KR-20260066476-A - Method And Apparatus for Determination of Autonomous Parking Route

Abstract

A method and apparatus for recognizing and controlling human behavior are disclosed. According to one aspect of the present disclosure, a computer-implemented method for setting an autonomous parking path comprises: a process of generating an autonomous parking path by a first path generation method; a process of generating an autonomous parking path by a second path generation method; a process of generating an autonomous parking path by a third path generation method; a process of performing a virtual driving of the paths generated by the first, second, and third path generation methods; and a process of determining a set path according to a pre-set path priority for a vehicle performing autonomous parking on the path passed when the virtual driving is passed.

Inventors

• 김장신
• 최아정

Assignees

• 현대자동차주식회사
• 기아 주식회사

Dates

Publication Date

20260512

Application Date

20241104

Claims (20)

1. As a computer implementation method for autonomous parking path setting, A process of generating an autonomous parking path by a first path generation method; A process of generating an autonomous parking path by a second path generation method; Process of generating an autonomous parking path by a third path generation method; A process of performing virtual driving of a path generated by the first, second, and third path generation methods; and A computer-implemented method comprising the process of determining a set path according to a preset path priority for a vehicle performing autonomous parking on the path passed when the above virtual driving is passed.
2. In Article 1, The first path generation method is, A process of generating a parking space and surrounding map, etc., using vehicle specifications, parking lot information, vehicle sensor information, vehicle specifications, etc., previously stored in the controller of the vehicle performing the autonomous parking above; A process of generating a parking path range by setting the nearest parking path and the outermost parking path based on the above parking space and surrounding map, etc.; The method includes a process for generating information regarding the locations of path points where a vehicle will move within the above-mentioned parking path range and the direction angle of each path point, and The above nearest parking route is, It is a path from the vehicle's current position to the parking completion point using the minimum turning radius while approaching the parking line adjacent to the parking space, and The above outermost parking route is, A computer implementation method for a path from the vehicle's current position to the parking completion point, which is a path using a minimum turning radius in close proximity to a parking line across from the parking space.
3. In Article 1, The second path generation method is, A computer implementation method in which a cloud center, having received a parking space and surrounding map generated using vehicle specifications, parking lot information, vehicle sensor information, and vehicle specifications stored in a controller of a vehicle performing autonomous parking, generates information regarding the locations of path points to which the vehicle will move and the direction angles of each path point using a pre-trained spatial identification deep learning model.
4. In Article 1, The third path generation method is, A computer implementation method that generates information regarding the locations of path points to which a vehicle will move and the direction angles of each path point by using a pre-trained path generation deep learning model installed in the controller of the vehicle performing the autonomous parking, based on a parking space and surrounding map generated using vehicle specifications, parking lot information, vehicle sensor information, vehicle specifications, etc., which are stored in the controller of the vehicle performing the autonomous parking.
5. In Article 2, A computer-implemented method further comprising a process of reducing the nearest parking path or the outermost parking path when an obstacle is found in the range of a parking path, thereby ensuring that the obstacle is not included in the range of a parking path.
6. In Article 2, A process of one-step parking in which circles are drawn from the center of the vehicle to fill the above parking path range, and the center points of the circles and the points of contact between the circles can become each path point of the parking path; and A computer implementation method that further includes a process of performing multi-step parking when one-step parking is not possible.
7. In Paragraph 3, Using a pre-trained spatial awareness deep learning model is, A process of transmitting information to a parking space determination module that recognizes and classifies people, vehicles, other objects, empty spaces, etc., from the parking space and surrounding map received from a vehicle communicating with the cloud center using the above-mentioned spatial recognition deep learning model; The above parking space determination module is, A process of transmitting information such as the location of my vehicle, available parking spaces, and empty spaces to a comparison algorithm module based on the information received from the above spatial identification deep learning model; and The above comparison algorithm module is, A computer-implemented method comprising the process of generating a path capable of autonomous parking by comparing the information received from the above-mentioned parking space determination module with the parking path generation history information managed by the cloud center, and transmitting the generated path to a vehicle communicating with the cloud center.
8. In Paragraph 3, Spatial awareness deep learning models are, A computer implementation method, which is a model that uses the parking space and surrounding map received from all vehicles communicating with the cloud center as input data among training data, and uses people, vehicles, other objects, empty spaces, etc. received from all vehicle types communicating with the cloud center as target data among training data.
9. In Paragraph 4, Path generation deep learning models are, Among the parking path generation history information received from all vehicles communicating with the cloud center, the parking path range is used as input data in the training data, and the parking-possible path vector is used as target data in the training data, and A computer implementation method in which a pre-trained path generation deep learning model installed in the controller of the above vehicle receives and stores updated weights as training results of the path generation deep learning model from the cloud center.
