Automatic parking method and device, vehicle control unit and storage medium

Abstract

The invention provides an automatic parking method and device, a vehicle control unit and a storage medium, and belongs to the technical field of automatic driving. The method comprises the steps that scene data are obtained, the scene data comprise ultrasonic data of multiple detection points corresponding to a vehicle, and the ultrasonic data of any detection point are data collected by an ultrasonic sensor for the detection points; the scene data are input into a target prediction model for prediction, prediction attribute values of all the detection points are obtained, the target prediction model is used for determining the prediction attribute values based on the ultrasonic data, and the prediction attribute value of any detection point indicates whether the predicted detection point is located in an area where an obstacle of the vehicle is located or not; determining a prediction boundary of the obstacle based on the prediction attribute value of each detection point; and based on the predicted boundary of the obstacle, the vehicle is controlled to be parked in the target parking space, and the target parking space is located outside the area where the obstacle is located. In this way, the accuracy and precision of determining the boundary of the obstacle are improved, and therefore the parking efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and particularly to an automatic parking method, device, vehicle controller, and storage medium. Background Art

[0002] With the continuous development of intelligent vehicles, the application of automatic parking is becoming more and more widespread. Vehicles equipped with automatic parking function can detect the boundary of obstacles through various vehicle-mounted sensors and vehicle controllers without manual intervention, and thus complete the process of parking into the garage. However, there is always a deviation between the detected obstacle boundary and the actual obstacle boundary, which will lead to too many times of parking adjustment and reduce the parking efficiency. Summary of the Invention

[0003] Embodiments of this application provide an automatic parking method, device, vehicle controller, and storage medium, which improve the accuracy of the detected obstacle boundary, and thus improve the parking efficiency. The technical solutions are as follows:

[0004] On the one hand, an automatic parking method is provided, and the method includes:

[0005] Obtain scene data, where the scene data includes ultrasonic data of a plurality of detection points corresponding to the vehicle, and the ultrasonic data of any detection point is the data collected by an ultrasonic sensor for the detection point;

[0006] Input the scene data into a target prediction model for prediction to obtain predicted attribute values of each detection point. The target prediction model is used to determine the predicted attribute values based on the ultrasonic data, and the predicted attribute value of any detection point indicates whether the predicted detection point is located in the area where the obstacle of the vehicle is located;

[0007] Based on the predicted attribute values of each detection point, determine the predicted boundary of the obstacle;

[0008] Based on the predicted boundary of the obstacle, control the vehicle to park into a target parking space, and the target parking space is located outside the area where the obstacle is located.

[0009] In some embodiments, the determining the predicted boundary of the obstacle based on the predicted attribute values of each detection point includes:

[0010] Based on the predicted attribute values of each detection point, screen the plurality of detection points to obtain a plurality of effective detection points, and the plurality of effective detection points are all located in the area where the obstacle is located;

[0011] Based on the plurality of effective detection points, determine the predicted boundary of the obstacle.

[0012] In some embodiments, screening the multiple detection points based on the predicted attribute values of the respective detection points to obtain multiple valid detection points includes:

[0013] When the predicted attribute value of any detection point is not greater than a preset threshold, discard the detection point;

[0014] When the predicted attribute value of any detection point is greater than the preset threshold, determine the detection point as a valid detection point.

[0015] In some embodiments, screening the multiple detection points based on the predicted attribute values of the respective detection points to obtain multiple valid detection points includes:

[0016] When the predicted attribute value of any detection point is a first preset value, discard the detection point;

[0017] When the predicted attribute value of any detection point is a second preset value, determine the detection point as a valid detection point.

[0018] In some embodiments, the training steps of the target prediction model are as follows:

[0019] Obtain a training data set, where the training data set includes ultrasonic data of multiple detection points corresponding to a training vehicle and label attribute values of the multiple detection points, and the label attribute value of any detection point indicates whether the detection point is located in the area where an obstacle of the vehicle is located;

[0020] Input the training data set into an initial prediction model for prediction to obtain the predicted attribute values of the respective detection points;

[0021] Update the parameters of the initial prediction model based on the predicted attribute values of the respective detection points and the label attribute values of the respective detection points to obtain the target prediction model.

[0022] In some embodiments, the method further includes:

[0023] Obtain the laser data of the training vehicle, where the laser data is data collected by a laser sensor for the surrounding environment of the training vehicle;

[0024] Based on the laser data, determine the laser boundary of the obstacle;

[0025] Based on the spatial relationship between the laser boundary and the multiple detection points, add respective label attribute values to the multiple detection points.

[0026] In some embodiments, adding respective label attribute values to the multiple detection points based on the spatial relationship between the laser boundary and the multiple detection points includes:

[0027] When any detection point is located outside the laser boundary, add a first attribute value to the detection point, where the first attribute value indicates that the detection point is outside the area where the obstacle is located;

[0028] When any detection point is located inside the laser boundary, add a second attribute value to the detection point, where the second attribute value indicates that the detection point is inside the area where the obstacle is located.

[0029] On the other hand, an automatic parking device is provided, and the device includes:

[0030] A first acquisition module, configured to acquire scene data, where the scene data includes ultrasonic data of a plurality of detection points corresponding to the vehicle, and the ultrasonic data of any detection point is data collected by an ultrasonic sensor for the detection point;

[0031] A prediction module, configured to input the scene data into a target prediction model for prediction to obtain prediction attribute values of each detection point, where the target prediction model is used to determine prediction attribute values based on ultrasonic data, and the prediction attribute value of any detection point indicates whether the predicted detection point is inside the area where the obstacle of the vehicle is located;

[0032] A first determination module, configured to determine a predicted boundary of the obstacle based on the prediction attribute values of each detection point;

[0033] A control module, configured to control the vehicle to park in a target parking space based on the predicted boundary of the obstacle, where the target parking space is outside the area where the obstacle is located.

