Adaptive Driver Following Method Considering Perception Process and Driver Behavior

Abstract

The present invention relates to the field of autonomous driving adaptive cruise control, and specifically relates to an adaptive driver following method that comprehensively considers the perception process and driver behavior, including obtaining the motion trajectory data of the current vehicle and the motion trajectory data of the vehicle following the current vehicle; estimating the speed of the following vehicle and the distance between the current vehicle and the following vehicle based on external perception disturbances according to the historical motion trajectory data of the current driver; identifying driver characteristic parameters from the distance between the current vehicle and the following vehicle and the driving speed of the current vehicle; establishing a following behavior model based on the perception process and driver behavior characteristics, taking the driver characteristic parameters and the speed of the current vehicle as inputs, and calculating the discrete value of the following distance; the present invention shows good performance in describing the following behavior of individual drivers, and the absolute value of the calculated average error compared with the prediction result of the classical FVD model is less than 80%, and the degree of coincidence with the actual following behavior of the driver is higher.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving adaptive cruise, and particularly relates to an adaptive driver following method that comprehensively considers the perception process and driver behavior. Background Art

[0002] As one of the driver models, the driver following model has always been a research hotspot in traffic flow theory, which describes the interaction between adjacent vehicles in a vehicle fleet traveling on a single lane with overtaking restricted. Most of the existing following models are established using foreign driving behavior test data sets, and the models reflect the following characteristics of foreign roads and drivers. There are certain differences in traffic, vehicles, driving styles and cultures among different countries, and these differences are very likely to cause obvious differences in driving behavior characteristics. Secondly, in terms of the drivers themselves, due to the individual differences in the reaction time and attention level of different drivers, and at the same time, the psychological characteristics, driving experience and road environment perception level of the drivers are highly subjective. The existing models ignore the state relationship between the driver characteristic parameters and the perception process, are difficult to adapt to the driving habits of different drivers, and cannot effectively improve the satisfaction and acceptance of human-machine interaction and cooperation. Summary of the Invention

[0003] In order to be able to simulate the driving habits of different drivers to complete vehicle following in different driving scenarios, the present invention proposes an adaptive driver following method that comprehensively considers the perception process and driver behavior, specifically including the following steps:

[0004] Obtain the motion trajectory data of the current vehicle and the motion trajectory data of the vehicle following the current vehicle;

[0005] Based on the external perception disturbance, estimate the speed of the following vehicle and the distance between the current vehicle and the following vehicle according to the historical motion trajectory data of the current driver;

[0006] Identify the driver characteristic parameters from the distance between the current vehicle and the following vehicle and the driving speed of the current vehicle;

[0007] Establish a following behavior model based on the perception process and driver behavior characteristics, take the driver characteristic parameters and the speed of the current vehicle as inputs, and calculate the discrete value of the following distance.

[0008] Further, the process of estimating the distance between the current vehicle and the following vehicle includes:

[0009] lns est -lns = V s w s (t);

[0010] Wherein, s estis the estimated value of the distance between the current vehicle and the following vehicle; s is the actual value of the distance between the current vehicle and the following vehicle collected by the vehicle-mounted sensing device; V s is the estimated value of the inter-vehicle distance s est is the standard relative error of the logarithm of the actual value s; w s (t) is the error of the estimated value of the distance between the current vehicle and the following vehicle.

[0011] Furthermore, the process of estimating the speed of the following vehicle includes:

[0012]

[0013] where is the estimated value of the speed of the following vehicle; v l is the actual speed provided by the vehicle-mounted device; s is the actual value of the distance between the current vehicle and the following vehicle collected by the vehicle-mounted sensing device; σ r is the standard deviation of the perspective change rate; w l (t) is the distribution of the estimated error of the leading vehicle's speed and its change over time.

[0014] Furthermore, the perspective change rate is expressed as:

[0015]

[0016] τ TTC = s / Δv;

[0017] where τ TTC is the time to collision; Δv is the difference between the speed of the host vehicle and the speed of the leading vehicle.