10. In Article 1, A computer-implemented method further comprising the process of transmitting, as a result of the above virtual driving, a parking path range, the generated path, and the actual parking result, etc., to a cloud center communicating with a vehicle performing the above autonomous parking.
11. at least one memory for storing instructions; and at least one processor, wherein the at least one processor executes the instructions, A process of generating an autonomous parking path by a first path generation method; A process of generating an autonomous parking path by a second path generation method; Process of generating an autonomous parking path by a third path generation method; A process of performing virtual driving of a path generated by the first, second, and third path generation methods; and A device that performs a process of determining a set path according to a preset path priority for a vehicle performing autonomous parking on the path passed when the above virtual driving is passed.
12. In Article 11, The first path generation method is, A process of generating a parking space and surrounding map, etc., using vehicle specifications, parking lot information, vehicle sensor information, vehicle specifications, etc., previously stored in the controller of the vehicle performing the autonomous parking above; A process of generating a parking path range by setting the nearest parking path and the outermost parking path based on the above parking space and surrounding map, etc.; The method includes a process for generating information regarding the locations of path points where a vehicle will move within the above-mentioned parking path range and the direction angle of each path point, and The above nearest parking path is a path that uses a minimum turning radius while approaching the parking line adjacent to the parking space, and is a path from the vehicle's current position to the parking completion point. The above outermost parking path is a path that uses a minimum turning radius in close proximity to a parking-available line across from the parking space, and is a path from the vehicle's current position to the parking completion point.
13. In Article 11, The second path generation method is, A device comprising a method in which a cloud center, having received a parking space and surrounding map generated using vehicle specifications, parking lot information, vehicle sensor information, vehicle specifications, etc., which are stored in a controller of a vehicle performing autonomous parking, generates information regarding the locations of path points to which the vehicle will move and the direction angles of each path point using a pre-trained spatial identification deep learning model.
14. In Article 11, The third path generation method is, A device, which is a method for generating information regarding the locations of path points to which a vehicle will move and the direction angles of each path point, based on a parking space and surrounding map generated using vehicle specifications, parking lot information, vehicle sensor information, vehicle specifications, etc., which are stored in the controller of the vehicle performing the autonomous parking, and using a pre-trained path generation deep learning model installed in the controller of the vehicle performing the autonomous parking.
15. In Article 12, A device that further performs a process of ensuring that, when an obstacle is found in the range of a parking path, the obstacle is not included in the range of a parking path by reducing the nearest parking path or the outermost parking path.
16. In Article 12, A device that further performs a process of one-step parking when circles are drawn from the center of the vehicle to fill the parking path range, and the center points of the circles and the points of contact between the circles can become each path point of the parking path; and a process of multi-step parking when one-step parking is not possible.
17. In Article 13, Using a pre-trained spatial awareness deep learning model is, A process of transmitting information to a parking space determination module that recognizes and classifies people, vehicles, other objects, empty spaces, etc., from the parking space and surrounding map received from a vehicle communicating with the cloud center using the above-mentioned spatial recognition deep learning model; The above parking space determination module transmits information such as the location of my vehicle, available parking space, and empty space to a comparison algorithm module based on the information received from the above space identification deep learning model; and A device that performs the process of generating a path capable of autonomous parking by comparing the above comparison algorithm module with the parking path generation history information managed by the cloud center based on the information received from the above parking space determination module, and transmitting it to a vehicle communicating with the cloud center.
18. In Article 13, Spatial awareness deep learning models are, A device, which is a model that uses the parking space and surrounding map received from all vehicles communicating with the cloud center as input data among training data, and uses people, vehicles, other objects, empty spaces, etc. received from all vehicle types communicating with the cloud center as target data among training data.
19. In Article 14, Path generation deep learning models are, Among the parking path generation history information received from all vehicles communicating with the cloud center, the parking path range is used as input data in the training data, and the parking-possible path vector is used as target data in the training data, and A device in which a pre-trained path generation deep learning model installed in the controller of the above vehicle receives and stores updated weights as training results of the path generation deep learning model from the cloud center.
20. In Article 11, A device that further performs the process of transmitting, as a result of the above virtual driving, a parking path range, the above generated path, and the actual parking result, etc., to a cloud center communicating with a vehicle performing the above autonomous parking.

Description

Method and Apparatus for Determination of Autonomous Parking Route The present disclosure relates to a method and apparatus for determining an autonomous parking path. The following description merely provides background information related to the present embodiment and does not constitute prior art. An autonomous vehicle is a vehicle capable of driving on the road by itself without human intervention. Autonomous vehicles utilize various sensors and control systems to perceive the environment, set a route, and control driving. Autonomous parking refers to the process in which an autonomous vehicle independently searches for a parking space, moves into it, and completes the parking process without driver intervention. Autonomous parking reduces the burden on drivers, particularly in tight spaces or complex environments, and enables safer and more efficient parking. The autonomous parking system can operate smoothly by utilizing communication between the vehicle and the parking facility control center. The control center monitors real-time parking lot information, the location of available spaces, and the location and status of vehicles, and allocates appropriate parking spaces to autonomous vehicles. The control center communicates with vehicles using communication technologies such as V2X (vehicle-to-everything) and enhances the safety of autonomous parking by providing information on obstacles or road conditions that may occur during the parking process. Because autonomous parking systems require advanced control capabilities, Advanced Driver Assistance System (ADAS) controllers are used. ADAS controllers manage the vehicle's speed, steering, and braking, and accurately move the vehicle along the parking path. ADAS controllers analyze vehicle sensor information collected in real time by ultrasonic sensors, cameras, radar, and other devices mounted on the vehicle, and control the vehicle's driving state based on the collected data. In particular, precise steering control and real-time obstacle avoidance are required for autonomous parking. Deep learning technology can optimize autonomous parking systems by interacting with a parking facility control center and an ADAS controller. The deep learning model receives location data of empty spaces and vehicles within the parking lot from the control center and can set an optimal parking path. The deep learning model receives real-time data such as vehicle speed and steering angle from the controller to finely control the parking process, and can modify or optimize the path in real-time using communication with the control center. FIG. 1 is a block diagram of a device for generating and following an autonomous parking path using a vehicle center and direction vector according to one embodiment of the present disclosure. FIG. 2 is a block diagram of a device for setting an autonomous parking path according to one embodiment of the present disclosure. FIG. 3 is an example diagram illustrating the process of setting a parking path using a spatial identification deep learning model when a cloud center according to one embodiment of the present disclosure acts as a parking lot control center. Figure 4 is a diagram illustrating the process of processing training data for a path generation deep learning model. Figure 5 is a diagram illustrating the turning center of a vehicle according to the steering condition of each wheel. FIG. 6 is a drawing for explaining a first path generation method according to one embodiment of the present disclosure. Figure 7 is a diagram illustrating the process of setting the parking path for sections C-D of Figure 6. Figure 8 is a diagram illustrating the process of calculating the turning center point of a vehicle. Figure 9 is a diagram illustrating the process of calculating the steering angle for each wheel of a vehicle. FIG. 10 is a diagram illustrating the process of setting the autonomous parking path range of the first path generation method. FIG. 11 is a drawing for explaining the area where the center of the vehicle can be located and the area where the outermost part of the vehicle can be located. FIG. 12 is a diagram illustrating the process of changing the parking path when an obstacle occurs inside the nearest parking path. Figure 13 is a diagram illustrating the process of changing the parking path when an obstacle occurs outside the vehicle in the outermost parking path. FIG. 14 is a drawing for explaining the process of setting a one-step parking path according to one embodiment of the present disclosure. FIG. 15 is a diagram illustrating the process of reducing the range of a parking path when an obstacle occurs in a one-step parking path. Figure 16 is a diagram illustrating a case where one-step parking is not possible. FIG. 17 is a diagram illustrating the multi-step parking process when there is an obstacle in front of the parking space. FIG. 18 is a diagram illustrating the multi-step parking process when there are no o