[0034] In some embodiments, the first determination module includes:

[0035] A screening unit, configured to screen the plurality of detection points based on the prediction attribute values of each detection point to obtain a plurality of valid detection points, and the plurality of valid detection points are all inside the area where the obstacle is located;

[0036] A determination unit, configured to determine a predicted boundary of the obstacle based on the plurality of valid detection points.

[0037] In some embodiments, the screening unit is configured to discard a detection point when the prediction attribute value of any detection point is not greater than a preset threshold; and determine the detection point as a valid detection point when the prediction attribute value of any detection point is greater than the preset threshold.

[0038] In some embodiments, the screening unit is configured to discard a detection point when the predicted attribute value at any detection point is a first preset value; and determine the detection point as a valid detection point when the predicted attribute value at any detection point is a second preset value.

[0039] In some embodiments, the training device of the target prediction model includes:

[0040] A second acquisition module, configured to acquire a training data set, where the training data set includes ultrasonic data of a plurality of detection points corresponding to a training vehicle and label attribute values of the plurality of detection points, and the label attribute value of any detection point indicates whether the detection point is located in the area where an obstacle of the vehicle is located;

[0041] A processing module, configured to input the training data set into an initial prediction model for prediction to obtain predicted attribute values of each detection point;

[0042] An update module, configured to update parameters of the initial prediction model based on the predicted attribute values of each detection point and the label attribute values of each detection point to obtain the target prediction model.

[0043] In some embodiments, the device further includes:

[0044] A third acquisition module, configured to acquire laser data of the training vehicle, where the laser data is data collected by a laser sensor for the surrounding environment of the training vehicle;

[0045] A second determination module, configured to determine a laser boundary of the obstacle based on the laser data;

[0046] An addition module, configured to add respective label attribute values to the plurality of detection points based on a spatial relationship between the laser boundary and the plurality of detection points.

[0047] In some embodiments, the addition module is configured to add a first attribute value to a detection point when the detection point is outside the laser boundary, where the first attribute value indicates that the detection point is outside the area where the obstacle is located; and add a second attribute value to the detection point when the detection point is inside the laser boundary, where the second attribute value indicates that the detection point is inside the area where the obstacle is located.

[0048] On the other hand, a vehicle controller is provided, and the vehicle controller includes a main control module. The main control module includes a processor and a memory. The memory is used to store at least one segment of computer program, and the at least one segment of computer program is loaded and executed by the processor to implement the automatic parking method in the embodiments of the present application.

[0049] On the other hand, a computer-readable storage medium is provided, in which at least one segment of computer program is stored, and the at least one segment of computer program is loaded and executed by a processor to implement the automatic parking method in the embodiments of the present application.

[0050] On the other hand, a computer program product is provided, including a computer program, and the computer program is executed by a processor to implement the automatic parking method in the embodiments of the present application.

[0051] The present application provides an automatic parking method. Through a target prediction model, prediction attribute values of each detection point are obtained based on ultrasonic data of each detection point, and thus each detection point is screened based on the prediction attribute values. This method can filter out noise points among multiple detection points, avoiding interference of noise points on the process of determining the obstacle boundary and improving the accuracy of the predicted obstacle boundary. Compared with the traditional method, the present application determines an obstacle boundary with a smaller actual deviation, greatly reducing the number of times of adjusting the steering wheel while parking, enabling the vehicle to park in the parking space with fewer adjustments, reducing the time consumption of the entire automatic parking process, and greatly improving the parking efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] To clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0053] Figure 1 is a schematic diagram of the implementation environment of an automatic parking method provided by an embodiment of the present application;

[0054] Figure 2 is a flowchart of an automatic parking method provided by an embodiment of the present application;

[0055] Figure 3 is a flowchart of another automatic parking method provided by an embodiment of the present application;

[0056] Figure 4 is a schematic diagram of the installation positions of ultrasonic sensors provided by an embodiment of the present application;

[0057] Figure 5 is a schematic diagram of an obstacle boundary provided by an embodiment of the present application;

[0058] Figure 6 is a flowchart of model training provided by an embodiment of the present application;

[0059] Figure 7It is a flowchart of boundary prediction provided by an embodiment of the present application;

[0060] Figure 8 It is a schematic diagram of predicting a boundary provided by an embodiment of the present application;

[0061] Figure 9 It is a block diagram of an automatic parking device provided by an embodiment of the present application;

[0062] Figure 10 It is a schematic structural diagram of a vehicle controller provided by an embodiment of the present application. Detailed implementation manners

[0063] To make the objectives, technical solutions and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0064] In the present application, terms such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and effects. It should be understood that there is no logical or chronological dependency between "first", "second", and "nth", nor are the quantity and execution order limited.

[0065] In the present application, the term "at least one" means one or more, and the meaning of "a plurality" means two or more.

[0066] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.) and signals involved in the present application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of relevant countries and regions. For example, the state parameters of the vehicle involved in the present application are obtained under full authorization.

[0067] Figure 1 It is a schematic diagram of the implementation environment of an automatic parking method provided by an embodiment of the present application. Refer to Figure 1 , this implementation environment includes a parameter acquisition device 101 and a vehicle controller 102, and this implementation environment is deployed in a vehicle. The parameter acquisition device 101 and the vehicle controller 102 can be directly or indirectly connected through wired or wireless communication methods, and the present application does not limit this here.

[0068] Among them, the parameter acquisition device 101 is used to acquire the scene data of the vehicle. The parameter acquisition device 101 includes an ultrasonic sensor (i.e., an ultrasonic probe), and may also include at least one of a laser sensor (i.e., a lidar) or a millimeter-wave sensor. Among them, the ultrasonic sensor measures the distance between the vehicle and surrounding obstacles by transmitting and receiving ultrasonic signals, and determines the position of the detection point on the obstacle. The laser sensor is used to emit a laser beam and measure the time when the laser is reflected back to create laser point cloud data of the vehicle's surrounding scene. The parameter acquisition device 101 may also include a camera for acquiring images of the vehicle's surrounding scene, that is, converting the vehicle's surrounding scene into a digital image signal. The parameter acquisition device 101 may also include a data transmission module for sending the vehicle's scene data to the vehicle controller 102.

[0069] Among them, the vehicle controller 102 is used to obtain the scene data of the vehicle, so as to control the vehicle to park in the parking space based on the scene data. It should be noted that the vehicle controller 102 obtains the scene data of the vehicle from the parameter acquisition device 101. This application does not limit the data acquisition method and data transmission method here.

[0070] Figure 2 is a flowchart of an automatic parking method provided by an embodiment of this application, and this method is executed by the vehicle controller. Refer to Figure 2 , this method includes the following steps:

[0071] 201. The vehicle controller obtains scene data, and the scene data includes ultrasonic data of multiple detection points corresponding to the vehicle. The ultrasonic data of any detection point is the data collected by the ultrasonic sensor for the detection point.

[0072] In the embodiment of this application, according to a preset acquisition frequency, the ultrasonic sensor on the vehicle continuously emits ultrasonic signals in the corresponding direction. When the ultrasonic signal encounters surrounding obstacles, it will be reflected. The ultrasonic sensor receives the reflected ultrasonic signal and records the time interval from transmission to reception. According to the reflection situation of the current ultrasonic signal, the position of the detection point is determined. According to the propagation speed of the ultrasonic signal in the air and the time interval from transmission to reception, the distance between the vehicle and the obstacle is determined. This belongs to the data collected by the ultrasonic sensor for the detection point.

[0073] The scene data is data related to the surrounding scene where the vehicle is currently located. When ultrasonic sensors are respectively configured at different positions on the vehicle, the scene data may include the data collected by all ultrasonic sensors; it may also only include the data collected by specific ultrasonic sensors. For example, ultrasonic sensors for the four directions of left front, right front, left rear, and right rear are respectively configured on the sides of the vehicle body, and the scene data is the data collected by the ultrasonic sensor for the left front direction.

[0074] By obtaining ultrasonic data at multiple detection points, the vehicle controller can comprehensively understand the environmental information around the vehicle. The ultrasonic sensor can measure the distance between the vehicle and surrounding obstacles in real time and accurately, providing basic data for subsequent data processing.

[0075] 202. The vehicle controller inputs the scenario data into the target prediction model for prediction to obtain the predicted attribute values of each detection point. The target prediction model is used to determine the predicted attribute values based on the ultrasonic data. The predicted attribute value of any detection point indicates whether the predicted detection point is located in the area where the obstacle of the vehicle is located.

[0076] In the embodiment of the present application, the target prediction model is a trained model. The target prediction model predicts the attribute values of each detection point based on the input ultrasonic data of each detection point, so as to determine whether each detection point is located at the position where the obstacle of the vehicle is located. The predicted attribute value is a numerical value corresponding to each detection point. The embodiment of the present application does not limit the range of the predicted attribute value and the meaning indicated by different predicted attribute values. For example, the predicted attribute value can be 0 or 1. When the predicted attribute value is 0, it means that the detection point is located outside the area where the obstacle of the vehicle is located. When the predicted attribute value is 1, it means that the detection point is located in the area where the obstacle of the vehicle is located. Another example is that the predicted attribute value is any value within the range of 0 to 1. When the predicted attribute value is not greater than 0.5, it means that the detection point is located outside the area where the obstacle of the vehicle is located. When the predicted attribute value is greater than 0.5, it means that the detection point is located in the area where the obstacle of the vehicle is located.

[0077] By processing the ultrasonic data through the target prediction model, it is possible to more accurately determine whether each detection point is located in the area where the obstacle is located. Compared with directly using the ultrasonic data, the predicted attribute value can more clearly reflect the distribution of the obstacles, filter out some noise and interference information, improve the accuracy of judging the position of the obstacles, and enable the subsequent more precise determination of the boundary of the obstacles.

[0078] 203. The vehicle controller determines the predicted boundary of the obstacle based on the predicted attribute values of each detection point.

[0079] In the embodiment of the present application, the predicted boundary of the obstacle refers to the contour boundary of the obstacle determined according to the predicted attribute values of each detection point. When determining the predicted boundary here, the interference of noise points is excluded, and the vehicle can more accurately understand the position and range of the obstacle, so as to plan a parking path to avoid the obstacle and achieve automatic parking.

[0080] Compared with the traditional method, the predicted boundary determined by predicting the attribute value is more accurate, with a smaller deviation from the actual boundary, reducing the collision risk caused by inaccurate boundary judgment and also reducing the number of adjustments during the process of parking the vehicle into the garage, thus greatly improving the parking efficiency.

[0081] 204. The vehicle controller controls the vehicle to park into the target parking space based on the predicted boundary of the obstacle, and the target parking space is located outside the area where the obstacle is located.

[0082] In the embodiment of the present application, the target parking space is the parking space into which the vehicle parks. There is no overlapping area between the target parking space and the obstacle, which ensures that the vehicle will not collide with the obstacle during the parking process. Due to the more accurate determination of the target parking space range, the vehicle can better avoid obstacles during the parking process, reducing unnecessary adjustments and collision risks, and improving the parking efficiency and safety.

