KR20200075918A - Vehicle and control method thereof - Google Patents

Abstract

The present invention relates to a vehicle and a control method thereof. An objective of the present invention is to efficiently assist in labeling of learning data and exclude unique property information of a vehicle in determining the driving tendency of a driver to accurately classify the driving tendency of the driver and select an automatic driving mode. To achieve the objective, according to the present invention, the control method of a vehicle comprises: a step of extracting a statistical property during a preset section at preset time intervals from vehicle driving information acquired by driving of a vehicle; a step of extracting a statistical property from the converted vehicle driving information, and performing labeling of the vehicle driving information to extract the driving tendency of a driver from the extracted statistical property; and a step of a generating classification logic to classify the driving tendency of the driver based on a result of the labeling.

Description

Vehicle and its control method{VEHICLE AND CONTROL METHOD THEREOF}

The present invention relates to a vehicle, and relates to a vehicle and a control method provided to set and operate a driving mode according to a driver's driving tendency.

With the development of sensor technology and automatic control technology, various types of automatic driving modes are also provided in vehicles, and automatic driving modes suitable for the driving environment are selected to change the vehicle's system settings to suit the selected driving mode, thereby improving driver convenience and safety. I came to plan.

However, the selection of the automatic driving mode only applies to the automatic driving mode determined by the manufacturer of the vehicle at a time, and a cumbersome process in which the driver directly selects the driving mode in consideration of his driving tendency is involved.

According to one aspect of the present invention, the vehicle system selects an optimal automatic driving mode suitable for the driving tendency of the driver through analysis of the driving tendency of the driver, but efficiently labels learning data in determining the driving tendency of the driver. The objective is to enable the driver to classify the driving tendency and select the automatic driving mode more accurately by assisting and excluding vehicle-specific characteristic information.

A vehicle control method according to the present invention for the above-described object includes: extracting statistical characteristics during a predetermined section at a preset time interval from vehicle driving information obtained through driving of the vehicle; Extracting statistical characteristics from the converted vehicle driving information, and performing labeling of the vehicle driving information to extract a driver's driving tendency from the extracted statistical characteristics; And generating classification logic for classifying a driver's driving tendency based on the result of the labeling.

The above-described vehicle control method further comprises converting the vehicle driving information obtained through driving of the vehicle into a form of driving information of a predetermined standard vehicle to exclude unique characteristics of the vehicle from the vehicle driving information. Includes.

The vehicle control method described above distinguishes driving information generated by a driver's operation from driving information generated by a driver's operation in order to exclude unique characteristics of the vehicle from the vehicle driving information.

In the above-described vehicle control method, the driving information generated by the driver's operation includes at least one of a damper speed, a vehicle speed, a longitudinal acceleration, a lateral acceleration, and a yaw rate.

In the above-described vehicle control method, the driving information generated by the driver's operation includes at least one of the steering angular speed, the accelerator pedal operation amount, and the brake pedal operation amount.

In the above-described vehicle control method, the inherent characteristics of the vehicle are excluded by selectively taking some of the damper speed, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angular velocity and accelerator pedal operation amount, and brake pedal operation amount.

In the above-described vehicle control method, in the step of extracting the statistical characteristics, each of the preset sections is made to have sections overlapped with each other.

In the aforementioned vehicle control method, the statistical characteristics include a feature vector, a maximum deviation, a minimum deviation, and a standard deviation.

In the vehicle control method described above, a machine learning-based neural network learning theory is applied for the labeling.

In the vehicle control method described above, an error matrix is used for the labeling.

In the above-described vehicle control method, the labeling generates a classification result of the data set of the vehicle driving information through driver classification logic; And generating the rearranged error matrix by rearranging the generated classification result through error matrix rearrangement logic.

In the above-described vehicle control method, the error matrix is represented as an image in which a cluster is expressed through reordering.

In the above-described vehicle control method, the method further includes constructing a database by obtaining the vehicle driving information through the test driving of the vehicle.

The above-described vehicle control method further includes receiving the vehicle driving information stored in the database and removing noise included in the vehicle driving information.

