CN112874456B - Intelligent vehicle adjusting method and system - Google Patents
Intelligent vehicle adjusting method and system Download PDFInfo
- Publication number
- CN112874456B CN112874456B CN202110036153.1A CN202110036153A CN112874456B CN 112874456 B CN112874456 B CN 112874456B CN 202110036153 A CN202110036153 A CN 202110036153A CN 112874456 B CN112874456 B CN 112874456B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- matrix
- characteristic
- driver
- adjustment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R16/00—Electric or fluid circuits specially adapted for vehicles and not otherwise provided for; Arrangement of elements of electric or fluid circuits specially adapted for vehicles and not otherwise provided for
- B60R16/02—Electric or fluid circuits specially adapted for vehicles and not otherwise provided for; Arrangement of elements of electric or fluid circuits specially adapted for vehicles and not otherwise provided for electric constitutive elements
- B60R16/037—Electric or fluid circuits specially adapted for vehicles and not otherwise provided for; Arrangement of elements of electric or fluid circuits specially adapted for vehicles and not otherwise provided for electric constitutive elements for occupant comfort, e.g. for automatic adjustment of appliances according to personal settings, e.g. seats, mirrors, steering wheel
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention provides an intelligent automobile adjusting method, which comprises the following steps: s1, collecting human body characteristic information, comparing the information, and realizing unlocking and starting functions: comparing the acquired fingerprints, and if the fingerprints pass the identification, unlocking; s2, collecting real-time pressure when the fingerprint is pressed, and starting the automobile by one key on the basis of fingerprint unlocking when the real-time pressure exceeds a preset pressure threshold; s3, primarily adjusting the in-vehicle facilities on the basis of fingerprint identification and one-key starting of the vehicle; s4, intelligent regulation of the vehicle is realized by continuously learning and comparing by adopting a machine learning method; and S5, realizing further adjustment of the in-vehicle facilities according to the result of the machine learning component. The method can establish a characteristic model through real-time learning based on habits of a driver, continuously learn and correct, achieve the best driving effect, intelligently adjust driving parameters according to weather, and guarantee the driving safety of the driver.
Description
Technical Field
The invention relates to the technical field of automobiles, in particular to an intelligent vehicle adjusting method and system.
Background
At present, the development of automobile technology is gradually mature from a starting stage, and along with the development of the automobile technology in China, consumers put higher requirements on the safety and the comfort of automobiles. With the progress of network technology and the development of artificial intelligence technology, the development of intelligent automobiles is confronted with new opportunities and challenges. An important link in the driving process of a vehicle is vehicle adjustment, which generally comprises an engine starting function, windshield defogging, windshield wipers, seat position, seat ventilation and heating, rearview air conditioning adjustment and the like. Traditional vehicle regulation mainly relies on driver's manual operation, changes driver every time and all needs readjustment, has brought unnecessary trouble for the driving, has also increased some potential safety hazards.
The existing intelligent vehicle adjusting method is that certain parameters of a vehicle, such as a seat position, a steering wheel position, a rearview mirror position and the like, are reset through the identification and comparison of a driver according to data originally set by a person; detailed parameters of the vehicle cannot be intelligently adjusted according to changes in the driver's state, such as whether a lot of sweating is occurring, whether weight changes occur, whether changes occur due to seasonal changes in clothes addition, and the like. Meanwhile, the driver changes, and the mainstream gearbox is familiar with the driving habit to adapt to the change, such as the response speed of an accelerator; at the moment, the driver is replaced again, the gearbox needs a certain time to adapt to the change, great discomfort is brought to the driver, and certain potential safety hazards are brought to the driving.
Disclosure of Invention
In order to solve the above-mentioned deficiencies of the prior art, an object of the present invention is to provide an intelligent vehicle adjusting method and system, which can adjust some parameters of an automobile on the basis of unlocking the automobile according to the comparison between the collected human body characteristic information and the stored information, and can establish a characteristic model based on the habit of the driver through real-time learning, continuously learn and correct to achieve the best driving effect and intelligently adjust the driving parameters according to the weather, thereby ensuring the driving safety of the driver.
The invention provides an intelligent vehicle adjusting method, which comprises the following steps:
s1, collecting human body characteristic information, comparing the information, and realizing unlocking and starting functions: comparing the acquired fingerprints, identifying whether the acquired human body characteristic information is a stored driver, unlocking if the acquired human body characteristic information is identified, and giving an alarm if the acquired human body characteristic information is not identified;
s2, collecting real-time pressure when the fingerprint is pressed, and starting the automobile by one key on the basis of fingerprint unlocking when the real-time pressure exceeds a preset pressure threshold;
s3, primarily adjusting the in-vehicle facility on the basis of fingerprint identification and one-key starting of the vehicle, wherein the primary adjustment comprises seat position adjustment, steering wheel position adjustment and rearview mirror position adjustment;
s4, according to the vehicle-mounted sensor, the stored personal information and the change of the driving habits of the driver, establishing an identification model of the behavior characteristics of the driver and the meteorological information by a convolutional neural network and a particle swarm algorithm model by adopting a machine learning method, and dividing the behavior characteristics of the driver into an acceleration characteristic, a braking characteristic, a positioning characteristic, a weight characteristic and a body temperature characteristic; through constantly studying and comparing, realize the intelligent regulation of vehicle, specifically include the following substep:
s41, dividing the characteristics of weather information into temperature, rain, snow and fog, collecting the operation behavior signal of the driver and the current state of the vehicle, and analyzing the behavior and weather condition of the driver;
extracting characteristic parameters capable of representing positioning, braking and accelerating behaviors of a driver, and establishing an initial driver behavior target matrix; extracting characteristic parameters capable of representing the temperature and the rain and snow state of meteorological information, and establishing an initial meteorological information matrix:
