KR102274806B1 - Method for automated parking and track generation method for automated parking - Google Patents

Abstract

Disclosed are an automatic parking method using an artificial neural network, and a trajectory generating method for automatic parking. The disclosed automatic parking method comprises: a step of generating a parking trajectory from a first location of a vehicle to a second location spaced at a preset distance from the parking space; a step of estimating the location and a size of an obstacle by using a front image of the vehicle and a first artificial neural network while moved along the parking trajectory; a step of determining whether the obstacle collides with the vehicle by using a second artificial neural network according to the estimated location and size; and a step of updating the parking trajectory according to a determination result.

Description

METHOD FOR AUTOMATED PARKING AND TRACK GENERATION METHOD FOR AUTOMATED PARKING

The present invention relates to an automatic parking method and a trajectory generating method for automatic parking.

Recently, a driver assistance system (ADAS) has been installed in many vehicles, and for a driver having difficulty in parking, a driver assistance system supporting automatic parking is being used.

It is common to use a vision sensor or an ultrasonic sensor mounted on a vehicle for automatic parking. It is a method in which parking is made in a parking space while searching for a parking space using a vision sensor and an ultrasonic sensor and avoiding a collision with a nearby vehicle.

Recently, research on an automatic parking method using more various sensors or using a wireless positioning method is being conducted.

As related prior literature, there are Korean Patent Registration Nos. 10-1970920, 10-1655290, and 10-1593839, which are patent documents.

The present invention is to provide an automatic parking method using an artificial neural network.

Another object of the present invention is to provide a trajectory generating method for automatic parking capable of generating an automatic parking trajectory with high accuracy through positioning of a vehicle and a parking space.

According to an embodiment of the present invention for achieving the above object, the method comprising: generating a parking trajectory from a first location of a vehicle to a second location spaced apart from a parking space by a preset distance; estimating the position and size of an obstacle by using a front image of the vehicle and a first artificial neural network while moving along the parking trajectory; determining whether the obstacle collides with the vehicle using a second artificial neural network according to the estimated location and size; and updating the parking trajectory according to the determination result.

In addition, according to another embodiment of the present invention for achieving the above object, generating a parking trajectory from a first position of the vehicle to a second position spaced apart by a preset distance from the parking space; moving to the second position along the parking trajectory; returning to the first position when an obstacle is detected during movement; and generating a corrected parking trajectory by updating the parking trajectory.

In addition, according to another embodiment of the present invention for achieving the above object, the method comprising: obtaining a plurality of coordinate values for a first position of a vehicle; generating a first trajectory from the first location to the second location by using the plurality of coordinate values and a coordinate value for a second location spaced apart by a preset distance from the parking space; and generating a second trajectory from the second position to a third position corresponding to the parking space.

According to an embodiment of the present invention, it is possible to reduce the risk of collision with an obstacle that may occur in the automatic parking process by more accurately detecting an obstacle using an artificial neural network.

In addition, according to an embodiment of the present invention, a more accurate automatic parking trajectory from the current location of the vehicle to the parking space is generated by generating an automatic parking trajectory based on the coordinate values for the vehicle and the parking space estimated through positioning. can do.

1 is a view for explaining an automatic parking system according to an embodiment of the present invention.
2 is a view for explaining an automatic parking method according to an embodiment of the present invention.
3 is a view for explaining an automatic parking method according to another embodiment of the present invention.
4 is a view for explaining a method for generating a trajectory for automatic parking according to an embodiment of the present invention.
5 is a view for explaining a trajectory for automatic parking according to an embodiment of the present invention.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention. In describing each figure, like reference numerals have been used for like elements.

The present invention proposes an automatic parking method utilizing machine learning.

According to an embodiment of the present invention, machine learning is utilized to detect obstacles present on the automatic parking trajectory of a vehicle. In particular, an embodiment of the present invention uses an artificial neural network that detects an obstacle and estimates the location and size of the obstacle and an artificial neural network that determines whether an obstacle collides with a vehicle, so that a lighter artificial neural network can be used. , the cost or time required for learning and inference of the artificial neural network can be reduced.