The vehicle according to the present invention for the above-described object comprises: feature extraction logic for each section that extracts statistical characteristics during a preset section at preset time intervals from vehicle driving information obtained through driving of the vehicle; Driving propensity labeling logic for extracting statistical characteristics from the converted vehicle driving information and performing labeling of the vehicle driving information to extract a driver's driving propensity from the extracted statistical characteristics; And learning logic for generating classification logic for classifying a driver's driving tendency based on the result of the labeling.

In the above-described vehicle, the vehicle characteristic standardization logic to exclude the unique characteristics of the vehicle from the vehicle driving information by converting the vehicle driving information obtained through the driving of the vehicle into a form of driving information of a predetermined standard vehicle It includes more.

In the above-described vehicle, the vehicle characteristic standardization logic selectively selects the driving information generated by the driver's manipulation and the driving information generated by the driving environment in order to exclude the unique characteristics of the vehicle from the vehicle driving information. To take.

In the above-described vehicle, the driving information generated by the driver's operation includes at least one of a damper speed, a vehicle speed, a longitudinal acceleration, a lateral acceleration, and a yaw rate.

In the above-described vehicle, the driving information generated by the driver's operation includes at least one of the steering angular speed, the accelerator pedal operation amount, and the brake pedal operation amount.

In the above-described vehicle, the inherent characteristics of the vehicle are excluded by selectively taking part of the damper speed, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angular velocity, accelerator pedal operation amount, and brake pedal operation amount.

In the above-described vehicle, the feature extraction logic for each section, in the step of extracting the statistical characteristics, each of the preset sections has a section overlapped with each other.

In the above-described vehicle, the statistical characteristics include a feature vector, a maximum deviation, a minimum deviation, and a standard deviation.

In the above-described vehicle, a machine learning-based neural network learning theory is applied for the labeling.

In the vehicle described above, an error matrix is used for the labeling.

In the above-described vehicle, the driving propensity labeling logic generates a classification result of the data set of the vehicle driving information through driver classification logic; The rearranged error matrix is generated by rearranging the generated classification result through error matrix rearrangement logic.

In the above-described vehicle, the error matrix is represented by an image in which a cluster is expressed through reordering.

In the above-mentioned vehicle, a database for storing the vehicle driving information obtained through the test driving of the vehicle is further included.

In the above-described vehicle, the vehicle may further include noise removal logic that receives the vehicle driving information stored in the database and removes noise included in the vehicle driving information.

Another vehicle control method according to the present invention for the above-described object, in the vehicle driving information of the test vehicle converted by converting the vehicle driving information obtained through the driving of the test vehicle into a form of driving information of a predetermined standard vehicle Allowing the unique characteristics of the test vehicle to be excluded; Extracting statistical characteristics during a predetermined section at predetermined time intervals from the converted vehicle driving information; Extracting statistical characteristics from the converted vehicle driving information and applying machine learning-based neural network learning theory to the extracted statistical characteristics to perform labeling of the vehicle driving information to extract a driver's driving propensity; And generating classification logic for classifying a driver's driving tendency based on the result of the labeling.

Another vehicle control method according to the present invention for the above-described object, in the vehicle driving information of the mass-produced vehicle converted by converting the vehicle driving information obtained through the driving of the mass-produced vehicle into the form of driving information of a predetermined standard vehicle Allowing the unique characteristics of the mass-production vehicle to be excluded; Extracting statistical characteristics during a predetermined section at predetermined time intervals from the converted vehicle driving information; Classifying a driving propensity of a driver driving the mass-produced vehicle using a classification logic provided through labeling of the vehicle driving information to extract a driving propensity of the driver; And determining an automatic driving mode of the vehicle based on the classified driving propensity.

According to one aspect of the present invention, the vehicle system selects an optimal automatic driving mode suitable for the driving tendency of the driver through analysis of the driving tendency of the driver, but efficiently labels learning data in determining the driving tendency of the driver. By assisting and excluding vehicle-specific characteristic information, it is possible to classify a driver's driving propensity and select an automatic driving mode more accurately.

1 is a view showing a control system for classifying a driving tendency of a vehicle according to an embodiment of the present invention.
2 is a diagram illustrating a flow of a learning step and a classification step for classification of driving propensity according to an embodiment of the present invention.
3 is a view showing a vehicle feature standardization method according to an embodiment of the present invention.
4 is a view showing an embodiment of vehicle feature standardization in an embodiment of the present invention.
5 is a diagram illustrating feature extraction for each section according to an embodiment of the present invention.
6 is a view showing a process of rearranging driver classification results using an error matrix for driving propensity labeling according to an embodiment of the present invention.
7 is a view showing a driver classification based on a driving tendency according to an embodiment of the present invention.
8 is a diagram illustrating an example of classification according to a sensor input in classification of driving propensity according to another embodiment of the present invention.

1 is a view showing a control system for classifying a driving tendency of a vehicle according to an embodiment of the present invention. The driving propensity determination unit 112 classifies and determines the driving propensity of the driver driving the vehicle and provides the result of the classification to the driving control unit 162, thereby allowing the driving control unit 162 to match the driving propensity of the driver in the mass-production vehicle. It is possible to select an appropriate driving mode and perform driving assist control. The driving propensity classification result may include ECO mode / COMPORT mode / NORMAL mode / SPORT mode / SPORT+ mode.

A sensor group 150 composed of a plurality of sensors for detecting state information (driving information) of a vehicle is connected to the driving propensity determination unit 112 so that communication is possible. The sensor group 150 includes a lateral acceleration sensor 132 and a longitudinal acceleration sensor 134, a steering angular velocity sensor 136, a yaw rate sensor 138, a navigation 140, a damper speed sensor 142, and various other types. It includes a kind of sensor 144. In addition, a database 122 is connected to the driving propensity determination unit 112 so that communication is possible. In this database 122, vehicle driving information of various types of test vehicles is stored in a plurality of data sets.

The driving control unit 162 includes various elements including the steering 172, the engine 174, the braking 176, and the battlefield 178 for selection of an appropriate driving mode and driving assistance control suitable for a driver's driving tendency in a mass-production vehicle. Control.

2 is a diagram illustrating a flow of a learning step and a classification step for classification of driving propensity according to an embodiment of the present invention.

In FIG. 2, the process indicated by the dotted arrow is a learning step for generating classification logic 214. The learning step is a process of generating a classification logic 214 based on vehicle driving information obtained while driving a plurality of test vehicles by a plurality of test drivers in the development stage of the vehicle. In the learning phase, vehicle driving information of the test vehicle that the test driver obtains while driving the test vehicle and stores it in the database 122 is used. In addition, in the learning step, the vehicle driving information stored in the database 122 includes noise removal logic 204, vehicle feature standardization logic 206, feature extraction logic for each section 208, driving propensity labeling logic 210, and learning Through the logic 212, the learned classification logic 214 based on the machine learning-based neural only model is completed. The completed classification logic 214 is used to classify and determine the actual driving tendency of the driver driving the vehicle in the actual classification step indicated by the solid arrow in FIG. 2.

In FIG. 2, the process indicated by the solid arrow is mass-produced through the classification logic 214 already generated in the development process based on the vehicle driving information obtained during actual driving of the vehicle that has been developed and produced and sold (hereinafter, mass-produced vehicle). This is a classification step that determines and classifies the actual driving tendency of the driver of the vehicle. Unlike the learning stage, in the classification stage, a driver's propensity is classified using vehicle driving information acquired in real time while a driver of a mass-produced vehicle that has been produced and sold while driving the vehicle. In the classification step, statistical characteristics are extracted while the vehicle driving information acquired in real time passes through the noise removal logic 204, the vehicle feature standardization logic 206, and the feature extraction logic 208 for each section, and is extracted. The driver's driving tendency is classified in the classification logic 214 using statistical characteristics. The classified driving propensity information is transmitted to the driving control unit 162 and is used to select an appropriate driving mode suitable for the driving propensity of the driver in the mass-production vehicle and to control driving assistance.

Hereinafter, a process of generating a classifier through a test driver and a test vehicle in a learning step will be first described, and then a process of classifying and determining a driver's driving tendency in real time through a classifier in which mass production vehicles are generated will be described.

<Database construction>

For the purpose of generating the machine learning-based learned classification logic 214, a plurality of test drivers are driven to drive a vehicle (hereinafter, a test vehicle), and vehicle driving information is collected through driving of the test vehicle by the test driver And accumulate and store it in the database 122.

The vehicle driving information stored in the database 122 includes vehicle driving data generated by driving a test driver having a driving tendency to be classified. For example, in order to classify driving tendencies for each driving mode of the ECO mode / COMPORT mode / NORMAL mode / SPORT mode / SPORT+ mode, the vehicle driving information for each driving tendency (by driving mode) is stored in the database. The vehicle driving information for each driving propensity includes driving speed information, lateral acceleration information, longitudinal acceleration information, and yaw rate information of the test vehicle. The'ECO mode' is a driving mode that reduces fuel consumption and enables economical driving. 'COMPORT mode' is a comfortable and comfortable driving mode. The'NORMAL mode' is a normal ordinary driving mode. 'SPORT mode' is a dynamic and cheerful driving mode. 'SPORT+ mode' is a more dynamic and fast driving mode than'SPORT mode'.

<Noise reduction>

The vehicle driving information stored in the database 122 is removed from the noise included in the vehicle driving information in the noise removal logic 204. A low pass filter may be used to remove noise included in vehicle driving information.

In addition, an exponential moving average expressed by Equation 1 below may be used to remove noise included in vehicle driving information. In Equation 1 below, St is performed as a result of performing the exponential moving average method using the raw data Yt.

<Equation 1>

In addition, a Gaussian Filter Convolution method may be used to remove noise included in vehicle driving information.

<Vehicle feature standardization>

Various different types of test vehicles may be used to collect vehicle driving information. In an embodiment of the present invention, a unique characteristic of a vehicle used for collecting vehicle driving information is removed from the vehicle driving information. Thus, by excluding the unique characteristics of the vehicle, it is possible to determine the driver's driving propensity.

The obtained vehicle driving information is converted into the standard driving information form of the standard vehicle Z. Here, the standard driving information is driving information that can be obtained when the same type of driving is performed using the standard vehicle Z. Vehicle feature standardization is, for example, when vehicle driving information is obtained by performing a specific type of driving with different types of vehicles CAR#1, CAR#2, CAR#3, and CAR#4, vehicle CAR#1, CAR By excluding the unique characteristics of each of #2, CAR#3, and CAR#4, it means to convert to standardized vehicle driving information that can be obtained when the same specific type of driving is performed with the standard vehicle Z.

3 is a view showing a vehicle feature standardization method according to an embodiment of the present invention.

As shown in FIG. 3, in the embodiment of the present invention, vehicle feature standardization is performed by selectively using the value of the sensor detection data of the vehicle according to the situation. However, in Fig. 3, No. Even if various information is used as in case 1, the accuracy of driving propensity determination is not always high. Rather, No. As in the case of 2, when removing some vehicle driving information (such as a damper speed), the accuracy of driving propensity determination may be higher.

Accordingly, in this case, since the'damper speed' is more affected by various road environments than the driver's driving tendency, it may be more desirable to selectively remove the'damper speed' to determine the driving tendency of the driver. A damper means a reducer provided to absorb vibrations in a vehicle suspension.

'Accel pedal' refers to the amount of operation of the accelerator pedal for acceleration. In the case of'longitudinal acceleration' and'accelerator pedal', the driver's propensity to drive may be affected, but since the effect from traffic lights or traffic jams is large, conditional standardization is necessary.

In the case of'lateral acceleration','yield rate', and'steering angular velocity', inevitable transverse behavior according to the road pattern occurs, so even in this case, conditional standardization is necessary.

4 is a view showing an embodiment of vehicle feature standardization in an embodiment of the present invention. 4 shows an embodiment in which a specific sensor value is excluded from vehicle driving information in a specific driving condition.

First, the detection value of the longitudinal acceleration sensor may be excluded based on the distance between the front vehicles and the traffic signal. The distance between the front vehicles can be measured through the Smart Cruise Control (SCC) function provided in the vehicle. The traffic light signal can be confirmed through the traffic light information provided from the traffic light information providing system. When the front inter-vehicle distance is greater than or equal to the preset reference inter-vehicle distance (Dscc) (YES in 412), and the signal of the traffic light is a stop signal (YES in 414), the longitudinal acceleration sensor 134 in the vehicle driving information at that time ) Is excluded. For example, values of vehicle speed, longitudinal acceleration, brake pedal, and accelerator pedal are excluded from vehicle driving information. The detected value of the longitudinal acceleration sensor generated by the stop signal of the traffic light when the distance between the front and rear is greater than the preset distance is a measure of the external driving environment rather than the driving tendency of the driver, and is used to determine the driver's unique driving tendency. Since it is insufficient to be used, it is preferable to exclude it when determining the driving tendency.

In addition, the detection value of the lateral acceleration sensor may be excluded based on the curvature of the road on which the vehicle is running. The curvature of the road can be confirmed through the map information of the navigation 140 installed in the vehicle. If the curvature of the road on which the vehicle is driving is less than or equal to the preset reference curvature Rnavi (YES in 422), the detection value of the lateral acceleration sensor is excluded (426). For example, the values of lateral acceleration, yaw rate and steering angular velocity are excluded. The detected value of the lateral acceleration sensor 132, which occurs when the curvature of the road is greater than the preset reference value (Rnavi), indicates the external driving environment rather than the driving tendency of the driver, and is used as data for determining the driver's unique driving tendency. Since it is not sufficient, it is desirable to exclude it when determining driving propensity.

In addition, the detection value of the damper speed sensor 142 may be excluded based on the presence or absence of a speed bump on the road on which the vehicle is running. The presence or absence of a speed bump on the road can be confirmed through map information of the navigation 140 installed in the vehicle. If there is no speed bump on the road on which the vehicle is driving (No at 432), the detection value of the damper speed sensor 142 is excluded (436). For example, the value of the vertical acceleration of the damper is excluded. The vertical acceleration value of the damper generated by the speed bumps on the road can be seen to indicate the driver's driving tendency when passing through the speed bump, but the change in the vertical acceleration of the damper when the speed bump is not present Represents only the roughness of the road surface (external driving environment). Therefore, it is preferable to exclude the value of the acceleration in the vertical direction of the damper in a place other than the speed bump ('No' of 432) when determining the driver's driving tendency.

In order to standardize vehicle characteristics, it is necessary to convert vehicle driving information into a form necessary for learning a transformation model. An encoder-decoder structure can be used for this conversion. In order to learn the transformation model, a large amount of datasets acquired from a standard vehicle Z in various environments (different driving patterns by multiple drivers with different driving inclinations, various road environments, various weather, etc.) are required.

Learning about the internal parameters of the encoder-decoder of the neural network structure using multiple data sets. At this time, an objective function that receives each data value of vehicle driving information that has passed through the encoder-decoder is used. A discriminator of the neural network structure is used as the objective function, and the identifier enables the transformation model to repeatedly learn the unique characteristics of the standard vehicle Z among a plurality of data sets. Although the transformation model does not extract only the characteristics of the standard vehicle Z, it enables data transformation in which the characteristics of the standard vehicle Z are reflected. Through iterative learning, a transform model of an encoder-decoder structure can be generated.

Through the standardization of vehicle characteristics, it is possible to solve a data deflection problem that may occur due to differences in the type of vehicle driving information acquisition means (ie, vehicle type) and the vehicle driving environment (road, weather, etc.).

<Extraction of statistical characteristics for each section>

5 is a diagram illustrating feature extraction for each section according to an embodiment of the present invention.

As shown in FIG. 5, in the driving propensity classification based on vehicle driving information according to an embodiment of the present invention, a section for extracting statistical characteristics during a predetermined interval at preset time intervals from vehicle driving information from which noise and vehicle characteristics are removed Star feature extraction is performed. When extracting features for each section, each section divided may have a section overlapped with each other.

For example, as illustrated in FIG. 5, statistical characteristics are extracted for a 4 second interval at 4 second intervals, but an overlapping interval of 2 seconds may exist for each interval. Statistical characteristics may include feature vectors, maximum deviations, minimum deviations, and standard deviations. Statistical characteristics are extracted for all sensor detection values to be input, but if necessary, each sensor detection value can have different kinds of statistical characteristics.

<Driving propensity labeling>

6 is a view showing a process of rearranging driver classification results using an error matrix for driving propensity labeling according to an embodiment of the present invention. In the driving propensity labeling step, a clustering algorithm is applied to the extracted statistical characteristics (feature vector) to assist with the labeling for extracting the propensity to drive.

As shown in FIG. 6, in the driving propensity classification according to an embodiment of the present invention, a data set 602 of vehicle driving information including a driver ID to label learning data for learning a neural network based on machine learning is classified as a driver Passing the logic 604 to generate a classification result 606 and rearranging the generated classification result 606 through the error matrix rearrangement logic 608 to obtain the rearranged error matrix 610. The error matrix is represented as an image in which a cluster is expressed through reordering. The rearrangement can use Maciejewski's method or Cuthill-McKee algorithm. In addition, the number of classes to be labeled may be determined by using the image expressed by the cluster, and a driving characteristic unique to the driver may be used as a standard for labeling.

7 is a view showing a driver classification based on a driving tendency according to an embodiment of the present invention.

7(A) is a view showing the result of performing driver classification with one driver (A) having a driving tendency of'COMPORT mode' and two drivers (B, C) having a driving tendency of'SPORT mode'. . In FIG. 7(A), the number without parentheses is a test result, and the number in parentheses is a normalized result for a true value.

As shown in Fig. 7(A), it can be seen that the ratio between the actual driving tendency and the classification result does not coincide and the block is generated between B-Cs. For example, the ratio of classifying drivers who actually tend to B to C and classifying drivers who actually tend to C to B is relatively high, so that a block is formed between B-C.

In this case, the driver's propensity to drive can be classified as'COMPORT-SPORT', and data from A can be used as COMPORT training data and data from B and C can be used as SPORT training data.

FIG. 7(B) shows 1 driver (a) having a driving tendency of'COMPORT mode' and 3 drivers (b, c, d) having a driving tendency of'SPORT mode', 1 having a driving tendency of SPORT+ mode. This is the result of the classification of the driver by the number of drivers (e). In Fig. 7B, numbers without parentheses indicate test results, and numbers in parentheses are normalized results for true values.

As shown in FIG. 7(B), it can be seen that a block is generated between b-c-d having a relatively low ratio in which the actual driving tendency and the classification result do not match.

In this case, the driver's driving tendency can be classified as'COMPORT-SPORT-SPORT+' to use the data of a as the COMPORT learning data, the data of b, c, and d can be used as the SPORT learning data, and the data of e can be used. Can be used as SPORT+ learning data.

<Learning>

The learning logic performs machine learning-based neural network learning using the statistical characteristics (feature vector) extracted in the statistical characteristic extraction step and the driving propensity extracted in the driving propensity labeling step, and generates a driving propensity classifier as a result of the neural network learning. .

The classification logic 214 for classifying the driving propensity may be convolutional neural networks, recurrent neural networks, or the like. The classification logic 214 may be designed by learning a feature vector obtained through noise reduction, vehicle characteristic removal, and feature extraction for each section using a neural network model. In the classification logic 214, sufficient learning is repeatedly performed until the weights converge. The weight in the classification logic 214 is updated according to the training data, and finally has a weight suitable for the training data.

<Driving propensity classification>

The feature vector extracted from the feature extraction logic for each section may be input to the classification logic 214 to obtain a result of the driver's driving propensity classification. A machine learning-based neural network model can be built using the Long Short-Term Memory (LSTM) model. As described above, the long-term dependency problem of RNN can be solved by constructing a machine learning-based neural network model using the LSTM model. In this case, it is determined what information is left using the internal cell state.

