CN117445901A - Parking path planning method and path node prediction model training method - Google Patents

Parking path planning method and path node prediction model training method Download PDF

Info

Publication number
CN117445901A
CN117445901A CN202210832136.3A CN202210832136A CN117445901A CN 117445901 A CN117445901 A CN 117445901A CN 202210832136 A CN202210832136 A CN 202210832136A CN 117445901 A CN117445901 A CN 117445901A
Authority
CN
China
Prior art keywords
parking
information
path
parking space
path node
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210832136.3A
Other languages
Chinese (zh)
Inventor
杜影丽
李睿哲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Haikang Auto Software Co ltd
Original Assignee
Hangzhou Haikang Auto Software Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Haikang Auto Software Co ltd filed Critical Hangzhou Haikang Auto Software Co ltd
Priority to CN202210832136.3A priority Critical patent/CN117445901A/en
Publication of CN117445901A publication Critical patent/CN117445901A/en
Pending legal-status Critical Current

Links

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
    • B60W30/06—Automatic manoeuvring for parking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Mechanical Engineering (AREA)
  • Transportation (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Automation & Control Theory (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Traffic Control Systems (AREA)

Abstract

The application provides a parking path planning method and a training method of a path node prediction model, and relates to the technical field of vehicle path planning. The planning method comprises the following steps: acquiring scene information of a target parking space of a vehicle to be parked; the scene information of the target parking space is used for describing the target parking space; inputting scene information of a target parking space into a path node prediction model, and determining parking path node information of a vehicle to be parked; the path node prediction model is used for representing the corresponding relation between parking path node information and scene information of a parking space, and the parking path node information comprises position information and posture information of a vehicle to be parked at the parking path node; and planning a path of the vehicle to be parked to enter the target parking space according to the parking path node information. The method is suitable for the path planning process when the vehicle parks, and is used for reducing the calculated amount and improving the parking efficiency.

