078] Figure 5 is the signal equipment potential field E S Schematic diagram of . DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] A vehicle risk potential field assessment method based on driving style coefficient, the framework is as follows Figure 1 As shown, the following steps are included:

[0081] A method for evaluating automobile risk potential field based on driving style coefficients comprises the following steps:

[0082] Step 1: Establish velocity risk function and acceleration risk function;

[0083] Based on the risk function developed by Kloeden and McLean on Australian roads, the speed risk function R v The formula for (v(t)) is:

[0084]

[0085] Where a(t) is the vehicle speed at time t. The acceleration risk function formula is:

[0086]

[0087] Step 2: Collect the driver's driving behavior data, including acceleration, speed data and historical trajectory in different time windows. Calculate the risk integral value of different time windows through the established speed risk function and acceleration risk function. Record the vehicle driving information in different time windows and calculate the normalized driving style coefficient on the time window in chronological order.

[0088] Based on the risk function established in step 1, the acceleration a(t) at each moment is calculated within a given time window [tw,t], and the formula is:

[0089] a(t)=F(t) / m

[0090] Where F(t) is the total external force that varies with time and m is the mass of the vehicle. Then the acceleration risk function R is applied a Calculate the acceleration risk value R at each moment a (a(t)).

[0091] For each moment in the time window [tw,t], perform the integral operation of the risk function:

[0092]

[0093] Where R v (a(t)) is the speed risk function, v(t) is the vehicle speed at time t, and a(t) is the vehicle acceleration at time t. This integral result comprehensively considers the impact of acceleration and speed on driving risk.

[0094] The calculated time window integration result R t (t) is normalized to eliminate the influence of different time window lengths on the evaluation results and obtain the overall driving style score:

[0095]

[0096] Among them, Rmax and R min are the maximum and minimum values ​​of the time window integration results, respectively, and N is the number of acquired time windows.

[0097] As the driver continues to drive, the generation of new time windows will affect the driving style score. After each new time window is generated, the formula for updating the driving style coefficient can be expressed as:

[0098]

[0099] Where N(t) is the number of valid time windows at time t, R max (t) and R min (t) is the maximum and minimum value in the current time window.

[0100] This comprehensive score R style It can accurately reflect the overall driving style of the vehicle and provide a basis for the improvement of the risk potential field.

[0101] Step 3: predicting the target vehicle and surrounding vehicles based on the hybrid neural network model;

[0102] Collect the historical state data of the driving vehicle, including position, speed, acceleration and steering angle. Before inputting into the hybrid neural network, preprocess the collected data, including denoising, normalization and filling missing values, to improve the performance of the model. Define the state of the vehicle, including normal driving, deceleration and lane change. Based on the existing state and historical data, establish a Markov chain model to describe the transition probability between vehicle states, that is, the state transition matrix.

[0103] Then the hybrid neural network model is established. Firstly, the historical trajectory, speed, acceleration data and state transition probability of the vehicle and surrounding vehicles are used as the input of the hybrid neural network.

[0104] Then, the convolutional neural network (CNN) extracts the spatial trajectory features of the moving vehicle and its surrounding vehicles; the long short-term memory network (LSTM) is used to process the time series data and extract the speed and acceleration features. The fully connected layer is used to integrate the features extracted from the CNN and LSTM, and a series of coordinate points of continuous paths are output. The probability of each possible path is output through the softmax layer to ensure that the sum of the probabilities of all paths is 1 and the cross entropy loss function is used to train the probability output part to ensure that the model can accurately predict the probability of each path. Finally, the similarity is used to measure the accuracy of the model output path.

