阅读说明：本技术 一种基于视觉感知的安全超车预警系统及方法 (Safety overtaking early warning system and method based on visual perception ) 是由 唐阳山 曹玉珠 魏丹 徐兆华 白艳 王琪 刘哲 于 2019-10-12 设计创作，主要内容包括：本发明公开了一种基于视觉感知的安全超车预警系统,包括：后向检测装置,其可拆卸设置在汽车后视镜上,能够获取本车道和邻车道的后车行驶图像；距离传感器,其设置在汽车前方和后侧,能够检测本车道和邻车道的前车和后车车距；车速传感器,能够检测位于检测本车道和邻车道的前车和后车行驶速度；计算处理单元,其连接所述后向检测装置、所述距离传感器和所述车速传感器,能够根据所述后向检测装置、所述距离传感器和所述车速传感器检测到的数据计算得到超车预警信息；超车预警系统控制器,其连接所述计算处理单元,能够根据所述计算处理单元计算得到的超车预警信息控制汽车行驶状态,本发明还公开了一种基于视觉感知的安全超车预警方法。(The invention discloses a safety overtaking early warning system based on visual perception, which comprises: the backward detection device is detachably arranged on the automobile rearview mirror and can acquire the rear automobile running images of the lane and the adjacent lane; distance sensors, which are arranged in front of and behind the automobile and can detect the distance between the front automobile and the rear automobile in the lane and the adjacent lane; the vehicle speed sensor can detect the running speeds of a front vehicle and a rear vehicle positioned on the detected lane and an adjacent lane; the calculation processing unit is connected with the backward detection device, the distance sensor and the vehicle speed sensor and can calculate and obtain overtaking early warning information according to data detected by the backward detection device, the distance sensor and the vehicle speed sensor; the overtaking early warning system controller is connected with the calculation processing unit and can control the running state of the automobile according to the overtaking early warning information calculated by the calculation processing unit.)

1. A safety overtaking early warning system based on visual perception is characterized by comprising:

the backward detection device is detachably arranged on the automobile rearview mirror and can acquire the rear automobile running images of the lane and the adjacent lane;

distance sensors, which are arranged in front of and behind the automobile and can detect the distance between the front automobile and the rear automobile in the lane and the adjacent lane;

vehicle speed sensors, which are provided in front of and behind the vehicle and are capable of detecting the traveling speeds of the front vehicle and the rear vehicle in the detection own lane and the adjacent lane;

the calculation processing unit is connected with the backward detection device, the distance sensor and the vehicle speed sensor and can calculate and obtain overtaking early warning information according to data detected by the backward detection device, the distance sensor and the vehicle speed sensor;

and the overtaking early warning system controller is connected with the calculation processing unit and can control the running state of the automobile according to the overtaking early warning information calculated by the calculation processing unit.

2. A safety overtaking early warning method based on visual perception is characterized by comprising the following steps:

the method comprises the following steps of firstly, sequentially carrying out pixel point compression, color image graying, gray level stretching, filtering and binarization on an acquired vehicle driving image to obtain a preprocessed vehicle driving image containing a rear vehicle image;

performing marginalization processing on the preprocessed vehicle driving image containing the vehicle image, positioning a rear vehicle image in the vehicle driving image, and calculating the area ratio of the rear vehicle image to the vehicle driving image;

thirdly, detecting the rear vehicle running speed of the lane and the rear vehicle running speed of the adjacent lane by a vehicle speed sensor, and calculating the vehicle distance between the rear vehicle and the vehicle in the vehicle running image according to the area ratio change rate of the rear vehicle image and the vehicle running image in the t time interval;

step four, calculating the collision probability of the front vehicles according to the running speed of the front vehicles in the lane, the running speed of the front vehicles in the adjacent lane, the distance between the front vehicles in the lane and the distance between the front vehicles in the adjacent lane;

step five, calculating the rear vehicle collision probability according to the rear vehicle running speed of the lane, the rear vehicle running speed of the adjacent lane, the rear vehicle distance of the lane and the rear vehicle distance of the adjacent lane;

inputting the collision probability of the front vehicle, the collision probability of the rear vehicle, the running speed of the vehicle and the running acceleration of the vehicle into a neural network controller to obtain a vector group representing the control of overtaking running parameters;

and outputting the vector group controlled by the overtaking driving parameters as a plagiarism and control strategy.