In addition, the present invention proposes a method of generating an automatic parking trajectory by using coordinate values for a vehicle and a parking space.

An embodiment of the present invention generates an automatic parking trajectory based on the coordinate values for the vehicle and the parking space estimated through positioning, so that a more accurate automatic parking trajectory can be generated. As an embodiment, the positioning may be performed by a UWB Real-Time Location System (RLTS), an inertial sensor, a vision sensor, etc. mounted on a vehicle.

The automatic parking method and the method for generating a trajectory for automatic parking according to an embodiment of the present invention may be performed in a computing device including a processor and a memory, and such a computing device may be mounted and operated in a general vehicle or an autonomous driving vehicle. have.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a view for explaining an automatic parking system according to an embodiment of the present invention.

Referring to FIG. 1 , an automatic parking system according to an embodiment of the present invention includes a vehicle 110 , a UWB tag, and anchors 191 to 194 .

The UWB tag is mounted on the vehicle 110 , and the vehicle 110 may estimate the location of the vehicle 110 using a radio signal transmitted/received between the UWB tag and an anchor. In addition, the UWB tag may be disposed in each parking space, and the automatic parking system may estimate the location of the parking space by using a wireless signal transmitted and received between the UWB tag and the anchor of the parking space.

When the location of the parked vehicle corresponds to the location of the parking space, it may be determined that the parking space has already been parked, and when the location of the parked vehicle does not correspond to the location of the parking space, the corresponding parking space is parked can be considered as a possible space. In the parking area 130 of FIG. 1 , a hatched parking space indicates a space in which parking has already been made, and a non-shaded parking space indicates a space where parking is possible.

The self-parking vehicle 110 may be parked in a parking space in two steps. The self-parking vehicle 110 may first move from a first position, which is a current position where the vehicle is stopped, to a second position 120 , and then move from the second position 120 to a third position. The second position 120 is a position spaced apart from the parking space 131 by a preset distance, and the third position is a position corresponding to the parking space 131 . The first position is a position more spaced apart from the parking space 131 than the second position 120 , and the automatic parking function is a position where the automatic parking vehicle 110 is further spaced apart from the parking space 131 than the second position 120 . It may be activated when it is stopped at the first position. That is, when the self-parking vehicle 110 stops at the first position, it may generate a parking trajectory and move from the first position through the second position 120 to the third position.

The self-parking vehicle 110 may determine whether the vehicle is stopped at the first position by using the estimated vehicle position and the position of the parking space 131 . When the estimated location of the vehicle is spaced apart from the location of the parking space 131 by a preset distance, it may be determined that the autonomously parked vehicle 110 is stopped at the first location. The first and second positions may be variously set according to the size of the parking lot, the size of the parking space, the type of vehicle, and the like.

After generating a trajectory for automatic parking, the self-parking vehicle 110 may move from the first position to the second position 120 and from the second position 120 to the third position. In the process of moving from the first position to the second position 120 , the self-parking vehicle 110 detects an obstacle. When an obstacle is detected, the self-parking trajectory is updated and according to the updated trajectory, the self-parking vehicle 110 is moved from the first position to the second position. Go to (120).

The self-parking vehicle 110 may use an artificial neural network to detect obstacles. A pre-learned artificial neural network may be mounted on the autonomous parking vehicle 110 , and the autonomous parking vehicle 110 may estimate the location and size of an obstacle using the artificial neural network and determine whether to collide with the obstacle.

2 is a view for explaining an automatic parking method according to an embodiment of the present invention.

Referring to FIG. 2 , the self-parking vehicle according to an embodiment of the present invention generates a parking trajectory from a first position of the vehicle to a second position spaced apart by a preset distance from the parking space (S210), and parks the vehicle. move along the trajectory. A method of generating a parking trajectory is described in more detail with reference to FIGS. 4 and 5 .

The autonomously parked vehicle estimates the position and size of the obstacle by using the front image of the vehicle and the first artificial neural network while moving along the parking trajectory ( S220 ). As an embodiment, the first artificial neural network may be a CNN.

The first artificial neural network may be learned based on images including various obstacles and training data in which positions and sizes of obstacles included in the images are labeled, and when the front image of the vehicle is input to the first artificial neural network, the first The artificial neural network may infer the position and size of the obstacle included in the forward image.