[0105] The target vehicle is predicted through the hybrid neural network model, generating k different trajectories and the generation probability p of each trajectory. kAt the same time, for the surrounding m vehicles, k different trajectories are generated and the generation probability p of each trajectory is mk , where m = 1, 2, ... M. Finally, the trajectory generation probabilities of the target vehicle and surrounding vehicles are normalized by softmax to ensure that the sum of the probabilities of all generated trajectories is 1, that is:

[0106] Σ k p k =1

[0107] Step 4: Generate a risk potential field with different driving vehicle emphases using the driving style coefficient to assess driving risk;

[0108] Establish a Gaussian automobile risk potential field and define the static risk potential field E static and the interactive vehicle potential field E V , where E static Including the lane marking potential field E L , road boundary potential field E B and signal equipment potential field E S ,Right now:

[0109] E static =E L +E B +E S

[0110] The specific operations are as follows:

[0111] A Gaussian automobile risk potential field is established. Through the definition of the Gaussian potential field, the potential fields from different sources are uniformly modeled. The automobile is defined as the charge in the Gaussian field, and its charge q is defined as:

[0112] q=k*m

[0113] Where m is the mass of the car and k is the constant of proportionality.

[0114] The general Gaussian potential field is defined as follows:

[0115]

[0116] Where N is the total number of potential field sources, A i is the field strength coefficient of the ith potential source, d i is the distance between the current vehicle and the i-th potential field source, σ i is the influence range of the i-th potential field source.

[0117] In order to prevent vehicles from changing lanes at will during driving, the lane marking potential field E is set L , which is defined as:

[0118]

[0119] Among them, A Li is the field strength coefficient of the i-th lane marking, d Li is the distance between the vehicle and the lane marking, σ L is the attenuation coefficient of the lane marking.

[0120] In order to ensure that the vehicle drives safely on the lane, the road boundary potential field E is set B , which is defined as:

[0121]

[0122] Among them, A Bj is the field strength coefficient at the jth road boundary, d Bj is the distance between the vehicle and the road boundary, σ B is the attenuation parameter of the road boundary.

[0123] To prevent collision with other vehicles, an interactive vehicle potential field E is set V , which is defined as:

[0124]

[0125] Among them, A Vk is the field strength coefficient of the kth interacting vehicle, d Vk is the distance between the vehicle and the kth interacting vehicle, σ V is the attenuation parameter of the interacting vehicle potential field.

[0126] To prevent danger caused by changes in road conditions, set up signal equipment potential field E S , which is defined as:

[0127]

[0128] Among them, A Sl is the field strength coefficient of the Lth signal device, d Sl is the distance between the vehicle and the lth signal device, σ S is the attenuation parameter of the signal device, and L is the total number of signal devices.

[0129] Therefore, the total potential field of all potential sources is E T is defined as:

[0130]

[0131] Where N is the total number of potential field sources.

[0132] Figure 2-Figure 5 They are the lane marking potential field E L 、Road boundary potential field E B 、Interactive vehicle potential field EV and signal equipment potential field E S Schematic diagram of .

[0133] Step 5: Based on the predicted trajectory of the target vehicle and surrounding vehicles and their generation probability, calculate the target vehicle in the interactive vehicle potential field E V The expected risk potential field E;

[0134] The predicted trajectories of the target vehicle and surrounding vehicles obtained in step 3 and their generation probabilities are calculated to calculate the target vehicle in the interactive vehicle potential field E V The expected risk value of the target vehicle is E. Suppose the risk value of the interaction between the target vehicle and the surrounding vehicles in the predicted trajectory after t seconds is E km ,have to:

[0135] E km =E V

[0136] Calculate the target vehicle's potential field E in the interactive vehicle V The expected risk value E is

[0137]

[0138] Among them, p k is the generation probability of the kth predicted trajectory of the target vehicle, p km is the probability of generating the kth predicted trajectory of the mth surrounding vehicle, E km is the risk value between the target vehicle and surrounding vehicles.

[0139] Step 6: Risk assessment and decision support. In order to make the risk potential field focus on different drivers with different driving styles, aggressive drivers pay more attention to the interaction with other vehicles, while conservative drivers pay more attention to the safety of lane keeping and road boundaries. style (t) and risk potential field E static and E V Couple and calculate the final risk potential field E final :

[0140] E final =E static +R style (t)*E

[0141] Based on E final When high-risk behavior is detected, the system can improve safety through sound prompts, visual warnings, or automatic adjustments to the vehicle control system, hoping to achieve the goal of reducing traffic accidents and improving road safety.

[0142] Step 7: When the driving risk is detected to exceed the preset threshold, the driver is reminded to improve his driving behavior and the status of the vehicle lights is used to remind the following vehicles to pay attention to the risk.

[0143] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. A method for evaluating automobile risk potential field based on driving style coefficient, characterized in that: The steps include: Step 1: Establish velocity risk function and acceleration risk function; Step 2: Collect the driver's driving behavior data, including acceleration, speed data and historical trajectory in different time windows, and calculate the risk integral value of different time windows through the established speed risk function and acceleration risk function; Recording vehicle driving information in different time windows and calculating normalized driving style coefficients on the time windows in chronological order; Step 3: predicting the target vehicle and surrounding vehicles based on the hybrid neural network model; Step 4: Generate a risk potential field with different driving vehicle emphases using the driving style coefficient to assess driving risk; Step 5: Based on the predicted trajectory of the target vehicle and surrounding vehicles and their generation probability, calculate the target vehicle in the interactive vehicle potential field E V The expected risk potential field E; Step 6: Couple the driving style coefficient, the predicted expected risk and the Gaussian vehicle risk potential field to obtain the final risk potential field to assess the risk of the vehicle trip.

2. The method for evaluating automobile risk potential field based on driving style coefficient according to claim 1, characterized in that: Step 1: Speed ​​risk function R v The formula for (v(t)) is: Where a(t) is the vehicle speed at time t; the acceleration risk function formula is:

3. The method for evaluating automobile risk potential field based on driving style coefficient according to claim 1, characterized in that: Step 2: Based on the risk function established in step 1, within a given time window [tw, t], calculate the acceleration a(t) at each moment, and the formula is: a(t)=F(t) / m Where F(t) is the total external force that varies with time, and m is the mass of the vehicle. Then the acceleration risk function R is applied a Calculate the acceleration risk value R at each moment a (a(t)); For each moment in the time window [tw,t], perform the integral operation of the risk function: Where R v (a(t)) is the speed risk function, v(t) is the vehicle speed at time t, and a(t) is the vehicle acceleration at time t; this integral result comprehensively considers the impact of acceleration and speed on driving risk; The calculated time window integration result R t (t) is normalized to eliminate the influence of different time window lengths on the evaluation results and obtain the overall driving style score: Among them, R max and R min are the maximum and minimum values ​​of the time window integration results, respectively, and N is the number of acquired time windows; As the driver continues to drive, the generation of new time windows will affect the driving style score; after each new time window is generated, the formula for updating the driving style coefficient can be expressed as: Where N(t) is the number of valid time windows at time t, R max (t) and R min (t) is the maximum and minimum value in the current time window.

4. The method for evaluating automobile risk potential field based on driving style coefficient according to claim 1, characterized in that: Step 3: Collect historical status data of the driving vehicle, including position, speed, acceleration and steering angle. Before inputting into the hybrid neural network, pre-process the collected data, including denoising, normalization and filling missing values, to improve the performance of the model; define the status of the vehicle, including normal driving, deceleration and lane change; Based on the existing state and historical data, a Markov chain model is established to describe the transition probability between vehicle states, namely the state transition matrix; Then a hybrid neural network model is established; first, the historical trajectory, speed, acceleration data and state transition probability of the vehicle and surrounding vehicles are used as the input of the hybrid neural network; Then, the convolutional neural network (CNN) extracts the spatial trajectory features of the moving vehicle and its surrounding vehicles; the long short-term memory network (LSTM) is used to process the time series data and extract the speed and acceleration features; the fully connected layer is used to integrate the features extracted from the CNN and LSTM, and output a series of coordinate points of continuous paths; the probability of each possible path is output through the softmax layer to ensure that the sum of the probabilities of all paths is 1 and the cross entropy loss function is used to train the probability output part to ensure that the model can accurately predict the probability of each path, and finally the similarity is used to measure the accuracy of the model output path; The target vehicle is predicted through the hybrid neural network model, generating k different trajectories and the generation probability p of each trajectory. k At the same time, for the surrounding m vehicles, k different trajectories are generated and the generation probability p of each trajectory is mk , where m = 1, 2, ... M; Finally, the trajectory generation probabilities of the target vehicle and surrounding vehicles are normalized by softmax to ensure that the sum of the probabilities of all generated trajectories is 1, that is: S k p k =1.

5. The method for evaluating automobile risk potential field based on driving style coefficient according to claim 1, characterized in that: Step 4: A Gaussian automobile risk potential field is established. Through the definition of the Gaussian potential field, the potential fields from different sources are uniformly modeled. The automobile is defined as the charge in the Gaussian field, and its charge q is defined as: q=k*m Where m is the mass of the car and k is the proportionality constant; The general Gaussian potential field is defined as follows: Where N is the total number of potential field sources, A i is the field strength coefficient of the ith potential source, d i is the distance between the current vehicle and the i-th potential field source, σ i is the influence range of the ith potential source; In order to prevent vehicles from changing lanes at will during driving, the lane marking potential field E is set L , which is defined as: Among them, A Li is the field strength coefficient of the i-th lane marking, d Li is the distance between the vehicle and the lane marking, σ L is the attenuation coefficient of the lane marking; In order to ensure that the vehicle drives safely on the lane, the road boundary potential field E is set B , which is defined as: Among them, A Bj is the field strength coefficient at the jth road boundary, d Bj is the distance between the vehicle and the road boundary, σ B is the attenuation parameter of the road boundary; To prevent collision with other vehicles, an interactive vehicle potential field E is set V , which is defined as: Among them, A Vk is the field strength coefficient of the kth interacting vehicle, d Vk is the distance between the vehicle and the kth interacting vehicle, σ V is the attenuation parameter of the interacting vehicle potential field; To prevent danger caused by changes in road conditions, set up signal equipment potential field E S , which is defined as: Among them, A Sl is the field strength coefficient of the lth signal device, d Sl is the distance between the vehicle and the lth signal device, σ S is the attenuation parameter of the signal device, and L is the total number of signal devices.

6. The method for evaluating automobile risk potential field based on driving style coefficient according to claim 4, characterized in that: Step 5: The predicted trajectories of the target vehicle and surrounding vehicles obtained in step 3 and their generation probabilities are calculated to calculate the target vehicle in the interactive vehicle potential field E V The expected risk value of the target vehicle is E; the risk value of the interaction between the target vehicle and the surrounding vehicles in the predicted trajectory after t seconds is E km ,have to: AND km =And V The target vehicle is in the interactive vehicle potential field E V The expected risk potential field E can be expressed as: Among them, p k is the generation probability of the kth predicted trajectory of the target vehicle, p km is the probability of generating the kth predicted trajectory of the mth surrounding vehicle, E km is the risk value between the target vehicle and surrounding vehicles.

7. The method for evaluating automobile risk potential field based on driving style coefficient according to claim 6, characterized in that: Step 6: The total potential field E is divided into the static risk potential field E static and expected risk potential field E, static risk potential field E static Defined as: AND static =And L +E B +E S Among them, E static Including lane marking potential field E L , road boundary potential field E B and signal equipment potential field E S ; In order to make the risk potential field focus on different drivers with different driving styles, drivers with aggressive driving styles pay more attention to the interaction with other vehicles, and drivers with conservative driving styles pay more attention to the safety of lane keeping and road boundaries; the driving style coefficient R style (t) is coupled with the risk potential field to obtain the final risk potential field E final , defined as: E final =E static +R style (t)*E。

Images