[0018] Furthermore, a stochastic differential equation is constructed, and a group of white noise initialized with a set of pseudo-random numbers is substituted into the constructed equation. The value obtained by solving is used as the error of the estimated value of the speed of the following vehicle or the error of the estimated value of the distance between the current vehicle and the following vehicle. The solution of the stochastic differential equation is expressed as:

[0019]

[0020] where w(t) represents the solution of a stochastic differential equation; a 0 、a 1 、a are intermediate parameters, expressed as t 0 represents the time at a certain moment; t represents the following behavior time; ξ(s) represents the standard white noise, s represents the actual value of the distance between the current vehicle and the following vehicle collected by the vehicle-mounted sensing device; τ represents the driver's estimated error time.

[0021] Furthermore, the standard white noise ξ(s) is expressed as:

[0022]

[0023] Among them, \(w\) represents \(w(t)\), which is the solution of a stochastic differential equation.

[0024] Furthermore, the driver characteristic parameters are identified through the ego-vehicle speed \(v\) f (t), and the specific steps are as follows:

[0025]

[0026] Among them, is the parameter vector, which is composed of the driver's time-distance discrete value parameter \(\lambda(k)\) and the discrete value \(L(k)\) of the driver's stopping distance, and is expressed as and is the process matrix; is the input matrix constructed according to the ego-vehicle speed \(v\) f (t), and is expressed as

[0027] Furthermore, during the \(N\) iterations, those that meet the stability criteria are selected as the candidate set of driver parameters, and the stability criteria are:

[0028] \(\Delta\lambda(k)=|(\lambda(k)-\lambda(k - 1)) / \lambda(k)|\);

[0029] \(\Delta L(k)=|(L(k)-L(k - 1)) / L(k)|\);

[0030] \(\Delta=\max\{\Delta\lambda(k),\Delta L(k)\}<\varepsilon\);

[0031] Among them, \(\varepsilon\) is a constant.

[0032] Furthermore, a linear relationship of the form \(Y = aX + b\) is constructed from the parameter characteristics of the candidate set, and the parameter vector with the correlation coefficient closest to 1 in the constructed linear relationship is selected. The correlation coefficient is expressed as:

[0033]

[0034] Among them, \(X\) i represents the \(i\)-th independent variable in the linear relationship; \(Y\) i represents the \(i\)-th dependent variable in the linear relationship; \(N\) represents the total number of samples. In this application, the two parameters of the following vehicle speed and the following vehicle distance are selected according to this correlation coefficient, and the coefficients \(a\) and the bias term \(b\) in the linear relationship \(Y = aX + b\) are identified as the driver characteristic parameters.

[0035] Furthermore, a following behavior model based on the perception process and the driver's behavior characteristics is established. Taking the driver characteristic parameters and the speed of the current vehicle as inputs, the discrete value of the following vehicle distance is calculated:

[0036] The vehicle dynamics model during the driver's longitudinal car - following process can be written as:

[0037]

[0038] The vehicle dynamics model during the driver's longitudinal car - following process is a third - order non - linear system. There is an error between the actual following distance and the ideal following distance s d The following error is described as follows:

[0039]

[0040] Substitute the controllable quantity u(t) into the vehicle longitudinal dynamics state equation, and we get:

[0041]

[0042] Among them, f(X) and g(X) are the nominal functions about the system state respectively. u(t) and y represent the control input of the system and the system output respectively. u(t) is the control generated according to the driver characteristic parameters; The state variable of the system, x f is the driving distance of the host vehicle, is the following speed of the host vehicle, is the following acceleration of the host vehicle, d(t) is the disturbance of the system; represents the vector about the following error and its derivative, x l represents the driving distance of the leading vehicle, s 0 is the initial distance between the two vehicles; represents the speed of the leading vehicle; represents the acceleration of the leading vehicle; represents the change rate of the following acceleration of the following vehicle; μ represents the lag time constant of the engine or braking system, M is the vehicle's own weight, F R is the vehicle rolling resistance, L is the stopping distance obtained by the driver characteristic identification function, λ is the time distance obtained by the driver characteristic identification function, c 1 、c 2 are the characteristic parameters satisfying the Hurwitz stability criterion, is the fourth - order derivative of the following speed of the host vehicle, represents the estimated value of the model switching gain, sat(S(X)) is the saturation function form of S(X), and S(X) is the sliding - mode function.

[0043] The present invention constructs a car-following model with adaptive learning ability that comprehensively considers the perception process and driver characteristics. The general Wiener process is used to simulate the independence and randomness of the driver's perception process, and the least recursive squares method is used to learn driving habits. Therefore, the car-following model proposed by the present invention has the advantages of strong interpretability and easy description of the uncertainty of driving behavior, and can provide a more accurate, anthropomorphic and driver-expected driver car-following model for the adaptive cruise control system, automatic following system, etc. of intelligent vehicles; In addition, the verification research based on natural driving behavior experimental data shows that the adaptive car-following model that comprehensively considers the perception process and driver behavior proposed by the present invention shows good performance in describing the car-following behavior of individual drivers. Compared with the prediction results of the classical FVD model, the absolute value of the average error calculated by the model is less than 80%, and the model has a higher degree of agreement with the actual car-following behavior of the driver. Description of the Drawings

[0044] Figure 1 Schematic diagram of the driver's longitudinal car-following behavior model in the present invention;

[0045] Figure 2 Relationship between the driver's perspective transformation and relative speed in the present invention;

[0046] Figure 3 Flowchart of the self-learning parameter identification algorithm for driver characteristics in a preferred embodiment of the present invention;

[0047] Figure 4 Flow of the driver characteristic parameter identification test scheme in a preferred embodiment of the present invention;

[0048] Figure 5 Schematic diagram of the comparison of car-following errors between the present invention and FVD Figure 1 ;

[0049] Figure 6 Schematic diagram of the comparison of car-following errors between the present invention and FVD Figure 2 ;

[0050] Figure 7 Schematic diagram of the comparison of safety indicators between the present invention and FVD Figure 1 ;

[0051] Figure 8 Schematic diagram of the comparison of safety indicators between the present invention and FVD Figure 2 。 Detailed Embodiments

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] The present invention proposes an adaptive driver following method that comprehensively considers the perception process and driver behavior, specifically including the following steps:

[0054] Obtain the motion trajectory data of the current vehicle and the motion trajectory data of the vehicle following the current vehicle;

[0055] Based on the external perception disturbance, estimate the speed of the following vehicle and the distance between the current vehicle and the following vehicle according to the historical motion trajectory data of the current driver;

[0056] Identify the driver characteristic parameters from the distance between the current vehicle and the following vehicle and the driving speed of the current vehicle;

[0057] Establish a following behavior model based on the perception process and driver behavior characteristics, take the driver characteristic parameters and the speed of the current vehicle as inputs, and calculate the discrete value of the following distance.

[0058] Embodiment

[0059] Research shows that drivers will try to maintain a desired following distance during the following process, and are affected by the driving environment, the driving states of the two vehicles (the distance between the two, the vehicle speed, etc.), and the driving style characteristics of the driver. Different drivers show obvious differences in following behavior. Therefore, in order to make the following model more adaptable, the present invention provides an adaptive following model that comprehensively considers external perception disturbance and driver characteristics. As Figure 1 , in this embodiment, the driver characteristic parameters are identified according to the speed of the leading vehicle, the speed of the following vehicle, and the distance between the two vehicles, and the speed of the following vehicle is controlled according to the driver characteristic parameters to achieve the purpose of the following vehicle following the leading vehicle. This method mainly includes two parts: external perception disturbance simulation and driver characteristic parameter adaptive identification method. This embodiment further illustrates and analyzes the solution of the present invention from the two aspects of external perception disturbance simulation and driver characteristic parameter adaptive identification method.

[0060] 1. External perception disturbance simulation

[0061] External perception disturbance simulation refers to simulating the subjective perception process of drivers towards the driving environment. During vehicle following, the driver of the host vehicle often estimates the speed and following distance information of the leading vehicle based on the relative position change of the leading vehicle in the field of vision and the change in the size of the rear of the leading vehicle, and this process has a certain degree of independence and randomness. Facing the above facts, if the accurate values of the following distance and the speed of the leading vehicle obtained by the sensor are directly input into the decision-making layer of the following model during the modeling process, the model results cannot essentially reflect the following behaviors of different types of drivers in real driving scenarios. In order to obtain a following model that conforms to the driving behavior characteristics of drivers, it is necessary to consider the independence and randomness of the perception process during the modeling process. In view of this, this module introduces the generalized Wiener process to describe the perception characteristics of drivers, and the output results obtained therefrom will be used as the input of the judgment and operation module, which to a certain extent reflects the influence of the perception process affected by environmental disturbances in the driver following model.

[0062] 1.1 Driver's perception process of following distance

[0063] In most driving situations, the relative error of the following distance can be considered a constant value, usually expressed by the logarithmic difference of the following distance. The evolution process of the driver's perception error of the following distance can be described as:

[0064] ln s est -ln s = V s w s (t)

[0065] Among them, V s represents the standard relative error of the logarithm of the estimated following distance s est and the actual value s, also known as the coefficient of statistical variation. Assuming that the estimation error is an unbiased error, w s (t) represents a random variable subject to the standard normal distribution.

[0066] 1.2 Driver's perception process of the speed of the leading vehicle

[0067] In real driving situations, drivers estimate the size of the leading vehicle through the relative position change of the leading vehicle in the field of vision and the change in the size of the rear of the vehicle in the field of vision. In the top view describing the driver's field of vision, the current leading vehicle angle φ is as Figure 2 shown, the angle change is dφ, and the relative position of the leading vehicle within the time dt is (v f - v l )dt.

[0068] Based on the above reasoning, the visual angle change rate r is defined as follows:

[0069]

[0070] As can be seen from the above equation, the relative perspective change rate r is the time-to-collision, τ TTC = the reciprocal of s / Δv, as Figure 2 , the value of Δv is v f -v l , w veh is the vehicle width, s is the actual value of the distance between the current vehicle and the following vehicle collected by the on-vehicle sensing device, is the perspective of the leading vehicle; τ TTC is an important safety indicator for measuring safety anti-collision. The leading vehicle speed error is defined as follows:

[0071]

[0072] Among them, σ r is the standard deviation of the perspective change rate, Δv est is the change in the estimated value of the vehicle speed; similar to the random variable w s (t), w l (t) describes the distribution of the leading vehicle speed estimation error and its change over time. w s (t), w l (t) are both random variables subject to the standard normal distribution, reflecting the change in the estimation error shown under the disturbance perception of comprehensive factors such as different drivers' driving conditions, psychological and physiological conditions, and external environment.

[0073] 1.3 Simulation of the Driver's Perception Disturbance Process

[0074] To simulate the change in the driver's perception error with perception characteristics, the introduced generalized Wiener process is used to describe the driver's perception characteristics. By different initializations, different drivers' perception characteristics can be simulated, and ultimately reflected in the change process of the driver's perception error of the leading vehicle speed and the vehicle distance. Its stochastic differential equation expression form is as follows:

[0075]

[0076] Among them, the driver's estimation error time τ is a model parameter, ξ(t) is the standardized white noise, and the solution of the stochastic differential equation can be expressed as:

[0077]

[0078] Select two groups of pseudo-random numbers to initialize the white noise and substitute it into the above equation, and solve a set of independent {w s (t), w l(t)} to represent the perceptual behavior characteristics of the driver under perturbed conditions. The obtained leading vehicle speed and following distance of the driver under perceptual perturbation And its result will be used as the input of the driver's judgment and operation model, and can simulate the influence of the driver's perception process on the car-following model to a certain extent.

[0079] 2. Adaptive Identification Method for Driver Characteristic Parameters

[0080] There are significant differences in the car-following driving styles of different drivers, that is, drivers have different expectations for the following speed and the distance between vehicles during the car-following process. Usually, the time headway (unit: s) and the stopping distance (unit: m) are selected to describe the expected following distance of the driver. Most current car-following models for drivers take the time headway and the stopping distance as constant values, ignoring the characteristic differences between drivers. The car-following models established in this way have poor adaptability to drivers. Therefore, in this paper, the time headway and the stopping distance are used as the driver's car-following behavior characteristic parameters, and they are divided into three categories to represent three types of drivers: aggressive, mild, and conservative. The car-following data collected in the driver perception module, including the speed of the host vehicle, the speed of the leading vehicle, and the following distance, indirectly reflects the car-following style of the driver. Therefore, it is used as the input of the driver characteristic parameter identification algorithm. At the same time, the identified result of the driver characteristic parameter will be used to represent the expected following distance of this type of driver characteristic and substituted into the design of the subsequent control module.

[0081] The adaptive driver characteristic parameter identification algorithm is based on the recursive least squares algorithm RLS (recursive least squares). In the process of parameter recursive estimation, each time new observation data is obtained, the recursive algorithm is used to correct the previous estimation result, and then a new parameter estimation value is recursively obtained.

[0082] The identification of the driver's car-following behavior characteristic parameters aims to identify the driver characteristic parameters λ and L according to the following distance s est (t) and the speed v of the host vehicle f (t), and use the identified parameters as the input parameters of the driver's longitudinal car-following behavior model to establish a car-following behavior model based on the perception process and the driver's behavior characteristics. The description of the recursive least squares algorithm used in the adaptive parameter identification algorithm of this model is as follows:

[0083] Parameter vector It is expressed as follows:

[0084]

[0085] Among them, λ(k) and L(k) respectively represent the discrete values of the driver characteristic parameters time headway and stopping distance, k = 0, 1, 2,..., N, and N is a positive integer.

[0086] Input vector is expressed as follows:

[0087]

[0088] where v f (k) represents the discrete value of the following - vehicle following speed, k = 0, 1, 2, …, N, and N is a positive integer.

[0089] Let the output variable be expressed as follows:

[0090]

[0091] where represents the discrete value of the following - distance, k = 0, 1, 2, …, N, and N is a positive integer.

[0092] If the driver fully controls the vehicle for following, the adaptive following model will collect the driving data during the driver's following process. Since these data indirectly represent the driver's following style or driver type, they will be used as inputs for the driver characteristic parameter identification algorithm; if the vehicle is not fully controlled by the driver, the driver characteristic parameters identified from the previous following record data will be called.

[0093] During the process of obtaining the driver parameters, the driver characteristic parameters are identified through the ego - vehicle speed v f (t), and the specific steps are as follows:

[0094]

[0095] where is the parameter vector, which is composed of the driver's time - distance discrete value parameter λ(k) and the discrete value L(k) of the driver's stopping distance, and is expressed as and is the process matrix; is the input matrix constructed according to the ego - vehicle speed v f (t), and is expressed as I is the identity matrix.

[0096] During the N - iteration process, those meeting the stability criteria are selected as the candidate set of driver parameters, and the stability criteria are:

[0097] Δλ(k)=|(λ(k)-λ(k - 1)) / λ(k)|;

[0098] ΔL(k)=|(L(k)-L(k - 1)) / L(k)|;

[0099] Δ = max{Δλ(k),ΔL(k)}<ε;

[0100] Among them, ε is a constant, and those skilled in the art select an appropriately small value as the value of this constant according to experience. This embodiment does not limit this too much.

[0101] In addition, in order to prove the correlation degree of the data to be identified, this embodiment constructs the driver characteristics into a linear relationship in the form of Y = aX + b, and calculates the correlation coefficient of this linear relationship, which is expressed as:

[0102]

[0103] Among them, X i represents the i-th independent variable in the linear relationship; Y i represents the i-th dependent variable in the linear relationship; N represents the total sample number, and N is the number of independent variables; if r is closer to 1, it proves that its linear correlation is better. In this embodiment, by selecting. When selecting the data for the best parameters for identifying driver characteristics, this application selects the parameter with the correlation coefficient closest to 1 from the existing parameters as the parameter for identifying driver parameters through the above parameters. The parameter selected in this embodiment is the following vehicle following speed v f of the linear relationship with the following distance, and two driver characteristic parameters are learned from it.

[0104] After obtaining the current driver parameters, the current identified driver parameters can be directly used for following, or the optimal driver parameters can be statistically calculated for the N driver parameters closest to the current moment. It can be the average value of the historical N driver parameters or the weighted average value. If the weighted average value is used, the driver parameters closer to the current moment have higher weights; as a preferred implementation manner, the similarity between the current identified driver parameters and the historical driver parameters can be calculated by means of cosine distance and other methods. If the similarity between the current driver parameters and the historical average value is within the set range, the statistical calculation value of the current identified driver parameters and the historical parameters is used as the value of the current driver characteristic parameters for following, otherwise the currently identified driver parameters are used for following.

[0105] According to the longitudinal following dynamics state equation of the driver and the identification algorithm based on the following behavior characteristics of the driver, the identified driver characteristic parameters λ and L are used to determine the ideal following distance s of different types of drivers d . This article uses the sliding mode variable structure control algorithm to approximate the ideal following distance. At the same time, aiming at the chattering problem of the variable structure control input quantity, the saturation function is used to replace the sign function, and the boundary layer is compensated by fuzzy inference to eliminate the chattering to a certain extent, so as to establish a longitudinal driver following behavior model based on the identification of driver characteristic parameters.

[0106] The vehicle dynamics model during the driver's longitudinal car - following process can be written as:

[0107]

[0108] Among them, f(X) and g(X) are the nominal functions of the system state respectively, u(t) and y represent the control input and the system output of the system. Assume that the state variables of the system are observable, that is, its state vector is where x f is the driving distance of the host vehicle, is the car - following speed of the host vehicle, is the car - following acceleration of the host vehicle, and d(t) is the disturbance of the system.

[0109] The vehicle dynamics model during the driver's longitudinal car - following process is a third - order non - linear system. There is an error between the actual car - following distance and the ideal car - following distance s d The car - following error is described as follows:

[0110]

[0111] Among them, represents the vector of the car - following error and its derivative, and x l represents the driving distance of the leading vehicle, s 0 is the initial distance between the two vehicles; represents the speed of the leading vehicle; correspondingly, represents the acceleration of the leading vehicle; represents the change rate of the car - following acceleration of the trailing vehicle.

[0112] Substitute the controllable quantity u into the vehicle longitudinal dynamics state equation, that is:

[0113]

[0114] A ρ = c w Aρ / 2;

[0115] Among them, m is the vehicle mass, g is the acceleration due to gravity, f R is the rolling resistance coefficient, α is the road surface gradient, c w is the air resistance coefficient, A is the vehicle frontal area, ρ is the air density; f(X) and g(X) are the nominal functions of the system state respectively, u(t) and y represent the control input and the system output of the system, and u(t) is the control generated according to the driver characteristic parameters; The state variables of the system, x f is the driving distance of the host vehicle, is the car - following speed of the host vehicle, is the car - following acceleration of the host vehicle, and d(t) is the disturbance of the system; Denote the vector regarding the following - distance error and its derivative as \(x\). l Denote the driving distance of the leading vehicle as \(s\). 0 \(d_0\) is the initial distance between the two vehicles; Denote the speed of the leading vehicle as \(v\); Denote the acceleration of the leading vehicle as \(a\); Denote the change rate of the following - vehicle acceleration. \(\mu\) represents the lag - time constant of the engine or braking system, \(M\) represents the vehicle's own weight, \(F\); R \(F_r\) is the vehicle rolling resistance, \(L\) is the stopping distance obtained from the driver - characteristic identification function, \(\lambda\) is the time - headway obtained from the driver - characteristic identification function, \(c_1\); 1 \(c_2\); 2 \(c_1\) and \(c_2\) are characteristic parameters satisfying the Hurwitz stability criterion; \(\dddot{\dot{v}}\) is the fourth - order derivative of the following - vehicle speed; Denote the estimated value of the model - switching gain. \(sat(S(X))\) is the saturation - function form of \(S(X)\), and \(S(X)\) is the sliding - mode function, expressed as:

[0116]

[0117] where \(\Delta=\max\{\Delta\lambda(k),\Delta L(k)\}<\varepsilon\).

