KR102050426B1 - Autonomous driving control apparatus and method based on driver model - Google Patents

Abstract

The present invention relates to an autonomous driving control apparatus and a method based on a driver simulation model. The autonomous driving control apparatus based on a driver simulation model includes: a surrounding condition collection unit collecting the surrounding conditions of a vehicle; a driver simulation model receiving unit receiving a productive hostile-trained driver simulation model based on a training population including a learning driving situation previously collected by a learning vehicle and a learning driving situation previously performed by the learning driver; a vehicle lane change determination unit determining whether a lane change of the vehicle is necessary; and a driving situation generation unit for providing a driving situation of the vehicle to the driver simulation model to generate a future driving situation, when it is determined that the lane change is necessary.

Description

Autonomous driving control apparatus and method based on driver simulation model {AUTONOMOUS DRIVING CONTROL APPARATUS AND METHOD BASED ON DRIVER MODEL}

The present invention relates to an autonomous driving control technology based on a driver simulation model, and more particularly, to an autonomous driving control device based on a driver simulation model capable of controlling driving based on a real driver model in a lane change situation of an autonomous vehicle. It is about a method.

Currently, research on autonomous driving technology is being actively conducted. Most platforms for autonomous driving are sensor-oriented. The development of sensor-oriented technology has the problems of increased cost due to sensor installation and reduced fuel economy due to increased tolerance weight. Therefore, there is a need to develop a technology that can replace a part of a sensor through an AI-based algorithm using vehicle-specific data.

In addition, since the sensor-oriented technology alone does not have sufficient ability to cope with unexpected situations that can occur in various situations, a technology that can effectively cope with a dangerous situation is required in the development of an advanced autonomous vehicle.

Korean Patent Registration No. 10-0992761 (2010.11.01)

An embodiment of the present invention is to provide an autonomous driving control apparatus and method based on a driver simulation model that can control the driving based on the actual driver model in the lane change situation of the autonomous vehicle.

One embodiment of the present invention is to provide a driver simulation model based autonomous driving control apparatus and method that can provide improved coping ability by utilizing the advanced artificial intelligence driver simulation model for autonomous driving by learning the driving of a variety of drivers in reality do.

An embodiment of the present invention is based on the driver simulation model that can flexibly cope with the lane change by applying the driving situation according to the concession of the surrounding vehicle provided through the driver simulation model in autonomous driving in the lane change situation automatically detected. An autonomous driving control apparatus and method are provided.

Among the embodiments, the autonomous driving control apparatus based on the driver simulation model may include an environment collection unit for collecting the environment of the vehicle, a learning driving situation previously collected by the learning vehicle, and a learning driving situation previously performed by the learning driver. The driver simulation model receiver for receiving a productive hostile-trained driver simulation model based on a learning population including a vehicle lane change determination unit and a lane change determination unit configured to determine a need for cut-in of the vehicle; If it is determined that it is necessary to include a driving situation generation unit for providing a driving situation of the vehicle to the driver simulation model to generate a future driving situation.

The surrounding situation collecting unit may collect, as the surrounding situation, driving conditions related to front, side, and rear of the vehicle and driving conditions related to steering, deceleration, and acceleration of the vehicle through at least one radar sensor.

The driver simulation model may be generated by applying the preprocessing process of performing synchronization with the learning driving situation and the learning driving condition and normalizing the learning driving condition with respect to the learning population, and then performing the productive hostile learning.

The driver simulation model may generate a candidate driving situation for the learning driving situation as a first learning process of the productive hostile learning, and generate a probability of discriminating the candidate driving situation as a second learning process.

The driver simulation model may reflect the candidate driving situation that has been successfully determined through the second learning process in the first learning process, and may reflect the candidate driving situation that has failed the discrimination in the second learning process.

The driving situation generation unit may generate the future driving situation by applying the weight of the first learning process as it is when the determination success rate of the second learning process exceeds a reference threshold value with respect to the driver simulation model.

If it is determined that the lane change is necessary, the driving situation generation unit may predict whether or not concession of a neighboring vehicle traveling near the vehicle in the target lane of the lane change is provided to the driver simulation model together with the driving situation of the vehicle. have.

The autonomous driving control device may further include an autonomous driving control unit that controls autonomous driving related to lane change of the vehicle based on the future driving situation.

The autonomous driving controller selects any one of a plurality of future driving conditions for the overlapping section when the future driving situation is continuously generated based on a specific time interval and overlapping sections between a plurality of future driving situations occur. Applicable to the autonomous driving.

Among the embodiments, the autonomous driving control method based on the driver simulation model includes collecting the surrounding situation of the vehicle, the learning driving situation previously collected by the learning vehicle and the learning driving situation previously performed by the learning driver. Receiving a productive hostile-trained driver simulation model based on a training population, determining whether the vehicle needs a cut-in, and determining that the lane change is necessary, Providing the driver simulation model to generate a future driving situation.

The disclosed technique can have the following effects. However, since a specific embodiment does not mean to include all of the following effects or only the following effects, it should not be understood that the scope of the disclosed technology is limited by this.

The autonomous driving control apparatus and method based on the driver simulation model according to an embodiment of the present invention can provide improved coping ability by utilizing advanced AI driver simulation models for learning by actually driving various drivers.

Autonomous driving control apparatus and method based on the driver simulation model according to an embodiment of the present invention by applying the driving situation according to the concession of the surrounding vehicle provided through the driver simulation model in the detected lane change situation to autonomous driving Flexibility to deal with lane changes

1 is a diagram illustrating an autonomous driving control system based on a driver simulation model according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating a functional configuration of the autonomous driving control device in FIG. 1.
3 is a flowchart illustrating an autonomous driving control process performed by the autonomous driving control apparatus of FIG. 1.
4 and 5 are exemplary diagrams illustrating a process of productive hostile learning for generating a driver simulation model.
FIG. 6 is an exemplary view illustrating a process of performing autonomous driving control through a driver simulation model in the autonomous driving control apparatus of FIG. 1.

Description of the present invention is only an embodiment for structural or functional description, the scope of the present invention should not be construed as limited by the embodiments described in the text. That is, since the embodiments may be variously modified and may have various forms, the scope of the present invention should be understood to include equivalents capable of realizing the technical idea. In addition, the objects or effects presented in the present invention does not mean that a specific embodiment should include all or only such effects, the scope of the present invention should not be understood as being limited thereby.

On the other hand, the meaning of the terms described in the present application should be understood as follows.

Terms such as "first" and "second" are intended to distinguish one component from another component, and the scope of rights should not be limited by these terms. For example, the first component may be named a second component, and similarly, the second component may also be named a first component.

When a component is referred to as being "connected" to another component, it should be understood that there may be other components in between, although it may be directly connected to the other component. On the other hand, when a component is referred to as being "directly connected" to another component, it should be understood that there is no other component in between. On the other hand, other expressions describing the relationship between the components, such as "between" and "immediately between" or "neighboring to" and "directly neighboring to", should be interpreted as well.

Singular expressions should be understood to include plural expressions unless the context clearly indicates otherwise, and terms such as "comprise" or "have" refer to a feature, number, step, operation, component, part, or feature thereof. It is to be understood that the combination is intended to be present and does not exclude in advance the possibility of the presence or addition of one or more other features or numbers, steps, operations, components, parts or combinations thereof.

In each step, an identification code (e.g., a, b, c, etc.) is used for convenience of description, and the identification code does not describe the order of the steps, and each step clearly indicates a specific order in context. Unless stated otherwise, they may occur out of the order noted. That is, each step may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

The present invention can be embodied as computer readable code on a computer readable recording medium, and the computer readable recording medium includes all kinds of recording devices in which data can be read by a computer system. . Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As used herein, "vehicle", "vehicular" or other similar terminology refers to automobiles, passenger automobiles, buses, generally including sport utility vehicles (SUVs). , Trucks, various commercial vehicles, ships including various boats and ships, airplanes, and the like, and hybrid vehicles, electric vehicles, hybrid electric vehicles, hydrogen powered vehicles and other alternative fuels (eg, petroleum). Fuels derived from non-resource resources). As mentioned herein, an electric vehicle (EV), as part of its locomotion capabilities, is a rechargeable energy storage device (eg, one or more rechargeable electrochemical cells or other types of batteries). It is a vehicle containing the electric power obtained from. EV is not limited to automobiles and may include motorcycles, carts, scooters, and the like. In addition, a hybrid vehicle is a vehicle having two or more power sources, for example gasoline based power and electric based power (eg, hybrid electric vehicle (HEV)).

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. Generally, the terms defined in the dictionary used are to be interpreted to coincide with the meanings in the context of the related art, and should not be interpreted as having ideal or excessively formal meanings unless clearly defined in the present application.

1 is a diagram illustrating an autonomous driving control system based on a driver simulation model according to an embodiment of the present invention.

Referring to FIG. 1, an autonomous driving control system 100 based on a driver simulation model may include an autonomous vehicle 110, an autonomous driving control device 130, and a database 150.

The autonomous vehicle 110 may correspond to a vehicle as a means of transporting passengers or cargo using power generated by the engine. The autonomous vehicle 110 is not necessarily limited thereto, and may include various vehicles that can move by using power. In one embodiment, the autonomous vehicle 110 may be implemented to include a plurality of sensors that can collect the relevant data in order to monitor various situations occurring during driving. For example, the autonomous vehicle 110 may include sensors for acquiring driving data such as steering, acceleration, and deceleration of the vehicle, and a situation such as the position and relative speed of the surrounding vehicle as the surrounding situation of the vehicle. And sensors for acquiring data.

In addition, the autonomous vehicle 110 may include a communication device capable of communicating with an external system, may be connected to the autonomous driving control device 130 through a network, the plurality of autonomous vehicles 110 It may be connected to the autonomous driving control device 130 at the same time.

The autonomous driving control device 130 uses the driver simulation model generated as a result of learning in advance in order to effectively cope with the interruption (or lane change) situation occurring in the autonomous driving process of the autonomous vehicle 110. The server 110 may be implemented as a server corresponding to a computer or a program capable of controlling the driving of the vehicle 110. The autonomous driving control device 130 may be connected to the autonomous vehicle 110 and a wired network or a wireless network such as Bluetooth, WiFi, or the like, and may communicate with the autonomous vehicle 110 through a wired or wireless network.

In an embodiment, the autonomous driving control device 130 may store information necessary for driving control of the autonomous driving vehicle 110 in association with the database 150. On the other hand, the autonomous driving control device 130 may be implemented by including a database 150 therein, unlike FIG. In addition, the autonomous driving control device 130 may be implemented as a physical component including a processor, a memory, a user input / output unit, and a network input / output unit.

The database 150 may store various pieces of information required by the autonomous driving control device 130 to control driving of the autonomous vehicle 110 using the driver simulation model. For example, the database 150 may store information about the surrounding situation in the driving process from the autonomous vehicle 110, and may store information about the driver simulation model generated as a result of learning in advance. Not limited to this, information collected or processed in various forms may be stored in the process of controlling the driving of the autonomous vehicle 110 using the driver simulation model.

FIG. 2 is a block diagram illustrating a functional configuration of the autonomous driving control device in FIG. 1.

Referring to FIG. 2, the autonomous driving control device 130 includes a surrounding situation collecting unit 210, a driver simulation model receiving unit 220, a vehicle lane change determining unit 230, a driving situation generating unit 240, and an autonomous driving control unit. 250 and the controller 260 may be included.

The surrounding situation collecting unit 210 may collect the surrounding situation of the vehicle. The surrounding situation collection unit 210 may collect data on the surrounding situation from a plurality of sensors installed in the vehicle, and may process the data into a form required for autonomous driving of the vehicle through preprocessing of the collected data. The surrounding situation collecting unit 210 may store the collected information in association with the database 150 and perform classification or sorting as necessary.

In one embodiment, the surroundings collection unit 210 is a vehicle steering situation measured from the plurality of detection sensors and the driving situation regarding the front, side and rear of the vehicle measured by the at least one radar sensor, Operational conditions relating to deceleration and acceleration can be collected as ambient conditions. In particular, the surrounding situation collecting unit 210 may collect the driving situation and the driving situation occurring in the lane change (or interrupt) situation as the surrounding situation of the vehicle. The driving situation may be collected through radar sensors installed in the vehicle as situation information generated during driving of the vehicle, and classified into front, side, and rear sides according to the position and the measurement area attached to the vehicle. The driving situation may correspond to control data, such as steering wheel angle, throttle, and brake pressure, which are directly generated during the driving process of the vehicle, and include a plurality of detection sensors installed in the vehicle. Can be measured through.

The driver simulation model receiving unit 220 may receive a productive hostile-learned driver simulation model based on a learning population including a learning driving situation previously collected by the learning vehicle and a learning driving situation previously performed by the learning driver. have. Here, the productive antagonistic learning (GAN) corresponds to a learning method in which the generator model that generates the data and the discriminator model that compares the generated data with the actual data to discriminate and advance the data are advanced. Can be. The discriminator model may use a learning population composed of a learning driving situation and a learning driving situation as a comparison target for determining data output by the generator model. As a result of the productive hostile learning, the generator model and the discriminator model may be generated, respectively, and the weight information of the generator model may be applied to the driver simulation model.

In one embodiment, the driver simulation model may be generated by applying a preprocessing process that performs synchronization with the learning driving situation and the learning driving condition and normalizes the learning driving condition to the learning population, and then performs productive hostile learning. The driver simulation model can be generated as a result of learning the driving situation and the driving situation occurring in the actual driving situation, and a preprocessing process for the collected information can be applied as a preliminary step for learning. The preprocessing process can be largely performed in two forms, and the synchronization process for synchronizing the sync of the driving situation and the driving situation and the normalization process for the driving situation can be performed to improve the performance of the learning algorithm.

More specifically, the synchronization process for driving and driving conditions is to process and store driving conditions corresponding to radar data at 20 fps, and to control the driving conditions corresponding to driver control data in a time series of 20 fps (0.05 seconds). And storing them by classifying the braking pressure and then synchronizing the sync of the driving and driving conditions to generate one data set (for example, one pickle type). The normalization process for the driving situation may correspond to an operation of converting the range of each sensor data corresponding to the driving situation into a predetermined range, for example, a value between -1 and 1.

In one embodiment, the driver simulation model may generate a candidate driving situation for the learning driving situation as a first learning process of productive hostile learning, and generate a probability of discriminating the candidate driving situation as a second learning process. Productive antagonistic learning can be performed through the process of interaction between the generator model and the discriminator model, the first learning process corresponds to the learning process for generating the generator model, and the second learning process for learning the discriminator model. It may correspond to.

That is, the first learning process may generate a candidate driving situation for the learning driving situation included in the learning population. The candidate driving situation may correspond to a temporary driving situation generated by the generator model during the productive hostile learning process. In addition, the second learning process may generate a discrimination probability by comparing the candidate driving situation generated by the first learning process with the learning driving situation corresponding to the learning driving situation. Here, the discrimination probability may correspond to a degree in which the candidate driving situation generated by the generator model is similar to the actual driving situation, and the discrimination probability may be higher as the candidate driving situation is similar to the actual driving situation.

In one embodiment, the first learning process of the driver simulation model may generate the driving situation (eg, driving control data) for the next t seconds based on the driving situation (eg, radar data) for t seconds. Can be. The second learning process of the driver simulation model may generate a discrimination probability for the corresponding driving situation by comparing the driving situation for t seconds with the actual driving situation at the same time. As a result of the second learning process, the driver simulation model ends when the probability that the discriminator model succeeds in determining is close to a specific criterion (for example, 0.5) and ends the learning and is generated based on the weight of the generator model at that time. Can be.

In an embodiment, the driver simulation model may reflect the candidate driving situation that has been successfully determined through the second learning process in the first learning process, and may reflect the candidate driving situation that has failed the discrimination in the second learning process. If the determination probability according to the second learning process is higher than the reference probability, the determination may be determined to be successful. For example, if the determination probability is higher than 0.5, the determination may be determined to be successful. The driver simulation model reflects the candidate driving situation that has been successfully determined through the second learning process in the first learning process, and the feedback driving loss that has been missed through the iterative learning process is reflected in the second learning process. ) May be performed in a direction that minimizes. The driver simulation model may use a loss function for calculating a feedback loss, and a loss function value may be transmitted to each learning process to update a weight applied to the learning process.

The vehicle lane change determination unit 230 may determine whether the vehicle lane change is necessary. The vehicle lane change determining unit 230 may determine whether the vehicle lane change is necessary based on the surrounding situation information of the vehicle collected by the surrounding situation collecting unit 210. For example, when the preceding vehicle exists in front of the vehicle and the preceding vehicle can be overtaken through the left or right lane of the driving lane, the vehicle lane change determination unit 230 may determine that the lane change is necessary. Whether the lane change is required may be expressed as a numerical value regarding the degree of necessity, and may be determined based on a result of comparison with a specific reference value.

If it is determined that the lane change is necessary, the driving situation generation unit 240 may provide the driving situation of the vehicle to the driver simulation model to generate a future driving situation. The driving situation generation unit 240 may provide the driving situation for the past t seconds from the current time point as an input to the driver simulation model, and the driver simulation model provides the driving state for the next t seconds from the current time point corresponding to the driving situation. It can be output as a future operating situation.

In one embodiment, the driving situation generation unit 240 generates a driving situation in the future by applying the weight of the first learning process as it is when the determination success rate by the second learning process exceeds the reference threshold for the driver simulation model. can do. Productive hostile learning may be repeatedly performed in a direction in which feedback loss is minimized until the discriminant success rate by the second learning process exceeds the reference threshold, and the discriminant success rate by the second learning process exceeds the reference threshold. The weight applied to the producer model at this point can be applied to the driver simulation model. The driving situation generation unit 240 may generate a future driving situation corresponding to the current driving situation through the driver simulation model.

In one embodiment, when it is determined that the lane change is necessary, the driving situation generation unit 240 predicts the concession of the surrounding vehicle that is driving in close proximity to the vehicle in the target lane of the lane change, and applies the driving simulation model to the driver simulation model. Can provide. The driver simulation model can determine the future driving situation of the vehicle by reflecting not only the driving situation of the vehicle but also the concession of surrounding vehicles. When it is determined that the lane change is necessary, the driving situation generator 240 may primarily determine the direction of the lane change, and may determine a trailing vehicle that is driving in the lane according to the determined lane change direction, and On the basis of the distance and the relative speed of the trailing vehicle, the concession of the trailing vehicle can be determined secondarily. The driving situation generation unit 240 may provide the driver simulation model with information about whether a trailing vehicle is yielded to the driver simulation model, and the driver simulation model may determine a future driving situation based on this.

In one embodiment, the driving situation generation unit 240 is a lane change direction determination module for determining the direction of the lane change based on the driving situation of the vehicle, the target lane for determining the target lane of the lane change based on the direction of the lane change Determination module, Peripheral vehicle determination module for determining the surrounding vehicle running in close proximity to the vehicle in the target lane of lane change, Peripheral vehicle driving status collection module for collecting the driving situation of the surrounding vehicle and Peripheral vehicle based on the driving situation of the surrounding vehicle It may include a concession prediction module for predicting the concession of.

The autonomous driving control unit 250 may control autonomous driving regarding lane change of the vehicle based on future driving conditions. The autonomous driving control unit 250 may control autonomous driving of the vehicle based on a future driving situation generated by the driving situation generation unit 240 in a lane change state of the vehicle. That is, the autonomous driving control unit 250 may control autonomous driving of the vehicle by controlling steering, acceleration, and deceleration of the vehicle by using information about the steering angle, the throttle, and the braking pressure of the vehicle included in the future driving situation.

In an embodiment, the autonomous driving control unit 250 may generate a plurality of future driving situations for the overlapping section when the future driving situation is continuously generated based on a specific time interval and an overlapping section between the plurality of future driving situations occurs. Any one can be selected and applied to autonomous driving. The driving situation generating unit 240 may solve the unnatural flow of the future driving situation generated by applying a time window sliding method at a predetermined time interval in consideration of the real time of autonomous driving.

For example, the driving situation generator 240 may continuously generate a future driving situation every 1 second, and the generated driving situation may correspond to a driving situation for 3 seconds based on the current time point. . Therefore, the plurality of future driving conditions generated by the driving situation generating unit 240 are each two in a section between 1 second and 2 seconds, between 2 seconds and 3 seconds, and between 3 seconds and 4 seconds, based on the current time point. , Three and two future operating situations may overlap each other. The autonomous driving control unit 250 may select any one of a plurality of future driving situations for sections overlapping each other and use the autonomous driving of the vehicle.

In one embodiment, the autonomous driving control unit 250 may calculate the average of the plurality of future driving conditions for the overlapping section and apply the autonomous driving when the overlapping section between the plurality of future driving situations occurs. That is, the autonomous driving control unit 250 may calculate an average of each of the steering angle, throttle, and braking pressures included in the corresponding driving conditions in a section where two or more driving conditions overlap, and the average steering angle and average Autonomous driving control for the vehicle can be performed based on the throttle and average braking pressure.

The controller 260 controls the overall operation of the autonomous driving control device 130, and includes a surrounding situation collecting unit 210, a driver simulation model receiving unit 220, a vehicle lane change determining unit 230, and a driving situation generating unit 240. ) And control flow or data flow between the autonomous driving control unit 250.

3 is a flowchart illustrating an autonomous driving control process performed by the autonomous driving control apparatus of FIG. 1.

Referring to FIG. 3, the autonomous driving control device 130 may collect the surrounding situation of the vehicle through the surrounding situation collecting unit 210 (step S310). The autonomous driving control device 130 is productive antagonist based on a learning population including a learning driving situation previously collected by the learning vehicle through the driver simulation model receiver 220 and a learning driving situation previously performed by the learning driver. The trained driver simulation model may be received (step S330). The autonomous driving control device 130 may determine whether the vehicle lane change is necessary through the vehicle lane change determination unit 230 (step S350).

When the autonomous driving control device 130 determines that the lane change is necessary through the driving condition generation unit 240, the autonomous driving control device 130 may generate the driving situation in the future by providing the driving situation of the vehicle to the driver simulation model (step S370). The autonomous driving control device 130 may control the autonomous driving related to the lane change of the vehicle based on the future driving situation through the autonomous driving control unit 250 (step S390).

4 and 5 are exemplary diagrams illustrating a process of productive hostile learning for generating a driver simulation model.

Referring to FIG. 4, productive antagonistic learning may be performed through a first learning process for generating a generator model and a second learning process for generating a discriminator model. The first learning process may be performed as a process of outputting a candidate driving situation (control data) corresponding to the control data of the vehicle by inputting information on the driving situation of the vehicle (sensor data) and the intention to yield the surrounding vehicles. The second learning process may be performed as a process of outputting a discrimination probability (validity) by inputting a candidate driving situation generated in the first learning process. In the second learning process, the collected sensor data and the actual driving situation (control data) may be used for comparison with the candidate driving situation. Productive hostile learning may be repeatedly performed through the interaction between the first learning process and the second learning process, and the weight applied to each learning process may be updated in a direction in which the feedback loss is minimized.

Referring to FIG. 5, the learning population used for the productive hostile learning may be composed of a driving situation and a driving situation regarding a vehicle. The driving situation may correspond to sensor data measured through the radar as information on the external condition of the vehicle, and the driving condition may correspond to control data measured through driving driven by the actual driver as information about the internal condition of the vehicle. Can be. In one embodiment, the pre-processing process may be applied to the learning population, and the pre-processing process may be processed by being divided into synchronization with the driving situation and the driving condition and normalization with respect to the driving condition.

Productive hostile learning may be repeatedly performed through interaction between a first learning process for generating a generator model and a second learning process for generating a discriminator model. In particular, productive hostile learning may be repeated in a direction in which feedback loss is minimized until the probability of success of discrimination by the discriminator model satisfies a specific criterion. If the discriminant success probability by the discriminant model satisfies a certain criterion and the training is completed, the weight applied to the generator model as a result of the first learning process may be stored as a result of the productive hostile learning, and is applied to the driver simulation model as it is Can be used for autonomous driving control.

FIG. 6 is an exemplary view illustrating a process of performing autonomous driving control through a driver simulation model in the autonomous driving control apparatus of FIG. 1.

Referring to FIG. 6, the autonomous driving control device 130 may collect the surrounding situation of the vehicle through the surrounding situation collecting unit 210. The surrounding situation may include driving conditions of the vehicle, and the driving situation may include front, side, and rear data of the vehicle measured by the radar. The autonomous driving control device 130 may predict whether or not concession of the surrounding vehicle is made through the driving situation generator 240 when it is determined whether the vehicle lane change is necessary through the vehicle lane change determination unit 230. The driving situation generator 240 may predict whether or not concessions are made based on the relative speed of the surrounding vehicles and the distance to the surrounding vehicles. The autonomous driving control device 130 may generate a future driving situation suitable for the concession of the surrounding vehicle through the driver simulation model by inputting the driving situation of the vehicle and the existence of the concession of the surrounding vehicle. The autonomous driving control device 130 may control the autonomous driving related to the lane change of the vehicle through the autonomous driving control unit 250 based on a future driving situation.

Although described above with reference to a preferred embodiment of the present invention, those skilled in the art will be variously modified and changed within the scope of the invention without departing from the spirit and scope of the invention described in the claims below I can understand that you can.

100: Autonomous driving control system based on driver simulation model
110: autonomous vehicle 130: autonomous driving control device
150: database
210: surrounding situation collecting unit 220: driver simulation model receiving unit
230: vehicle lane change determination unit 240: driving situation generation unit
250: autonomous driving control unit 260: control unit

Claims (10)

Surrounding situation collecting unit for collecting the surrounding situation of the vehicle;
A driver simulation model receiver for receiving a productive hostile-trained driver simulation model based on a learning population including a learning driving situation previously collected by the learning vehicle and a learning driving situation previously performed by the learning driver;
A vehicle lane change determination unit determining whether a lane change of the vehicle is necessary; And
If it is determined that the lane change is necessary to include a driving situation generation unit for providing a driving situation of the vehicle to the driver simulation model to generate a future driving situation,
The driver simulation model generates a candidate driving situation for the learning driving situation as a first learning process of the productive hostile learning, and generates a discrimination probability for the candidate driving situation as a second learning process,
An autonomous driving control apparatus based on a driver simulation model, characterized in that the candidate driving situation that has been discriminated through the second learning process is reflected in the first learning process and the candidate driving situation that has failed the determination is reflected in the second learning process. .
The method of claim 1, wherein the ambient situation collecting unit
Based on at least one radar sensor, driving conditions related to the front, side, and rear of the vehicle and driving conditions related to steering, deceleration, and acceleration of the vehicle are collected as the surrounding situation. Autonomous driving control device.
The method of claim 1, wherein the driver simulation model
Based on the driver simulation model, the productive host learning is generated after applying the preprocessing process of performing synchronization for the learning driving situation, the learning driving situation, and normalization of the learning driving situation to the learning population. Autonomous driving control device.
delete delete The method of claim 1, wherein the driving situation generating unit
Based on the driver simulation model, the future driving situation is generated by applying the weight of the first learning process as it is when the determination success rate of the second learning process exceeds the reference threshold value. Autonomous driving control device.
The method of claim 1, wherein the driving situation generating unit
If it is determined that the lane change is necessary, the driver's simulation of predicting the concession of the surrounding vehicle traveling near the vehicle in the target lane of the lane change and providing the driver simulation model together with the driving situation of the vehicle is performed. Model-based autonomous driving control device.
The method of claim 1,
An autonomous driving control apparatus based on a driver simulation model, further comprising an autonomous driving control unit for controlling autonomous driving related to the lane change of the vehicle based on the future driving situation.
The method of claim 8, wherein the autonomous driving control unit
When the future driving situation is continuously generated based on a specific time interval and an overlapping section between a plurality of future driving situations occurs, one of a plurality of future driving situations is selected for the overlapping section and applied to the autonomous driving. Autonomous driving control device based on the driver simulation model, characterized in that.
In the autonomous driving control method performed in the autonomous driving control device,
Collecting ambient conditions of the vehicle;
Receiving a productive hostile-learned driver simulation model based on a learning population comprising a learning driving situation previously collected by the learning vehicle and a learning driving situation previously performed by the learning driver;
Determining whether the vehicle needs to be cut-in; And
If it is determined that the lane change is necessary, providing a driving situation of the vehicle to the driver simulation model to generate a future driving situation,
The driver simulation model generates a candidate driving situation for the learning driving situation as a first learning process of the productive hostile learning, and generates a discrimination probability for the candidate driving situation as a second learning process,
Autonomous driving control method based on a driver simulation model, characterized in that the candidate driving situation that has been successfully determined through the second learning process is reflected in the first learning process and the candidate driving situation that has failed the determination is reflected in the second learning process. .

Images