US-20260125048-A1 - METHOD AND APPARATUS FOR DETERMINATION OF AUTONOMOUS PARKING ROUTE

US20260125048A1US 20260125048 A1US20260125048 A1US 20260125048A1US-20260125048-A1

Abstract

An embodiment a computer-implemented method for setting an autonomous parking route includes generating an autonomous parking route by a first route generation method using a parking space and a surrounding space; generating an autonomous parking route by a second route generation method using a cloud center and a pre-trained spatial recognition deep learning model; generating an autonomous parking route by a third route generation method using a pre-trained route generation deep learning model installed in a controller of a vehicle; performing virtual driving along the routes generated by the first, second, and third route generation methods; and determining set routes in accordance with a preset route priority in the vehicle performing autonomous parking for passed routes when the virtual driving is passed.

Inventors

Jang Shin Kim
A Jeong Choi

Assignees

HYUNDAI MOTOR COMPANY
KIA CORPORATION

Dates

Publication Date: 20260507
Application Date: 20250609
Priority Date: 20241104

Claims (20)

1 . A computer-implemented method for setting an autonomous parking route, the method comprising: generating an autonomous parking route by a first route generation method using a parking space and a surrounding space; generating an autonomous parking route by a second route generation method using a cloud center and a pre-trained spatial recognition deep learning model; generating an autonomous parking route by a third route generation method using a pre-trained route generation deep learning model installed in a controller of a vehicle; performing virtual driving along the routes generated by the first, second, and third route generation methods; and determining set routes in accordance with a preset route priority in the vehicle performing autonomous parking for passed routes when the virtual driving is passed.
2 . The computer-implemented method of claim 1 , wherein the first route generation method comprises: creating a map corresponding to the parking space and the surrounding space based at least on using vehicle specifications, parking facility information, vehicle sensor information, and vehicle specifications stored in advance in a controller of the vehicle performing autonomous parking; generating a parking route range by setting a nearest parking route and an outermost parking route on the basis of at least the parking space and surrounding map; and generating information about locations of waypoints along which the vehicle will move in the parking route range, and information about a directional angle of each of the waypoints, wherein the nearest parking route is a route that is close to a parking line, which is in contact with a parking space, and that uses a minimum turning radius, and the outermost parking route is a route that is close to an allowable parking line at the opposite side of the parking space and uses a minimum turning radius, and is a route from a current location of the vehicle to a parking completion point.
3 . The computer-implemented method of claim 1 , wherein the second route generation method comprises the cloud center, which receives a map corresponding to a parking space and a surrounding space based at least on using vehicle specifications, parking facility information, vehicle sensor information, and vehicle specifications stored in advance in a controller of the vehicle performing autonomous parking, and generates information about locations of waypoints along which the vehicle will move and information about a directional angle of each of the waypoints, using the pre-trained spatial recognition deep learning model.
4 . The computer-implemented method of claim 1 , wherein the third route generation method is a method of generating information about locations of waypoints along which the vehicle performing autonomous parking will move and information about a directional angle of each of the waypoints, using the pre-trained route generation deep learning model installed in a controller of the vehicle performing autonomously on the basis of a map corresponding to a parking space and a surrounding space based at least on using vehicle specifications, parking facility information, vehicle sensor information, and vehicle specifications stored in advance in the controller of the vehicle.
5 . The computer-implemented method of claim 2 , further comprising reducing the nearest parking route or the outermost parking route when an obstacle is detected in the parking route range such that the obstacle is not included in a range of an available parking route.
6 . The computer-implemented method of claim 2 , further comprising: performing one-step parking including determining circles with respect to a vehicle center to completely fill the parking route range, wherein waypoints of a parking route are respectively based at least on center points of the circles and points of tangency of the circles; and performing multi-step parking when the one-step parking is not possible.
7 . The computer-implemented method of claim 3 , wherein using the pre-trained spatial recognition deep learning model includes transmitting information determined based on perceiving and classifying people, vehicles, other objects, and empty spaces from the parking space and surrounding map received from a vehicle communicating with the cloud center using the spatial recognition deep learning model, to an available parking space determination module; the available parking space determination module performs transmitting information of a location of a subject vehicle, available parking spaces, and the empty spaces to a comparison algorithm module on the basis of the information received from the spatial recognition deep learning model; and the comparison algorithm module comprises a process of generating an available autonomous parking route by comparing the information received from the available parking space determination module with a parking route generation history that the cloud center manages, and of transmitting the available autonomous parking route to the vehicle with which the cloud center communicates.
8 . The computer-implemented method of claim 3 , wherein the spatial recognition deep learning model is a model that uses, as input of training data, maps corresponding to the parking space and surrounding spaces received from all of vehicles with which the cloud center communicates, and uses, as a target of the training data, people, vehicles, other objects, and empty spaces received from all kinds of vehicles with which the cloud center communicates.
9 . The computer-implemented method of claim 4 , wherein the route generation deep learning model uses a parking route range of parking route generation history information received from all of vehicles, with which a cloud center communicates, as input data of training data, and uses available route vectors as target data of the training data; and a route generation deep learning model installed in the controller of the vehicle receives and stores weights updated as a training result of the route generation deep learning model from the cloud center.
10 . The computer-implemented method of claim 1 , further comprising transmitting a parking route range, the created routes, and actual parking results, as results of the virtual driving, to a cloud center that communicates with the vehicle performing autonomous parking.
11 . An apparatus comprising: at least one memory storing commands; and at least one processor, wherein the at least one processor, by executing the commands, performs: generating an autonomous parking route by a first route generation method using a parking space and a surrounding space; generating an autonomous parking route by a second route generation method using a cloud center and a pre-trained spatial recognition deep learning model; generating an autonomous parking route by a third route generation method using a pre-trained route generation deep learning model installed in a controller of a vehicle; performing virtual driving along the routes generated by the first, second, and third route generation methods; and determining set routes in accordance with a preset route priority in the vehicle performing autonomous parking for passed routes when the virtual driving is passed.
12 . The apparatus of claim 11 , wherein the first route generation method comprises: creating a map corresponding to the parking space and the surrounding space based at least on using vehicle specifications, parking facility information, vehicle sensor information, and vehicle specifications stored in advance in a controller of the vehicle performing autonomous parking; generating a parking route range by setting a nearest parking route and an outermost parking route on the basis of at least the parking space and surrounding map; and generating information about locations of waypoints along which the vehicle will move in the parking route range, and information about a directional angle of each of the waypoints, wherein the nearest parking route is a route that is close to a parking line, which is in contact with a parking space, and that uses a minimum turning radius, and the outermost parking route is a route that is close to an allowable parking line at the opposite side of the parking space and uses a minimum turning radius, and is a route from a current location of the vehicle to a parking completion point.
13 . The apparatus of claim 11 , wherein the second route generation method comprises the cloud center, which receives a map corresponding to a parking space and surrounding space based at least on using vehicle specifications, parking facility information, vehicle sensor information, and vehicle specifications stored in advance in a controller of the vehicle performing autonomous parking, and generates information about locations of waypoints along which the vehicle will move and information about a directional angle of each of the waypoints, using the pre-trained spatial recognition deep learning model.
14 . The apparatus of claim 11 , wherein the third route generation method is a method of generating information about locations of waypoints along which the vehicle performing autonomous parking will move and information about a directional angle of each of the waypoints, using the pre-trained route generation deep learning model installed in a controller of the vehicle performing autonomously on the basis of a map corresponding to a parking space and a surrounding space using vehicle specifications, parking facility information, vehicle sensor information, and vehicle specifications stored in advance in the controller of the vehicle.
15 . The apparatus of claim 12 , further performing reducing the nearest parking route or the outermost parking route when an obstacle is detected in the parking route range such that the obstacle is not included in a range of an available parking route.
16 . The apparatus of claim 12 , further performing: performing one-step parking including determining circles with respect to a vehicle center to completely fill the parking route range, wherein waypoints of a parking route are respectively based at least on center points of the circles and points of tangency of the circles; and performing multi-step parking when the one-step parking is not possible.
17 . The apparatus of claim 13 , wherein using the pre-trained spatial recognition deep learning model includes transmitting information determined based on perceiving and classifying people, vehicles, other objects, and empty spaces from the parking space and surrounding map received from a vehicle communicating with the cloud center using the spatial recognition deep learning model, to an available parking space determination module; the available parking space determination module performs transmitting information of a location of a subject vehicle, available parking spaces, and the empty spaces to a comparison algorithm module on the basis of the information received from the spatial recognition deep learning model; and the comparison algorithm module performs generating an available autonomous parking route by comparing the information received from the available parking space determination module with a parking route generation history that the cloud center manages, and of transmitting the available autonomous parking route to the vehicle with which the cloud center communicates.