3. The safety overtaking early warning method based on visual perception as claimed in claim 2, wherein the image compression ratio of the pixel point compression process in the step one is:

wherein R is_imgThe image compression rate is delta, the threshold pixel height is delta, and the pixel height of an automobile image sample containing a license plate is h;

the image binarization process comprises the following steps:

and obtaining the optimal threshold value of binarization by adopting the following formula:

wherein, V_maxIs the maximum value of gray value of image pixel point, V_minIs the minimum value of gray value of image pixel point, V_bestIs the best threshold value for binarization.

4. The safety overtaking early warning method based on visual perception as claimed in claim 3, wherein the calculation formula of the distance between the rear vehicle and the vehicle in the third step is as follows:

wherein S is_iThe distance between the rear vehicle and the vehicle, S₀Is a standard vehicle distance, delta_iThe area ratio change rate, delta, of the rear vehicle image in the driving image₀Is the area-to-area average rate of change, v_iIs the rear vehicle running speed v₀The vehicle running speed.

5. The safety overtaking warning method based on visual perception as claimed in claim 4, wherein the formula for calculating the collision probability of the preceding vehicle is as follows:

wherein S is_qThe distance between the front vehicle and the current vehicle, S₀Is a standard vehicle distance, v_qThe running speed v of the front vehicle₀The running speed of the vehicle;

the rear vehicle collision probability calculation formula is as follows:

wherein, P_iAs rear vehicle collision probability, S_iThe distance between the rear vehicle and the vehicle, S₀Is a standard vehicle distance, v_iIs the rear vehicle running speed v₀The vehicle running speed.

6. The safety overtaking warning method based on visual perception as claimed in claim 4, wherein the calculation process of the neural network controller in the sixth step is as follows:

step 1, sequentially normalizing parameters including a front vehicle collision probability, a rear vehicle collision probability, a vehicle running speed and a vehicle running acceleration;

step 2, determining an input layer neuron vector x ═ x of the three-layer BP neural network₁,x₂,x₃,x₄In which x₁Is the front vehicle collision probability coefficient, x₂Is the rear vehicle collision probability coefficient, x₃Is the running speed index, x, of the vehicle₄The vehicle running acceleration coefficient;

and 3, mapping the input layer vector to a hidden layer, wherein the hidden layer vector y is { y ═ y₁,y₂,…,y_mM is the number of hidden nodes;

and 4, obtaining an output layer neuron vector o ═ o₁,o₂,o₃,o₄}; wherein o is₁For the next period of speed control coefficient, o₂For the next period of acceleration control coefficient, o₃For the next time period angle control coefficient, o₄The emergency braking coefficient of the next time period;

step 5, the controller outputs a control strategy relational expression of the next period;

wherein the number m of hidden layer nodes satisfies:

7. The safety overtaking warning method based on visual perception as claimed in claim 6, wherein the probability of collision of the preceding vehicle is Y_tThe rear vehicle collision probability Y_pThe running speed v of the vehicle_iAnd the running acceleration a of the vehicle_iThe normalization formula of (a) is:

wherein x is_jFor parameters in the input layer vector, X_jRespectively is a measurement parameter Y_t、Y_p、v_i、a_i，j＝1,2,3,4；X_jmaxAnd X_jminRespectively, a maximum value and a minimum value in the corresponding measured parameter.

8. The safety overtaking warning method based on visual perception as claimed in claim 5, wherein the controller outputs the vehicle control strategy relation for the next time period as:

wherein v is_i+1Is t_i～(t_i+t₀) Driving speed of the vehicle a_i+1Is t_i～(t_i+t₀) Acceleration of travel of the vehicle, beta_i+1Is t_i～(t_i+t₀) Time interval of vehicle turning angle, P_i+1Is t_i～(t_i+t₀) Braking commands for the vehicle are timed.

9. The safety overtaking warning method based on visual perception as claimed in claim 8 wherein the output layer o₄Has a neuron value of 0 or 1 when o₄At 1, when the vehicle is braking, when o₄When 0, no braking is performed.

10. The safety overtaking early warning method based on visual perception as claimed in claim 6, wherein the excitation functions of the hidden layer and the output layer both adopt S-shaped functions f_j(x)＝1/(1+e^-x)。

Technical Field

The invention relates to the field of safe driving of automobiles, in particular to a safe overtaking early warning system based on visual perception and a safe overtaking early warning method based on visual perception.

Background

With the rapid development of computer technology and artificial intelligence technology, intelligent robot technology has become a hot spot for research of numerous scholars at home and abroad. Among them, the service robot opens up a new field of robot application, and the human wants to get rid of the repetitive work. The current express mail needs manual dispatch, the manual demand is large, the efficiency is low, and the manual dispatch is replaced by intelligent unmanned automatic dispatch. The research of automatic driving becomes a big thing for more convenient differentiation and definition of automatic driving technology.

Disclosure of Invention

The invention designs and develops a safety overtaking early warning system based on visual perception, a backward detection device is arranged on an automobile rearview mirror, and rear automobile running images of a current lane and an adjacent lane can be obtained.

The invention also designs and develops a safe overtaking early warning method based on visual perception, which can calculate the collision probability of the front vehicle and the collision probability of the rear vehicle according to the vehicle running information, and the collision probability of the front vehicle and the collision probability of the rear vehicle as well as the running speed and the running acceleration of the vehicle are input into a vehicle steering control strategy obtained by a neural network controller to realize automatic safe driving.

The technical scheme provided by the invention is as follows:

a safety overtaking early warning system based on visual perception comprises:

the backward detection device is detachably arranged on the automobile rearview mirror and can acquire the rear automobile running images of the lane and the adjacent lane;

distance sensors, which are arranged in front of and behind the automobile and can detect the distance between the front automobile and the rear automobile in the lane and the adjacent lane;

vehicle speed sensors, which are provided in front of and behind the vehicle and are capable of detecting the traveling speeds of the front vehicle and the rear vehicle in the detection own lane and the adjacent lane;

the calculation processing unit is connected with the backward detection device, the distance sensor and the vehicle speed sensor and can calculate and obtain overtaking early warning information according to data detected by the backward detection device, the distance sensor and the vehicle speed sensor;

and the overtaking early warning system controller is connected with the calculation processing unit and can control the running state of the automobile according to the overtaking early warning information calculated by the calculation processing unit.

A safety overtaking early warning method based on visual perception comprises the following steps:

the method comprises the following steps of firstly, sequentially carrying out pixel point compression, color image graying, gray level stretching, filtering and binarization on an acquired vehicle driving image to obtain a preprocessed vehicle driving image containing a rear vehicle image;

performing marginalization processing on the preprocessed vehicle driving image containing the vehicle image, positioning a rear vehicle image in the vehicle driving image, and calculating the area ratio of the rear vehicle image to the vehicle driving image;

thirdly, detecting the rear vehicle running speed of the lane and the rear vehicle running speed of the adjacent lane by a vehicle speed sensor, and calculating the vehicle distance between the rear vehicle and the vehicle in the vehicle running image according to the area ratio change rate of the rear vehicle image and the vehicle running image in the t time interval;

step four, calculating the collision probability of the front vehicles according to the running speed of the front vehicles in the lane, the running speed of the front vehicles in the adjacent lane, the distance between the front vehicles in the lane and the distance between the front vehicles in the adjacent lane;

step five, calculating the rear vehicle collision probability according to the rear vehicle running speed of the lane, the rear vehicle running speed of the adjacent lane, the rear vehicle distance of the lane and the rear vehicle distance of the adjacent lane;

inputting the collision probability of the front vehicle, the collision probability of the rear vehicle, the running speed of the vehicle and the running acceleration of the vehicle into a neural network controller to obtain a vector group representing the control of overtaking running parameters;

and outputting the vector group controlled by the overtaking driving parameters as a plagiarism and control strategy.

Preferably, the image compression ratio of the pixel point compression process in the step one is as follows:

wherein R is_imgThe image compression rate is delta, the threshold pixel height is delta, and the pixel height of an automobile image sample containing a license plate is h;

the image binarization process comprises the following steps:

and obtaining the optimal threshold value of binarization by adopting the following formula:

wherein, V_maxIs the maximum value of gray value of image pixel point, V_minIs the minimum value of gray value of image pixel point, V_bestIs the best threshold value for binarization.

Preferably, the calculation formula of the distance between the rear vehicle and the vehicle in the third step is as follows:

wherein S is_iThe distance between the rear vehicle and the vehicle, S₀Is a standard vehicle distance, delta_iThe area ratio change rate, delta, of the rear vehicle image in the driving image₀Is the area-to-area average rate of change, v_iIs the rear vehicle running speed v₀The vehicle running speed.

Preferably, the formula for calculating the collision probability of the front vehicle is as follows:

wherein S is_qThe distance between the front vehicle and the current vehicle, S₀Is a standard vehicle distance, v_qThe running speed v of the front vehicle₀The running speed of the vehicle;

the rear vehicle collision probability calculation formula is as follows:

wherein, P_iAs rear vehicle collision probability, S_iThe distance between the rear vehicle and the vehicle, S₀Is a standard vehicle distance, v_iIs the rear vehicle running speed v₀The vehicle running speed.

Preferably, the calculation process of the neural network controller in the step six is as follows:

step 1, sequentially normalizing parameters including a front vehicle collision probability, a rear vehicle collision probability, a vehicle running speed and a vehicle running acceleration;

step 2, determining an input layer neuron vector x ═ x of the three-layer BP neural network₁,x₂,x₃,x₄In which x₁Is the front vehicle collision probability coefficient, x₂Is the rear vehicle collision probability coefficient, x₃Is the running speed index, x, of the vehicle₄The vehicle running acceleration coefficient;

and 3, mapping the input layer vector to a hidden layer, wherein the hidden layer vector y is { y ═ y₁,y₂,…,y_mM is the number of hidden nodes;

and 4, obtaining an output layer neuron vector o ═ o₁,o₂,o₃,o₄}; wherein o is₁For the next period of speed control coefficient, o₂For the next period of acceleration control coefficient, o₃For the next time period angle control coefficient, o₄The emergency braking coefficient of the next time period;

step 5, the controller outputs a control strategy relational expression of the next period;

wherein the number m of hidden layer nodes satisfies:

wherein n is the number of nodes of the input layer, and p is the number of nodes of the output layer.

Preferably, the front vehicle collision probability Y_tThe rear vehicle collision probability Y_pThe running speed v of the vehicle_iAnd the running acceleration a of the vehicle_iThe normalization formula of (a) is:

wherein x is_jFor parameters in the input layer vector, X_jRespectively is a measurement parameter Y_t、Y_p、v_i、a_i，j＝1,2,3,4；X_jmaxAnd X_jminRespectively, a maximum value and a minimum value in the corresponding measured parameter.

Preferably, the controller outputs the vehicle control strategy relation for the next time period as:

wherein v is_i+1Is t_i～(t_i+t₀) Driving speed of the vehicle a_i+1Is t_i～(t_i+t₀) Acceleration of travel of the vehicle, beta_i+1Is t_i～(t_i+t₀) Time interval of vehicle turning angle, P_i+1Is t_i～(t_i+t₀) Braking commands for the vehicle are timed.

Preferably, the output layer o₄Has a neuron value of 0 or 1 when o₄At 1, when the vehicle is braking, when o₄When 0, no braking is performed.

Preferably, the excitation functions of the hidden layer and the output layer both adopt S-shaped functions f_j(x)＝1/(1+e^-x)。

The invention has the advantages of

The invention designs and develops a safety overtaking early warning system based on visual perception, a backward detection device is arranged on an automobile rearview mirror, and rear automobile running images of a current lane and an adjacent lane can be obtained.

The invention also designs and develops a safe overtaking early warning method based on visual perception, which can calculate the collision probability of the front vehicle and the collision probability of the rear vehicle according to the vehicle running information, and the collision probability of the front vehicle and the collision probability of the rear vehicle as well as the running speed and the running acceleration of the vehicle are input into a vehicle steering control strategy obtained by a neural network controller to realize automatic safe driving.

Detailed Description

The present invention is described in further detail below to enable those skilled in the art to practice the invention with reference to the description.

The invention provides a safety overtaking early warning system based on visual perception, which comprises: the system comprises a backward detection device, a distance sensor, a vehicle speed sensor, a calculation unit and an overtaking early warning system controller.

The backward detection device is detachably arranged on the automobile rearview mirror and can acquire the rear automobile running images of the lane and the adjacent lane; the distance sensors are arranged in front of and behind the automobile and can detect the distances between the front automobile and the rear automobile in the lane and the adjacent lane; the vehicle speed sensors are arranged in front of and behind the vehicle and can detect the running speeds of the front vehicle and the rear vehicle in the detected lane and the adjacent lane; the calculation processing unit is connected with the backward detection device, the distance sensor and the vehicle speed sensor and can calculate and obtain overtaking early warning information according to data detected by the backward detection device, the distance sensor and the vehicle speed sensor; the overtaking early warning system controller is connected with the calculation processing unit and can control the running state of the automobile according to the overtaking early warning information calculated by the calculation processing unit.

The invention also provides a safety overtaking early warning method based on visual perception, which comprises the following steps:

the method comprises the following steps of firstly, sequentially carrying out pixel point compression, color image graying, gray level stretching, filtering and binarization on an acquired vehicle driving image to obtain a preprocessed vehicle driving image containing a rear vehicle image;

the image compression ratio of the pixel point compression process is as follows:

wherein R is_imgThe image compression rate is delta, the threshold pixel height is delta, and the pixel height of an automobile image sample containing a license plate is h;

the image binarization process comprises the following steps:

and obtaining the optimal threshold value of binarization by adopting the following formula:

wherein, V_maxIs the maximum value of gray value of image pixel point, V_minIs the minimum value of gray value of image pixel point, V_bestThe threshold value is the best threshold value of binarization;

performing marginalization processing on the preprocessed vehicle driving image containing the vehicle image, positioning a rear vehicle image in the vehicle driving image, and calculating the area ratio of the rear vehicle image to the vehicle driving image;

adopting a rectangular template with the same size, performing expansion operation on a vehicle driving image containing a vehicle image, filling holes, communicating an automobile image area, then corroding by using the rectangular template with the same size, eliminating an isolated small area, reserving a large communicated area, namely positioning the vehicle area, and calculating the area ratio of the communicated area in the driving image;

thirdly, detecting the rear vehicle running speed of the lane and the rear vehicle running speed of the adjacent lane by a vehicle speed sensor, and calculating the vehicle distance between the rear vehicle and the vehicle in the vehicle running image according to the area ratio change rate of the rear vehicle image and the vehicle running image in the t time interval;

wherein S is_iThe distance between the rear vehicle and the vehicle, S₀Is a standard vehicle distance, delta_iThe area ratio change rate, delta, of the rear vehicle image in the driving image₀Is the area-to-area average rate of change, v_iIs the rear vehicle running speed v₀The vehicle running speed.

Step four, calculating the collision probability of the front vehicles according to the running speed of the front vehicles in the lane, the running speed of the front vehicles in the adjacent lane, the distance between the front vehicles in the lane and the distance between the front vehicles in the adjacent lane;

wherein, P_qAs the probability of collision of the preceding vehicle, S_qThe distance between the front vehicle and the current vehicle, S₀Is a standard vehicle distance, v_qThe running speed v of the front vehicle₀The vehicle running speed.

Step five, calculating the rear vehicle collision probability according to the rear vehicle running speed of the lane, the rear vehicle running speed of the adjacent lane, the rear vehicle distance of the lane and the rear vehicle distance of the adjacent lane;

wherein, P_iAs rear vehicle collision probability, S_iThe distance between the rear vehicle and the vehicle, S₀Is a standard vehicle distance, v_iFor rear vehicleSpeed of travel, v₀The vehicle running speed.

Step six, calculating the collision probability P of the front vehicle_qRear vehicle collision probability P_iThe running speed v of the vehicle₀And the running acceleration a of the vehicle₀Inputting the vector group into a neural network controller to obtain the vector group representing the control of the overtaking driving parameters;

and establishing a BP neural network model.

Fully interconnected connections are formed among neurons of each layer on the BP model, the neurons in each layer are not connected, and the output and the input of neurons in an input layer are the same, namely o_i＝x_i. The operating characteristics of the neurons of the intermediate hidden and output layers are

o_pj＝f_j(net_pj)

Where p represents the current input sample, ω_jiIs the connection weight from neuron i to neuron j, o_piIs the current input of neuron j, o_pjIs the output thereof; f. of_jIs a non-linear, slightly non-decreasing function, generally taken as a sigmoid function, i.e. f_j(x)＝1/(1+e^-x)。

The BP network system structure adopted by the invention consists of three layers, wherein the first layer is an input layer, n nodes are provided in total, n detection signals representing the working state of the equipment are correspondingly provided, and the signal parameters are given by a data preprocessing module; the second layer is a hidden layer with m nodes,

wherein m is the number of intermediate layer nodes, n is the number of input layer nodes, and p is the number of output layer nodes; the third layer is an output layer, p nodes are provided in total, and the output is determined by the response actually needed by the system.

The mathematical model of the network is:

inputting a layer vector: x ═ x₁,x₂,…,x_n)^T

Intermediate layer vector: y ═ y₁,y₂,…,y_m)^T

Outputting a layer vector: z is (z)₁,z₂,…,z_p)^T

According to the sampling period, the number of nodes of an input layer is n-4, the number of nodes of an output layer is p-4, and the number of nodes of a hidden layer is estimated according to the following formula:

the data acquired by the sensors belong to different physical quantities, and the dimensions of the data are different. Therefore, the data needs to be normalized to a number between 0-1 before it is input into the artificial neural network.

The normalized formula is

Wherein x is_jFor parameters in the input layer vector, X_jRespectively, is a measurement parameter, X_jAre respectively P_q、P_i、v₀And a₀，j＝1,2,3,4；X_jmaxAnd X_jminAnd respectively adopting S-shaped functions for the maximum value and the minimum value in the corresponding measurement parameters.

Specifically, with respect to the front vehicle collision probability P_qNormalized to obtain the front vehicle collision probability coefficient x₁：

Wherein, max (P)_q) And min (P)_q) Respectively the maximum value and the minimum value of the collision probability of the front vehicle.

Likewise, rear vehicle collision probability P_iNormalizing by the following formula to obtain a rear vehicle collision probability coefficient x₂：

Wherein, max (P)_i) And min (P)_i) Respectively the maximum value and the minimum value of the collision probability of the rear vehicle.

Vehicle running speed v₀After normalization, the driving speed coefficient x of the vehicle is obtained₃：

Wherein, max (v)₀) And min (v)₀) Respectively the maximum value and the minimum value of the running speed of the vehicle.

Running acceleration a of the vehicle₀After normalization, the driving acceleration coefficient x of the vehicle is obtained₄：

Therein, max (a)₀) And min (a)₀) Respectively, the maximum acceleration and the minimum acceleration of the running of the vehicle.

Obtaining output layer neuron vector o ═ o₁,o₂,o₃,o₄}; wherein o is₁For the next period of speed control coefficient, o₂For the next period of acceleration control coefficient, o₃For the next time period angle control coefficient, o₄The emergency braking coefficient of the next period.

And (5) training the BP neural network.

After the BP neural network node model is established, the training of the BP neural network can be carried out. Obtaining training samples according to empirical data of the product, and giving a connection weight w between an input node i and a hidden layer node j_ijConnection weight w between hidden layer node j and output layer node k_jkThreshold value theta of hidden layer node j_jThreshold value w of node k of output layer_ij、w_jk、θ_j、θ_kAre all random numbers between-1 and 1.

(1) Training method

Each subnet adopts a separate training method; when training, firstly providing a group of training samples, wherein each sample consists of an input sample and an ideal output pair, and when all actual outputs of the network are consistent with the ideal outputs of the network, the training is finished; otherwise, the ideal output of the network is consistent with the actual output by correcting the weight; the output samples for each subnet training are shown in table 1.

TABLE 1 output samples for network training

(2) Training algorithm

The BP network is trained by using a back Propagation (Backward Propagation) algorithm, and the steps can be summarized as follows:

the first step is as follows: and selecting a network with a reasonable structure, and setting initial values of all node thresholds and connection weights.

The second step is that: for each input sample, the following calculations are made:

(a) forward calculation: for j unit of l layer

In the formula (I), the compound is shown in the specification,

for the weighted sum of the j unit information of the l layer at the nth calculation,is the connection weight between the j cell of the l layer and the cell i of the previous layer (i.e. the l-1 layer),

is the previous layer (i.e. l-1 layer, node number n)_l-1) The operating signal sent by the unit i; when i is 0, order

Is the threshold of the j cell of the l layer.

If the activation function of the unit j is a sigmoid function, then

And is

If neuron j belongs to the first hidden layer (l ═ 1), then there are

If neuron j belongs to the output layer (L ═ L), then there are

(b) And (3) calculating the error reversely:

for output unit

Pair hidden unit

(c) Correcting the weight value:

η is the learning rate.

The third step: inputting a new sample or a new period sample until the network converges, and randomly re-ordering the input sequence of the samples in each period during training.

The BP algorithm adopts a gradient descent method to solve the extreme value of a nonlinear function, and has the problems of local minimum, low convergence speed and the like. A more effective algorithm is a Levenberg-Marquardt optimization algorithm, which enables the network learning time to be shorter and can effectively inhibit the network from being locally minimum. The weight adjustment rate is selected as

Δω＝(J^TJ+μI)^-1J^Te

Wherein J is a Jacobian (Jacobian) matrix of error to weight differentiation, I is an input vector, e is an error vector, and the variable mu is a scalar quantity which is self-adaptive and adjusted and is used for determining whether the learning is finished according to a Newton method or a gradient method.

When the system is designed, the system model is a network which is only initialized, the weight needs to be learned and adjusted according to data samples obtained in the using process, and therefore the self-learning function of the system is designed. Under the condition of appointing learning samples and quantity, the system can carry out self-learning so as to continuously improve the network performance.

The vehicle control strategy has the relational expression:

wherein v is_i+1Is t_i～(t_i+t₀) Driving speed of the vehicle a_i+1Is t_i～(t_i+t₀) Acceleration of travel of the vehicle, beta_i+1Is t_i～(t_i+t₀) Time interval of vehicle turning angle, P_i+1Is t_i～(t_i+t₀) Braking commands for the vehicle are timed.

Output layer o₄Has a neuron value of 0 or 1 when o₄At 1, when the vehicle is braking, when o₄When 0, no braking is performed.

And the vector group controlled by the overtaking driving parameters is used as a plagiarism and control strategy to output beneficial effects

While embodiments of the invention have been described above, it is not limited to the applications set forth in the description and the embodiments, which are fully applicable to various fields of endeavor for which the invention may be embodied with additional modifications as would be readily apparent to those skilled in the art, and the invention is therefore not limited to the details given herein and to the embodiments shown and described without departing from the generic concept as defined by the claims and their equivalents.