Description

Parking path planning method and path node prediction model training method
Technical Field
The present disclosure relates to the field of vehicle path planning technologies, and in particular, to a parking path planning method and a training method of a path node prediction model.
Background
The automatic parking system (automated parking assist, APA) refers to a driving assistance system that automatically parks a vehicle by controlling acceleration and deceleration, steering angle, and forward and reverse gear of the vehicle.
When an automatic parking system plans an automatic parking path of a vehicle, the whole path planning process is usually divided into a plurality of sections of paths for planning due to limitations of perception errors, vehicle control errors, calculated amount and the like. For example, an automatic parking system may first plan a forward path, control the vehicle to advance to a shift position, stop at the shift position, shift the forward gear to a reverse gear, and then plan a reverse path, control the vehicle to be dumped into a parking space.
However, current automatic parking systems have low parking efficiency.
Disclosure of Invention
Based on the technical problems, the application provides a parking path planning method and a training method of a path node prediction model, wherein the planning method can predict parking path node information matched with scene information of a target parking space by using the parking path node prediction model, and the quality of a parking path planned according to the parking path node information is higher, so that the automatic parking efficiency is improved.
In a first aspect, the present application provides a method for planning an automatic parking path, the method comprising: acquiring scene information of a target parking space of a vehicle to be parked; the scene information of the target parking space is used for describing the target parking space; inputting scene information of a target parking space into a path node prediction model, and determining parking path node information of a vehicle to be parked; the path node prediction model is used for representing the corresponding relation between parking path node information and scene information of a parking space, and the parking path node information comprises position information and posture information of a vehicle to be parked at the parking path node; and planning a path of the vehicle to be parked to enter the target parking space according to the parking path node information.
According to the parking path planning method, the scene information of the target parking space can be obtained, the target parking path node information corresponding to the scene information of the target parking space is predicted according to the path node prediction model for representing the corresponding relation between the parking path node information and the scene information of the parking space, the target parking path node information and the scene information of the target parking space have a pre-trained matching relation, the path quality of a planned parking path is higher from the position corresponding to the parking path node information according to the gesture corresponding to the parking path node information, and the efficiency of vehicle parking is improved as a whole.
Optionally, the scene information of the target parking space includes at least one of the following: boundary information of obstacles around the target parking space, boundary information of the target parking space or final state pose information of the vehicle corresponding to the target parking space; boundary information of obstacles around the target parking space is used for describing a drivable area around the target parking space; the boundary information of the target parking space and the final pose information of the vehicle corresponding to the target parking space are used for describing the position of the target parking space.
It is understood that when the path node prediction model is utilized to predict the parking path node information, the parking path planning method provided by the application can also consider the boundaries of the obstacles around the target parking space, and the boundaries of different obstacles can correspond to different parking path node information and different parking paths, so that the types of planned parking paths are enriched, and the universality of the path node prediction model is improved.
Optionally, the park path node comprises a shift node of the vehicle to be parked. The method further comprises the steps of: controlling a vehicle to be parked to travel to a position corresponding to the parking path node information, and controlling the vehicle to be parked to be in a reverse gear state at the position and to be in a posture corresponding to the parking path node information; and controlling the vehicle to be parked to enter the target parking space according to the planned path under the reverse gear state and the gesture corresponding to the parking path node information.
Optionally, the path node prediction model is obtained by training a preset model based on a training sample set, wherein the training sample set comprises a plurality of training samples, each training sample comprises scene information of a parking space and a label corresponding to the scene information of the parking space, and the label comprises parking path node information of the vehicle.
Optionally, the label corresponding to each parking space is obtained by screening a parking planning path corresponding to each candidate parking path node information from a plurality of candidate parking path node information corresponding to each parking space; the parking planning path corresponding to the node information of each candidate parking path is a path of a vehicle driving into a parking space, which is planned based on the node information of each candidate parking path; the plurality of candidate parking path node information corresponding to each parking space comprises a plurality of position information in the candidate area and preset posture information of the vehicle corresponding to each position information in a parking state.
Optionally, the label corresponding to each parking space includes candidate parking path node information corresponding to the situation that the quality of the parking planning path is greater than a preset quality threshold or the quality is highest; the quality of the parking plan path is determined based on at least one of a number of shifts of the vehicle, a path length, and a path curvature during entry into the parking space according to the parking plan path.
It should be understood that when the label in the training sample is obtained, the candidate parking path node information corresponding to the parking planning path with higher path quality can be selected as the label, and the parking path node information predicted by the path node prediction model obtained by training the training sample with the label also has the characteristic of the label, that is, the path quality of the parking path corresponding to the predicted parking path node information is higher (the number of gear shifts is less, the parking path is short, and the curvature of the parking path is small).
Optionally, the path node prediction model is a neural network model; the neural network model comprises a plurality of network modules and an output layer, wherein each network module comprises a full-connection layer, an activation layer and a normalization layer; the full-connection layer is used for extracting the characteristics of scene information of the target parking space; the activation layer is used for carrying out nonlinear processing on the characteristics of the scene information of the target parking space extracted by the full-connection layer to obtain nonlinear characteristics; the normalization layer is used for normalizing the nonlinear characteristics; and the output layer is used for outputting parking path node information of the vehicle to be parked according to the nonlinear characteristics after normalization processing of the normalization layer.
In a second aspect, the present application provides a parking path planning apparatus comprising respective modules for the method of the first aspect described above.
In a third aspect, the present application provides a method for training a path node prediction model, where the method includes: acquiring a training sample set, wherein the training sample set comprises a plurality of training samples, each training sample comprises scene information of a parking space and a label corresponding to the scene information of the parking space, the label comprises parking path node information in the process of parking a vehicle, and the parking path node information comprises position information and gesture information of the vehicle at the parking path node; and training a preset model based on the training sample set to obtain a path node information prediction model.
In one possible implementation, obtaining a training sample set includes: acquiring scene information of a plurality of parking spaces, and determining candidate areas of each parking space; selecting a plurality of position information and preset posture information of a vehicle corresponding to each position information in a parking state from candidate areas of each parking space as candidate parking path node information corresponding to each parking space; planning a parking planning path corresponding to each candidate parking path node information based on each candidate parking path node information; and screening the labels corresponding to the parking spaces from the plurality of candidate parking path node information corresponding to the parking spaces based on the parking planning paths corresponding to the candidate parking path node information.
Optionally, after planning the parking planning path corresponding to each candidate parking path node information based on each candidate parking path node information, the method further includes: and determining the quality of the parking planning path corresponding to each candidate parking path node information based on at least one of the shift times, the path length and the path curvature of the vehicle in the process of driving into the parking space according to the parking planning path corresponding to each candidate parking path node information. Based on the parking planning path corresponding to each candidate parking path node information, the label corresponding to each parking space is obtained by screening from the plurality of candidate parking path node information corresponding to each parking space, and the method comprises the following steps: and determining the candidate parking path node information corresponding to the condition that the quality of the parking planning path is greater than a preset quality threshold value or the quality is highest from the plurality of candidate parking path node information corresponding to each parking space as a label of each parking space.
Optionally, the preset model is a neural network model; the neural network model comprises a plurality of network modules and an output layer, wherein each network module comprises a full-connection layer, an activation layer and a normalization layer; the full-connection layer is used for extracting the characteristics of the scene information of the parking spaces in the training samples; the activation layer is used for carrying out nonlinear processing on the characteristics of the scene information of the parking spaces in the training samples extracted by the full-connection layer to obtain nonlinear characteristics; the normalization layer is used for normalizing the nonlinear characteristics; and the output layer is used for outputting parking path node information according to the nonlinear characteristics after normalization processing of the normalization layer.
In a fourth aspect, the present application provides a training apparatus for a path node prediction model, the apparatus comprising respective modules for the method described in the third aspect above.
In a fifth aspect, the present application provides a computer program product which, when run on a computer, causes the computer to perform the steps of the related method of the first aspect described above to implement the method of the first or third aspect described above.
In a sixth aspect, the present application provides an electronic device, including: a processor and a memory; the memory stores instructions executable by the processor; the processor is configured to execute instructions to cause the electronic device to implement the method according to the first or third aspect described above.
In a seventh aspect, the present application provides a computer-readable storage medium comprising: computer software instructions; the computer software instructions, when run in an electronic device, cause the electronic device to implement the method of the first or third aspect described above.
In an eighth aspect, the present application provides a chip comprising a processor and an interface, the processor being coupled to a memory through the interface, which when executed by the processor, causes the method of the first or third aspect described above to be performed.
The advantageous effects of the second aspect to the eighth aspect described above may be described with reference to the first aspect, and will not be repeated.
Drawings
FIG. 1 is a schematic view of a side parking scenario;
FIG. 2 is a schematic diagram of a reverse warehouse entry scenario;
FIG. 3 is a schematic view of a parking scenario in a diagonal direction;
fig. 4 is a schematic diagram of the composition of an automatic parking system according to an embodiment of the present application;
fig. 5 is a schematic diagram of the composition of an electronic device according to an embodiment of the present application;
fig. 6 is a flow chart of a parking path planning method according to an embodiment of the present application;
fig. 7 is a schematic view of scene information of a target parking space according to an embodiment of the present application;
fig. 8 is a schematic view of scene information of another target parking space according to an embodiment of the present application;
fig. 9 is a schematic view of scene information of another target parking space according to an embodiment of the present application;
fig. 10 is a flowchart of a training method of a path node prediction model according to an embodiment of the present application;
FIG. 11 is a schematic diagram of a candidate region according to an embodiment of the present disclosure;
FIG. 12 is a schematic view of another candidate region provided in an embodiment of the present application;
FIG. 13 is a schematic view of another candidate region provided in an embodiment of the present application;
fig. 14 is a schematic diagram of a method for acquiring a label of a training sample according to an embodiment of the present application;
Fig. 15 is a schematic diagram of a parking path planning apparatus according to an embodiment of the present application;
fig. 16 is a schematic diagram of a training device of a path node prediction model according to an embodiment of the present application.
Detailed Description
Currently, when an automatic parking system plans a vehicle parking route, the whole path planning process is usually divided into multiple paths for planning due to limitations such as sensing errors, vehicle control errors, calculated amount and the like. Current automated parking systems are less efficient at automatically parking vehicles.
On the basis, the embodiment of the application provides a parking path planning method, device, equipment and storage medium, which can utilize a trained path node prediction model to predict a parking space map scene to obtain a target gear shifting node adapting to the parking space map scene, and improve the quality of a parking path, thereby improving the parking efficiency.
Fig. 1 is a schematic view of a side parking scenario. As shown in fig. 1, taking a lateral parking space beside a non-motor vehicle lane as an example, a vehicle to be parked can travel to a certain shift position in the non-motor vehicle lane by using a forward gear, and the forward gear is switched to a reverse gear at the shift position, and finally the vehicle is backed into a target parking space from the side of the shift position, and the position can also be understood as a lateral parking (starting) position or a position where the vehicle to be parked starts to park into the target parking space, and in fig. 1, the black arrow is taken as an example, and the moving track of the vehicle to be parked is shown.
Fig. 2 is a schematic diagram of a reverse warehouse entry scenario. As shown in fig. 2, taking a parking space in a garage as an example, a vehicle to be parked can travel to a certain shift position in a garage passageway by using a forward gear, and the forward gear is switched to a reverse gear at the shift position, and finally the vehicle is backed into a target parking space from the shift position, where the position can also be understood as a (initial) parking position in the reverse parking space, or a position where the vehicle to be parked starts to park into the target parking space, and in fig. 2, a black arrow is also taken as an example, and a moving track of the vehicle to be parked is shown.
Fig. 3 is a schematic view of a parking scene in a diagonal direction. As shown in fig. 3, taking an oblique parking space of a parking lot as an example, a vehicle to be parked can travel to a certain shift position in a parking lot aisle by using a forward gear, and the forward gear is switched to a reverse gear at the shift position, and finally, the vehicle is obliquely reversed from the shift position into a target parking space, and the position can also be understood as an (initial) parking position of oblique parking, or a position where the vehicle to be parked starts to park into the target parking space, and in fig. 3, a black arrow is also taken as an example, and a moving track of the vehicle to be parked is shown.
Fig. 4 is a schematic diagram illustrating the composition of an automatic parking system according to an embodiment of the present application. As shown in fig. 4, the automatic parking system may include an image pickup device 100 and a calculation processing device 200. The image acquisition apparatus 100 and the calculation processing apparatus 200 may be connected via a wired network or a wireless network.
The image capturing mechanism 100 may be one or more vehicle-mounted cameras provided on the vehicle to be parked in fig. 1 to 3 described above. For example, a back-up video camera or a 360 panoramic camera, etc. Alternatively, the image acquisition apparatus 100 may be an ultrasonic sensing device. For example, the ultrasonic sensing device may include an ultrasonic radar module and an imaging module, the ultrasonic radar module may include an APA ultrasonic radar or a park assist (ultrasonic parking assist, UPA) ultrasonic radar, or the like. Alternatively, the image acquisition apparatus 100 may also be a lidar.
The image capturing device 100 may be used to capture a monitoring image of the surroundings of a vehicle to be parked.
The computing processing device 200 may be an electronic device having a computing processing function, such as a computer or a server. The server may be a single server, or may be a server cluster formed by a plurality of servers. In some implementations, the server cluster may also be a distributed cluster. Optionally, the server may also be implemented on a cloud platform, which may include, for example, a private cloud, public cloud, hybrid cloud, community cloud (community cloud), distributed cloud, inter-cloud, multi-cloud (multi-cloud), and the like, or any combination thereof.
Alternatively, the electronic device may also be a driving computer provided on the vehicle to be parked in fig. 1 to 3, which may also be referred to as an electronic control unit (electronic control unit, ECU). In this case, the ECU may be connected to the above-described in-vehicle camera through an automobile harness (e.g., CAN bus).
Optionally, the computing processing device 200 may also be an Application (APP) installed in the above electronic apparatus to provide an automatic parking function; alternatively, the computing processing device 200 may be a central processing unit (central processing unit, CPU) in the electronic apparatus; alternatively, the computing device 200 may be a functional module for executing the parking path planning method in the electronic apparatus. The embodiments of the present application are not limited in this regard.
The computing device 200 may be configured to determine, according to the monitoring image around the vehicle to be parked collected by the image collecting device 100, scene information of a target parking space of the vehicle to be parked, input the scene information of the target parking space into the path node prediction model by the path node information path node prediction model, and plan a path for the vehicle to be parked to drive into the target parking space according to the path node information output by the path node prediction model. Specific processes may be described in the following method embodiments, and are not described herein.
For simplicity of description, the computing processing device 200 will be described below by taking the above-mentioned electronic device as an example.
Fig. 5 is a schematic diagram of the composition of an electronic device according to an embodiment of the present application. As shown in fig. 5, the electronic device may include a processor 10, a memory 20, a communication line 30, a communication interface 40, and an input-output interface 50.
The processor 10, the memory 20, the communication interface 40, and the input/output interface 50 may be connected by a communication line 30.
The processor 10 is configured to execute instructions stored in the memory 20 to implement a parking path planning method provided in the following embodiments of the present application. The processor 10 may be a central processing unit (central processing unit, CPU), a general purpose processor network processor (network processor, NP), a digital signal processor (digital signal processing, DSP), a microprocessor, a microcontroller, a programmable logic device (programmable logic device, PLD), or any combination thereof. The processor 10 may also be any other apparatus having a processing function, such as a circuit, a device, or a software module, which is not limited in this embodiment. In one example, processor 10 may include one or more CPUs, such as CPU0 and CPU1 in fig. 5. As an alternative implementation, the electronic device may include multiple processors, for example, and may include a processor 60 (illustrated in phantom in fig. 5) in addition to the processor 10.
Memory 20 for storing instructions. For example, the instructions may be a computer program. Alternatively, memory 20 may be a read-only memory (ROM) or other type of static storage device that may store static information and/or instructions, an access memory (random access memory, RAM) or other type of dynamic storage device that may store information and/or instructions, an electrically erasable programmable read-only memory (electrically erasable programmable read-only memory, EEPROM), a compact disc read-only memory (compact disc read-only memory, CD-ROM) or other optical storage, optical storage (including compact disc, laser disc, optical disc, digital versatile disc, blu-ray disc, etc.), magnetic disk storage media, or other magnetic storage devices, etc., as examples of which are not limited in this application.
It should be noted that, the memory 20 may exist separately from the processor 10 or may be integrated with the processor 10. The memory 20 may be located inside the electronic device or outside the electronic device, which is not limited in this embodiment of the present application.
A communication line 30 for communicating information between the components comprised by the electronic device.
A communication interface 40 for communicating with other devices (e.g., the above-described in-vehicle cameras) or other communication networks. The other communication network may be an ethernet, a radio access network (radio access network, RAN), a wireless local area network (wireless local area networks, WLAN), etc. The communication interface 40 may be a module, a circuit, a transceiver, or any device capable of enabling communication.
And an input-output interface 50 for implementing man-machine interaction between the user and the electronic device. Such as enabling action interactions, text interactions, voice interactions, etc. between the user and the electronic device.
The input/output interface 50 may be a touch screen, a keyboard, a mouse, physical keys of a console in a vehicle, or a center control screen of a vehicle, etc., through which action interaction or text interaction between a user and an electronic device may be achieved.
The input output interface 50 may also be an audio module, which may include a speaker, microphone, etc., through which voice interaction between the user and the electronic device may be achieved, for example.
It should be noted that the structure shown in fig. 5 does not constitute a limitation of the electronic device, and the electronic device may include more or less components than those shown in fig. 5, or a combination of some components, or a different arrangement of components.