[0083] The embodiment of the present application provides an automatic parking method. Through the target prediction model, the predicted attribute values of each detection point are obtained based on the ultrasonic data of each detection point, and thus each detection point is screened based on the predicted attribute values. This method can filter out the noise points among multiple detection points, avoiding the interference of the noise points on the process of determining the obstacle boundary and improving the accuracy of the predicted obstacle boundary. Compared with the traditional method, the present application determines an obstacle boundary with a smaller actual deviation, greatly reducing the number of times of adjusting the vehicle position during parking, enabling the vehicle to park into the parking space with fewer adjustments, reducing the time consumption of the entire automatic parking process, and greatly improving the parking efficiency.

[0084] Based on the above simple introduction of the automatic parking method, the following will introduce the automatic parking method in more detail. Figure 3 is a flowchart of another automatic parking method provided by the embodiment of the present application, and this method is executed by the vehicle controller. Refer to Figure 3 and this method includes the following steps:

[0085] 301. The vehicle controller acquires scene data, and the scene data includes the ultrasonic data of multiple detection points corresponding to the vehicle. The ultrasonic data of any detection point is the data collected by the ultrasonic sensor for the detection point.

[0086] In the embodiments of the present application, the ultrasonic sensors on the vehicle continuously emit ultrasonic signals at a preset acquisition frequency. When the ultrasonic signals encounter surrounding obstacles, they will be reflected. The ultrasonic sensors receive the reflected ultrasonic signals and record the time interval from transmission to reception. According to the reflection situation of the current ultrasonic signal, the position of the current detection point is determined. According to the propagation speed of the ultrasonic signal in the air and the time interval from transmission to reception, the distance between the vehicle and the obstacle is determined. By obtaining the ultrasonic data of multiple detection points, the vehicle controller can comprehensively understand the environmental information around the vehicle. The ultrasonic sensors can measure the distance between the vehicle and the surrounding obstacles in real time and accurately, providing basic data for subsequent data processing.

[0087] In some embodiments, the ultrasonic data includes the first echo distance, the first echo energy value, the first echo width, the second echo distance, the second echo energy value, and the second echo width. Among them, the first echo distance refers to the distance from the emission point of the ultrasonic signal to the first-reflected obstacle, that is, the distance corresponding to the echo reflected by the first obstacle after the ultrasonic sensor emits the ultrasonic signal. The first echo energy value represents the energy intensity of the first echo signal and is used to indicate the material or surface characteristics of the obstacle, etc. For example, a larger energy value may indicate that the surface of the obstacle is relatively flat; while a smaller energy value indicates that the surface of the obstacle is relatively rough. The first echo width represents the continuous width of the first echo signal in time. The second echo distance refers to the distance from the emission point of the ultrasonic signal to the second-reflected obstacle, that is, the distance corresponding to the echo that the ultrasonic signal emitted by the ultrasonic sensor is reflected by the first obstacle and then continues to propagate and is reflected by the second obstacle again. The second echo energy value represents the energy intensity of the second echo signal. The second echo width represents the continuous width of the second echo signal in time.

[0088] In some embodiments, the scene data is collected by the ultrasonic sensors on the side of the vehicle body. The scene data can be the data collected by all the ultrasonic sensors; or it can be the data collected by specific ultrasonic sensors. For the convenience of description, see Figure 4 as shown Figure 4 is a schematic diagram of the installation positions of an ultrasonic sensor provided by the embodiments of the present application. Among them, the ultrasonic sensor 401 is arranged on the side body of the vehicle.

[0089] 302. The vehicle controller inputs the scene data into the target prediction model for prediction to obtain the predicted attribute values of each detection point. The target prediction model is used to determine the predicted attribute values based on the ultrasonic data, and the predicted attribute value of any detection point indicates whether the predicted detection point is located in the area where the obstacle of the vehicle is located.

[0090] In the embodiments of the present application, the target prediction model is a trained model. The target prediction model predicts the attribute values of each detection point based on the ultrasonic data of each input detection point, so as to determine whether each detection point is located at the position of an obstacle of the vehicle. The predicted attribute value is a numerical value corresponding to each detection point. By processing the ultrasonic data through the target prediction model, it is possible to more accurately determine whether each detection point is located within the area where the obstacle is located. Compared with directly using the ultrasonic data, the predicted attribute value can more clearly reflect the distribution of the obstacles, filter out some noise and interference information, improve the accuracy of judging the position of the obstacles, and enable the subsequent more accurate determination of the boundaries of the obstacles.

[0091] In some embodiments, the training steps of the target prediction model are as follows: Obtain a training data set, where the training data set includes the ultrasonic data of multiple detection points corresponding to a training vehicle and the label attribute values of the multiple detection points, and the label attribute value of any detection point indicates whether the detection point is located within the area where the obstacle of the vehicle is located; Input the training data set into the initial prediction model for prediction to obtain the predicted attribute values of each detection point; Based on the predicted attribute values of each detection point and the label attribute values of each detection point, update the parameters of the initial prediction model to obtain the target prediction model. For example, select an appropriate kernel function of the initial prediction model, use a support vector machine (SVM, Support Vector Machine) to train the initial prediction model, and update the parameters corresponding to the kernel function, so as to obtain the target prediction model.

[0092] In some embodiments, the label attribute values are automatically added based on the laser data. Correspondingly, obtain the laser data of the training vehicle, where the laser data is the data collected by the laser sensor for the surrounding environment of the training vehicle; Based on the laser data, determine the laser boundary of the obstacle; Based on the spatial relationship between the laser boundary and the multiple detection points, add the respective label attribute values to the multiple detection points. Using the laser data as the actual value of the obstacle boundary provides a basis for subsequently adding the respective label attribute values to each detection point and improves the reliability of the data.

[0093] In some embodiments, a laser sensor and an ultrasonic sensor are provided at the same height of the vehicle body. See Figure 4, the laser sensor 402 and the ultrasonic sensor 401 are set at the same height on the side body of the vehicle. This ensures the consistency between the laser data and the ultrasonic data, enabling the laser data to be used as a subsequent reference. In some embodiments, the preset acquisition frequencies of the laser sensor and the ultrasonic sensor are the same. At the same moment, both the laser sensor and the ultrasonic sensor collect data. That is, when collecting the data of the two sensors, it is ensured that the timestamps are aligned, thereby avoiding introducing errors during data collection, ensuring the reliability of the laser data, and making the attribute tag values added to each detection point more accurate.

[0094] Optionally, the laser sensor is a single-line lidar (Light Detection and Ranging). The single-line lidar includes a laser emission and reception channel. During operation, the single-line lidar emits a beam of laser, measures the time it takes for the laser to travel from emission to reflection from an object, and calculates the distance between the single-line lidar and the object based on the speed of light. In some embodiments, based on the distance between the single-line lidar and the object, the position information of the single-line lidar, the angle information when the laser is emitted, and the vehicle body attitude information, etc., the position information of the detection points on the obstacle is determined, such as the coordinates of the detection points, which will not be elaborated here.

[0095] In some embodiments, by determining whether a detection point is inside or outside the laser boundary, a label attribute value is added to the detection point. Correspondingly, in the case where any detection point is outside the laser boundary, a first attribute value is added to the detection point, and the first attribute value indicates that the detection point is outside the area where the obstacle is located; in the case where any detection point is inside the laser boundary, a second attribute value is added to the detection point, and the second attribute value indicates that the detection point is inside the area where the obstacle is located. For example, the first attribute value is 0 and the second attribute value is 1. Here, outside the laser boundary refers to the area other than the obstacle, and inside the laser boundary refers to the area where the obstacle is located. In this way, respective label attribute values are added to each detection point, distinguishing the valid detection points and noise points among the multiple detection points, providing a basis for the subsequent training process, and improving the reliability of the data.

[0096] For ease of description, refer to Figure 5 as shown in Figure 5It is a schematic diagram of an obstacle boundary provided by an embodiment of the present application. Among them, there are multiple detection points corresponding to ultrasonic data around obstacle 501, obstacle 502, and obstacle 503. Due to the characteristics of ultrasonic waves, detection points will also be formed at positions without obstacles, and these detection points will interfere with the processing process. For example, the points in area 504 and area 505 are noise points, which will cause the detected parking space area to be smaller than the actual parking space area. In the embodiment of the present application, a first attribute value is added to the detection points (i.e., noise points) in area 504 and area 505, and a second attribute value is added to the detection points other than those in area 504 and area 505.

[0097] The above content is described by taking the computing device automatically adding attribute values to each detection point as an example. Among them, the computing device refers to the device for training the model. In some embodiments, the label attribute values of each detection point can be manually added by technicians, and the embodiments of the present application do not limit this.

[0098] In some embodiments, the ultrasonic data in the training dataset includes, in addition to the first echo distance, the first echo energy value, the first echo width, the second echo distance, the second echo energy value, and the second echo width, a timestamp, so as to align the ultrasonic data and the laser data in the time dimension. For the convenience of describing the data in the training dataset, see Table 1 below. Among them, the units of the first echo distance and the second echo distance are millimeters (mm), and the units of the first echo width and the second echo width are milliseconds (ms), and this is used as an example for description. Among them, the first echo energy is 5100, indicating that during the acquisition process corresponding to the corresponding timestamp, the emitted ultrasonic wave did not reflect within the effective range, that is, there is no obstacle within the effective range in the corresponding direction.

[0099] Table 1 Data in the training dataset

[0100]

[0101]

[0102] For the convenience of describing the model training process, see Figure 6 as shown Figure 6 It is a flowchart of model training provided by an embodiment of the present application. First, sensors are set on the vehicle, and the sensors include a laser sensor and an ultrasonic sensor. A data acquisition program is configured, and this program is used to control the laser sensor and the ultrasonic sensor to respectively acquire laser data and ultrasonic data. The data is acquired and label attribute values are added to obtain a training dataset. The initial prediction model is trained based on the training dataset to obtain a target prediction model.

[0103] 303. The vehicle controller screens multiple detection points based on the predicted attribute values of each detection point, obtains multiple effective detection points, and all the multiple effective detection points are located within the area where the obstacle is located.

[0104] In the embodiment of the present application, the multiple detection points are divided into effective detection points and noise points. Through the predicted attribute values of each detection point, the effective detection points and the noise points can be distinguished, so as to filter out the noise points and retain the effective detection points.

[0105] In some embodiments, the noise points and the effective detection points are distinguished based on the relationship between the predicted attribute value and a preset threshold. Correspondingly, when the predicted attribute value of any detection point is not greater than the preset threshold, the detection point is discarded; when the predicted attribute value of any detection point is greater than the preset threshold, the detection point is determined as an effective detection point. For example, the predicted attribute value is any value within the range of 0 to 1, and the preset threshold is 0.5. When the predicted attribute value is not greater than 0.5, it means that the detection point is located outside the area where the obstacle of the vehicle is located, and the detection point is discarded; when the predicted attribute value is greater than 0.5, it means that the detection point is located within the area where the obstacle of the vehicle is located, and the detection point is retained to obtain an effective detection point. Or, the predicted attribute value is any value within the range of 0 to 10, and the preset threshold is 8. When the predicted attribute value is not greater than 8, it means that the detection point is located outside the area where the obstacle of the vehicle is located, and the detection point is discarded; when the predicted attribute value is greater than 8, it means that the detection point is located within the area where the obstacle of the vehicle is located, and the detection point is retained to obtain an effective detection point.

[0106] In this way, the predicted attribute values are distributed within a continuous numerical range. After being compared with the preset threshold, the detection points can be screened more carefully. The degree of closeness between the predicted attribute value and the preset threshold can reflect the reliability of the detection point, can more accurately filter out unreliable noise points, and thus more accurately locate the area where the obstacle is located.

[0107] In some embodiments, the noise points and the effective detection points are distinguished based on the numerical value of the predicted attribute value. Correspondingly, when the predicted attribute value of any detection point is the first preset value, the detection point is discarded; when the predicted attribute value of any detection point is the second preset value, the detection point is determined as an effective detection point. For example, the predicted attribute value can be 0 or 1. When the predicted attribute value is 0, it means that the detection point is located outside the area where the obstacle of the vehicle is located, and the detection point is discarded; when the predicted attribute value is 1, it means that the detection point is located within the area where the obstacle of the vehicle is located, and the detection point is retained to obtain an effective detection point.

[0108] In this way, the judgment logic is simple. It only needs to determine whether the predicted attribute value is a specific preset value, without the need for complex numerical comparison and threshold setting, which reduces the calculation amount and the complexity of the algorithm, and further improves the efficiency.

[0109] The above preset threshold, first preset value, and second preset value are all values that can be set by technicians themselves, and will not be elaborated here.

[0110] For ease of describing the process of processing based on the target prediction model, refer to Figure 7 as shown in Figure 7 FIG. 8 is a flowchart of boundary prediction provided by an embodiment of the present application. First, a detection point is determined, and the ultrasonic data of the detection point is input into the target prediction model to obtain the predicted attribute value corresponding to the detection point. Then, the relationship between the predicted attribute value and the preset threshold is judged. If the predicted attribute value is greater than the preset threshold, the detection point is retained to obtain an effective detection point; if the predicted attribute value is not greater than the preset threshold, the detection point is discarded.

[0111] 304. The vehicle controller determines the predicted boundary of the obstacle based on multiple effective detection points.

[0112] In the embodiment of the present application, the predicted boundary of the obstacle refers to the contour boundary of the obstacle determined according to the predicted attribute values of each detection point. The process of determining the predicted boundary here avoids the interference of noise points, and the vehicle can more accurately understand the position and range of the obstacle, so as to plan a parking path to avoid the obstacle and realize automatic parking. Compared with the traditional method, the predicted boundary determined by the predicted attribute value is more accurate, with a smaller deviation from the actual boundary, reducing the collision risk caused by inaccurate boundary judgment, and also reducing the number of adjustments during the process of parking into the garage, greatly improving the parking efficiency.

[0113] For ease of description, refer to Figure 8 as shown in Figure 8 FIG. 9 is a schematic diagram of a predicted boundary provided by an embodiment of the present application. Among them, there are multiple detection points corresponding to ultrasonic data around the obstacle 801, obstacle 802, and obstacle 803. In the traditional method, the predicted boundary of the obstacle directly determined based on the ultrasonic data is 804. The predicted boundary of the obstacle determined based on multiple effective detection points is 805. By comparing the two and combining the actual boundary of the obstacle, it can be seen that after screening the detection points based on the predicted attribute value to obtain effective detection points, the accuracy of the predicted boundary of the obstacle further determined is higher, which can effectively reduce the number of times of adjusting the vehicle during parking, and can also greatly improve the success rate of parking into a narrow parking space, thus improving the parking efficiency as a whole.

[0114] 305. The vehicle controller controls the vehicle to park into the target parking space based on the predicted boundary of the obstacle, and the target parking space is located outside the area where the obstacle is located.

[0115] In the embodiments of the present application, the target parking space is the parking space where the vehicle parks. There is no overlapping area between the target parking space and the obstacle, which ensures that the vehicle will not collide with the obstacle during the parking process. Since the determination of the obstacle boundary is more accurate, the vehicle can better avoid obstacles during the parking process, reducing unnecessary adjustments and collision risks, and improving the efficiency and safety of parking.

[0116] To facilitate the description of the effects of the embodiments of the present application, refer to Table 2 below.

[0117] Table 2 Deviation Range Statistics

[0118]

[0119] Among them, the deviation range is the range of the deviation between the predicted boundary and the actual boundary. The number of statistics is the number of times the deviation value is counted within the corresponding deviation range. The absolute mean of the accuracy improvement is the average value of the accuracy improvement within the corresponding deviation range. The standard deviation of the accuracy improvement is the degree of dispersion of the accuracy improvement values within the corresponding deviation range.

[0120] The embodiments of the present application provide an automatic parking method. Through the target prediction model, the predicted attribute values of each detection point are obtained based on the ultrasonic data of each detection point, and thus each detection point is screened based on the predicted attribute values. This method can filter out the noise points among multiple detection points, which avoids the interference of the noise points on the process of determining the obstacle boundary and improves the accuracy of the predicted obstacle boundary. In the traditional method, the obstacle boundary directly determined based on the ultrasonic data is inaccurate, and the range of this obstacle boundary is larger than the actual obstacle boundary. Therefore, the predicted parking space area obtained by this method is usually narrower than the actual parking space area, which may cause the vehicle to abandon parking in this space or the obtained parking space boundary is inaccurate, resulting in too many parking adjustment times or parking failure. Compared with the traditional method, the present application determines an obstacle boundary with a smaller actual deviation, greatly reducing the number of adjustment times, enabling the vehicle to park in the parking space with fewer adjustments, reducing the time consumption of the entire automatic parking process, and greatly improving the parking efficiency.

[0121] Figure 9 It is a block diagram of an automatic parking device provided by the embodiments of the present application. This device is used to execute the steps when the above automatic parking method is executed. Refer to Figure 9 This automatic parking device includes: a first acquisition module 901, a prediction module 902, a first determination module 903, and a control module 904.

[0122] The first acquisition module 901 is used to acquire scene data. The scene data includes the ultrasonic data of multiple detection points corresponding to the vehicle. The ultrasonic data of any detection point is the data collected by the ultrasonic sensor for the detection point;

[0123] A prediction module 902, configured to input the scenario data into a target prediction model for prediction to obtain predicted attribute values of each detection point, where the target prediction model is used to determine the predicted attribute values based on ultrasonic data, and the predicted attribute value of any detection point indicates whether the predicted detection point is located in the area where an obstacle of the vehicle is located;

[0124] A first determination module 903, configured to determine a predicted boundary of the obstacle based on the predicted attribute values of each detection point;

[0125] A control module 904, configured to control the vehicle to park in a target parking space based on the predicted boundary of the obstacle, where the target parking space is located outside the area where the obstacle is located.

[0126] In some embodiments, the first determination module 903 includes:

[0127] A screening unit, configured to screen the multiple detection points based on the predicted attribute values of each detection point to obtain multiple valid detection points, and all the multiple valid detection points are located in the area where the obstacle is located;

[0128] A determination unit, configured to determine the predicted boundary of the obstacle based on the multiple valid detection points.

[0129] In some embodiments, the screening unit is configured to discard a detection point when the predicted attribute value of any detection point is not greater than a preset threshold; and determine the detection point as a valid detection point when the predicted attribute value of any detection point is greater than the preset threshold.

[0130] In some embodiments, the screening unit is configured to discard a detection point when the predicted attribute value of any detection point is a first preset value; and determine the detection point as a valid detection point when the predicted attribute value of any detection point is a second preset value.

[0131] In some embodiments, the training device of the target prediction model includes:

[0132] A second acquisition module, configured to acquire a training data set, where the training data set includes ultrasonic data of multiple detection points corresponding to a training vehicle and label attribute values of the multiple detection points, and the label attribute value of any detection point indicates whether the detection point is located in the area where an obstacle of the vehicle is located;

[0133] A processing module, configured to input the training data set into an initial prediction model for prediction to obtain predicted attribute values of each detection point;

[0134] An update module, configured to update parameters of the initial prediction model based on predicted attribute values of each detection point and labeled attribute values of each detection point, so as to obtain the target prediction model.

[0135] In some embodiments, the apparatus further includes:

[0136] A third acquisition module, configured to acquire lidar data of the training vehicle, where the lidar data is data collected by a lidar sensor for the surrounding environment of the training vehicle;

[0137] A second determination module, configured to determine a lidar boundary of the obstacle based on the lidar data;

[0138] An addition module, configured to add respective labeled attribute values to the plurality of detection points based on a spatial relationship between the lidar boundary and the plurality of detection points.

[0139] In some embodiments, the addition module is configured to, when any detection point is outside the lidar boundary, add a first attribute value to the detection point, where the first attribute value indicates that the detection point is outside the area where the obstacle is located; when any detection point is inside the lidar boundary, add a second attribute value to the detection point, where the second attribute value indicates that the detection point is inside the area where the obstacle is located.

[0140] An automatic parking apparatus provided in this application obtains predicted attribute values of each detection point based on ultrasonic data of each detection point through a target prediction model, and thus filters each detection point based on the predicted attribute values. This method can filter out noise points among multiple detection points, which avoids interference of noise points on the process of determining the obstacle boundary and improves the accuracy of the predicted obstacle boundary. Compared with the traditional method, this application determines an obstacle boundary with a smaller actual deviation, greatly reduces the number of times of parking maneuvers, enables the vehicle to park in the parking space after fewer adjustments, reduces the time consumption of the entire automatic parking process, and greatly improves the parking efficiency.

[0141] It should be noted that: when the automatic parking apparatus provided in the above embodiment runs an application program, only the above division of each functional module is used for illustration. In actual application, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the electronic device is divided into different functional modules to complete all or part of the functions described above. In addition, the automatic parking apparatus provided in the above embodiment and the embodiment of the automatic parking method belong to the same concept, and the implementation process is shown in the method embodiment, which will not be elaborated here.

[0142] Figure 10 It is a schematic structural diagram of a vehicle controller provided in an embodiment of this application.

[0143] Generally, the vehicle controller 1000 includes: a main control module 1001, a CAN interface 1002, a hard-wired input interface 1003, and a hard-wired output interface 1004. The main control module 1001 is respectively connected to the CAN interface 1002, the hard-wired input interface 1003, and the hard-wired output interface 1004. The main control module 1001 generally includes a processor and a memory.

[0144] Among them, the processor may include one or more processing cores, such as a 4-core processor, a 10-core processor, etc. The processor can be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor can also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for the rendering and drawing of the content to be displayed on the vehicle display screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0145] Among them, the memory may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory is used to store at least one computer program, and the at least one computer program is used to be executed by the processor to implement the automatic parking method provided in the method embodiments of the present application.

[0146] The CAN interface 1002 may include a power CAN interface, a motor CAN interface, and a collection CAN interface. Among them, the power CAN interface is used to communicate with the vehicle's power system module, the motor CAN interface is used to communicate with the vehicle's motor controller, and the collection CAN interface is used to communicate with the vehicle's parameter collection device.

[0147] The hardwire input interface 1003 is used to receive hardwire control signals. The hardwire output interface 1004 is used to send control instructions to the electronic control components of the vehicle, so that the electronic control components of the vehicle perform corresponding actions. Among them, the electronic control components of the vehicle include a power management system, a motor controller, an on-vehicle charger, a body control system, etc.

[0148] The main control module 1001 can communicate with the vehicle's power system module, motor controller, and parameter acquisition device through the CAN interface 1002, and generate control instructions according to the hardwire control signals received by the hardwire input interface 1003, so as to send control instructions to the electronic control components of the vehicle through the hardwire output interface 1004, thereby realizing data transmission and vehicle control.

[0149] Those skilled in the art can understand that Figure 10 the structure shown in does not constitute a limitation on the vehicle controller 1000, and may include more or fewer components than shown in the figure, or combine certain components, or adopt different component arrangements.

[0150] The embodiment of the present application also provides a computer-readable storage medium, in which at least one segment of computer program is stored, and the at least one segment of computer program is loaded and executed by a processor to implement the automatic parking method in the above embodiment. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0151] The embodiment of the present application also provides a computer program product, including a computer program, and the computer program is executed by a processor to implement the automatic parking method in the embodiment of the present application.

[0152] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiment can be completed by hardware, or can be completed by a program instructing relevant hardware. The program can be stored in a computer-readable storage medium, and the above-mentioned storage medium can be a read-only memory, a magnetic disk, or an optical disc, etc.

[0153] The above are only optional embodiments of the present application, and are not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An automatic parking method, characterized in that: The method comprises: Acquire scene data, where the scene data includes ultrasonic data of multiple detection points corresponding to the vehicle, where the ultrasonic data of any detection point is data collected by the ultrasonic sensor for the detection point; Inputting the scene data into a target prediction model for prediction to obtain a predicted attribute value of each detection point, wherein the target prediction model is used to determine the predicted attribute value based on the ultrasonic data, and the predicted attribute value of any detection point indicates whether the predicted detection point is located in the area where the obstacle of the vehicle is located; Determining a predicted boundary of the obstacle based on the predicted attribute values ​​of each detection point; Based on the predicted boundary of the obstacle, the vehicle is controlled to park in a target parking space, where the target parking space is outside the area where the obstacle is located.

2. The method according to claim 1, characterized in that The step of determining the predicted boundary of the obstacle based on the predicted attribute values ​​of each detection point includes: Based on the predicted attribute value of each detection point, the multiple detection points are screened to obtain multiple valid detection points, and the multiple valid detection points are all located in the area where the obstacle is located; Based on the multiple valid detection points, a predicted boundary of the obstacle is determined.

3. The method according to claim 2, characterized in that The screening of the plurality of detection points based on the predicted attribute value of each detection point to obtain a plurality of valid detection points includes: If the predicted attribute value of any detection point is not greater than a preset threshold, the detection point is discarded; When the predicted attribute value of any detection point is greater than a preset threshold, the detection point is determined as a valid detection point.

4. The method according to claim 2, characterized in that: The screening of the plurality of detection points based on the predicted attribute value of each detection point to obtain a plurality of valid detection points includes: When the predicted attribute value of any detection point is the first preset value, discard the detection point; When the predicted attribute value of any detection point is the second preset value, the detection point is determined as a valid detection point.

5. The method according to claim 1, characterized in that The training steps of the target prediction model are: Acquire a training data set, the training data set including ultrasonic data of multiple detection points corresponding to a training vehicle and label attribute values ​​of the multiple detection points, the label attribute value of any detection point indicating whether the detection point is located in an area where an obstacle of the vehicle is located; Inputting the training data set into the initial prediction model for prediction to obtain the predicted attribute value of each detection point; Based on the predicted attribute value of each detection point and the label attribute value of each detection point, the parameters of the initial prediction model are updated to obtain the target prediction model.

6. The method according to claim 5, characterized in that The method further comprises: Acquire laser data of the training vehicle, where the laser data is data collected by a laser sensor for the surrounding environment of the training vehicle; Based on the laser data, determining a laser boundary of the obstacle; Based on the spatial relationship between the laser boundary and the multiple detection points, respective label attribute values ​​are added to the multiple detection points.

7. The method according to claim 6, characterized in that The adding respective label attribute values ​​to the multiple detection points based on the spatial relationship between the laser boundary and the multiple detection points includes: In the case where any detection point is outside the laser boundary, adding a first attribute value to the detection point, wherein the first attribute value indicates that the detection point is outside the area where the obstacle is located; In the case that any detection point is located within the laser boundary, a second attribute value is added to the detection point, where the second attribute value indicates that the detection point is located within the area where the obstacle is located.

8. An automatic parking device, characterized in that: The device comprises: A first acquisition module is used to acquire scene data, wherein the scene data includes ultrasonic data of multiple detection points corresponding to the vehicle, and the ultrasonic data of any detection point is data collected by the ultrasonic sensor for the detection point; A prediction module, used for inputting the scene data into a target prediction model for prediction, and obtaining a predicted attribute value of each detection point, wherein the target prediction model is used for determining the predicted attribute value based on the ultrasonic data, and the predicted attribute value of any detection point indicates whether the predicted detection point is located in the area where the obstacle of the vehicle is located; A first determination module, configured to determine a predicted boundary of the obstacle based on the predicted attribute values ​​of each detection point; A control module is used to control the vehicle to park in a target parking space based on the predicted boundary of the obstacle, and the target parking space is located outside the area where the obstacle is located.

9. A vehicle controller, characterized in that: The vehicle controller includes a main control module, which includes a processor and a memory. The memory is used to store at least one computer program. The at least one computer program is loaded by the processor and executes the automatic parking method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store at least one computer program, and the at least one computer program is used to execute the automatic parking method according to any one of claims 1 to 7.

Images