8 is a diagram illustrating an example of classification according to a sensor input in classification of driving propensity according to another embodiment of the present invention. 8 shows a graph of the detection curves of the lateral acceleration sensor 132, the longitudinal acceleration sensor 134, the steering angular velocity sensor 136, and the yaw rate sensor 138 of the vehicle. As a model for the classification logic 214, a long short-term memory (LSTM) based on a cyclic neural network as described above may be used.

The task of classifying a driver's driving tendency from detection values of various sensors is performed every 60 seconds, and the classification result of observing the past 120 seconds can be used.

As shown in FIG. 8, most of the sections were predicted as'COMPORT', but large changes were detected in all sensors in the vicinity of 1750 seconds (802) of FIG. 8, and the corresponding section 804 could be classified as'SPORT'. have.

Returning to FIG. 2 again, when the classification logic is completed through the series of learning steps (dotted arrow process in FIG. 2), the mass production vehicle equipped with the classification logic is not a test driver but a real driver (for example, the owner of the vehicle). ) A classification step (solid arrow process in FIG. 2) for classifying and determining a driving propensity may be performed. In the classification step, statistical characteristics are extracted while the vehicle driving information acquired in real time passes through the noise removal logic 204, the vehicle feature standardization logic 206, and the feature extraction logic 208 for each section, and is extracted. The statistical characteristics are classified in the driving logic in the classification logic 214. The classified driving propensity information is transmitted to the driving control part 162 and is used to select an appropriate driving mode suitable for the driving propensity of the driver in the mass-production vehicle and to control driving assistance.

The above description is merely illustrative of the technical idea, and those of ordinary skill in the art of the present invention will be able to make various modifications, changes, and substitutions without departing from essential characteristics. Therefore, the above-described embodiments and the accompanying drawings are not intended to limit the technical spirit, but to explain, and the scope of the technical spirit is not limited by the embodiments and the accompanying drawings. The scope of protection should be interpreted by the claims below, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of rights.

112: driving propensity determination unit
122: database
132: Lateral acceleration sensor
134: longitudinal acceleration sensor
136: steering angle speed sensor
138: yaw rate sensor
140: navigation
142: damper speed sensor
150: sensor group
162: driving control
172: steering
174: engine
176: braking
178: Battlefield
204: noise cancellation logic
206: vehicle feature standardization logic
208: Feature extraction logic for each section
210: driving propensity labeling logic
212: learning logic
214: classification logic

Claims (30)

Extracting statistical characteristics during a predetermined section at a preset time interval from vehicle driving information obtained through driving of the vehicle;
Extracting statistical characteristics from the converted vehicle driving information, and performing labeling of the vehicle driving information to extract a driver's driving tendency from the extracted statistical characteristics;
And generating classification logic for classifying a driving tendency of the driver based on the result of the labeling. According to claim 1,
And converting the vehicle driving information obtained through the driving of the vehicle into a predetermined standard vehicle driving information to exclude the unique characteristics of the vehicle from the vehicle driving information. According to claim 2,
A vehicle control method for distinguishing driving information generated by a driver's operation from driving information generated by a driving environment in order to exclude unique characteristics of the vehicle from the vehicle driving information. According to claim 2,
The driving information generated by the driver's manipulation includes at least one of a damper speed, a vehicle speed, a longitudinal acceleration, a lateral acceleration, and a yaw rate. The method of claim 4,
The driving information generated by the operation of the driver includes a steering angular speed, an accelerator pedal operation amount, and a brake pedal operation amount. The method of claim 5,
A vehicle control method that excludes the unique characteristics of the vehicle by selectively taking some of the damper speed, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angular speed, accelerator pedal operation amount, and brake pedal operation amount. According to claim 1,
In the step of extracting the statistical characteristics, each of the preset sections has a section overlapping each other. The method of claim 7,
The statistical characteristic includes a feature vector, a maximum deviation, a minimum deviation, and a standard deviation. According to claim 1,
A vehicle control method using a machine learning-based neural network learning theory for the labeling. According to claim 1,
A vehicle control method in which an error matrix is used for the labeling. The method of claim 10, wherein the labeling,
Generating a classification result of the vehicle driving information data set through driver classification logic;
And generating the rearranged error matrix by rearranging the generated classification result through error matrix rearrangement logic. The method of claim 11,
The error matrix is a vehicle control method represented by an image in which a cluster is expressed through reordering. According to claim 1,
And acquiring the vehicle driving information through a test driving of the vehicle to construct a database. The method of claim 13,
And receiving the vehicle driving information stored in the database to remove noise included in the vehicle driving information. Feature extraction logic for each section that extracts statistical characteristics during a predetermined section at a preset time interval from vehicle driving information obtained through driving of the vehicle;
Driving propensity labeling logic for extracting statistical characteristics from the converted vehicle driving information and performing labeling of the vehicle driving information to extract a driver's driving propensity from the extracted statistical characteristics;
A vehicle including learning logic that generates classification logic for classifying a driving tendency of a driver based on the result of the labeling. The method of claim 15,
A vehicle further comprising vehicle characteristic standardization logic to exclude the unique characteristics of the vehicle from the vehicle driving information by converting the vehicle driving information obtained through the driving of the vehicle into a predetermined standard vehicle driving information. The method of claim 16, wherein the vehicle feature standardization logic,
In order to exclude the unique characteristics of the vehicle from the vehicle driving information, a vehicle selectively taking the driving information generated by the driver's manipulation and the driving information generated by the driving environment. The method of claim 16,
The vehicle driving information generated by the driver's operation includes at least one of a damper speed, a vehicle speed, a longitudinal acceleration, a lateral acceleration, and a yaw rate. The method of claim 18,
The vehicle driving information generated by the driver's operation includes at least one of a steering angular speed, an accelerator pedal operation amount, and a brake pedal operation amount. The method of claim 19,
A vehicle that excludes the unique characteristics of the vehicle by selectively taking a part of the damper speed, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angular speed, accelerator pedal operation amount, and brake pedal operation amount. 21. The method of claim 20, The feature extraction logic for each section,
In the step of extracting the statistical characteristics, the vehicle so that each of the preset section has a section overlapped with each other. The method of claim 21,
The statistical characteristic is a vehicle including a feature vector, a maximum deviation, a minimum deviation, and a standard deviation. The method of claim 15,
Vehicle applying machine learning based neural network learning theory for the labeling. The method of claim 15,
A vehicle in which an error matrix is used for the labeling. 16. The method of claim 15, wherein the driving propensity labeling logic,
Generating a classification result of the vehicle driving information data set through driver classification logic;
A vehicle that generates a rearranged error matrix by rearranging the generated classification result through an error matrix rearrangement logic. The method of claim 25,
The error matrix is a vehicle represented by an image in which a cluster is expressed through reordering. The method of claim 15,
The vehicle further includes a database that stores the vehicle driving information obtained through the test driving of the vehicle. The method of claim 27,
A vehicle further comprising noise removal logic for removing the noise included in the vehicle driving information by receiving the vehicle driving information stored in the database. Converting the vehicle driving information obtained through the driving of the test vehicle into a predetermined standard vehicle driving information to exclude unique characteristics of the test vehicle from the vehicle driving information of the converted test vehicle;
Extracting statistical characteristics during a predetermined section at predetermined time intervals from the converted vehicle driving information;
Extracting statistical characteristics from the converted vehicle driving information and applying machine learning-based neural network learning theory to the extracted statistical characteristics to perform labeling of the vehicle driving information to extract a driver's driving propensity;
And generating classification logic for classifying a driving tendency of the driver based on the result of the labeling. Converting vehicle driving information obtained through driving of a mass-produced vehicle into a form of driving information of a predetermined standard vehicle to exclude unique characteristics of the mass-produced vehicle from the vehicle driving information of the converted mass-produced vehicle;
Extracting statistical characteristics during a predetermined section at predetermined time intervals from the converted vehicle driving information;
Classifying a driving propensity of a driver driving the mass-produced vehicle using a classification logic provided through labeling of the vehicle driving information to extract a driving propensity of the driver;
And determining an automatic driving mode of the vehicle based on the classified driving propensity.

Images