[0118] The system output is \(y = x\); f which represents the following - driving distance; \(\dot{y}\) is the following - vehicle speed of the own vehicle; \(\ddot{y}\) is the following - vehicle acceleration of the own vehicle. All of these can be obtained from the system output. Furthermore, the output of the following - vehicle model can be the kinematic physical quantities such as the following - driving distance, following - vehicle speed, and following - vehicle acceleration. At the same time, according to the motion state of the leading vehicle, derived quantities such as the following - distance, relative speed, and acceleration difference can be obtained. These physical quantities describe the driver's following - vehicle behavior from the perspective of describing the vehicle's motion state, enabling the simulation of the driver's longitudinal following - vehicle behavior model to be realized.

[0119] In the specific implementation process, the vehicle operation process is as Figure 3 shown, and specifically includes the following steps:

[0120] System initialization;

[0121] Judge whether it is in the following - vehicle state at the current time. If it is in the following - vehicle state, then judge whether the driver is in the driving state. If in the driving state, collect the data of the following - vehicle state, and judge whether the accuracy of the collected parameters meets the requirements, that is, judge whether \(\Delta=\max\{\Delta\lambda(k),\Delta L(k)\}<\varepsilon\). If it meets the requirements, save the parameters, statistically distribute the parameters according to the history, and re - extract the parameters;

[0122] If the vehicle is in a following state but the driver is not in a driving state, the saved parameters are called and used for following control;

[0123] The above-mentioned parameters are driver characteristic parameters.

[0124] Such as Figure 4 , in the process of obtaining the driver characteristic parameters, first obtain the following speed and following distance according to the driving behavior of the driver at the position. During the driver parameter identification process, construct a relevant expression, that is, S d =λv f +L. According to this expression, obtain the driver characteristic parameters λ and L. In the process of calculating the correlation between the two parameters, also construct the driver parameter in the form of S d =λv f +L. The correlation between S d and v f in the linear relationship is obtained. That is, in this embodiment, in the linear relationship Y = aX + b, Y represents the value of S d , and X represents the value of v f .

[0125] Such as Figures 5 to 8 The error comparison between this embodiment and the FVD scheme and the safety index comparison show that it can be seen from the figure that the adaptive car-following model proposed by the present invention that comprehensively considers the perception process and the driver's behavior shows good performance in describing the individual driver's car-following behavior. The comparison with the prediction results of the classical FVD model shows that the absolute value of the average error calculated by the model is less than 80%, and the coincidence degree with the actual car-following behavior of the driver is higher.

[0126] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirits of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive driver following method that comprehensively considers the perception process and driver behavior, characterized in that, it includes the following steps: Obtain the motion trajectory data of the current vehicle and the motion trajectory data of the vehicle following the current vehicle; Based on the external perception disturbance, estimate the speed of the following vehicle and the distance between the current vehicle and the following vehicle according to the historical motion trajectory data of the current driver. Estimating the distance between the current vehicle and the following vehicle includes: lns est -lns = V s w s (t); Among them, s est is the estimated value of the distance between the current vehicle and the following vehicle; s is the actual value of the distance between the current vehicle and the following vehicle collected by the vehicle-mounted sensing device; V s is the estimated value of the vehicle distance s est the standard relative error of the logarithm of the actual value s; w s (t) is the error of the estimated value of the distance between the current vehicle and the following vehicle; Identify the driver characteristic parameters from the distance between the current vehicle and the following vehicle and the driving speed of the current vehicle, and obtain the optimal driver characteristic parameters according to the statistical history driver characteristic parameters; Establish a following behavior model based on the perception process and driver behavior characteristics, take the driver characteristic parameters and the speed of the current vehicle as inputs, and calculate the discrete value of the following distance.

2. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 1, characterized in that, the process of estimating the speed of the following vehicle includes: Among them, is the estimated value of the speed of the following vehicle; v l is the actual speed provided by the on-vehicle device; s is the actual value of the distance between the current vehicle and the following vehicle collected by the vehicle-mounted perception device; σ r is the standard deviation of the perspective change rate; w l (t) is the distribution of the estimated error of the leading vehicle's speed and its change over time.

3. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 2, characterized in that, the perspective change rate is expressed as: τ TTC = s / Δv; where τ TTC is the collision time; Δv is the difference between the speed of the host vehicle and the speed of the leading vehicle.

4. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 1 or 2, characterized in that, Construct a stochastic differential equation, select a group of white noise initialized with pseudo-random numbers and substitute it into the constructed equation, and the obtained value is used as the error of the estimated value of the speed of the following vehicle or the error of the estimated value of the distance between the current vehicle and the following vehicle. The solution of the stochastic differential equation is expressed as: where, w(t) represents the solution of a stochastic differential equation; a 0 , a 1 , a are intermediate parameters, expressed as t 0 represents the time at a certain moment; t represents the car-following behavior time; ξ(s) represents the standard white noise, and s represents the actual value of the distance between the current vehicle and the following vehicle collected by the vehicle-mounted sensing device; τ represents the driver's estimation error time.

5. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 4, characterized in that, the standard white noise ξ(s) is expressed as: where w represents w(t), which represents the solution of a stochastic differential equation.

6. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 1, characterized in that, Based on the speed v of the host vehicle f (t) identify the driver characteristic parameters, specifically including the following steps: Among them, is a parameter vector, which is composed of the driver's time-distance discrete value parameter λ(k) and the discrete value L(k) of the driver's stopping distance, and is expressed as and is a process matrix; is an input matrix constructed according to the ego-vehicle speed v f (t), and is expressed as I is an identity matrix.

7. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 6, characterized in that, In the N -th iteration process, select those that meet the stability criteria as the candidate set of driver parameters. The stability criteria are: Δλ(k) = |(λ(k) - λ(k - 1)) / λ(k)|; ΔL(k) = |(L(k) - L(k - 1)) / L(k)|; Δ = max{Δλ(k), ΔL(k)} < ε; where ε is a constant.

8. The adaptive driver following method that comprehensively considers the perception process and driver behavior according to claim 6, characterized in that, Establish a following behavior model based on the perception process and driver behavior characteristics, take the driver characteristic parameters and the speed of the current vehicle as inputs, and calculate the discrete value of the following distance: The vehicle dynamics model during the driver's longitudinal following process can be written as: The vehicle dynamics model during the driver's longitudinal car-following process is a third-order nonlinear system. There is an error between the actual car-following distance and the ideal car-following distance s d The car-following error is described as follows: Substitute the controllable quantity u(t) into the vehicle longitudinal dynamics state equation to obtain: Among them, f(X) and g(X) are respectively nominal functions regarding the system state, u(t) and y represent the control input of the system and the system output respectively, and u(t) is the control generated according to the driver characteristic parameters; The state variable of the system, x f Is the driving distance of the host vehicle, Is the following - vehicle speed of the host vehicle, Is the following - vehicle acceleration of the host vehicle, and d(t) is the disturbance of the system; Represents a vector regarding the following - vehicle error and its derivative, x l Represents the driving distance of the leading vehicle, s 0 Is the initial distance between the two vehicles; Represents the speed of the leading vehicle; Represents the acceleration of the leading vehicle; Represents the change rate of the following - vehicle acceleration of the rear vehicle; μ represents the lag time constant of the engine or brake system, M is the vehicle's own weight, F R Is the vehicle rolling resistance, L is the stopping distance obtained from the driver characteristic identification function, λ is the time gap obtained from the driver characteristic identification function, c 1 、c 2 Are characteristic parameters that satisfy the Hurwitz stability criterion, Is the fourth - order derivative of the following - vehicle speed of the host vehicle, Represents the estimated value of the model switching gain, sat(S(X)) is the saturation function form of S(X), and S(X) is the sliding - mode function.

Images