The self-parking vehicle determines whether or not the vehicle collides with an obstacle by using the second artificial neural network according to the location and size estimated in step S210 (S230). When the estimated obstacle is far away from the parking trajectory or the estimated obstacle is very small, the probability of collision is significantly reduced, so the autonomous parking vehicle does not determine whether the vehicle collides with the obstacle. However, when the location of the estimated obstacle is close to the parking trajectory or the size of the estimated obstacle is large, the probability of collision increases. Therefore, the autonomous parking vehicle uses a second artificial neural network to determine the exact probability of collision between the obstacle and the vehicle. .

The second artificial neural network can be learned based on training data labeled with various positions and sizes of obstacles and whether or not they collide. The position and size of obstacles inferred through the first artificial neural network are input to the second artificial neural network. If so, the second artificial neural network can infer whether the vehicle collides with the obstacle.

According to an embodiment of the present invention, it is possible to more accurately detect an obstacle using an artificial neural network, thereby reducing the risk of collision with an obstacle that may occur in the automatic parking process.

In particular, an embodiment of the present invention distinguishes between an artificial neural network for detecting obstacles and an artificial neural network for detecting whether a vehicle collides, thereby reducing the cost and time required for learning and reasoning through weight reduction of the artificial neural network. can In addition, the automatic parking time can be shortened by selectively using an artificial neural network to determine whether a collision exists according to the estimated location and size of the obstacle.

Meanwhile, in step S230 , when the obstacle is an obstacle in a stopped state, it may be determined whether the obstacle collides with the vehicle. In the case of a dynamic obstacle, such as a vehicle or a person, while moving between the parking space and the self-parking vehicle, there is a high probability that the vehicle deviates from the parking trajectory, so the self-parking vehicle determines whether the vehicle collides with the vehicle for the static obstacle. When the position of the obstacle estimated in step S210 does not change with time, the autonomous parking vehicle may determine the detected obstacle as an obstacle in a stopped state.

The self-parking vehicle updates the parking trajectory according to the determination result of step S230 (S240). When an obstacle determined to be collided is detected, the self-parking vehicle may update a parking trajectory to avoid the obstacle, and may move to a parking space along the updated parking trajectory.

3 is a view for explaining an automatic parking method according to another embodiment of the present invention.

Referring to FIG. 3 , the self-parking vehicle according to an embodiment of the present invention generates a parking trajectory from a first position of the vehicle to a second position spaced apart by a preset distance from the parking space (S310), It moves to the second position along the trajectory (S320). When an obstacle is detected while moving, the self-parking vehicle returns to the first position (S330) and updates the parking trajectory to generate a corrected parking trajectory (S340).

The self-parking vehicle may move to a parking space along a corrected parking trajectory.

As an embodiment, the parking trajectory may be a circular trajectory, and the self-parking vehicle may correct the rotation angle and speed of the vehicle in consideration of the driving state of the vehicle moving along the circular trajectory. In the process of moving the vehicle along the circular trajectory, since the vehicle may deviate from the circular trajectory due to centrifugal force or road surface conditions, in step S320, the self-parking vehicle monitors the rotation angle and speed of the moving vehicle, and the vehicle follows the circular trajectory. In order to do so, the rotation angle and speed of the vehicle are corrected, and the vehicle can move to the second position.

)를 이용하여 차량의 속도(v)를 계산할 수 있다. The self-parking vehicle calculates the rotation angle (δ) of the vehicle by using the radius (R) for the circular trajectory and the length (L) of the vehicle as in [Equation 1], and the radius (R) and the displacement (

) can be used to calculate the vehicle's speed (v).

For example, the self-parking vehicle may deviate from the circular trajectory by turning or sliding at a rotation angle smaller than the rotation angle for following the circular trajectory, in which case the self-parking vehicle may increase the rotation angle and reduce the vehicle speed. have. The self-parking vehicle may periodically estimate the current location of the vehicle using a wireless tag and an inertial sensor while the vehicle is moving, and determine whether the vehicle deviated from the original trajectory.

As will be described later, the self-parking vehicle may generate first and second coordinate values on the xy plane for the current location of the vehicle, and use one of the first and second coordinate values to move the vehicle to the original location. It may be determined whether the vehicle deviated from the trajectory or whether the vehicle deviated from the original trajectory using an average value of the first and second coordinate values.

4 is a diagram for explaining a method for generating a trajectory for automatic parking according to an embodiment of the present invention, and FIG. 5 is a diagram for explaining a trajectory for automatic parking according to an embodiment of the present invention.

Referring to FIG. 4 , the self-parking vehicle according to an embodiment of the present invention acquires a plurality of coordinate values for a first position of the vehicle ( S410 ). As an embodiment, the plurality of coordinate values may include a first coordinate value estimated through a wireless tag mounted on the vehicle and a second coordinate value estimated through a vision sensor or an inertial sensor mounted on the vehicle. The wireless tag may be, for example, a UWB tag, and the vision sensor and the inertial sensor may be mounted at different locations from the wireless tag. For example, the wireless tag may be mounted on the rear of the vehicle, and the vision sensor and the inertial sensor may be mounted on the front of the vehicle.

The self-parking vehicle generates a first trajectory from the first position to the second position by using the plurality of coordinate values obtained in step S410 and the coordinate values for the second position spaced apart by a preset distance from the parking space (S420). )do. As one embodiment, as shown in FIG. 5 , the self-parking vehicle 510 is a first position from two coordinate values 511 and 512 for a first position and one coordinate value for a second position 521 , A circle passing through the first position and the second position may be created, and in this circle, some trajectories from the first position to the second position 521 may be a circle trajectory 541 for automatic parking. have.

The second location 521 may exist on a virtual straight line 520 spaced apart by a preset distance from the third location 531 corresponding to the parking space, and the virtual straight line 520 is arranged to be parallel to the parking space. can be In addition, the coordinates for the second position 521 may be calculated through the relationship between the third position 530 and the virtual straight line 520 .

Also, the second position 521 may be determined according to the first position. As shown in FIG. 5 , when the first position at which the self-parking vehicle is stopped is located on the right side with respect to the third position 531 , the second position 521 is located on the left side with respect to the third position 531 . can be located in Conversely, when the first position is on the left with respect to the third position 531 , the second position 521 may be located on the right side with respect to the third position 531 .

Returning to FIG. 4 , in step S420 , the self-parking vehicle may update the original trajectory when an obstacle is detected while the vehicle is moving along the circular trajectory, as described above. After the self-parking vehicle returns to the first position, the self-parking vehicle may update the original trajectory according to the changed second position. The changed second position may be determined as an arbitrary point on the virtual straight line. Alternatively, according to an embodiment, the self-parking vehicle may select a new parking space to update the original trajectory.

And the self-parking vehicle according to an embodiment of the present invention generates a second trajectory from the second position to the third position corresponding to the parking space (S430). As an embodiment, the self-parking vehicle may generate the second trajectory such that the heading angle of the vehicle at the second position is aligned with the parking angle relative to the parking space in the parking space.

Referring to FIG. 5 , the parking angle for the parking space may be an angle perpendicular to an imaginary straight line 520 arranged parallel to the parking space, that is, 90 degrees. The second trajectory 542 may be generated so that the heading angle of the vehicle is 90 degrees in the parking space. The self-parking vehicle may generate the second trajectory 542 using a cubic function model as in Equation 2 as an embodiment. Here, α is a constant determined in various ways according to embodiments.

The self-parking vehicle acquires a plurality of coordinate values, ie, first and second coordinate values, of the vehicle through a wireless tag and an inertial sensor at a second location 521 on the xy plane, and obtains a plurality of coordinate values for the second location 521 The second trajectory 542 may be generated by calculating a and b values by substituting the coordinate values of and the coordinate values for the third position 531 into the cubic function model.

According to an embodiment of the present invention, by generating an automatic parking trajectory based on the coordinate values for the vehicle and the parking space estimated through positioning, a more accurate automatic parking trajectory from the current location of the vehicle to the parking space can be generated. can

The technical contents described above may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiments, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. A hardware device may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, the present invention has been described with specific matters such as specific components and limited embodiments and drawings, but these are only provided to help a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , various modifications and variations are possible from these descriptions by those of ordinary skill in the art to which the present invention pertains. Therefore, the spirit of the present invention should not be limited to the described embodiments, and not only the claims to be described later, but also all those with equivalent or equivalent modifications to the claims will be said to belong to the scope of the spirit of the present invention. .

Claims (13)

generating a parking trajectory from the first location to a second location spaced apart from a parking space by a preset distance by using a plurality of coordinate values for the first location of the vehicle;
estimating the position and size of an obstacle by using a front image of the vehicle and a first artificial neural network while moving along the parking trajectory;
determining whether the obstacle collides with the vehicle using a second artificial neural network according to the estimated location and size; and
and updating the parking trajectory according to the determination result,
The plurality of coordinate values are
a first coordinate value estimated through a wireless tag mounted on the vehicle; and
It is mounted on the vehicle and includes a second coordinate value estimated through a vision sensor or an inertial sensor mounted at a different location from the wireless tag,
The parking trajectory, which is a circular trajectory, is
In a circle generated from a plurality of coordinate values for the first position and coordinate values for the second position, a locus from the first position to the second position is
automatic parking method.
The method of claim 1,
The step of determining whether the obstacle and the vehicle collide
When the obstacle is an obstacle in a stationary state, determining whether the obstacle collides with the vehicle
automatic parking method.
generating a parking trajectory from the first location to a second location spaced apart from a parking space by a preset distance by using a plurality of coordinate values for the first location of the vehicle;
moving to the second position along the parking trajectory;
returning to the first position when an obstacle is detected during movement; and
updating the parking trajectory to generate a corrected parking trajectory,
The plurality of coordinate values are
a first coordinate value estimated through a wireless tag mounted on the vehicle; and
It is mounted on the vehicle and includes a second coordinate value estimated through a vision sensor or an inertial sensor mounted at a different location from the wireless tag,
The parking trajectory of the circular trajectory is
In a circle generated from a plurality of coordinate values for the first position and coordinate values for the second position, a locus from the first position to the second position is
Automated parking method comprising.
delete delete 4. The method of claim 3,
The step of moving to the second position is
calculating a rotation angle of the vehicle using a radius of the circular trajectory and a length of the vehicle, and calculating a speed of the vehicle using the radius and displacement of the vehicle; and
correcting the rotation angle and the speed so that the vehicle follows a circular trajectory.
Automated parking method comprising.
obtaining a plurality of coordinate values for a first position of the vehicle;
generating a first trajectory from the first location to the second location by using the plurality of coordinate values and a coordinate value for a second location spaced apart by a preset distance from the parking space; and
generating a second trajectory from the second location to a third location corresponding to the parking space,
The plurality of coordinate values are
a first coordinate value estimated through a wireless tag mounted on the vehicle; and
It is mounted on the vehicle and includes a second coordinate value estimated through a vision sensor or an inertial sensor mounted at a different location from the wireless tag,
The first trajectory, which is a circular trajectory, is
In a circle generated from a plurality of coordinate values for the first position and coordinate values for the second position, a locus from the first position to the second position is
How to create a trajectory for automatic parking.
delete delete 8. The method of claim 7,
The step of generating the first trajectory is
When an obstacle is detected while the vehicle is moving according to the first trajectory, the original trajectory is updated according to the changed second position.
How to create a trajectory for automatic parking.
8. The method of claim 7,
The second position is
It exists on an imaginary straight line spaced apart by the distance from the third position, and is determined according to the first position.
How to create a trajectory for automatic parking.
8. The method of claim 7,
The step of generating the second trajectory is
generating the second trajectory so that the heading angle of the vehicle at the second position is aligned with the parking angle with respect to the parking space in the parking space
How to create a trajectory for automatic parking.
13. The method of claim 12,
The step of generating the second trajectory is
generating the second trajectory by substituting a plurality of coordinate values of the vehicle obtained at the second position and coordinate values for the third position into a cubic function model
How to create a trajectory for automatic parking.

Images