The following describes a parking path planning method provided in an embodiment of the present application with reference to the accompanying drawings.
Fig. 6 is a flow chart of a parking path planning method according to an embodiment of the present application. Alternatively, the method may be performed by an electronic device having the hardware structure shown in fig. 5, and as shown in fig. 6, the method may include S101 to S103.
S101, the electronic equipment acquires scene information of a target parking space of a vehicle to be parked.
The target parking space can be any parking space without a parked vehicle (or idle). The scene information of the target parking space is used for describing the target parking space.
Optionally, the scene information of the target parking space may include, but is not limited to, at least one of the following: boundary information of obstacles around the target parking space, information of the target parking space, or final state and pose information of the vehicle corresponding to the target parking space, and the like.
The boundary information of the obstacles around the target parking space can be used for describing the drivable area around the target parking space. Boundary information of the target parking space and final vehicle pose information corresponding to the target parking space can be used for describing the position of the target parking space. For example, the vehicle final state pose information corresponding to the target parking space may include the position coordinates (x, y) when the vehicle is parked in the target parking space and the pose information theta of the vehicle body (or referred to as heading information of the vehicle body), that is, the vehicle final state information may be expressed as (x, y, theta) at this time, or the vehicle final state pose information may include only the position coordinates (x, y) when the vehicle is parked in the target parking space, and the x-y plane is parallel to the ground. When the vehicle to be parked is in the final state position information of the vehicle, the vehicle to be parked can be considered to be parked in the target parking space. The vehicle position coordinates may be coordinates of one feature point on the vehicle. For example, the coordinates of the geometric center of the vehicle body or the coordinates of the center of the rear axle (or rear axle differential), etc. Alternatively, the vehicle position coordinates may also be coordinates of a plurality of feature points on the vehicle. For example, the coordinates of the center points of the four wheels, the coordinates of the points of contact of the four wheels with the ground, or the coordinates of points on the boundary line of the vehicle body, etc. Depending on the coordinate system chosen, the (x, y, theta) may have different representations, which are not particularly limited in this embodiment of the present application, and may, for example, choose the coordinate system of the vehicle itself or the world coordinate system.
Illustratively, obstacles around the target parking space may include, but are not limited to, vehicles parked in the parking space adjacent to the target parking space, curb, warning sign, trash can, wall, and road marking, etc.
Fig. 7 is a schematic view of scene information of a target parking space according to an embodiment of the present application. As shown in fig. 7, taking the side parking space scene shown in fig. 1 as an example, the scene information of the target parking space acquired by the electronic device may include boundary information of the target parking space (shown by a dashed box in fig. 7), boundary information of an obstacle around the target parking space (shown by a black thick line in fig. 7), and vehicle final state pose information corresponding to the target parking space (shown by coordinates of a dot with an arrow in fig. 7). The boundary information of the target parking space can be obtained according to road scribing of the target parking space. The obstacles around the target parking space may include demarcation lines of the parked vehicles and non-motor lanes and motor lanes in the parking space around the target parking space, and the like. The final vehicle pose information corresponding to the target parking space may include coordinates (the coordinates include an abscissa x and an ordinate y) of a rear axle center of the vehicle on a plane parallel to the ground, and a body heading theta.
Fig. 8 is a schematic view of scene information of another target parking space according to an embodiment of the present application. As shown in fig. 8, taking the garage parking space scene shown in fig. 2 as an example, the electronic device may also obtain boundary information of the target parking space, boundary information of obstacles around the target parking space, and final state pose information of the vehicle corresponding to the target parking space, which may be specifically described with reference to fig. 7, and will not be described herein.
Fig. 9 is a schematic view of scene information of another target parking space according to an embodiment of the present application. As shown in fig. 9, taking the above-mentioned oblique parking space scenario shown in fig. 3 as an example, the electronic device may also obtain the boundary of the target parking space, the boundary of the obstacle around the target parking space, and the final pose information of the vehicle to be parked, which may be specifically described with reference to fig. 7, and will not be described herein.
It should be understood that when the path node prediction model is utilized to predict the parking path node information, the parking path planning method provided by the embodiment of the application can also consider the boundaries of the obstacles around the target parking space, and the boundaries of different obstacles can correspond to different parking path node information and different parking paths, so that the types of planned parking paths are enriched, and the universality of the path node prediction model is improved. And this scheme can be applicable to various types of parking stall scenes, for example transversely park, vertically park or rhombus parking stall etc.. Compared with the existing path node searching strategy, the method provided by the embodiment of the application can also remarkably reduce the calculated amount.
In a possible implementation manner, as described above, the electronic device may determine the scene information of the target parking space according to the monitoring image around the vehicle to be parked acquired by the image acquisition device 100. For example, the electronic device may identify the monitoring image around the vehicle to be parked acquired by the image acquisition apparatus 100 using a machine vision algorithm, and determine the scene information of the target parking space.
In some possible embodiments, the electronic device obtains the scene information of the target parking space after finding (or determining) the target parking space.
In one possible implementation, as described above, the electronic device may include the input/output interface 50, where the input/output interface 50 may be a physical key of a vehicle center console or a center screen of the vehicle, and the electronic device may receive a click operation of the physical key of the vehicle center console or the center screen of the vehicle by a user, and in response to the click operation, search for an idle parking space, and determine the target parking space according to the idle parking space.
In another possible implementation, as described above, the input/output interface 50 in the electronic device may also be an audio module, where the electronic device may receive a voice command from a user through the audio module, and in response to the voice command, search for an idle parking space, and determine the target parking space according to the idle parking space.
In still another possible implementation manner, the electronic device may be connected to a speed sensor of the vehicle to be parked, and the electronic device may obtain a current speed of the vehicle to be parked through the speed sensor, and when the current speed is lower than a preset speed threshold, automatically trigger the electronic device to search for an idle parking space, and determine the target parking space according to the idle parking space.
Optionally, when the electronic device finds a plurality of idle parking spaces, the electronic device may further display the plurality of idle parking spaces on a central control screen of the vehicle, receive a selection operation of a user on the plurality of idle parking spaces displayed on the central control screen of the vehicle, and determine the target parking space from the plurality of idle parking spaces in response to the selection operation.
S102, the electronic equipment inputs scene information of the target parking space into a path node prediction model, and parking path node information of the vehicle to be parked is determined.
The parking path node information may include position information and posture information of the vehicle to be parked at the parking path node. A park path node is a node that a vehicle to be parked passes through when parking, such as the start of a park path. The position information of the vehicle to be parked at the parking path node may be represented by position coordinates. The attitude information theta of the vehicle to be parked at the parking path node may include a yaw angle (for example, an x-y plane may be a plane parallel to the ground, and a z-axis perpendicular to the x-y plane) of the x, y, z directions of the vehicle to be parked in a preset three-dimensional coordinate system (for example, a three-dimensional world coordinate system), which may also be understood as a heading. The path node prediction model is preset in the electronic equipment, and can be used for representing the corresponding relation between the parking path node information and the scene information of the parking space, or can be used for predicting the parking path node information corresponding to the scene information of the target parking space according to the scene information of the target parking space. The training process of the model can be described with reference to fig. 10 below, and will not be described here.
Alternatively, the path node prediction model may be a neural network model. The neural network model may include a plurality of network modules, and an output layer. For example, the plurality of network modules, and the output layer, may be connected in series.
Wherein each network module includes a full connectivity layer, an activation layer, and a normalization layer. And the full connection layer is used for extracting the characteristics of the scene information of the target parking space. And the activation layer is used for carrying out nonlinear processing on the scene information of the target parking space extracted by the full-connection layer to obtain nonlinear characteristics. And the normalization layer is used for normalizing the nonlinear characteristics. And the output layer is used for outputting parking path node information of the vehicle to be parked according to the nonlinear characteristics after normalization processing of the normalization layer.
Illustratively, the correspondence between parking path node information and scene information of a parking space may be as shown in table 1 below.
TABLE 1
As shown in table 1, the table may include scene information items, parking path node information items, location information items, and attitude information items of the parking space. The parking space scene information items may include information such as scene 1, scene 2, scene 3, etc., the parking path node information items may include information 1, information 2, information 3, etc., the position information items may include position information such as position 1, position 2, position 3, etc., and the posture information items may include posture information such as direction 1, direction 2, direction 3, etc. Scene 1, information 1, position 1, and direction 1 have a correspondence. Scene 2, information 2, position 2, and direction 2 have a correspondence. Scene 3, information 3, position 3, and direction 3 have a correspondence. Scene 4, information 4, position 4, and direction 4 have a correspondence.
S103, the electronic equipment plans a path of the vehicle to be parked to enter the target parking space according to the parking path node information.
Optionally, the electronic device may utilize a planning algorithm to plan the path of the vehicle to be parked to drive into the target parking space. The planning algorithm may be described with reference to the related art, and will not be described here again.
In one possible implementation, the park path node comprises a shift node for the vehicle to be parked. In this case, the method may further include: the electronic equipment controls the vehicle to be parked to travel to a position corresponding to the parking path node information, and controls the vehicle to be parked to be in a reverse gear state at the position and to be in a posture corresponding to the parking path node information; and under the reverse gear state and the gesture corresponding to the parking path node information, the electronic equipment controls the vehicle to be parked to drive into the target parking space according to the planned path.
According to the parking path planning method provided by the embodiment of the application, the scene information of the target parking space can be obtained, the target parking path node information corresponding to the scene information of the target parking space is predicted according to the path node prediction model for representing the corresponding relation between the parking path node information and the scene information of the parking space, the path node can be a gear shifting node (for example, a first gear shifting node in the parking process), the target parking path node information and the scene information of the target parking space have a pre-trained matching relation, the path quality of a planned parking path is higher from the position corresponding to the parking path node information according to the gesture corresponding to the parking path node information, and the efficiency of vehicle parking is improved as a whole.
In some possible embodiments, the electronic device may also obtain a path node prediction model prior to S101.
The path node prediction model may be obtained by training the neural network in advance by the electronic device according to a training sample. Alternatively, the path node prediction model may be obtained by training the neural network in advance by any computing device with computing processing capability according to a training sample and sent to the electronic device, or transferred to the electronic device through an intermediate storage medium. The embodiments of the present application are not limited in this regard.
Alternatively, taking an example that the electronic device trains the neural network according to the training sample in advance to obtain the path node prediction model, referring to fig. 10, fig. 10 is a flowchart of a training method of the path node prediction model according to the embodiment of the present application. As shown in fig. 10, the method may further include S201 to S202 before S101.
S201, the electronic equipment acquires a training sample set.
Wherein the training sample set may comprise a plurality of training samples. Each training sample may include scene information for a parking space and a tag corresponding to the scene information for the parking space. The tag includes parking path node information during parking of the vehicle. The park path node information includes position information and attitude information of the vehicle at the park path node.
In a possible implementation manner, S201 may be specifically implemented as: the method comprises the steps that electronic equipment obtains scene information of a plurality of parking spaces and determines candidate areas of each parking space; the electronic equipment selects a plurality of pieces of position information and preset posture information of the vehicle corresponding to each piece of position information in a parking state from candidate areas of each parking space to serve as candidate parking path node information corresponding to each parking space; the method comprises the steps that the electronic equipment plans a parking planning path corresponding to each candidate parking path node information based on the candidate parking path node information; and the electronic equipment screens out labels corresponding to the parking spaces from the plurality of candidate parking path node information corresponding to the parking spaces based on the parking planning paths corresponding to the candidate parking path node information.
Wherein the candidate region may be set by a manager based on human experience.
Exemplary, fig. 11 is a schematic diagram of a candidate region provided in an embodiment of the present application. As shown in fig. 11, also taking the side parking space scenario shown in fig. 7 as an example, the candidate area may be set at a position flush with the previous parking space of the target parking space, and a plurality of positions may be included in the candidate area. X is x 0 An abscissa, y, representing a position within the candidate region 0 Representing the ordinate of the position theta 0 And the preset gesture information (heading) of the vehicle corresponding to the position in the parking state is indicated.
It should be noted that, for a certain position (i.e., a dot in fig. 11) in the candidate area, the position and different preset posture information corresponding to the position (i.e., an arrow connected to the dot in fig. 11) may constitute different candidate parking path node information (different arrows connected to the same dot are illustrated in fig. 11).
Illustratively, fig. 12 is a schematic diagram of another candidate region provided in an embodiment of the present application. As shown in fig. 12, fig. 12 shows a schematic diagram of candidate areas in a garage parking space scene, and specific details thereof may be described with reference to fig. 11, which is not repeated herein.
Illustratively, fig. 13 is a schematic diagram of yet another candidate region provided in an embodiment of the present application. As shown in fig. 13, fig. 13 shows a schematic diagram of a candidate area in a rhombic parking space scene, and specific details may also be described with reference to fig. 11, which is not repeated here.
Optionally, the electronic device acquiring scene information of the plurality of parking spaces may include: the electronic equipment randomly generates scene information of a plurality of parking spaces. For example, the electronic device may randomly generate scene information of a plurality of parking spaces according to manual experience information input by a manager.
Optionally, after planning the parking planning path corresponding to each candidate parking path node information based on each candidate parking path node information, the electronic device may further perform quality evaluation on each parking planning path, that is, the electronic device may determine the quality of the parking planning path corresponding to each candidate parking path node information based on at least one of the shift times, the path length, the path curvature, and other information related to the reflected path quality of the vehicle during the process of driving into the parking space according to the parking planning path corresponding to each candidate parking path node information. In this case, the electronic device may screen, based on the parking plan path corresponding to each of the candidate parking path node information, a tag corresponding to each of the parking spaces from the plurality of candidate parking path node information corresponding to each of the parking spaces, and may include: and the electronic equipment determines the candidate parking path node information corresponding to the condition that the quality of the parking planning path is greater than a preset quality threshold value or the quality is highest from the plurality of candidate parking path node information corresponding to each parking space as a label of each parking space.
Optionally, the quality of the parking planning path corresponding to each candidate parking path node information may be determined according to the number of gear shifts, the path length, and the path curvature of the vehicle during the parking space driving, and weights corresponding to the number of gear shifts, the path length, and the path curvature. For example, the relationship between the path quality of the parking planned path and the number of shifts, the path length, and the path curvature of the vehicle during the entry into the parking space according to the parking planned path may satisfy the following equation (1).
Q=α (-T) +β (-L) +γ (-K) formula (1)
In the formula (1), Q represents the path quality of the parking path. T represents the number of shifts of the parking path. Alpha represents the weight of the number of shifts of the parking path in the path quality of the parking path. L denotes the length of the parking path. Beta represents the weight that the length of the parking path occupies in the path quality of the parking path. K represents the curvature of the parking path. Gamma denotes the weight of the curvature of the parking path in the path quality of the parking path. The α, β, and γ may be preset in the electronic apparatus by a manager, and each of the α, β, and γ is greater than 0 and less than 1, and the sum of the α, β, and γ is 1.
It should be understood that, when obtaining the label in the training sample, the embodiment of the application may select the candidate parking path node information corresponding to the parking planning path with higher path quality as the label, and the parking path node information predicted by using the path node prediction model obtained by training the training sample with the label may also have the characteristic of the label, that is, the path quality of the parking path corresponding to the predicted parking path node information is higher (for example, the path quality may include at least one of a small number of gear shifts, a short parking path and a small curvature of the parking path).
In the embodiment of the application, the path node prediction model is obtained through training and is used for predicting the parking path node information on the parking path, compared with the existing path node search strategy, the calculation amount can be remarkably reduced, the types of the parking path which can be planned are enriched, the quality of path planning (comprising the reduction of gear shifting times, the shortening of path length and the like) is facilitated to be improved, and the model has strong universality for various parking space scenes.
Fig. 14 is a schematic diagram of a method for obtaining a label of a training sample according to an embodiment of the present application. As shown in fig. 14, the method may include: generating scene information of the parking space, setting a candidate area, randomly generating node information of candidate parking paths, planning the parking paths, evaluating path quality of the parking paths and outputting labels, wherein the six parts can be described by referring to the above step S201 and are not repeated here.
S202, the electronic equipment trains a preset model based on the training sample set to obtain a path node prediction model.
The preset model may also be a neural network model, and the structure of the neural network model may be described with reference to the path node information prediction model structure, which is not described herein.
Alternatively, as described above, the training sample set may include a plurality of training samples. The electronic device may input one or more training samples into the neural network each time to obtain a predicted value (parking path node information corresponding to scene information of a certain parking space predicted by the neural network), calculate a loss function (loss) according to the predicted value and a label in the training sample (parking path node information corresponding to scene information of the certain parking space), and adjust parameters of a preset model. And in this way, a plurality of training samples in the training sample set are input into a preset model one by one, and the training is iterated until the preset model converges.
Alternatively, the conditions for the preset model convergence (end training) may include: the electronic equipment inputs the training sample to the preset model for times reaching a preset time threshold, or the error of the predicted value and the label is smaller than a preset error threshold.
The foregoing description of the solution provided in the embodiments of the present application has been mainly presented in terms of a method. To achieve the above functions, it includes corresponding hardware structures and/or software modules that perform the respective functions. Those of skill in the art will readily appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein may be implemented as hardware or combinations of hardware and computer software. Whether a function is implemented as hardware or computer software driven hardware depends upon the particular application and design constraints imposed on the solution. The technical aim may be to use different methods to implement the described functions for each particular application, but such implementation should not be considered beyond the scope of the present application.
In an exemplary embodiment, the embodiment of the present application further provides an automatic parking device, which may be applied to the electronic apparatus shown in fig. 5 described above. Fig. 15 is a schematic diagram of a parking path planning apparatus according to an embodiment of the present application. As shown in fig. 15, the apparatus may include an acquisition module 1501 and a processing module 1502. The acquisition module 1501 is connected to the processing module 1502.
An obtaining module 1501, configured to obtain scene information of a target parking space of a vehicle to be parked; the scene information of the target parking space is used for describing the target parking space. A processing module 1502, configured to input scene information of a target parking space into a path node prediction model, and determine parking path node information of a vehicle to be parked; the path node prediction model is used for representing the corresponding relation between parking path node information and scene information of a parking space, and the parking path node information comprises position information and posture information of a vehicle to be parked at the parking path node; and planning a path of the vehicle to be parked to enter the target parking space according to the parking path node information.
In some possible embodiments, the scene information of the target parking space includes at least one of the following: boundary information of obstacles around the target parking space, boundary information of the target parking space or final state pose information of the vehicle corresponding to the target parking space; boundary information of obstacles around the target parking space is used for describing a drivable area around the target parking space; the boundary information of the target parking space and the final pose information of the vehicle corresponding to the target parking space are used for describing the position of the target parking space.
In other possible embodiments, the park path node comprises a shift node for a vehicle to be parked. The processing module 1502 is further configured to control the vehicle to be parked to a position corresponding to the parking path node information, and control the vehicle to be parked to be in a reverse gear state at the position and to be in a posture corresponding to the parking path node information; and controlling the vehicle to be parked to enter the target parking space according to the planned path under the reverse gear state and the gesture corresponding to the parking path node information.
In still other possible embodiments, the path node prediction model is obtained by training a preset model based on a training sample set, where the training sample set includes a plurality of training samples, each training sample includes scene information of a parking space and a tag corresponding to the scene information of the parking space, and the tag includes parking path node information of the vehicle.
In still other possible embodiments, the label corresponding to each parking space is obtained by screening a parking planning path corresponding to each candidate parking path node information from a plurality of candidate parking path node information corresponding to each parking space; the parking planning path corresponding to the node information of each candidate parking path is a path of a vehicle driving into a parking space, which is planned based on the node information of each candidate parking path; the plurality of candidate parking path node information corresponding to each parking space comprises a plurality of position information in the candidate area and preset posture information of the vehicle corresponding to each position information in a parking state.
In still other possible embodiments, the label corresponding to each parking space includes candidate parking path node information corresponding to a case that the quality of the parking planning path is greater than a preset quality threshold or the quality is highest; the quality of the parking plan path is determined based on at least one of a number of shifts of the vehicle, a path length, and a path curvature during entry into the parking space according to the parking plan path.
In still other possible embodiments, the path node prediction model is a neural network model; the neural network model comprises a plurality of network modules and an output layer, wherein each network module comprises a full-connection layer, an activation layer and a normalization layer; the full-connection layer is used for extracting the characteristics of scene information of the target parking space; the activation layer is used for carrying out nonlinear processing on the characteristics of the scene information of the target parking space extracted by the full-connection layer to obtain nonlinear characteristics; the normalization layer is used for normalizing the nonlinear characteristics; and the output layer is used for outputting parking path node information of the vehicle to be parked according to the nonlinear characteristics after normalization processing of the normalization layer.
In an exemplary embodiment, the embodiment of the present application further provides a training device for a path node prediction model, where the device may be applied to the electronic device shown in fig. 5. Fig. 16 is a schematic diagram of a training device of a path node prediction model according to an embodiment of the present application. As shown in fig. 16, the apparatus includes an acquisition module 1601 and a processing module 1602.
The obtaining module 1601 is configured to obtain a training sample set, where the training sample set includes a plurality of training samples, each training sample includes scene information of a parking space and a tag corresponding to the scene information of the parking space, the tag includes parking path node information in a parking process of a vehicle, and the parking path node information includes position information and posture information of the vehicle at the parking path node.
The processing module 1602 is configured to train a preset model based on the training sample set to obtain a path node information prediction model.
In some possible embodiments, the acquiring module 1601 is specifically configured to acquire scene information of a plurality of parking spaces, and determine a candidate area of each parking space; selecting a plurality of position information and preset posture information of a vehicle corresponding to each position information in a parking state from candidate areas of each parking space as candidate parking path node information corresponding to each parking space; planning a parking planning path corresponding to each candidate parking path node information based on each candidate parking path node information; and screening the labels corresponding to the parking spaces from the plurality of candidate parking path node information corresponding to the parking spaces based on the parking planning paths corresponding to the candidate parking path node information.
In other possible embodiments, the obtaining module 1601 is further configured to determine a quality of the parking plan path corresponding to each candidate parking path node information based on at least one of a shift number, a path length, and a path curvature of the vehicle during driving into the parking space according to the parking plan path corresponding to each candidate parking path node information. The obtaining module 1601 is specifically configured to determine, from a plurality of candidate parking path node information corresponding to each parking space, candidate parking path node information corresponding to a case where the quality of the parking planning path is greater than a preset quality threshold or the quality is highest, as a label of each parking space.
In still other possible embodiments, the predetermined model is a neural network model; the neural network model comprises a plurality of network modules and an output layer, wherein each network module comprises a full-connection layer, an activation layer and a normalization layer; the full-connection layer is used for extracting the characteristics of the scene information of the parking spaces in the training samples; the activation layer is used for carrying out nonlinear processing on the characteristics of the scene information of the parking spaces in the training samples extracted by the full-connection layer to obtain nonlinear characteristics; the normalization layer is used for normalizing the nonlinear characteristics; and the output layer is used for outputting parking path node information according to the nonlinear characteristics after normalization processing of the normalization layer.
It should be noted that the division of the modules in fig. 15 and 16 is illustrative, and is merely a logic function division, and other division manners may be implemented in practice. For example, two or more functions may also be integrated in one processing module. The integrated modules may be implemented in hardware or in software functional modules.
In an exemplary embodiment, a computer-readable storage medium is also provided, including computer-executable instructions that, when executed on an electronic device, cause the electronic device to perform any of the methods provided by the above embodiments.
In an exemplary embodiment, the present application also provides a computer program product comprising computer-executable instructions, which, when run on an electronic device, cause the electronic device to perform any of the methods provided by the above embodiments.
In an exemplary embodiment, the present application further provides a chip, including: a processor and an interface, the processor being coupled to the memory through the interface, the processor, when executing the computer program or the electronic device in the memory, executing instructions, causing any one of the methods provided by the embodiments described above to be performed.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using a software program, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer-executable instructions. When the computer-executable instructions are loaded and executed on a computer, the processes or functions in accordance with embodiments of the present application are fully or partially produced. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer-executable instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, from one website, computer, server, or data center by wired (e.g., coaxial cable, fiber optic, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). Computer readable storage media can be any available media that can be accessed by a computer or data storage devices including one or more servers, data centers, etc. that can be integrated with the media. The usable medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a Solid State Disk (SSD)), or the like.
Although the present application has been described herein in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed application, from a review of the figures, the disclosure, and the appended claims. In the claims, the word "Comprising" does not exclude other elements or steps, and the "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.
Although the present application has been described in connection with specific features and embodiments thereof, it will be apparent that various modifications and combinations can be made without departing from the spirit and scope of the application. Accordingly, the specification and drawings are merely exemplary illustrations of the present application as defined in the appended claims and are considered to cover any and all modifications, variations, combinations, or equivalents that fall within the scope of the present application. It will be apparent to those skilled in the art that various modifications and variations can be made in the present application without departing from the spirit or scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims and the equivalents thereof, the present application is intended to cover such modifications and variations.
The foregoing is merely a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions within the technical scope of the present disclosure should be covered in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (15)

1. A method of parking path planning, the method comprising:
acquiring scene information of a target parking space of a vehicle to be parked; the scene information of the target parking space is used for describing the target parking space;
inputting the scene information of the target parking space into a path node prediction model, and determining parking path node information of the vehicle to be parked; the path node prediction model is used for representing the corresponding relation between parking path node information and scene information of a parking space, and the parking path node information comprises position information and posture information of the vehicle to be parked at the parking path node;
and planning a path of the vehicle to be parked to enter the target parking space according to the parking path node information.
2. The method of claim 1, wherein the scene information of the target parking space comprises at least one of: boundary information of obstacles around the target parking space, boundary information of the target parking space or final state and pose information of the vehicle corresponding to the target parking space;
The boundary information of the obstacles around the target parking space is used for describing a drivable area around the target parking space;
and the boundary information of the target parking space and the final pose information of the vehicle corresponding to the target parking space are used for describing the position of the target parking space.
3. The method of claim 1, wherein the park path node comprises a shift node of the vehicle to be parked; the method further comprises the steps of:
controlling the vehicle to be parked to travel to a position corresponding to the parking path node information, and controlling the vehicle to be parked to be in a reverse gear state at the position and to be in a posture corresponding to the parking path node information;
and controlling the vehicle to be parked to enter the target parking space according to the planned path under the reverse gear state and the gesture corresponding to the parking path node information.
4. The method of claim 1, wherein the path node prediction model is obtained by training a preset model based on a training sample set, the training sample set including a plurality of training samples, each of the training samples including scene information of a parking space and a tag corresponding to the scene information of the parking space, the tag including parking path node information of a vehicle.
5. The method of claim 4, wherein the label corresponding to each parking space is obtained by screening a parking planning path corresponding to each candidate parking path node information from a plurality of candidate parking path node information corresponding to each parking space;
the parking planning path corresponding to the node information of each candidate parking path is a path of a vehicle entering a parking space, which is planned based on the node information of each candidate parking path;
the plurality of candidate parking path node information corresponding to each parking space comprises a plurality of position information in a candidate area and preset posture information of the vehicle corresponding to each position information in a parking state.
6. The method according to claim 5, wherein the label corresponding to each parking space includes candidate parking path node information corresponding to a case where the quality of the parking planned path is greater than a preset quality threshold or the quality is highest;
the quality of the parking planned path is determined based on at least one of a number of shifts of the vehicle, a path length, and a path curvature during driving into the parking space according to the parking planned path.
7. The method of any one of claims 1-6, wherein the path node prediction model is a neural network model;
The neural network model comprises a plurality of network modules and an output layer, wherein each network module comprises a full connection layer, an activation layer and a normalization layer;
the full connection layer is used for extracting the characteristics of the scene information of the target parking space;
the activation layer is used for carrying out nonlinear processing on the characteristics of the scene information of the target parking space extracted by the full connection layer to obtain nonlinear characteristics;
the normalization layer is used for normalizing the nonlinear characteristics;
and the output layer is used for outputting the parking path node information of the vehicle to be parked according to the nonlinear characteristics after normalization processing of the normalization layer.
8. A method of training a path node predictive model, the method comprising:
acquiring a training sample set, wherein the training sample set comprises a plurality of training samples, each training sample comprises scene information of a parking space and a label corresponding to the scene information of the parking space, the label comprises parking path node information in the process of parking a vehicle, and the parking path node information comprises position information and gesture information of the vehicle at the parking path node;
And training a preset model based on the training sample set to obtain a path node information prediction model.
9. The method of claim 8, wherein the acquiring a training sample set comprises:
acquiring scene information of a plurality of parking spaces, and determining candidate areas of each parking space;
selecting a plurality of position information and preset posture information of a vehicle corresponding to each position information in a parking state from the candidate area of each parking space as candidate parking path node information corresponding to each parking space;
planning a parking planning path corresponding to each candidate parking path node information based on each candidate parking path node information;
and screening a label corresponding to each parking space from the plurality of candidate parking path node information corresponding to each parking space based on the parking planning path corresponding to each candidate parking path node information.
10. The method of claim 9, wherein after planning a parking plan path corresponding to each candidate parking path node information based on the each candidate parking path node information, the method further comprises:
Determining the quality of a parking planning path corresponding to each candidate parking path node information based on at least one of the shift times, the path length and the path curvature of a vehicle in the process of entering a parking space according to the parking planning path corresponding to each candidate parking path node information;
the step of screening the label corresponding to each parking space from the plurality of candidate parking path node information corresponding to each parking space based on the parking planning path corresponding to each candidate parking path node information comprises the following steps:
and determining the candidate parking path node information corresponding to the situation that the quality of the parking planning path is greater than a preset quality threshold or the quality is highest from the plurality of candidate parking path node information corresponding to each parking space as a label of each parking space.
11. The method of claim 8, wherein the predetermined model is a neural network model;
the neural network model comprises a plurality of network modules and an output layer, wherein each network module comprises a full connection layer, an activation layer and a normalization layer;
the full connection layer is used for extracting the characteristics of scene information of the parking spaces in the training samples;
The activation layer is used for carrying out nonlinear processing on the characteristics of the scene information of the parking space in the training sample extracted by the full connection layer to obtain nonlinear characteristics;
the normalization layer is used for normalizing the nonlinear characteristics;
and the output layer is used for outputting parking path node information according to the nonlinear characteristics after normalization processing of the normalization layer.
12. A parking path planning apparatus, the apparatus comprising: the device comprises an acquisition module and a processing module;
the acquisition module is used for acquiring scene information of a target parking space of the vehicle to be parked; the scene information of the target parking space is used for describing the target parking space;
the processing module is used for inputting the scene information of the target parking space into a path node prediction model and determining parking path node information of the vehicle to be parked; the path node prediction model is used for representing the corresponding relation between parking path node information and scene information of a parking space, and the parking path node information comprises position information and posture information of the vehicle to be parked at the parking path node; and planning a path of the vehicle to be parked to enter the target parking space according to the parking path node information.
13. A training apparatus for a path node predictive model, the apparatus comprising:
the system comprises an acquisition module, a storage module and a storage module, wherein the acquisition module acquires a training sample set, the training sample set comprises a plurality of training samples, each training sample comprises scene information of a parking place and a label corresponding to the scene information of the parking place, the label comprises parking path node information in the process of parking a vehicle, and the parking path node information comprises position information and gesture information of the vehicle at the parking path node;
and the processing module is used for training a preset model based on the training sample set to obtain a path node information prediction model.
14. An electronic device, the electronic device comprising: a processor and a memory;
the memory stores instructions executable by the processor;
the processor is configured to, when executing the instructions, cause the electronic device to implement the method of any of claims 1-7 or claims 8-11.
15. A computer-readable storage medium, the computer-readable storage medium comprising: computer software instructions;
the computer software instructions, when run in an electronic device, cause the electronic device to implement the method of any one of claims 1-7 or claims 8-11.
CN202210832136.3A 2022-07-15 2022-07-15 Parking path planning method and path node prediction model training method Pending CN117445901A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210832136.3A CN117445901A (en) 2022-07-15 2022-07-15 Parking path planning method and path node prediction model training method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210832136.3A CN117445901A (en) 2022-07-15 2022-07-15 Parking path planning method and path node prediction model training method

Publications (1)

Publication Number Publication Date
CN117445901A true CN117445901A (en) 2024-01-26

Family

ID=89589732

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210832136.3A Pending CN117445901A (en) 2022-07-15 2022-07-15 Parking path planning method and path node prediction model training method

Country Status (1)

Country Link
CN (1) CN117445901A (en)

Similar Documents

Publication Publication Date Title
US11726477B2 (en) Methods and systems for trajectory forecasting with recurrent neural networks using inertial behavioral rollout
WO2022052406A1 (en) Automatic driving training method, apparatus and device, and medium
EP3819194B1 (en) Automatic parking method and apparatus, and electronic device
US20220340127A1 (en) Automatic parking control method and apparatus
CN111506058B (en) Method and device for planning a short-term path for autopilot by means of information fusion
CN112133089B (en) Vehicle track prediction method, system and device based on surrounding environment and behavior intention
EP4152204A1 (en) Lane line detection method, and related apparatus
CN107284443A (en) Detect available parking places
CN108602509A (en) The method and system of automatic driving vehicle is operated based on motion planning
CN112249032B (en) Automatic driving decision method, system, equipment and computer storage medium
JP2021515724A (en) LIDAR positioning to infer solutions using 3DCNN network in self-driving cars
CN110920611A (en) Vehicle control method and device based on adjacent vehicles
RU2750243C2 (en) Method and system for generating a trajectory for a self-driving car (sdc)
CN113631452B (en) Lane change area acquisition method and device
US11403947B2 (en) Systems and methods for identifying available parking spaces using connected vehicles
RU2757234C2 (en) Method and system for calculating data for controlling the operation of a self-driving car
CN114475656A (en) Travel track prediction method, travel track prediction device, electronic device, and storage medium
CN117445901A (en) Parking path planning method and path node prediction model training method
US20230194268A1 (en) Longitudinal-limited lane marking localizer
US20230192077A1 (en) Adjustment of object trajectory uncertainty by an autonomous vehicle
CN114511834A (en) Method and device for determining prompt information, electronic equipment and storage medium
CN114435347B (en) Method, device, equipment and storage medium for determining parking track
US20230221408A1 (en) Radar multipath filter with track priors
US20220172062A1 (en) Measuring confidence in deep neural networks
EP4102473A1 (en) Method and device for operating a self-driving car

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination