Car washing path planning method and robot

Abstract

The invention provides a car washing path planning method and a robot, and is applied to a car washing robot. And according to the surface contour point cloud and the feature semantic information of the to-be-washed vehicle, determining a semantic point cloud. And according to a cleaning process, carrying out recombination and normal vector clustering on the semantic point clouds to obtain point clouds of a plurality of areas to be cleaned, and carrying out gridding to obtain a semantic grid map. Path searching is carried out according to the semantic grid map, and coordinate transformation is carried out on a searched target path to obtain a target car washing path. Semantic point clouds are determined based on surface contour point clouds, modeling characteristics and smudginess characteristics of a vehicle, point clouds of a to-be-washed area are determined in combination with a cleaning process, the point clouds of the to-be-washed area are gridded to obtain a semantic grid map, refined modeling of the surface of the vehicle is achieved, path planning is conducted on the basis of modeling, and the method is simple and convenient to implement. The car washing robot can achieve self-adaptive and contact type refined car washing according to the planned path, the intelligent cleaning capacity is achieved for dirty areas, and the expected car washing effect is achieved.

Description

Technical Field

[0001] The present application relates to the field of robotics technology, and in particular to a car wash path planning method and robot. Background Art

[0002] Current contact-type automatic car washing equipment, in addition to using high-pressure water for flushing, also uses a roller brush to wipe along the surface of the vehicle to achieve a certain cleaning effect.

[0003] The roller brush of this type of car wash equipment usually moves along a fixed path, or is simply modeled on a sensor to adapt to vehicles of different sizes and appearances. The path is planned in a two-dimensional plane in the direction of vehicle travel and height, so that the roller brush moves as close to the outer contour of the vehicle as possible, and then passes through the front of the vehicle, hood, front window, roof, rear window and trunk in sequence.

[0004] However, this solution has low requirements for the accuracy of vehicle modeling, and the path planning of the roller brush is often rough, leaving a large margin of error to prevent the roller from accidentally touching the vehicle surface and damaging the vehicle. In addition, the vehicle can only be cleaned according to a fixed process, and it cannot perform intelligent cleaning for special dirt on the vehicle, and the expected cleaning effect cannot be achieved. Summary of the Invention

[0005] The present application provides a car wash path planning method and robot, which are used to solve the technical problems that the existing car wash path planning is limited and rough, leaving a large fault tolerance margin to prevent the roller from accidentally touching the vehicle surface and damaging the vehicle, and failing to identify dirty areas for targeted cleaning, thereby failing to achieve the expected cleaning effect.

[0006] In a first aspect, the present application provides a car wash path planning method, which is applied to a car wash robot, and the method comprises:

[0007] Determining a semantic point cloud of the vehicle to be washed based on a surface contour point cloud and feature semantic information of the vehicle to be washed, wherein the feature semantic information includes semantic information of various components of the vehicle to be washed and semantic information of dirty areas;

[0008] Reorganizing the semantic point cloud and performing normal vector clustering according to the cleaning process to obtain point clouds of multiple areas to be washed, and gridding the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed;

[0009] A path search is performed according to the semantic grid map, and a coordinate transformation is performed on the searched target path to obtain a target car wash path.

[0010] In a possible implementation, after obtaining the target car wash route, the method further includes:

[0011] Generate a path file corresponding to the target car washing path according to different process configurations corresponding to the cleaning process;

[0012] The corresponding path file is called according to the process configuration, and the car washing execution component of the car washing robot is controlled to clean the vehicle to be washed according to the called path file and the process configuration.

[0013] In a possible implementation, controlling the car wash execution component of the car wash robot to clean the vehicle to be washed according to the called path file and the process configuration includes:

[0014] Determining motion information of the car wash execution component according to the called path file;

[0015] Generate and issue control instructions to the car wash execution component;

[0016] The car wash execution component is controlled to respond to the control instruction to perform a car wash action according to the called path file, the process configuration and the motion information.

[0017] In a possible implementation, performing the car washing action according to the called path file, the process configuration, and the motion information includes:

[0018] Through the joint angles of the robotic arm and the posture of the mobile chassis, the cleaning tool is guided to perform the car washing action corresponding to the process configuration according to the target car washing path in the called path file;

[0019] The motion information of the car wash execution component includes the joint angle of the robotic arm and the posture of the chassis, and the cleaning tool is installed at the end of the robotic arm.

[0020] In a possible implementation, before the step of analyzing the surface contour point cloud and feature semantic information of the vehicle to be washed, the following steps are further included:

[0021] Acquire a surface point cloud of the vehicle to be washed in a vehicle coordinate system, and preprocess the surface point cloud to obtain the surface contour point cloud, wherein the preprocessing at least includes reducing the point cloud density and eliminating irrelevant point cloud operations;

[0022] A 2D image of the vehicle to be washed is acquired, and detection and segmentation are performed on the 2D image to determine semantic information of various parts of the vehicle to be washed and semantic information of the dirty area.

[0023] In a possible implementation, the reorganizing the semantic point cloud and clustering the normal vectors according to the cleaning process to obtain point clouds of multiple areas to be cleaned includes:

[0024] Reorganize the regions corresponding to the semantic point cloud according to the cleaning process according to a preset region reorganization strategy to obtain a plurality of component-level point clouds;

[0025] Each component-level point cloud is clustered by the normal vector to obtain a point cloud of each area to be cleaned.

[0026] In a possible implementation, gridding the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed includes:

[0027] Performing coordinate transformation on the point cloud of each area to be cleaned according to a preset projection transformation perspective to obtain a grid point cloud;

[0028] Performing averaging processing on the grid point cloud and determining the grid center depth to obtain an intermediate grid map;

[0029] Obtain the semantic grid map according to the intermediate grid map and the semantics of the point cloud of each area to be cleaned corresponding thereto;

[0030] The grid point cloud refers to the point cloud of each area to be cleaned in a grid coordinate system.

[0031] In a possible implementation, performing coordinate transformation on the point cloud of each area to be cleaned according to a preset projection transformation perspective to obtain a grid point cloud includes:

[0032] The coordinate transformation of the point cloud of each to-be-cleaned area is performed according to a first transformation matrix to obtain the grid point cloud, wherein the first transformation matrix is ​​obtained according to the preset projection transformation perspective.

[0033] In a possible implementation, performing a path search according to the semantic grid map includes:

[0034] Starting from a predefined position, a forward search of a bow-shaped path is performed on the semantic grid map along a preset direction according to a sampling grid and a preset inerasable area to obtain a preset amount of grid data;

[0035] The point clouds in the grid data are associated to obtain the searched target path.

[0036] In a possible implementation, the method further includes:

[0037] In the forward path search, the sampling grid is adaptively adjusted so that the sparsity of the grid data meets the contact surface requirement;

[0038] The contact surface requirement refers to the requirement for the contact area between the car washing robot and the vehicle to be washed when cleaning the vehicle to be washed.

[0039] In a second aspect, the present application provides a car wash path planning device, which is applied to a car wash robot, and the device includes:

[0040] a semantic processing module for determining a semantic point cloud of the vehicle to be washed based on a surface contour point cloud and feature semantic information of the vehicle to be washed, wherein the feature semantic information includes semantic information of various components of the vehicle to be washed and semantic information of dirty areas;

[0041] A gridding module is used to reorganize the semantic point cloud and cluster normal vectors according to the cleaning process to obtain point clouds of multiple areas to be washed, and grid the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed;

[0042] The path search module is used to perform path search according to the semantic grid map and perform coordinate transformation on the searched target path to obtain a target car wash path.

[0043] In a possible implementation, the device further includes a car wash control module; the car wash control module is configured to:

[0044] Generate a path file corresponding to the target car washing path according to different process configurations corresponding to the cleaning process;

[0045] The corresponding path file is called according to the process configuration, and the car washing execution component of the car washing robot is controlled to clean the vehicle to be washed according to the called path file and the process configuration.

[0046] In a possible implementation, the car wash control module is further configured to:

[0047] Determining motion information of the car wash execution component according to the called path file;

[0048] Generate and issue control instructions to the car wash execution component;

[0049] The car wash execution component is controlled to respond to the control instruction to perform a car wash action according to the called path file, the process configuration and the motion information.

[0050] In a possible implementation, the car wash control module is further configured to:

[0051] Through the joint angles of the robotic arm and the posture of the mobile chassis, the cleaning tool is guided to perform the car washing action corresponding to the process configuration according to the target car washing path in the called path file;

[0052] The motion information of the car wash execution component includes the joint angle of the robotic arm and the posture of the chassis, and the cleaning tool is installed at the end of the robotic arm.

[0053] In a possible implementation, the apparatus further includes: an acquisition module; the acquisition module is configured to:

[0054] Acquire a surface point cloud of the vehicle to be washed in a vehicle coordinate system, and preprocess the surface point cloud to obtain the surface contour point cloud, wherein the preprocessing at least includes reducing the point cloud density and eliminating irrelevant point cloud operations;

[0055] A 2D image of the vehicle to be washed is acquired, and detection and segmentation are performed on the 2D image to determine semantic information of various parts of the vehicle to be washed and semantic information of the dirty area.

[0056] In a possible implementation, the semantic processing module is specifically configured to:

[0057] Reorganize the regions corresponding to the semantic point cloud according to the cleaning process according to a preset region reorganization strategy to obtain a plurality of component-level point clouds;

[0058] Each component-level point cloud is clustered by the normal vector to obtain a point cloud of each area to be cleaned.

[0059] In a possible implementation, the gridding module is specifically configured to:

[0060] Performing coordinate transformation on the point cloud of each area to be cleaned according to a preset projection transformation perspective to obtain a grid point cloud;

[0061] Performing averaging processing on the grid point cloud and determining the grid center depth to obtain an intermediate grid map;

[0062] Obtain the semantic grid map according to the intermediate grid map and the semantics of the point cloud of each area to be cleaned corresponding thereto;

[0063] The grid point cloud refers to the point cloud of each area to be cleaned in a grid coordinate system.

[0064] In a possible implementation, the gridding module is further configured to:

[0065] The coordinate transformation of the point cloud of each to-be-cleaned area is performed according to a first transformation matrix to obtain the grid point cloud, wherein the first transformation matrix is ​​obtained according to the preset projection transformation perspective.

[0066] In a possible implementation, the path search module is specifically configured to:

[0067] Starting from a predefined position, a forward search of a bow-shaped path is performed on the semantic grid map along a preset direction according to a sampling grid and a preset inerasable area to obtain a preset amount of grid data;

[0068] The point clouds in the grid data are associated to obtain the searched target path.

[0069] In a possible implementation, the path search module is further configured to:

[0070] In the forward path search, the sampling grid is adaptively adjusted so that the sparsity of the grid data meets the contact surface requirement;

[0071] The contact surface requirement refers to the requirement for the contact area between the car washing robot and the vehicle to be washed when cleaning the vehicle to be washed.

[0072] In a third aspect, the present application provides a controller, comprising: a memory, a processor;

[0073] The memory stores computer-executable instructions;

[0074] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0075] In a fourth aspect, the present application provides a car washing robot, comprising a sensor, a car washing execution component, a cleaning tool, and the controller provided in the third aspect above;

[0076] The sensor is used to obtain a surface point cloud and a 2D image of the vehicle to be washed in a vehicle coordinate system;

[0077] The car wash execution component is used to guide the cleaning tool to perform a car wash action. The cleaning tool is installed at the end of the mechanical arm of the car wash execution component. The car wash execution component also includes a mobile chassis.

[0078] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementation methods of the first aspect.

[0079] In a sixth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the first aspect and / or various possible implementation methods of the first aspect.

[0080] The present application provides a car wash path planning method and robot, which is applied to a car wash robot. First, the semantic point cloud of the vehicle to be washed is determined based on the surface contour point cloud and feature semantic information of the vehicle to be washed. Then, the semantic point cloud is reorganized and normal vector clustered according to the cleaning process to obtain point clouds of multiple areas to be washed, and the point cloud of each area to be washed is gridded to obtain a semantic grid map of the vehicle to be washed. Path search is then performed based on the semantic grid map, and the searched target path is coordinate-transformed to obtain the target car wash path. The semantic point cloud of the vehicle is determined based on the surface contour point cloud, vehicle styling characteristics, and dirt characteristics of the vehicle, and the point cloud of the area to be washed is determined in combination with the cleaning process, and then the point cloud is gridded to obtain a semantic grid map, thereby realizing fine modeling of the surface of the vehicle to be washed, and then path planning is performed based on the modeling, so that the car wash robot can realize adaptive, contact-based fine car washing according to the planned path, and has intelligent cleaning capabilities for dirty areas, and can achieve the expected cleaning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0082] Figure 1 Schematic diagram of the scenario of the car wash path planning method provided in this application;

[0083] Figure 2 Schematic diagram of the process of car wash path planning method provided in this application Figure 1 ;

[0084] Figure 3 Schematic diagram of the process of car wash path planning method provided in this application Figure 2 ;

[0085] Figure 4 Schematic diagram of the grid data provided for this application;

[0086] Figure 5 A schematic diagram of the grid data in the vehicle coordinate system provided for this application;

[0087] Figure 6 A visualization diagram of the target car wash path and normal vector in the vicinity of the vehicle's rearview mirror provided in this application, avoiding the rearview mirror;

[0088] Figure 7 A schematic diagram of the structure of the car wash path planning device provided in this application;

[0089] Figure 8 Schematic diagram of the structure of the controller provided in this application.

[0090] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0091] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0092] Currently, contact-type automated car wash equipment uses a fixed path for its roller brushes or a two-dimensional planar car wash path based on simple sensor modeling. This can only achieve basic surface cleaning of the vehicle's contours, resulting in limitations in path planning. Furthermore, this low-precision modeling results in coarse path planning, leaving a large margin of error to prevent damage to the vehicle during the cleaning process, which in turn affects cleaning coverage and creates a conflict between modeling accuracy and safety. Furthermore, this planned path can only clean vehicles according to a fixed process and cannot identify dirty areas of the vehicle for targeted cleaning. This lack of intelligence results in poor cleaning results.

[0093] To address the aforementioned issues in the prior art, the present application provides a car wash path planning method and robot. The inventive concept of this application is as follows: based on the surface contour point cloud and feature semantic information of the vehicle to be washed acquired by sensors, a semantic point cloud of the vehicle to be washed is first determined. The semantic point cloud is then reorganized and normal vectors clustered according to the cleaning process to obtain point clouds of multiple areas to be washed. The point cloud of each area to be washed is then gridded to obtain a semantic grid map of the vehicle to be washed. This achieves refined modeling of the vehicle to be washed. Furthermore, a path search is performed on the semantic grid map to plan a target car wash path. Path planning based on this refined modeling enables the car wash robot to perform adaptive, contact-based, refined car washes according to the planned path. Furthermore, the feature semantic information includes semantic information about the various components of the vehicle to be washed, as well as semantic information about the dirty areas on the vehicle to be washed. This allows for consideration of uneven and dirty areas on the vehicle to be washed, effectively resolving conflicts between modeling and safety in the prior art and enabling targeted cleaning of dirty areas. This intelligent cleaning capability achieves the desired cleaning effect.

[0094] Figure 1 This is a schematic diagram of the scenario of the car wash path planning method provided in this application. Figure 1As shown, the application scenario of the embodiment of the present application is that the car washing robot 10 is configured to execute a car washing path planning method, and then clean the vehicle 20 to be washed according to the determined target car washing path, thereby completing the automatic car washing task of the car washing robot 10.

[0095] In some embodiments, the car wash robot 10 may include a sensor, a car wash execution component, a cleaning tool 13 and a controller 14 .

[0096] like Figure 1 As shown, the sensors may include a laser radar 111 and a camera 112. The laser radar 111 can be a multi-line laser radar, and the camera 112 can be a visible light camera. The laser radar 111 is used to acquire a surface point cloud of the vehicle 20 to be washed in the vehicle coordinate system for 3D reconstruction of the vehicle 20 to be washed. The camera 112 is used to acquire 2D images of the components and dirt of the vehicle 20 to be washed for component-level semantic detection and dirt detection. The laser radar 111 and camera 112 can be calibrated using internal and external parameters to achieve temporal and spatial alignment of the perceived information, thereby constructing a semantic point cloud of the vehicle 20 to be washed.

[0097] The car wash execution component includes a robotic arm 121 and a chassis 122. For example, the robotic arm 121 can have six or more axes, allowing for a larger arm span and workspace, covering areas such as the roof and side skirts of the vehicle 20 being washed. This also takes into account the flexibility required of the robotic arm 121 due to the distribution of wiping areas on the vehicle 20 being washed. The chassis 122 can be movable.

[0098] In some embodiments, different cleaning tools 1211 , such as a spray gun, a brush, or a sponge, can be installed at the end of the robotic arm 121 . The embodiment of the present application does not limit the material of the cleaning tool 1211 .

[0099] In some embodiments, the base of the robotic arm 121 may be mounted on a chassis or rails that can move autonomously, for example Figure 1 The mobile chassis shown enables the robotic arm 121 to move around the vehicle 20 to be washed under the action of the chassis 122, thereby covering different areas such as the front, rear, left and right of the vehicle 20 to be washed.

[0100] The controller 14 serves as a carrier of the algorithm software to execute the car wash path planning method provided in the embodiment of the present application. The controller 14 may be an embedded processor including a GPU, a CPU, a memory, a hard disk, and various IO interfaces. The embodiment of the present application does not limit the specific content of the controller 14.

[0101] It should be noted that Figure 1The car wash robot shown is only an illustrative example and does not limit the structure of the car wash robot. The car wash path planning method provided in the embodiment of the present application is applicable to any car wash equipment that can implement the method, including but not limited to Figure 1 The car washing robot shown.

[0102] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0103] Figure 2 Schematic diagram of the process of car wash path planning method provided in this application Figure 1 ,like Figure 2 As shown, the method includes:

[0104] S101 : Determine a semantic point cloud of the vehicle to be washed based on the surface contour point cloud and feature semantic information of the vehicle to be washed.

[0105] The feature semantic information includes semantic information about each vehicle component to be washed and semantic information about dirty areas. This semantic information includes information about the front, hood, front windows, roof, rear windows, trunk, rear, left and right front and rear doors, front and rear windows, and front and rear wheel suspensions. Semantic information about dirty areas describes the type and state of dirt in the area.

[0106] Based on the surface contour point cloud of the vehicle to be washed, semantic information is assigned to the surface contour point cloud based on the visual detection results, i.e., the semantic information of the vehicle's features to be washed, thereby generating a semantic point cloud of the vehicle to be washed. For example, using LiDAR and camera extrinsic calibration, the detected 2D dirty areas are back-projected into a 3D surface contour point cloud, and the semantic information corresponding to the point cloud of the dirty areas is assigned to form a semantic point cloud.

[0107] S102 , reorganizing the semantic point cloud and clustering the normal vectors according to the cleaning process to obtain point clouds of multiple areas to be washed, and gridding the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed.

[0108] The semantic point cloud is reorganized according to different cleaning processes, and the reorganized point cloud is clustered by normal vectors to obtain point clouds of multiple areas to be washed. For example, if the glass and non-glass of the vehicle to be washed are made of different materials, the cleaning process may be different. The cleaning process may be wiping, spraying, etc., so that the overall semantic point cloud can be split into corresponding multiple component-level point clouds, and the point cloud of the dirty area can exist separately. Normal vector clustering is to combine point clouds with surface normal vectors of similar directions or characteristics. The embodiment of the present application does not limit the process of normal vector clustering. For example, point clouds of multiple windows on a single side are obtained through reorganization and normal vector clustering.

[0109] Furthermore, the point cloud of each area to be cleaned is gridded to process the continuous point cloud into a grid structure to obtain a semantic grid map, wherein the semantic grid map is endowed with corresponding semantic information.

[0110] It can be seen from the description of the above steps that in the car wash path planning method provided in the embodiment of the present application, the semantic point cloud of the vehicle to be washed is first determined based on the acquired surface contour point cloud and feature semantic information of the vehicle to be washed, and then the semantic point cloud is reorganized and the normal vectors are clustered according to the cleaning process to obtain point clouds of multiple areas to be washed, and then the point cloud of each area to be washed is gridded to obtain a semantic grid map of the vehicle to be washed, thereby realizing three-dimensional refined modeling of the surface of the vehicle to be washed, which is conducive to covering the concave and convex areas and dirty areas of the vehicle to be washed during subsequent path search, and is also combined with the cleaning process, so that specific areas of the vehicle to be washed, such as gaps, dirt, etc., can be targetedly cleaned, such as multiple spraying, scrubbing and other cleaning operations, to achieve refined contact car washing and achieve the expected cleaning effect.

[0111] S103: Perform path search according to the semantic grid map, and perform coordinate transformation on the searched target path to obtain a target car wash path.

[0112] Path search, based on a semantic grid map, ensures full coverage while avoiding specific non-wipeable areas to meet the obstacle avoidance requirements of the end-of-arm tool, effectively resolving the conflict between modeling and safety. The resulting target path is then transformed into a target path in the vehicle coordinate system, which is then saved as the target car wash path.

[0113] From the above description, it can be seen that the car wash path planning method provided in the embodiment of the present application determines the semantic point cloud of the vehicle based on the surface contour point cloud, vehicle styling characteristics and dirt characteristics of the vehicle, and determines the area to be washed of the vehicle in combination with the cleaning process, and then grids the area to be washed to obtain a semantic grid map, thereby realizing fine-grained modeling of the surface of the vehicle to be washed, and then performing path planning based on the modeling, so that the car wash robot can realize adaptive and contact-based fine-grained car washing according to the planned path, and can identify dirty areas for targeted cleaning, so that the car wash robot has intelligent cleaning capabilities for dirty areas and can achieve the expected cleaning effect.

[0114] Furthermore, in order to facilitate the car washing robot to control the car washing execution component according to the cleaning process to achieve the car washing action, the method further includes:

[0115] S104: Generate a path file corresponding to the target car wash path according to different process configurations corresponding to the cleaning process.

[0116] After completing the above car wash path planning, the target car wash path can be saved as a separate file. For example, the target car wash path can be saved according to business needs. The business needs can be different process configurations, such as water spraying (non-contact type) or wiping (contact type) and other different process configurations.

[0117] S105 , calling the corresponding path file according to the process configuration, and controlling the car wash execution component of the car wash robot to clean the vehicle to be washed according to the called path file and the process configuration.

[0118] Furthermore, the execution order of the path files during a car wash can be determined by the process configuration. That is, the corresponding path files are called based on the process configuration. For example, the order of wiping or spraying areas can be to wipe the roof first, or to wipe a certain area multiple times, such as a dirty area. The path file is then used as input, and the car wash execution component is controlled according to the target car wash path saved in the path file and the corresponding process configuration to clean the vehicle to be washed, thus completing the car wash action.

[0119] In a possible implementation, step S105 may include:

[0120] First, the motion information of the car wash actuator is determined based on the invoked path file. For example, based on the target car wash path saved in the invoked path file, the robot arm's joint angles and the chassis's position are calculated using inverse kinematics. Then, control instructions are generated and issued to the car wash actuator, such as the robot arm and chassis. The car wash actuator then responds to the control instructions and performs the car wash according to the invoked path file, process configuration, and motion information.

[0121] In some embodiments, the motion information of the car wash execution component may include the joint angles of the robotic arm and the position of the chassis. A possible implementation method for performing a car wash action based on the invoked path file, process configuration, and motion information may be to respond to control instructions and, through the robotic arm's joint angles and the position of the mobile chassis, guide the cleaning tool installed in the robotic arm module to execute the car wash action corresponding to the process configuration according to the target car wash path in the invoked path file, thereby completing the detailed cleaning of the vehicle to achieve the desired cleaning effect.

[0122] In some embodiments, there may be specific spraying and wiping requirements when performing a car wash. For example, some areas can only be sprayed with water in the air and are not suitable for contact wiping, and there are also certain requirements for the direction of the water spray. In this case, it is necessary to offset the position and normal of the end of the robot arm. For example, the following implementation can be used to offset the elevation and normal of the target path in the vehicle coordinate system:

[0123]

[0124] Figure 3 Schematic diagram of the process of car wash path planning method provided in this application Figure 2 ,like Figure 3 As shown, the method includes:

[0125] S201a, obtaining a surface point cloud of the vehicle to be washed in a vehicle coordinate system, and preprocessing the surface point cloud to obtain a surface contour point cloud.

[0126] For example, simultaneous localization and mapping (SLAM) technology, based on sensors such as multi-line lidar or depth cameras, can be used to obtain a surface point cloud of the vehicle being washed in the vehicle coordinate system. This surface point cloud is then preprocessed to obtain point cloud data of the vehicle's surface, also known as a surface contour point cloud.

[0127] In some embodiments, preprocessing includes at least reducing point cloud density and removing irrelevant point clouds. For example, raw point clouds are generally large, making it difficult for downstream processing to be performed in full and in real time on embedded processing units. Therefore, downsampling can be used to reduce point cloud density. Furthermore, due to measurement noise from sensors such as lidar and cameras, as well as environmental noise, surface point clouds cannot truly reflect the contours of the vehicle's exterior surface and often contain many stray points, such as point clouds of the ground or other objects. Therefore, irrelevant point clouds need to be removed, retaining only the point cloud representing the contours of the vehicle.

[0128] S201b: Acquire a 2D image of the vehicle to be washed, and perform detection and segmentation on the 2D image to determine semantic information of various components of the vehicle to be washed and semantic information of dirty areas.

[0129] For example, a 2D image of a vehicle to be washed is acquired through a visible light camera, and segmentation is performed on the 2D image to detect semantic information of various components of the vehicle to be washed and dirty areas. For example, semantic detection of areas such as the front of the vehicle, hood, front window, roof, rear window, trunk, rear, left and right front and rear doors, front and rear windows, front and rear wheel suspensions, as well as semantic detection of dirty areas on the surface of the vehicle to be washed is performed, thereby obtaining semantic information of various components of the vehicle to be washed and semantic information of dirty areas.

[0130] S202 : Determine a semantic point cloud of the vehicle to be washed based on the surface contour point cloud and feature semantic information of the vehicle to be washed.

[0131] Based on the surface contour point cloud of the vehicle to be washed, corresponding semantic information is assigned to the surface contour point cloud according to the visual detection result, that is, the feature semantic information of the vehicle to be washed, to obtain the semantic point cloud of the vehicle to be washed.

[0132] S203 , reorganizing the regions corresponding to the semantic point cloud according to a preset region reorganization strategy based on the cleaning process to obtain a plurality of component-level point clouds.

[0133] S204 , clustering each component-level point cloud by normal vectors to obtain a point cloud for each area to be cleaned.

[0134] For example, if the glass and non-glass components of the vehicle to be washed are made of different materials, the cleaning process may be different, such as wiping, spraying, etc., so that the overall semantic point cloud can be split according to the preset region reconstruction strategy to obtain multiple corresponding component-level point clouds, and the point clouds of the dirty areas can exist separately. Among them, the preset region reconstruction strategy is a point cloud reconstruction strategy customized according to conditions such as the cleaning process and the material of the vehicle components. The embodiment of this application does not limit the specific content of the preset region reconstruction strategy. Its purpose is to split the entire point cloud into component-level point clouds based on the components of the vehicle to be washed. The components of the vehicle to be washed can be, for example, windows, doors, etc.

[0135] Normal vector clustering is performed on the component-level point cloud to generate point clouds for multiple areas to be cleaned. Normal vector clustering combines point clouds with surface normal vectors that have similar directions or characteristics. This embodiment of the application does not limit the normal vector clustering process. For example, point clouds for multiple windows on a single side can be generated through reorganization and normal vector clustering.

[0136] S205 , performing coordinate transformation on the point cloud of each area to be cleaned according to a preset projection transformation perspective to obtain a grid point cloud.

[0137] S206 , performing averaging processing on the grid point cloud and determining the grid center depth to obtain an intermediate grid map.

[0138] S207 , obtaining a semantic grid map based on the intermediate grid map and the semantics of the point cloud of each corresponding area to be cleaned.

[0139] The point cloud of each area to be washed is meshed, and the point cloud of each area to be washed in the vehicle coordinate system is converted into a point cloud of each area to be washed in the grid coordinate system according to a preset projection transformation perspective, and the point cloud of each area to be washed in the grid coordinate system is defined as a grid point cloud. Optionally, the grid coordinate system can be, for example, a global coordinate system.

[0140] In some embodiments, the first transformation matrix is ​​derived from a preset projection transformation perspective. Different transformation perspectives correspond to different first transformation matrices, such as those from the front, back, left, right, and top of the vehicle. For ease of calculation, overlapping coordinate axes can be selected. For example, from a perspective on the left side of the vehicle being washed, the corresponding first transformation matrix is ​​as follows:

[0141]

[0142] The conversion relationship between the vehicle coordinate system and the grid coordinate system is as follows:

[0143] Pgk=transform_left_side*Pvk

[0144] Ngk=transform_left_side*Nvk

[0145] The vehicle coordinate system typically uses the XOY plane parallel to the ground and the Z axis perpendicular to the ground, ensuring right-handed rotation. The origin O is the center of the front axle of the vehicle being washed, the X axis points toward the rear of the vehicle, and the Y axis points from the left side of the front axle to the right. The position of the point cloud in the k-th vehicle coordinate system is represented by Pvk = [Pvk_x, Pvk_y, Pvk_z], and the normal vector is represented by Nvk = [Nvk_x, Nvk_y, Nvk_z].

[0146] In this step, the 3D point cloud is converted to 2.5D grid coordinates, which can significantly reduce the complexity of solving the subsequent path planning problem. The grid coordinate system is defined as a right-handed coordinate system. The point cloud in the k-th vehicle coordinate system can be converted to the grid coordinate system as Pgk = [Pgk_x, Pgk_y, Pgk_z], and the normal vector is represented as Ngk = [Ngkx_x, Ngk_y, Ngk_z].

[0147] The point cloud in the vehicle coordinate system is converted to a point cloud in the grid coordinate system. After obtaining the grid point cloud, the grid point cloud is averaged and the grid center point value is calculated to determine the grid center depth. The grid point cloud after averaging and grid center point value calculation is determined as the intermediate grid map. The intermediate grid map is then assigned corresponding semantic information to form a 2.5D semantic grid map.

[0148] In addition to position and normal information, semantic grid maps also contain semantic information Sk. For example, the 2D grid point cloud, Pgk_x and Pgk_y, represents the position of a point on the XY plane, while Pgk_z, Ngk, and Sk represent the elevation, normal, and semantic information of the point, respectively.

[0149] During gridding, the three-dimensional planning problem is converted into a two-dimensional planning problem through projection transformation, which reduces the complexity of problem processing and makes the cleaning path planning method provided in the embodiment of the present application simpler, more practical and universal.

[0150] S208 , starting from the predefined position, performing a bow-shaped path forward search on the semantic grid map along a preset direction according to the sampling grid and the preset inerasable area, and obtaining a preset amount of grid data.

[0151] For example, starting from a predefined position, such as the predefined upper left corner of the front of the vehicle to be washed, a bow-shaped path is forward searched along the X-axis of the grid in units of sampling grids, while avoiding the preset non-erasable area, and the preset number of grids searched are obtained as grid data. Figure 4 As shown. Figure 4 In the grid shown, the polygonal domain box represents the area that needs to be cleaned, which is assumed to be a convex polygon; the circle represents the cleaning waypoint. After grid division, if the point cloud in the cleaning area falls within the grid area, it means that the grid is valid; the triangular area box represents the area that needs to be avoided in this cleaning area, which can be determined based on the geometric shape (such as elevation, normal vector and semantic information).

[0152] Among them, the preset non-wipeable area is to meet the obstacle avoidance requirements of the cleaning tool at the end of the robotic arm. The non-wipeable area can be set according to the size specifications of the cleaning tool, and this embodiment of the present application does not limit this.

[0153] In some embodiments, the size of the sampling grid can be pre-set according to the modeling accuracy and the computing power of the processor, for example, to a 1 cm square grid or a 1 cm×2 cm rectangular grid.

[0154] In some embodiments, since the exterior shape of the vehicle to be washed has both large, relatively flat surfaces (such as glass) and more concave, convex, and angular surfaces, the shape or contact surface of the part of the cleaning tool that contacts the vehicle to be washed is usually relatively fixed. Therefore, in order to take into account both full coverage and cleaning efficiency, it is necessary to adaptively and maximize the use of the contact surface between the cleaning tool and the vehicle to be washed. Therefore, in the forward path search, the sampling grid needs to be adaptively adjusted so that the sparsity of the acquired grid data can meet the contact surface requirements. The contact surface requirements refer to the requirements for the contact area between the car washing robot and the vehicle to be washed when cleaning the vehicle to be washed, such as the requirements for the contact surface between the cleaning tool and the vehicle to be washed.

[0155] For example, a possible implementation of adaptive mesh adjustment might include: when a point cloud falls into a mesh, determining the normal vectors of each mesh, obtaining the gradient of the normal vector change in two directions (forward and downward), and then sampling horizontally. If the gradient change is less than a preset threshold, the mesh is sampled sparsely; if it is greater than the threshold, the mesh is sampled densely. The preset threshold is determined based on the contact surface requirements.

[0156] Furthermore, elevation information processing is also performed on the grid data. For example, a grid may fall into multiple point clouds. Due to the lack of depth information in the grid coordinate system, the point clouds falling into the grid need to be screened and averaged, and the depth of the grid center point needs to be calculated.

[0157] S209 , associating the point clouds in the grid data to obtain the searched target path, and performing coordinate transformation on the searched target path to obtain the target car wash path.

[0158] The point clouds falling into the grid are sequentially associated by searching to form a bow-shaped path and obtain the target path.

[0159] The following shows a possible implementation of traversing all point clouds in the semantic grid map and traversing all grids:

[0160]

[0161]

[0162] After the target path is formed in the grid coordinate system, a coordinate transformation is performed on the target path to determine the target path in the vehicle coordinate system, thereby obtaining the target car wash path. The coordinate transformation from the grid coordinate system to the vehicle coordinate system is achieved using a second transformation matrix, which is the inverse of the first transformation matrix.

[0163] The target car wash path is saved in a separate file based on the different process configurations corresponding to the cleaning process, forming a path file for the vehicle to be washed. The car wash execution component responds to the control command, calls the path file, reads the corresponding saved target car wash path, and executes the corresponding car wash action, enabling the car wash robot to intelligently wash the vehicle to be washed.

[0164] In some embodiments, the grid data in the vehicle coordinate system obtained by the car wash path planning method provided by the embodiment of the present application is as follows: Figure 5 As shown, Figure 6 Schematic diagram of the target car washing path and normal vector visualization for the area near the rearview mirror of the vehicle to be washed, avoiding the rearview mirror.

[0165] The car wash path planning method provided in the embodiment of the present application can be applied to a car wash robot. This method can determine an adaptive car wash path based on the appearance and styling characteristics of the vehicle to be washed and the surface dirt characteristics in combination with the cleaning process. Therefore, based on the planned car wash path, the cleaning tool can be guided to move in accordance with the surface of the vehicle to be washed, thereby achieving adaptive contact car washing. And by identifying dirty areas, the dirty areas can be cleaned in a targeted manner, or even multiple times. During the path search, the grid is used to achieve adaptive adjustment of the path sparsity through adaptive adjustment. While the path planning can achieve full coverage of the vehicle body surface, the concave and convex areas of the vehicle to be washed are cleaned in a targeted manner to ensure that the cleaning effect meets the expectations.

[0166] Figure 7 This is a schematic diagram of the structure of the car wash path planning device provided in this application, which is applied to a car wash robot. Figure 7 As shown, the car wash path planning device 40 provided in the embodiment of the present application includes

[0167] The semantic processing module 401 is used to determine the semantic point cloud of the vehicle to be washed based on the surface contour point cloud and feature semantic information of the vehicle to be washed, wherein the feature semantic information includes semantic information of each component of the vehicle to be washed and semantic information of the dirty area;

[0168] A gridding module 402 is configured to reorganize the semantic point cloud and cluster normal vectors according to the cleaning process to obtain point clouds of multiple areas to be washed, and grid the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed;

[0169] The path search module 403 is used to perform path search according to the semantic grid map and perform coordinate transformation on the searched target path to obtain a target car wash path.

[0170] In a possible implementation, the car wash path planning device 40 further includes a car wash control module, which is configured to:

[0171] Generate a path file corresponding to the target car wash path according to different process configurations corresponding to the cleaning process;

[0172] The corresponding path file is called according to the process configuration, and the car washing execution component of the car washing robot is controlled to clean the vehicle to be washed according to the called path file and the process configuration.

[0173] In one possible implementation, the car wash control module is further configured to:

[0174] Determine the motion information of the car wash execution component according to the called path file;

[0175] Generate and send control instructions to the car wash execution component;

[0176] The car wash execution component is controlled to respond to the control instruction to perform the car wash action according to the called path file, process configuration and motion information.

[0177] In one possible implementation, the car wash control module is further configured to:

[0178] Through the joint angles of the robotic arm and the position of the mobile chassis, the cleaning tool is guided to perform the car washing action corresponding to the process configuration according to the target car washing path in the called path file;

[0179] Among them, the motion information of the car wash execution component includes the joint angle of the robotic arm and the position of the chassis, and the cleaning tool is installed at the end of the robotic arm.

[0180] In a possible implementation, the car wash route planning device further includes: an acquisition module; the acquisition module is configured to:

[0181] Obtaining a surface point cloud of the vehicle to be washed in a vehicle coordinate system, and preprocessing the surface point cloud to obtain a surface contour point cloud, wherein the preprocessing at least includes reducing the point cloud density and eliminating irrelevant point cloud operations;

[0182] A 2D image of the vehicle to be washed is obtained, and detection and segmentation are performed on the 2D image to determine semantic information of various parts of the vehicle to be washed and semantic information of dirty areas.

[0183] In a possible implementation, the semantic processing module 401 is specifically configured to:

[0184] According to the cleaning process, the regions corresponding to the semantic point cloud are reorganized according to a preset region reorganization strategy to obtain multiple component-level point clouds;

[0185] Each component-level point cloud is clustered by normal vectors to obtain the point cloud of each area to be cleaned.

[0186] In a possible implementation, the gridding module 402 is specifically configured to:

[0187] Perform coordinate transformation on the point cloud of each area to be cleaned according to the preset projection transformation perspective to obtain a grid point cloud;

[0188] Perform averaging on the grid point cloud and determine the grid center depth to obtain the intermediate grid map;

[0189] According to the semantics of the point cloud of each area to be cleaned and the intermediate grid map, a semantic grid map is obtained;

[0190] The grid point cloud refers to the point cloud of each area to be cleaned in the grid coordinate system.

[0191] In a possible implementation, the gridding module 402 is further configured to:

[0192] The coordinates of the point cloud of each area to be cleaned are transformed according to the first transformation matrix to obtain a grid point cloud. The first transformation matrix is ​​obtained according to a preset projection transformation perspective.

[0193] In a possible implementation, the path search module 403 is specifically configured to:

[0194] Starting from a predefined position, a forward search of a bow-shaped path is performed on the semantic grid map along a preset direction according to the sampling grid and the preset inerasable area to obtain a preset amount of grid data;

[0195] The point cloud in the grid data is associated to obtain the searched target path.

[0196] In a possible implementation, the path search module 403 is further configured to:

[0197] In the forward path search, the sampling grid is adaptively adjusted so that the sparsity of the grid data meets the contact surface requirements;

[0198] The contact surface requirement refers to the requirement for the contact area between the car washing robot and the vehicle to be washed when cleaning the vehicle to be washed.

[0199] The car wash path planning device provided in the embodiment of the present application can execute the method provided in the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0200] Figure 8 The structural diagram of the controller provided in this application is as follows: Figure 8 As shown, the controller 50 provided in this embodiment includes: at least one processor 501 and a memory 502 .

[0201] Optionally, the controller 50 further includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected via a bus.

[0202] In a specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that the at least one processor 501 performs the above method.

[0203] The specific implementation process of the processor 501 can be found in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here in this embodiment.

[0204] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly executed by a hardware processor or by a combination of hardware and software modules in the processor.

[0205] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0206] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be classified into address buses, data buses, and control buses. For ease of illustration, the buses in the drawings of this application are not limited to just one bus or just one type of bus.

[0207] The present application also provides a car washing robot, comprising a sensor, a car washing execution component, a cleaning tool, and a controller, wherein the controller is configured to execute the method of the above method embodiment;

[0208] A sensor is used to obtain a surface point cloud and a 2D image of the vehicle to be washed in a vehicle coordinate system;

[0209] The car wash execution component is used to guide the cleaning tool to achieve the car wash action. The cleaning tool is installed at the end of the mechanical arm of the car wash execution component. The car wash execution component also includes a mobile chassis.

[0210] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0211] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0212] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0213] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist in a device as discrete components.

[0214] The division of units is merely a logical functional division; actual implementations may employ alternative divisions, such as combining or integrating multiple units or components into another system, or omitting or disabling certain features. Furthermore, any direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units, either through an interface, electrical, mechanical, or other means.

[0215] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0216] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0217] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0218] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0219] Finally, it should be noted that those skilled in the art will readily identify other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. The present invention is not limited to the precise structure described above and illustrated in the accompanying drawings, and various modifications and variations may be made without departing from the scope thereof. The scope of the present invention is limited solely by the appended claims.

Claims

1. A car wash path planning method, characterized in that: Applied to a car washing robot, the method comprises: Determining a semantic point cloud of the vehicle to be washed based on a surface contour point cloud and feature semantic information of the vehicle to be washed, wherein the feature semantic information includes semantic information of various components of the vehicle to be washed and semantic information of dirty areas; Reorganizing the semantic point cloud and performing normal vector clustering according to the cleaning process to obtain point clouds of multiple areas to be washed, and gridding the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed; A path search is performed according to the semantic grid map, and a coordinate transformation is performed on the searched target path to obtain a target car wash path.

2. The method according to claim 1, characterized in that After obtaining the target car wash path, the method further includes: Generate a path file corresponding to the target car washing path according to different process configurations corresponding to the cleaning process; The corresponding path file is called according to the process configuration, and the car washing execution component of the car washing robot is controlled to clean the vehicle to be washed according to the called path file and the process configuration.

3. The method according to claim 2, characterized in that The step of controlling the car washing execution component of the car washing robot to clean the vehicle to be washed according to the called path file and the process configuration includes: Determining motion information of the car wash execution component according to the called path file; Generate and issue control instructions to the car wash execution component; The car wash execution component is controlled to respond to the control instruction to perform a car wash action according to the called path file, the process configuration and the motion information.

4. The method according to claim 3, characterized in that The performing of the car washing action according to the called path file, the process configuration, and the motion information includes: Through the joint angles of the robotic arm and the posture of the mobile chassis, the cleaning tool is guided to perform the car washing action corresponding to the process configuration according to the target car washing path in the called path file; The motion information of the car wash execution component includes the joint angle of the robotic arm and the posture of the chassis, and the cleaning tool is installed at the end of the robotic arm.

5. The method according to any one of claims 1 to 4, characterized in that Before the step of washing the vehicle based on the surface contour point cloud and feature semantic information, the method further includes: Acquire a surface point cloud of the vehicle to be washed in a vehicle coordinate system, and preprocess the surface point cloud to obtain the surface contour point cloud, wherein the preprocessing at least includes reducing the point cloud density and eliminating irrelevant point cloud operations; A 2D image of the vehicle to be washed is acquired, and detection and segmentation are performed on the 2D image to determine semantic information of various parts of the vehicle to be washed and semantic information of the dirty area.

6. The method according to claim 5, characterized in that The step of reorganizing the semantic point cloud and clustering normal vectors according to the cleaning process to obtain point clouds of multiple areas to be cleaned includes: Reorganize the regions corresponding to the semantic point cloud according to the cleaning process according to a preset region reorganization strategy to obtain a plurality of component-level point clouds; Each component-level point cloud is clustered by the normal vector to obtain a point cloud of each area to be cleaned.

7. The method according to claim 6, characterized in that The step of gridding the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed includes: Performing coordinate transformation on the point cloud of each area to be cleaned according to a preset projection transformation perspective to obtain a grid point cloud; Performing averaging processing on the grid point cloud and determining the grid center depth to obtain an intermediate grid map; Obtain the semantic grid map according to the intermediate grid map and the semantics of the point cloud of each area to be cleaned corresponding thereto; The grid point cloud refers to the point cloud of each area to be cleaned in a grid coordinate system.

8. The method according to claim 7, characterized in that The coordinate transformation of the point cloud of each area to be cleaned is performed according to a preset projection transformation perspective to obtain a grid point cloud, including: The coordinate transformation of the point cloud of each area to be cleaned is performed according to a first transformation matrix to obtain the grid point cloud, wherein the first transformation matrix is ​​obtained according to the preset projection transformation perspective.

9. The method according to claim 8, characterized in that The performing path search according to the semantic grid map includes: Starting from a predefined position, a forward search is performed on the semantic grid map along a preset direction in a bow-shaped path according to a sampling grid and a preset inerasable area to obtain a preset amount of grid data; The point clouds in the grid data are associated to obtain the searched target path.

10. The method according to claim 9, characterized in that The method further comprises: In the forward path search, the sampling grid is adaptively adjusted so that the sparsity of the grid data meets the contact surface requirement; The contact surface requirement refers to the requirement for the contact area between the car washing robot and the vehicle to be washed when cleaning the vehicle to be washed.

11. A car wash path planning device, characterized in that: Applied to a car washing robot, the device comprises: a semantic processing module for determining a semantic point cloud of the vehicle to be washed based on a surface contour point cloud and feature semantic information of the vehicle to be washed, wherein the feature semantic information includes semantic information of various components of the vehicle to be washed and semantic information of dirty areas; A gridding module is used to reorganize the semantic point cloud and cluster normal vectors according to the cleaning process to obtain point clouds of multiple areas to be washed, and grid the point cloud of each area to be washed to obtain a semantic grid map of the vehicle to be washed; The path search module is used to perform path search according to the semantic grid map and perform coordinate transformation on the searched target path to obtain a target car wash path.

12. A controller, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 10.

13. A car washing robot, characterized in that: comprising a sensor, a car wash execution component, a cleaning tool, and the controller according to claim 12; The sensor is used to obtain a surface point cloud and a 2D image of the vehicle to be washed in a vehicle coordinate system; The car wash execution component is used to guide the cleaning tool to perform a car wash action. The cleaning tool is installed at the end of the mechanical arm of the car wash execution component. The car wash execution component also includes a mobile chassis.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 10 when executed by a processor.

15. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 10 when executed by a processor.

Images