18 . The apparatus of claim 13 , wherein the spatial recognition deep learning model is a model that uses, as input of training data, maps corresponding to the parking space and surrounding spaces received from all of vehicles, with which the cloud center communicates, and uses, as a target of the training data, people, vehicles, other objects, and empty spaces received from all kinds of vehicles, with which the cloud center communicates.
19 . The apparatus of claim 14 , wherein the route generation deep learning model uses a parking route range of parking route generation history information received from all of vehicles, with which a cloud center communicates, as input data of training data, and uses available route vectors as target data of the training data; and a route generation deep learning model installed in the controller of the vehicle receives and stores weights updated as a training result of the route generation deep learning model from the cloud center.
20 . The apparatus of claim 11 , further performing transmitting a parking route range, the created routes, and actual parking results, as results of the virtual driving, to a cloud center that communicates with the vehicle performing autonomous parking.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of and priority to Korean Patent Application No. 10-2024-0154528, filed on Nov. 4, 2024, in the Korea Intellectual Property Office, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD The present disclosure relates to a method and apparatus for determination of an autonomous parking route. BACKGROUND The following description simply provides only the background information related to the present embodiment without configuring the related art. An autonomous vehicle is a vehicle that can drive on the road by itself without human intervention. An autonomous vehicle perceives environments, sets routes, and controls driving using various sensors and control systems. Autonomous parking refers to a process in which an autonomous vehicle searches for a parking space, moves into the parking space, and completes parking by itself without human intervention. Autonomous parking reduces the driver's parking burden, especially in narrow parking spaces or complex environments, and enables safer and more efficient parking. An autonomous parking system can operate smoothly by utilizing communication between a vehicle and a parking facility control center. A parking facility control center monitors real-time information of a parking facility, the location of empty spaces, and the location and status of vehicles, and assigns appropriate parking spaces to autonomous vehicles. A parking facility control center communicates with vehicles using a communication technology such as vehicle-to-everything (V2X), and provides information on obstacles or road conditions during a parking process, thereby enhancing safety of autonomous parking. An autonomous parking system requires advanced control capabilities, so an Advanced Driver Assistance System (ADAS) controller is used. The ADAS controller controls the vehicle's speed, steering, and barking and accurately moves the vehicle along a parking route. The ADAS controller controls the driving status of a vehicle on the basis of collected data by analyzing, in real time, vehicle sensor information collected by an ultrasonic sensor, a camera, a radar, etc. mounted on the vehicle. Autonomous parking requires particularly precise steering control and real-time obstacle avoidance. Deep learning technology can optimize an autonomous parking system by interacting with a parking facility control center and an ADAS controller. A deep learning model can receive data on empty parking spaces and vehicle locations in a parking facility from a control center and can set an optimal parking route. A deep learning model can precisely control a parking process by receiving data, such as the speed, the steering angle, etc. of a vehicle, in real time from a controller, and can correct or optimize a route in real time using communication with a control center. SUMMARY Embodiments provide a method and apparatus for determination of an autonomous parking route. In detail, as long as unit vector information for movement of a vehicle is provided, it is possible to derive a center of rotation of the vehicle and it is possible to control movement of the vehicle, whereby it is possible to set a route for autonomous parking. Further embodiments provide setting an autonomous parking route using deep learning. According to an embodiment of the present disclosure, a computer-implemented method for setting an autonomous parking route, the method comprising: generating an autonomous parking route by a first route generation method using a parking space and a surrounding space; generating an autonomous parking route by a second route generation method using a cloud center and a pre-trained spatial recognition deep learning model; generating an autonomous parking route by a third route generation method using a pre-trained route generation deep learning model installed in a controller of a vehicle; performing virtual driving along the routes generated by the first, second, and third route generation methods; and determining set routes in accordance with a preset route priority in the vehicle performing autonomous parking for passed routes when the virtual driving is passed. The vehicle can then be autonomously driven according to the determining. According to another embodiment of the present disclosure, an apparatus comprises at least one memory storing commands and at least one processor. The at least one processor, by executing the commands, performs: generating an autonomous parking route by a first route generation method using a parking space and a surrounding space; generating an autonomous parking route by a second route generation method using a cloud center and a pre-trained spatial recognition deep learning model; generating an autonomous parking route by a third route generation method using a pre-trained route generation deep learning model installed in a controller of a vehicle; performing virtual drivin