wherein the initial driver behavior target matrix is:
D＝(d1,d2,……,dN) Wherein d is1To locate the vector, d2As braking vector, d3For acceleration vectors, dNIs the Nth vector;
the initial weather information matrix is:
M＝(m1，m2,……,mN) Wherein m is1Is a temperature vector, m2Is a rain and snow information vector, mNIs the Nth vector;
s42, extracting key features of the initial driver behavior target matrix and the initial meteorological information matrix by using a particle swarm algorithm, and specifically comprising the following substeps:
s421, assuming that N particles form a matrix in the first D-dimensional target search space, wherein the ith particle is an N-dimensional vector:
Xi＝(xi1,xi2,xi3,...xiD),i＝1,2,3...,N
each particle has its own velocity, which is an N-dimensional vector:
Vi＝(vi1,vi2,vi3,...viD),i＝1,2,3...,N
s422, searching an individual extreme value of each particle in the matrix, wherein the individual extreme value is an optimal position searched by the traversal of the particle, and the individual extreme value is recorded as:
Pbest＝(Pi1,Pi2,...,PiD),i＝1,2,...,N
s423, searching a global extreme value of the whole matrix, wherein the global extreme value is an optimal position searched by the whole particle swarm in a traversal way, and the global extreme value is recorded as:
gbest＝(Pg1,Pg2...PgN)
s424, updating the speed formula and the position formula of the example based on the individual extreme values searched in the step S421 and the global extreme values searched in the step S422:
velocity formula: v. ofid(t+1)＝w×vid(t)+c1 r1(Pid-xid)+c2r2(Pgd-xid)
Position formula: x is the number ofid(t+1)＝xid(t)+v(t+1)
Wherein, c1And c2Is a learning factor, i.e. an acceleration constant, adjusting the maximum step size of learning, r1And r2Is a random number with a value range of [0,1]]W is the inertial weight, w is a non-negative number used to adjust the search range for the solution space, vidRepresenting the speed, xidRepresenting position, t is time, representing number of iterations, PidIs the optimal value, P, traversed by the particlegdIs the optimal value traversed by the population;
s425, extracting key features according to a speed formula and a position formula of the particles:
expressing the weight of the eigenvector in the characteristic matrix in S41 by a matrix R, and using R as the optimal solution of the matrix R*Represents;
set k granulesChild { l }1,l2,…,lkEach particle liThe position of (a) is a solution of R, and R × D and R × M respectively represent fitness functions, which are accuracy, where D and M are matrices in S41, and the position obtained when each particle is solved is a local optimal solution, i.e., P in the formula in step S424idThe solution obtained corresponding to the highest fitness is the global optimal solution, i.e. P in the formula in step S424gdAfter all the particles are subjected to iterative computation, the optimal weight matrix R can be obtained*For training data, calculate R*×DtrainAnd R*×MtrainFor test data, R is calculated*×DtestAnd R*×MtestThe distance between the training data characteristic matrix and the test data characteristic matrix is as follows:
D1＝dis(R*×Dtest,R*×Dtrain)
D2＝dis(R*×Mtest,R*×Mtrain)
D1and D2Distance differences respectively representing driver behavior and meteorological information, where the distance dis is calculated as:
wherein X, Y ∈ Cn×k，Cn×kExpressing a characteristic matrix, wherein X and Y are dimensions, and obtaining key characteristics of an initial driver behavior target matrix and an initial meteorological information matrix according to the fitness function meaning;
s43, establishing a vehicle regulation behavior characteristic model related to the driver behavior, the windshield wiper, the heating, the demisting and the air conditioner switch based on the convolutional neural network, and realizing the active regulation and setting of vehicle facilities; the establishment of the vehicle regulation behavior feature model comprises the following sub-steps:
s431, forward propagation of the convolutional neural network:
al＝σ(zl)＝σ(al-1*Wl+bl)
wherein, the superscript represents the number of layers, the asterisk represents the convolution, W is the linear coefficient matrix, b is the bias vector, σ is the ReLU activation function, and the matrix M represents the initial information matrix, so the above formula can be written as the form of adding corresponding positions after the convolution of M sub-matrices, that is:
the standard of pooling uses maximum pooling operation, and the activation function uses sigmoid activation function;
s432, carrying out convolution layer back propagation calculation:
the forward propagation algorithm uses max to pool the inputs, and the backward propagation algorithm will reduce the error deltalThe size of all sub-matrixes is reduced to the size before pooling, and the process is called upsample, so that the size of all sub-matrixes can be obtained
Thus for tensor δl-1Then delta can be obtainedl-1＝upsample(δl)⊙σ′(zl-1)
δl-1And deltalCan be based on calculations
Is derived from a gradient expression of, wherein zlAnd zl-1The relationship of (1) is:
zl＝al-1*Wl+bl＝σ(zl-1)*Wl+bl
obtaining:
deducing the gradient error delta of each layer according to the above formulalFor convolutional layer z and the correlation of linear coefficient matrix W and bias vector bIs represented as:
zl＝al-1*Wl+b
thus for the ith layer, the derivative of the convolution sum matrix W is expressed as:
due to deltalIs a high-dimensional tensor, b is a constant, usually will be δlThe terms of the respective sub-matrices are summed to obtain an error vector to obtain the gradient of b:
for the back propagation of convolutional layers, δ can be obtained by cross-over calculationl+1And deltalThe recurrence relation of (A) is as follows:
calculating a characteristic matrix by using the convolutional neural network to obtain a vehicle regulation behavior characteristic model;
s5, after the inertia weight w value of the preliminarily optimized driver behavior target matrix and the weather information matrix obtained in the step S42 are optimized, extracting key features by further utilizing the particle swarm algorithm in the step S42 to obtain final key features, wherein the method specifically comprises the following substeps:
s51, describing the influence of the previous generation speed of the particles on the current generation speed by inertia weight w, wherein the larger the w value is, the stronger the global optimization capability is, and the weaker the local optimization capability is; conversely, the smaller the w value is, the strong local optimization capability is obtained, and the inertia weight w value optimization formula is as follows:
w is the inertial weight, wmaxIs the maximum inertial weight, wminRun is the current iteration number, run for the minimum inertial weightmaxThe total number of iterations of the algorithm;
s52, in order to prevent the particle swarm optimization from falling into local minimization, optimizing the inertia weight w by using a formula in S51, wherein w belongs to [0,1], and substituting the optimized inertia weight w value into the step S42 to extract key features to obtain the optimal key features;
s6, further optimizing the identification model and the vehicle regulation model obtained in the step S43 based on the final key features obtained in the step S5, and specifically comprising the following steps:
s61, excitation function optimization:
for the Relu activation function, the value of model fitting loss function is small, the Relu activation function is adopted during the optimization of the convolutional neural network, and the Relu activation function is as follows:
f(x)＝max(x，0)
s62, selection and optimization of a gradient descent algorithm:
the model fitting process of the neural network can be divided into two stages, wherein in the first stage, a predicted value is obtained through a forward propagation algorithm, the predicted value and a training data label are compared to obtain a difference between the predicted value and the training data label, in the second stage, the gradient of a loss function to each parameter is calculated through a backward propagation algorithm, and then all weights are updated through a gradient descent algorithm;
first, an assumption function and a loss function of an optimization model are determined, and in supervised learning, the assumption function used for fitting an input sample is denoted as hθ(x) Where the fitting function is hθ(x)＝θ0+θ1x; to evaluate the goodness of the model fit, the degree of fit is typically measured by a loss function, which is typically the square of the difference between the sample output and the hypothesis function, for m samples (x)i,yi) (i ═ 1,2,. m), the loss function is
Wherein xiExpressed as the i-th sample feature, yiRepresenting the output corresponding to the ith sample;
let the function be hθ(x)＝θ0+θ1x1+...+θnxnWherein thetai(i ═ 0,1, 2.., n) is a model parameter, and x is a model parameteri(i ═ 0,1, 2.. times, n) for n feature values of each sample, add feature x01, so that
For the hypothesis function, a loss function of
For thetaiGradient is expressed as
Order to
The distance of the current position descent can be obtained by multiplying the step length alpha by the gradient of the loss function, and
s7, further adjustment of the in-vehicle facilities is realized according to the vehicle adjustment behavior characteristic model obtained by the machine learning component, wherein the further adjustment of the in-vehicle facilities comprises intelligent gearbox adjustment, engine start and stop functions, windshield defogging adjustment, windshield wiper adjustment, seat position fine adjustment, seat ventilation and heating, rearview defogging and air conditioning; wherein the closing and the opening of the starting and stopping functions of the engine can be reminded by voice.
Preferably, in step S41, D ═ D (diet, smoking, starting, making a call, turning around, turning on a light, closing a door, adjusting a seat, shifting a speed); m ═ fog (fog, rain, snow, temperature, air cleaning, ultraviolet, wind direction, wind force, season).
Preferably, the key feature Dkey(start, turn, close, adjust seat, shift), MkeyFog, rain, snow, temperature, wind).
The invention also provides an intelligent automobile adjusting system which comprises a controller, a fingerprint identification module, a pressure identification module, a first adjusting component, a machine learning component, a second adjusting component and an on-board sensor,
the fingerprint identification module is used for acquiring human body characteristic information and comparing the information to realize unlocking and starting functions; the fingerprint identification module is provided with a fingerprint identifier, the fingerprint identifier is arranged at a vehicle door handle of a driving position, whether the vehicle door handle is a stored driver or not is identified through fingerprint comparison, and if the vehicle door handle passes the identification, unlocking is realized;
the pressure identification module is a pressure sensor, the pressure sensor is integrated in the fingerprint identifier, the pressure sensor can detect the real-time pressure of fingerprint pressing and send the real-time pressure to the controller, and when the real-time pressure detected by the pressure sensor exceeds the pressure threshold value in the controller, the controller controls the automobile to start one key on the basis of fingerprint unlocking;
the first adjusting assembly is used for realizing the primary adjustment of facilities in the automobile after comparing the human body characteristic information acquired by the fingerprint identification module with the personal information stored in the controller, and the primary adjustment comprises seat position adjustment, steering wheel position adjustment and rearview mirror position adjustment;
the machine learning component establishes an identification model of the behavior characteristic and the meteorological information of the driver by adopting a machine learning method through a convolutional neural network and a particle swarm algorithm model according to the vehicle-mounted sensor, the stored personal information and the change of the driving habit of the driver, and divides the behavior characteristic of the driver into an acceleration characteristic, a braking characteristic, a positioning characteristic, a weight characteristic and a body temperature characteristic; the intelligent adjustment of the vehicle is realized through continuous learning and comparison;
the second adjusting component realizes further adjustment of the in-vehicle facilities according to the result of the machine learning component, and the further adjustment of the in-vehicle facilities comprises intelligent gearbox adjustment, engine starting and stopping functions, windshield defogging adjustment, windshield wiper adjustment, seat position fine adjustment, seat ventilation and heating, rearview defogging and air conditioning; wherein the closing and the opening of the starting and stopping functions of the engine can be reminded by voice.
Preferably, the vehicle-mounted sensor comprises a seat sensor and a weather sensor, the seat sensor comprises a pressure sensor, a temperature sensor and a moisture sensor, and the seat sensor is used for judging the change of the body state of the driver; the weather sensor comprises a humidity sensor and a temperature sensor, and is used for judging the change of the inside and outside environment of the vehicle.
Compared with the prior art, the invention has the following beneficial effects:
(1) the invention provides an intelligent automobile adjusting method and system, which can adjust some parameters of an automobile on the basis of unlocking the automobile according to the comparison between collected human body characteristic information and stored information, can establish a characteristic model through real-time learning and based on the habit of a driver, continuously learn and correct, achieve the best driving effect, intelligently adjust driving parameters according to weather and ensure the driving safety of the driver.
(2) The invention can improve the control and driving level of the vehicle and ensure the safe, smooth and efficient running of the vehicle. The continuous research and perfection of the intelligent vehicle control system are equivalent to the extension of the control, vision and sense functions of a driver, and the safety of road traffic can be greatly promoted.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of a smart vehicle tuning method according to an embodiment of the invention;
fig. 3 is a block diagram schematically illustrating the structure of the present invention.
Detailed Description
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
Specifically, the invention also provides an intelligent automobile adjusting method, which comprises the following steps:
s1, collecting human body characteristic information, comparing the information, and realizing unlocking and starting functions: comparing the acquired fingerprints, identifying whether the acquired human body characteristic information is a stored driver, unlocking if the acquired human body characteristic information is identified, and giving an alarm if the acquired human body characteristic information is not identified;
s2, collecting real-time pressure when the fingerprint is pressed, and starting the automobile by one key on the basis of fingerprint unlocking when the real-time pressure exceeds a preset pressure threshold;
s3, primarily adjusting the in-vehicle facility on the basis of fingerprint identification and one-key starting of the vehicle, wherein the primary adjustment comprises seat position adjustment, steering wheel position adjustment and rearview mirror position adjustment;
s4, according to the vehicle-mounted sensor, the stored personal information and the change of the driving habits of the driver, establishing an identification model of the behavior characteristics of the driver and the meteorological information by a convolutional neural network and a particle swarm algorithm model by adopting a machine learning method, and dividing the behavior characteristics of the driver into an acceleration characteristic, a braking characteristic, a positioning characteristic, a weight characteristic and a body temperature characteristic; through constantly studying and comparing, realize the intelligent regulation of vehicle, specifically include the following substep:
s41, dividing the characteristics of weather information into temperature, rain, snow and fog, collecting the operation behavior signal of the driver and the current state of the vehicle, and analyzing the behavior and weather condition of the driver;
extracting characteristic parameters capable of representing positioning, braking and accelerating behaviors of a driver, and establishing an initial driver behavior target matrix; extracting characteristic parameters capable of representing the temperature and the rain and snow state of meteorological information, and establishing an initial meteorological information matrix:
wherein the initial driver behavior target matrix is:
D＝(d1,d2,……,dN) Wherein d is1To locate the vector, d2As braking vector, d3For acceleration vectors, dNIs the Nth vector;
the initial weather information matrix is:
M＝(m1，m2,……,mN) Wherein m is1Is a temperature vector, m2Is a rain and snow information vector, mNIs the Nth vector;
s42, extracting key features of the initial driver behavior target matrix and the initial meteorological information matrix by using a particle swarm algorithm, and specifically comprising the following substeps:
s421, assuming that N particles form a matrix in the first D-dimensional target search space, wherein the ith particle is an N-dimensional vector:
Xi＝(xi1,xi2,xi3,...xiD),i＝1,2,3...,N
each particle has its own velocity, which is an N-dimensional vector:
Vi＝(vi1,vi2,vi3,...viD),i＝1,2,3...,N
s422, searching an individual extreme value of each particle in the matrix, wherein the individual extreme value is an optimal position searched by the traversal of the particle, and the individual extreme value is recorded as:
Pbest＝(Pi1,Pi2,...,PiD),i＝1,2,...,N
s423, searching a global extreme value of the whole matrix, wherein the global extreme value is an optimal position searched by the whole particle swarm in a traversal way, and the global extreme value is recorded as:
gbest＝(Pg1,Pg2...PgN)
s424, updating the speed formula and the position formula of the example based on the individual extreme values searched in the step S421 and the global extreme values searched in the step S422:
velocity formula: v. ofid(t+1)＝w×vid(t)+c1 r1(Pid-xid)+c2r2(Pgd-xid)
Position formula: x is the number ofid(t+1)＝xid(t)+v(t+1)
Wherein, c1And c2Is a learning factor, i.e.Acceleration constant, adjustment of maximum step size of learning, r1And r2Is a random number with a value range of [0,1]]W is the inertial weight, w is a non-negative number used to adjust the search range for the solution space, vidRepresenting the speed, xidRepresenting position, t is time, representing number of iterations, PidIs the optimal value, P, traversed by the particlegdIs the optimal value traversed by the population;
s425, extracting key features according to a speed formula and a position formula of the particles:
expressing the weight of the eigenvector in the characteristic matrix in S41 by a matrix R, and using R as the optimal solution of the matrix R*Represents;
let k particles { l }1,l2,…,lkEach particle liThe position of (a) is a solution of R, and R × D and R × M respectively represent fitness functions, which are accuracy, where D and M are matrices in S41, and the position obtained when each particle is solved is a local optimal solution, i.e., P in the formula in step S424idThe solution obtained corresponding to the highest fitness is the global optimal solution, i.e. P in the formula in step S424gdAfter all the particles are subjected to iterative computation, the optimal weight matrix R can be obtained*For training data, calculate R*×DtrainAnd R*×MtrainFor test data, R is calculated*×DtestAnd R*×MtestThe distance between the training data characteristic matrix and the test data characteristic matrix is as follows:
D1＝dis(R*×Dtest,R*×Dtrain)
D2＝dis(R*×Mtest,R*×Mtrain)
D1and D2Distance differences respectively representing driver behavior and meteorological information, where the distance dis is calculated as:
wherein X, Y ∈ Cn×k，Cn×kExpressing a characteristic matrix, wherein X and Y are dimensions, and obtaining key characteristics of an initial driver behavior target matrix and an initial meteorological information matrix according to the fitness function meaning;
s43, establishing a vehicle regulation behavior characteristic model related to the driver behavior, the windshield wiper, the heating, the demisting and the air conditioner switch based on the convolutional neural network, and realizing the active regulation and setting of vehicle facilities; the establishment of the vehicle regulation behavior feature model comprises the following sub-steps:
s431, forward propagation of the convolutional neural network:
al＝σ(zl)＝σ(al-1*Wl+bl)
wherein, the superscript represents the number of layers, the asterisk represents the convolution, W is the linear coefficient matrix, b is the bias vector, σ is the ReLU activation function, and the matrix M represents the initial information matrix, so the above formula can be written as the form of adding corresponding positions after the convolution of M sub-matrices, that is:
the standard of pooling uses maximum pooling operation, and the activation function uses sigmoid activation function;
s432, carrying out convolution layer back propagation calculation:
the forward propagation algorithm uses max to pool the inputs, and the backward propagation algorithm will reduce the error deltalThe size of all sub-matrixes is reduced to the size before pooling, and the process is called upsample, so that the size of all sub-matrixes can be obtained
Thus for tensor δl-1Then delta can be obtainedl-1＝upsample(δl)⊙σ′(zl-1)
δl-1And deltalCan be based on calculations
Is derived from a gradient expression of, wherein zlAnd zl-1The relationship of (1) is:
zl＝al-1*Wl+bl＝σ(zl-1)*Wl+bl
obtaining:
deducing the gradient error delta of each layer according to the above formulalThe relationship between the convolution layer z and the linear coefficient matrix W and the bias vector b is expressed as:
zl＝al-1*Wl+b
thus for the ith layer, the derivative of the convolution sum matrix W is expressed as:
due to deltalIs a high-dimensional tensor, b is a constant, usually will be δlThe terms of the respective sub-matrices are summed to obtain an error vector to obtain the gradient of b:
For the back propagation of convolutional layers, δ can be obtained by cross-over calculationl+1And deltalThe recurrence relation of (A) is as follows:
calculating a characteristic matrix by using the convolutional neural network to obtain a vehicle regulation behavior characteristic model;
s5, after the inertia weight w value of the preliminarily optimized driver behavior target matrix and the weather information matrix obtained in the step S42 are optimized, extracting key features by further utilizing the particle swarm algorithm in the step S42 to obtain final key features, wherein the method specifically comprises the following substeps:
s51, describing the influence of the previous generation speed of the particles on the current generation speed by inertia weight w, wherein the larger the w value is, the stronger the global optimization capability is, and the weaker the local optimization capability is; conversely, the smaller the w value is, the strong local optimization capability is obtained, and the inertia weight w value optimization formula is as follows:
w is the inertial weight, wmaxIs the maximum inertial weight, wminRun is the current iteration number, run for the minimum inertial weightmaxThe total number of iterations of the algorithm;
s52, in order to prevent the particle swarm optimization from falling into local minimization, optimizing the inertia weight w by using a formula in S51, wherein w belongs to [0,1], and substituting the optimized inertia weight w value into the step S42 to extract key features to obtain the optimal key features;
s6, further optimizing the identification model and the vehicle regulation model obtained in the step S43 based on the final key features obtained in the step S5, and specifically comprising the following steps:
s61, excitation function optimization:
for the Relu activation function, the value of model fitting loss function is small, the Relu activation function is adopted during the optimization of the convolutional neural network, and the Relu activation function is as follows:
f(x)＝max(x，0)
s62, selection and optimization of a gradient descent algorithm:
the model fitting process of the neural network can be divided into two stages, wherein in the first stage, a predicted value is obtained through a forward propagation algorithm, the predicted value and a training data label are compared to obtain a difference between the predicted value and the training data label, in the second stage, the gradient of a loss function to each parameter is calculated through a backward propagation algorithm, and then all weights are updated through a gradient descent algorithm;
first, an assumption function and a loss function of an optimization model are determined, and in supervised learning, the assumption function used for fitting an input sample is denoted as hθ(x) Where the fitting function is hθ(x)＝θ0+θ1x. To evaluate the goodness of the model fit, the degree of fit is typically measured by a loss function, which is typically the square of the difference between the sample output and the hypothesis function, for m samples (x)i,yi) (i ═ 1,2,. m), the loss function is
Wherein xiExpressed as the i-th sample feature, yiRepresenting the output corresponding to the ith sample;
let the function be hθ(x)＝θ0+θ1x1+...+θnxnWherein thetai(i ═ 0,1, 2.., n) is a model parameter, and x is a model parameteri(i ═ 0,1, 2.. times, n) for n feature values of each sample, add feature x01, so that
For the hypothesis function, a loss function of
For thetaiGradient is expressed as
Order to
The distance of the current position descent can be obtained by multiplying the step length alpha by the gradient of the loss function, and
s7, further adjustment of the in-vehicle facilities is realized according to the vehicle adjustment behavior characteristic model obtained by the machine learning component, wherein the further adjustment of the in-vehicle facilities comprises intelligent gearbox adjustment, engine start and stop functions, windshield defogging adjustment, windshield wiper adjustment, seat position fine adjustment, seat ventilation and heating, rearview defogging and air conditioning; wherein the closing and the opening of the starting and stopping functions of the engine can be reminded by voice.
Preferably, the present invention further provides an intelligent vehicle regulation system, as shown in fig. 3, which includes a controller 1, a fingerprint recognition module 2, a pressure recognition module 3, a first regulation component 4, a machine learning component 5, a second regulation component 6, and an on-board sensor 7.
The fingerprint identification module 2 is used for collecting human body characteristic information and comparing the information to realize unlocking and starting functions; fingerprint identification module is provided with fingerprint identification ware, and fingerprint identification ware sets up in driver's seat door handle department, compares through the fingerprint, and the discernment is the driver that has stored, if the discernment passes through, then realizes the unblock.
Pressure identification module 3 is pressure sensor, and pressure sensor is integrated inside fingerprint identification ware, and pressure sensor can detect the real-time pressure that the fingerprint pressed and send to the controller, and when the real-time pressure that pressure sensor detected exceeded the inside pressure threshold value of controller, controller control car was a key start on the basis of fingerprint unblock.
The first adjusting component 4 is used for realizing the primary adjustment of facilities in the vehicle after comparing the human body characteristic information acquired by the fingerprint identification module with the personal information stored in the controller, and the primary adjustment comprises seat position adjustment, steering wheel position adjustment and rearview mirror position adjustment;
the machine learning component 5 establishes an identification model of the behavior characteristics and the meteorological information of the driver by adopting a machine learning method through a convolutional neural network and a particle swarm algorithm model according to the vehicle-mounted sensor, the stored personal information and the change of the driving habits of the driver, and divides the behavior characteristics of the driver into acceleration characteristics, braking characteristics, positioning characteristics, weight characteristics and body temperature characteristics; the intelligent adjustment of the vehicle is realized through continuous learning and comparison;
the second adjusting component 6 realizes further adjustment of the in-vehicle facilities according to the result of the machine learning component, wherein the further adjustment of the in-vehicle facilities comprises intelligent gearbox adjustment, engine starting and stopping functions, windshield defogging adjustment, windshield wiper adjustment, seat position fine adjustment, seat ventilation and heating, rearview defogging and air conditioning; the voice reminding is carried out when the starting and stopping functions of the engine are closed and started;
the vehicle-mounted sensor comprises a seat sensor and a weather sensor, the seat sensor comprises a pressure sensor, a temperature sensor and a moisture sensor, and the seat sensor is used for judging the change of the body state of the driver; the weather sensor comprises a humidity sensor and a temperature sensor, and the weather sensor is used for judging the change of the inside and outside environment of the vehicle.
The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements made to the technical solution of the present invention by those skilled in the art without departing from the spirit of the present invention shall fall within the protection scope defined by the claims of the present invention.
Claims (4)
1. An intelligent automobile adjusting method is characterized in that: which comprises the following steps:
s1, collecting human body characteristic information, comparing the information, and realizing unlocking and starting functions: comparing the acquired fingerprints, identifying whether the acquired human body characteristic information is a stored driver, unlocking if the acquired human body characteristic information is identified, and giving an alarm if the acquired human body characteristic information is not identified;
s2, collecting real-time pressure when the fingerprint is pressed, and starting the automobile by one key on the basis of fingerprint unlocking when the real-time pressure exceeds a preset pressure threshold;
s3, primarily adjusting the in-vehicle facility on the basis of fingerprint identification and one-key starting of the vehicle, wherein the primary adjustment comprises seat position adjustment, steering wheel position adjustment and rearview mirror position adjustment;
s4, according to the vehicle-mounted sensor, the stored personal information and the change of the driving habits of the driver, establishing an identification model of the behavior characteristics of the driver and the meteorological information by a convolutional neural network and a particle swarm algorithm model by adopting a machine learning method, and dividing the behavior characteristics of the driver into an acceleration characteristic, a braking characteristic, a positioning characteristic, a weight characteristic and a body temperature characteristic; through constantly studying and comparing, realize the intelligent regulation of vehicle, specifically include the following substep:
s41, dividing the characteristics of weather information into temperature, rain, snow and fog, collecting the operation behavior signal of the driver and the current state of the vehicle, and analyzing the behavior and weather condition of the driver;
extracting characteristic parameters capable of representing positioning, braking and accelerating behaviors of a driver, and establishing an initial driver behavior target matrix; extracting characteristic parameters capable of representing the temperature and the rain and snow state of meteorological information, and establishing an initial meteorological information matrix:
wherein the initial driver behavior target matrix is:
D＝(d1,d2,……,dN) Wherein d is1To locate the vector, d2As braking vector, d3For acceleration vectors, dNIs the Nth vector;
the initial weather information matrix is:
M＝(m1，m2,……,mN) Wherein m is1Is a temperature vector, m2Is a rain and snow information vector, mNIs the Nth vector;
s42, extracting key features of the initial driver behavior target matrix and the initial meteorological information matrix by using a particle swarm algorithm, and specifically comprising the following substeps:
s421, assuming that in the first D-dimensional target search space, a matrix X is formed by N particles, where the ith particle is an N-dimensional vector:
Xi＝(xi1,xi2,xi3,...xiD),i＝1,2,3...,N
each particle has its own velocity V, which is an N-dimensional vector:
Vi＝(vi1,vi2,vi3,...viD),i＝1,2,3...,N
s422, searching an individual extreme value P of each particle in the matrix, wherein the individual extreme value is the optimal position searched by the traversal of the particle so far, and the individual extreme value is recorded as PbestComprises the following steps:
Pbest＝(Pi1,Pi2,...,PiD),i＝1,2,...,N
s423, searching a global extreme value g of the whole matrix, wherein the global extreme value g is the optimal position searched by the whole particle swarm in a traversal way so farbestIs recorded as:
gbest＝(Pg1,Pg2...PgN)
s424, based on the individual extreme values searched in the step S422 and the global extreme values searched in the step S423, updating the speed formula and the position formula of the example:
velocity formula: v. ofid(t+1)＝w×vid(t)+c1r1(Pid-xid)+c2r2(Pgd-xid)
Position formula: x is the number ofid(t+1)＝xid(t)+v(t+1)
Wherein, c1And c2Is a learning factor, i.e. an acceleration constant, for adjusting the maximum step size of learning, r1And r2Is a random number with a value range of [0,1]]W is an inertial weight, which is a non-negative number, used to adjust the search range for the solution space, vidRepresenting the speed, xidRepresenting position, t time, representing number of iterations, v iteration speed, PidIs the optimal value, P, traversed by the particlegdIs the optimal value traversed by the population;
s425, extracting key features according to a speed formula and a position formula of the particles:
expressing the weight of the eigenvector in the characteristic matrix in S41 by a matrix R, and using R as the optimal solution of the matrix R*Represents;
let k particles { l }1,l2,…,lkEach particle liThe position of (a) is a solution of R, and R × D and R × M respectively represent fitness functions, which are accuracy, where D and M are matrices in S41, and the position obtained when each particle is solved is a local optimal solution, i.e., P in the formula in step S424idThe solution obtained corresponding to the highest fitness is the global optimal solution, i.e. P in the formula in step S424gdAfter all the particles are subjected to iterative computation, the optimal weight matrix R can be obtained*For training data, calculate R*×DtrainAnd R*×MtrainFor test data, R is calculated*×DtestAnd R*×MtestThe distance between the training data characteristic matrix and the test data characteristic matrix is as follows:
D1＝dis(R*×Dtest,R*×Dtrain)
D2＝dis(R*×Mtest,R*×Mtrain)
D1and D2Distance differences respectively representing driver behavior and meteorological information, where distance dis is calculated as:
wherein X, Y ∈ Cn×k，Cn×kExpressing a characteristic matrix, wherein X and Y are dimensions, and obtaining key characteristics of an initial driver behavior target matrix and an initial meteorological information matrix according to the fitness function significance;
s43, establishing a vehicle regulation behavior characteristic model related to the driver behavior, the windshield wiper, the heating, the demisting and the air conditioner switch based on the convolutional neural network, and realizing the active regulation and setting of vehicle facilities; the establishment of the vehicle regulation behavior feature model comprises the following sub-steps:
s431, forward propagation of the convolutional neural network:
al＝σ(zl)＝σ(al-1*Wl+bl)
the superscript in the formula represents the number of layers, the asterisk represents convolution, W is a linear coefficient matrix, a is an initial input matrix, b is a bias vector, sigma is a ReLU activation function, and the matrix M represents an initial information matrix, so that the above formula can be written into a form of adding corresponding positions after convolution of M submatrices, that is:
the standard of pooling uses maximum pooling operation, and the activation function uses sigmoid activation function;
s432, carrying out convolution layer back propagation calculation:
the forward propagation algorithm uses max to pool the input, and the reduced error delta is obtained when the input is reversely propagatedlAll the sub-matrix sizes are reduced to the sizes before the pooling to obtain
Thus for tensor δl-1To obtain deltal-1＝upsample(δl)⊙σ′(zl-1)；
δl-1And deltalAccording to the calculation of
Is derived, wherein zlAnd zl-1The relationship of (1) is:
zl＝al-1*Wl+bl＝σ(zl-1)*Wl+bl
obtaining:
deducing the gradient error delta of each layer according to the above formulalFor convolutional layer z and the relation table of linear coefficient matrix W and bias vector bShown as follows:
zl＝al-1*Wl+b
for layer l, the derivative of the convolution sum matrix W is expressed as:
where i, j is the number of input matrix layers, p, q is the number of matrix layers W, δlIs a high-dimensional tensor, b is a constant, willlThe terms of the respective sub-matrices are summed to obtain an error vector to obtain the gradient of b:
for the back propagation of convolutional layers, δ can be obtained by cross-over calculationl+1And deltalThe recurrence relation of (A) is as follows:
calculating a characteristic matrix by using the convolutional neural network to obtain a vehicle regulation behavior characteristic model;
s5, after the inertia weight w value of the preliminarily optimized driver behavior target matrix and the weather information matrix obtained in the step S42 are optimized, extracting key features by further utilizing the particle swarm algorithm in the step S42 to obtain final key features, wherein the method specifically comprises the following substeps:
s51, describing the influence of the previous generation speed of the particles on the current generation speed by inertia weight w, wherein the larger the w value is, the stronger the global optimization capability is, and the weaker the local optimization capability is; on the contrary, the smaller the w value is, the strong local optimization capability is,
the inertia weight w value optimization formula is as follows:
wherein w is the inertial weight, wmaxIs the maximum inertial weight, wminRun is the current iteration number, run for the minimum inertial weightmaxThe total number of iterations of the algorithm;
s52, in order to prevent the particle swarm optimization from falling into local minimization, optimizing the inertia weight w by using a formula in S51, wherein w belongs to [0,1], and substituting the optimized inertia weight w value into the step S42 to extract key features to obtain the optimal key features;
s6, further optimizing the identification model and the vehicle regulation model obtained in the step S43 based on the final key features obtained in the step S5, and specifically comprising the following steps:
s61, excitation function optimization:
for the Relu activation function, the value of model fitting loss function is small, the Relu activation function is adopted during the optimization of the convolutional neural network, and the Relu activation function is as follows:
f(x)＝max(x，0)
s62, selection and optimization of a gradient descent algorithm:
first, an assumption function and a loss function of an optimization model are determined, and in supervised learning, the assumption function used for fitting an input sample is denoted as hθ(x) The fitting function is hθ(x)＝θ0+θ1x；
The degree of fit is measured by a loss function, which is typically the square of the difference between the sample output and the hypothesis function, for m samples (x)i,yi) (i ═ 1,2,. m), the loss function is
Wherein xiExpressed as the i-th sample feature, yiRepresenting the output corresponding to the ith sample;
let the function be hθ(x)＝θ0+θ1x1+...+θnxnWherein thetai(i ═ 0,1, 2.., n) is a model parameter, and x is a model parameteri(i ═ 0,1, 2.. times, n) for n feature values of each sample, add feature x01, so that
For the hypothesis function, a loss function of
For thetaiGradient is expressed as
Order to
The distance of the current position descent can be obtained by multiplying the step length alpha by the gradient of the loss function, and
s7, further adjustment of the in-vehicle facilities is achieved according to the vehicle adjustment behavior characteristic model obtained by the machine learning component, and the further adjustment of the in-vehicle facilities comprises intelligent gearbox adjustment, engine starting and stopping functions, windshield defogging adjustment, windshield wiper adjustment, seat position fine adjustment, seat ventilation and heating, rearview defogging and air conditioning.
2. An intelligent vehicle adjusting system based on the intelligent vehicle adjusting method of claim 1, characterized in that: which comprises a controller, a fingerprint identification module, a pressure identification module, a first adjusting component, a machine learning component, a second adjusting component and a vehicle-mounted sensor,
the fingerprint identification module is used for acquiring human body characteristic information and comparing the information to realize unlocking and starting functions; the fingerprint identification module is provided with a fingerprint identifier, the fingerprint identifier is arranged at a vehicle door handle of a driving position, whether the vehicle door handle is a stored driver or not is identified through fingerprint comparison, and if the vehicle door handle passes the identification, unlocking is realized;
the pressure identification module is a pressure sensor, the pressure sensor is integrated in the fingerprint identifier, the pressure sensor can detect the real-time pressure of fingerprint pressing and send the real-time pressure to the controller, and when the real-time pressure detected by the pressure sensor exceeds the pressure threshold value in the controller, the controller controls the automobile to start one key on the basis of fingerprint unlocking;
the first adjusting assembly is used for realizing the primary adjustment of facilities in the automobile after comparing the human body characteristic information acquired by the fingerprint identification module with the personal information stored in the controller, and the primary adjustment comprises seat position adjustment, steering wheel position adjustment and rearview mirror position adjustment;
the machine learning component establishes an identification model of the behavior characteristic and the meteorological information of the driver by adopting a machine learning method through a convolutional neural network and a particle swarm algorithm model according to the vehicle-mounted sensor, the stored personal information and the change of the driving habit of the driver, and divides the behavior characteristic of the driver into an acceleration characteristic, a braking characteristic, a positioning characteristic, a weight characteristic and a body temperature characteristic; the intelligent adjustment of the vehicle is realized through continuous learning and comparison;
the second adjusting component realizes further adjustment of the in-vehicle facilities according to the result of the machine learning component, and the further adjustment of the in-vehicle facilities comprises intelligent gearbox adjustment, engine start and stop functions, windshield defogging adjustment, windshield wiper adjustment, seat position fine adjustment, seat ventilation and heating, rearview defogging and air conditioning.
3. The intelligent vehicle conditioning system of claim 2, wherein: the vehicle-mounted sensor comprises a seat sensor and a weather sensor, the seat sensor comprises a pressure sensor, a temperature sensor and a moisture sensor, and the seat sensor is used for judging the change of the body state of a driver; the weather sensor comprises a humidity sensor and a temperature sensor, and is used for judging the change of the inside and outside environment of the vehicle.
4. The intelligent vehicle conditioning system of claim 2, wherein: wherein the closing and the opening of the starting and stopping functions of the engine can be reminded by voice.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110036153.1A CN112874456B (en) | 2021-01-12 | 2021-01-12 | Intelligent vehicle adjusting method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110036153.1A CN112874456B (en) | 2021-01-12 | 2021-01-12 | Intelligent vehicle adjusting method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112874456A CN112874456A (en) | 2021-06-01 |
| CN112874456B true CN112874456B (en) | 2022-04-19 |
Family
ID=76044241
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110036153.1A Active CN112874456B (en) | 2021-01-12 | 2021-01-12 | Intelligent vehicle adjusting method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112874456B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP4316937A1 (en) * | 2022-08-04 | 2024-02-07 | e.solutions GmbH | Electronic apparatus for trainable vehicle component setting and vehicle |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113954703B (en) * | 2021-11-29 | 2023-04-28 | 东风汽车有限公司东风日产乘用车公司 | Vehicle space adjustment method, electronic device and storage medium |
| CN114572151A (en) * | 2022-01-25 | 2022-06-03 | 中国第一汽车股份有限公司 | Driving position seat control method and device based on fingerprint unlocking and vehicle |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5653462A (en) * | 1992-05-05 | 1997-08-05 | Automotive Technologies International, Inc. | Vehicle occupant position and velocity sensor |
| CN104590130A (en) * | 2015-01-06 | 2015-05-06 | 上海交通大学 | Rearview mirror self-adaptive adjustment method based on image identification |
| CN110110846A (en) * | 2019-04-24 | 2019-08-09 | 重庆邮电大学 | Auxiliary driver's vehicle exchange method based on convolutional neural networks |
| CN111071187A (en) * | 2018-10-19 | 2020-04-28 | 上海商汤智能科技有限公司 | Driving environment intelligent adjustment and driver registration method and device, vehicle and equipment |
| CN111717147A (en) * | 2019-03-22 | 2020-09-29 | 上海博泰悦臻网络技术服务有限公司 | Automatic adjusting system and method for vehicle seat |
| CN112026686A (en) * | 2019-06-04 | 2020-12-04 | 上海汽车集团股份有限公司 | Method and device for automatically adjusting position of vehicle seat |
| CN112070823A (en) * | 2020-08-28 | 2020-12-11 | 武汉亘星智能技术有限公司 | Video identification-based automobile intelligent cabin adjusting method, device and system |
-
2021
- 2021-01-12 CN CN202110036153.1A patent/CN112874456B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5653462A (en) * | 1992-05-05 | 1997-08-05 | Automotive Technologies International, Inc. | Vehicle occupant position and velocity sensor |
| CN104590130A (en) * | 2015-01-06 | 2015-05-06 | 上海交通大学 | Rearview mirror self-adaptive adjustment method based on image identification |
| CN111071187A (en) * | 2018-10-19 | 2020-04-28 | 上海商汤智能科技有限公司 | Driving environment intelligent adjustment and driver registration method and device, vehicle and equipment |
| CN111717147A (en) * | 2019-03-22 | 2020-09-29 | 上海博泰悦臻网络技术服务有限公司 | Automatic adjusting system and method for vehicle seat |
| CN110110846A (en) * | 2019-04-24 | 2019-08-09 | 重庆邮电大学 | Auxiliary driver's vehicle exchange method based on convolutional neural networks |
| CN112026686A (en) * | 2019-06-04 | 2020-12-04 | 上海汽车集团股份有限公司 | Method and device for automatically adjusting position of vehicle seat |
| CN112070823A (en) * | 2020-08-28 | 2020-12-11 | 武汉亘星智能技术有限公司 | Video identification-based automobile intelligent cabin adjusting method, device and system |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP4316937A1 (en) * | 2022-08-04 | 2024-02-07 | e.solutions GmbH | Electronic apparatus for trainable vehicle component setting and vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112874456A (en) | 2021-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112874456B (en) | Intelligent vehicle adjusting method and system | |
| CN109624986B (en) | Driving style learning cruise control system and method based on mode switching | |
| CN111433102B (en) | Method and system for collective vehicle control prediction in an autonomous vehicle | |
| CN101158496B (en) | Air conditioner for vehicle and controlling method thereof | |
| US20100087987A1 (en) | Apparatus and Method for Vehicle Driver Recognition and Customization Using Onboard Vehicle System Settings | |
| CN110949398A (en) | Method for detecting abnormal driving behavior of first-vehicle drivers in vehicle formation driving | |
| CN108482481B (en) | Four-wheel steering control method for four-wheel independent drive and steering electric automobile | |
| CN110119714B (en) | Driver fatigue detection method and device based on convolutional neural network | |
| WO2021129156A1 (en) | Control method, device and system of intelligent car | |
| CN111391831B (en) | Automobile following speed control method and system based on preceding automobile speed prediction | |
| CN110516380B (en) | Deep reinforcement test method and system based on vehicle driving simulation data | |
| CN115848398B (en) | Lane departure early warning system assessment method based on learning and considering driver behavior characteristics | |
| CN111783943B (en) | LSTM neural network-based driver braking strength prediction method | |
| JP6924629B2 (en) | Automatic control device and its control method | |
| CN115186594A (en) | Energy-saving speed optimization method under influence of man-vehicle-road coupling | |
| CN107367931A (en) | Adaptive vehicle controls | |
| CN111196124B (en) | In-vehicle environment regulation method and device, electronic equipment and storage medium | |
| CN111160089B (en) | Track prediction system and method based on different vehicle types | |
| CN116310542A (en) | Image classification method based on improved cross entropy loss function | |
| KR20170133213A (en) | Intelectual driver assistance system and method for automated driving | |
| CN112829755B (en) | System and method for recognizing state of passenger through pressure distribution of foot position of passenger | |
| US20220188621A1 (en) | Generative domain adaptation in a neural network | |
| CN113954844A (en) | Intelligent automobile man-machine driving mode switching system | |
| EP3967565B1 (en) | Methods and apparatuses for estimating an environmental condition | |
| CN113255725B (en) | Automobile sensor attack detection and repair method based on two-stage LSTM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |