Environmental perception method, device, electronic device and storage medium
By dividing the target task into sub-tasks and determining the voxel resolution according to the accuracy requirements, the problem of accuracy and calculation balance in environmental perception by general robots and specific functional robots is solved, and efficient environmental perception capabilities are achieved.
Patent Information
- Application Number
- CN202411096642.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-12
AI Technical Summary
In the prior art, general robots and specific functional robots find it difficult to balance task accuracy and data calculation amount in terms of environmental perception. If the voxel resolution is set too low, the accuracy will be low. If the setting is too high, the calculation amount will increase exponentially.
Divide the target task into sub-tasks, and determine the appropriate voxel resolution according to the accuracy requirements of the sub-tasks, divide the molecular spatial area into the target voxels, and perform environment perception based on the characteristic information of these voxels to achieve the accuracy requirements of different sub-tasks. Matching with voxel resolution.
The optimal balance between task accuracy and data calculation amount is achieved through appropriate voxel resolution, and the environmental perception capability of smart devices is improved.
Smart Images

Figure CN118629029B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to an environment perception method, device, electronic device and storage medium. Background Art
[0002] Environmental perception capability is a basic capability that general-purpose robots, robots with specific functions such as programmable industrial robots and other intelligent devices need to complete designated tasks. For example, they can obtain information such as the location, shape, type, speed, and driving trajectory of obstacles around them through environmental perception capabilities, and then complete designated tasks based on this information, such as moving to a shelf to pick up goods.
[0003] In the related technology, general robots, robots with specific functions and other intelligent devices can obtain environmental perception capabilities based on occupancy network technology to complete designated tasks. Among them, the core idea of occupancy network technology is to learn the probability of each voxel being occupied in three-dimensional space through neural networks, that is, to judge whether a spatial position is occupied by an entity. However, if the voxel resolution is set low, the intelligent device can only complete tasks with low precision requirements. If the voxel resolution is set high in order to complete tasks with high precision requirements, the amount of data calculation will increase exponentially. Therefore, how to achieve environmental perception based on occupancy network technology with appropriate voxel resolution, so as to balance the accuracy and data calculation amount of tasks that can be completed by intelligent devices, is an urgent problem to be solved. Summary of the invention
[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.
[0005] To this end, the first objective of this application is to propose an environment perception method.
[0006] The second objective of this application is to provide an environment sensing device.
[0007] The third objective of the present application is to provide an electronic device.
[0008] A fourth objective of the present application is to provide a computer-readable storage medium.
[0009] A fifth object of the present application is to provide a computer program product.
[0010] To achieve the above-mentioned purpose, the first aspect of the present application proposes an environmental perception method, including: dividing a target task to be performed in a target space area into at least one subtask; for any subtask in the at least one subtask, determining the subspace area required to be perceived to perform the arbitrary subtask in the target space area, and based on the accuracy requirement of the arbitrary subtask, determining the voxel resolution corresponding to the arbitrary subtask; dividing the subspace area corresponding to the arbitrary subtask into at least one target voxel according to the corresponding voxel resolution; and performing environmental perception on the target space area based on the feature information corresponding to the at least one target voxel into which each of the subspace areas is divided.
[0011] To achieve the above-mentioned purpose, the second aspect of the present application proposes an environmental perception device, including: a first processing module, used to divide the target task performed in the target space area into at least one subtask; a determination module, used to determine, for any subtask in the at least one subtask, the subspace area required to be perceived in the target space area to perform the arbitrary subtask, and based on the accuracy requirement of the arbitrary subtask, determine the voxel resolution corresponding to the arbitrary subtask; a second processing module, used to divide the subspace area corresponding to the arbitrary subtask into at least one target voxel according to the corresponding voxel resolution; and a third processing module, used to perform environmental perception of the target space area based on feature information corresponding to the at least one target voxel into which each of the subspace areas is divided.
[0012] To achieve the above-mentioned purpose, the third aspect of the embodiment of the present application proposes an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the environmental perception method as described in the first aspect of the embodiment of the present application.
[0013] To achieve the above-mentioned purpose, the fourth aspect of the embodiment of the present application proposes a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by the processor, they are used to implement the environmental perception method as described in the first aspect of the embodiment of the present application.
[0014] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes a computer program product, including a computer program, which, when executed by a processor, implements the environmental perception method as described in the first aspect of the embodiment of the present application.
[0015] The technical solution provided by this application brings at least the following beneficial effects:
[0016] The target task to be performed in the target space region is divided into at least one subtask, and for any subtask in the at least one subtask, the subspace region required to be perceived for performing any subtask in the target space region is determined, and based on the accuracy requirement of any subtask, the voxel resolution corresponding to any subtask is determined, and the subspace region corresponding to any subtask is divided into at least one target voxel according to the corresponding voxel resolution, and the environment perception of the target space region is performed based on the feature information corresponding to the at least one target voxel into which each subspace region is divided, so that the voxel resolution suitable for different subtasks is determined according to the accuracy requirements of different subtasks of the target task performed in the target space region, and then based on the suitable voxel resolution, the subspace regions corresponding to different subtasks are voxel-divided and the environment perception is realized, for example, for subtasks that require high-precision perception, high voxel resolution is used to improve the perception accuracy, and for subtasks that do not require high-precision perception, low voxel resolution is used to reduce the amount of data calculation, so that the accuracy of the tasks that can be completed by the smart device and the amount of data calculation can be optimally balanced.
[0017] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0019] Figure 1 A schematic diagram of a flow chart of an environment perception method provided in one embodiment of the present application;
[0020] Figure 2 A schematic diagram of a flow chart of an environment perception method provided by another embodiment of the present application;
[0021] Figure 3 A schematic diagram of a process flow of an environment perception process provided by an embodiment of the present application;
[0022] Figure 4 A schematic diagram of the structure of an environment sensing device provided in one embodiment of the present application;
[0023] Figure 5 It is a block diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION
[0024] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0025] The following describes the environment perception method, device, electronic device and storage medium of the embodiments of the present application with reference to the accompanying drawings.
[0026] Figure 1 A flowchart of an environment perception method provided in an embodiment of the present application.
[0027] It should be noted that the environmental perception method provided in the embodiments of the present application can be performed by an environmental perception device. Among them, the environmental perception device can be an electronic device, and can also be configured in an electronic device so that the electronic device can perceive the environment, and the present application does not limit this. Among them, the electronic device can be any intelligent device with environmental perception capabilities, such as a general-purpose robot, a robot with specific functions, an autonomous driving vehicle, a controller of an autonomous driving vehicle, an unmanned aerial vehicle, a smart home device, a wearable device, etc., and the present application does not limit this. The embodiments of the present application are described by taking the electronic device as a general-purpose robot and the environmental perception device as a device configured in the general-purpose robot to enable the general-purpose robot to have environmental perception capabilities as an example.
[0028] like Figure 1 As shown, the environment perception method includes the following steps:
[0029] Step 101 : dividing a target task to be performed in a target space region into at least one subtask.
[0030] Among them, the target task can be any task that a general robot can perform, for example, the target task can be "move to the front of the table and pick up the water cup on the table", "go to shelf A, pick up product B, and deliver it to warehouse C", etc.
[0031] The target space area is the space area that the general robot needs to perceive to perform the target task. For example, the target space area can be a three-dimensional area of 20*20*1.6 cubic meters. The first 20 in 20*20*1.6 represents the size of the target space area in the x-axis direction, the second 20 represents the size of the target space area in the y-axis direction, and 1.6 represents the size of the target space area in the z-axis direction.
[0032] In some embodiments, the target task can be divided into at least one stage according to the characteristics of the target task, and each stage corresponds to a subtask. The number of subtasks into which the target task is divided can be one or more, and can be adjusted according to the characteristics of the target task, and this application does not limit this.
[0033] For example, taking the target task of "moving to the table and picking up the water cup on the table" as an example, the target task can be divided into two subtasks: "moving to the table" and "picking up the water cup on the table" according to different task accuracy requirements.
[0034] Taking the target task of "go to shelf A, pick up product B, and deliver it to warehouse C" as an example, the target task can be divided into three subtasks: "go to shelf A", "pick up product B", and "deliver product B to warehouse C" according to different task accuracy requirements.
[0035] In some embodiments, the user can pre-set which subtasks the target task is to be divided into, so that the environment perception device can divide the target task into at least one subtask according to the preset task division information. Alternatively, the environment perception device can divide the target task into at least one subtask according to the characteristics of the target task. This application does not limit the method of dividing the target task into at least one subtask.
[0036] Step 102, for any subtask in at least one subtask, determine the subspace region required to be sensed in the target space region to perform any subtask, and determine the voxel resolution corresponding to any subtask based on the accuracy requirement of any subtask.
[0037] Among them, voxel, derived from the combination of "volume" and "pixel", is a discrete unit in three-dimensional space, similar to a pixel in a two-dimensional image. It is usually used to represent and process three-dimensional data, such as medical imaging (CT, MRI), three-dimensional reconstruction, volume rendering, and scene representation in computer graphics.
[0038] A voxel is a small cube or volume unit in three-dimensional space that occupies a fixed position in a three-dimensional grid. Each voxel has a definite size and position and can contain multiple attributes, such as density, color, or other physical properties.
[0039] Voxel resolution can represent the size and density of voxels in three-dimensional space, and determines the level of detail and computational complexity of the three-dimensional model. The higher the voxel resolution, the more voxels are contained in a unit volume, the finer the details of the three-dimensional model, and the higher the computational complexity.
[0040] In some embodiments, for any subtask among the at least one subtask into which the target task is divided, a subspace region required to be sensed to execute any subtask in the target space region may be determined.
[0041] For example, the target task is "move to the table and pick up the water cup on the table". After dividing the target task into two subtasks, "move to the table" and "pick up the water cup on the table", for the "move to the table" subtask, the subspace area corresponding to the subtask can be the area corresponding to the path from the current position to the position of the table; for the "pick up the water cup on the table" subtask, the subspace area corresponding to the subtask can be the area where the water cup is.
[0042] Take the target task of "go to shelf A, pick up item B, and deliver to warehouse C" as an example. After dividing the target task into three subtasks of "go to shelf A", "pick up item B" and "deliver item B to warehouse C", for the "go to shelf A" subtask, the subspace area corresponding to the subtask can be the area corresponding to the path from the current position to the location of shelf A; for the "pick up item B" subtask, the subspace area corresponding to the subtask can be the area where item B is located; for the "deliver item B to warehouse C" subtask, the subspace area corresponding to the subtask can be the area corresponding to the path from the current location of shelf A where the robot is located to the location of warehouse C.
[0043] In some embodiments, the correspondence between the degree of precision requirement of a task or subtask and the voxel resolution can be preset. The lower the precision requirement, the lower the corresponding voxel resolution. For example, the precision requirement of a task can be divided into three levels: low, medium, and high, wherein the low-level precision requirement corresponds to voxel resolution a, the medium-level precision requirement corresponds to voxel resolution b, and the high-level precision requirement corresponds to voxel resolution c. The voxel resolutions of voxel resolution a, voxel resolution b, and voxel resolution c increase in sequence. The higher the voxel resolution, the smaller the corresponding voxel size. For example, the size of the voxel corresponding to voxel resolution a can be 1 cubic meter, the size of the voxel corresponding to voxel resolution b can be 1 cubic decimeter, and the size of the voxel corresponding to voxel resolution c can be 1 cubic centimeter. Thus, for any subtask in at least one subtask into which the target task is divided, the voxel resolution corresponding to the arbitrary subtask can be determined according to the precision requirement of the arbitrary subtask and the preset correspondence.
[0044] For example, taking the target task of "go to shelf A, pick up product B, and deliver it to warehouse C" as an example, since the two subtasks of "go to shelf A" and "deliver product B to warehouse C" do not have high requirements for accuracy, the voxel resolution corresponding to the subtasks of "go to shelf A" and "deliver product B to warehouse C" can be determined to be b. As for the subtask of "pick up product B", since this subtask has high requirements for accuracy, the voxel resolution corresponding to the subtask of "pick up product B" can be determined to be c.
[0045] In some embodiments, the accuracy requirements corresponding to each of the multiple task types can be pre-set, so that for any subtask, the environmental perception device can determine the accuracy requirement corresponding to the arbitrary subtask based on the task type to which the subtask belongs and the correspondence between the pre-set task type and the accuracy requirement.
[0046] Step 103: Divide the subspace region corresponding to any subtask into at least one target voxel according to the corresponding voxel resolution.
[0047] The number of target voxels into which the subspace region corresponding to any subtask is divided is related to the size of the subspace region and the size of the voxel resolution.
[0048] The sizes of target voxels divided into different subspace regions may be the same or different, and this application does not impose any limitation on this.
[0049] Step 104 : Perform environmental perception on the target space region based on feature information corresponding to at least one target voxel into which each subspace region is divided.
[0050] In some embodiments, for each subspace region, feature information corresponding to each target voxel divided into the subspace region may be obtained, and the feature information may be obtained based on visual data of the target space region, for example, it may include features such as color, texture, material, edge, etc. Then, based on the feature information corresponding to each target voxel divided into all subspace regions, environmental perception of the target space region may be performed.
[0051] Among them, the visual data may include one or more combinations of the following: at least one two-dimensional picture, at least one point cloud data, at least one depth map, etc.
[0052] In some embodiments, step 104 may be implemented by:
[0053] For any target voxel divided into any subspace region, the arbitrary target voxel is classified based on the feature information corresponding to the arbitrary target voxel to obtain the classification information of the arbitrary target voxel, wherein the classification information includes: whether the arbitrary target voxel is occupied, and when the arbitrary target voxel is occupied, the type of object occupying the arbitrary target voxel belongs; based on the classification information of each target voxel divided into each subspace region, environmental perception is performed on the target space region.
[0054] Among them, the type of object may include tree type, lane type, vehicle type, pedestrian type, etc., which can be set as needed and this application does not impose any restrictions on this.
[0055] Here, taking the case where any target voxel is occupied and the type of the object occupying the target voxel is a pedestrian as an example, it means that the pedestrian appears at the location of the target voxel.
[0056] In some embodiments, a neural network may be configured in an environment sensing device or electronic device, and any target voxel may be classified through a classification network in the neural network to obtain classification information of the target voxel. In addition, one or more other network layers of the neural network may form a resolution adjustment network, which may perform a process of dividing a target task into subtasks, determining a subspace region and a voxel resolution corresponding to a subtask, and dividing each subspace region into at least one target voxel.
[0057] The environmental perception method provided in the embodiment of the present application divides the target task performed in the target space area into at least one subtask, determines the subspace area required to be perceived to perform any subtask in the target space area for any subtask in the at least one subtask, and determines the voxel resolution corresponding to any subtask based on the accuracy requirement of any subtask, divides the subspace area corresponding to any subtask into at least one target voxel according to the corresponding voxel resolution, and performs environmental perception on the target space area based on feature information corresponding to at least one target voxel divided into each subspace area, thereby achieving the determination of suitable voxel resolutions for different subtasks according to the accuracy requirements of different subtasks of the target task performed in the target space area, and then performs voxel division of the subspace areas corresponding to different subtasks based on the suitable voxel resolution and realizes environmental perception, for example, for subtasks that require high-precision perception, high voxel resolution is used to improve perception accuracy, and for subtasks that do not require high-precision perception, low voxel resolution is used to reduce the amount of data calculation, so that the accuracy of the tasks that can be completed by the smart device and the amount of data calculation can be optimally balanced.
[0058] Combine the following Figure 2, the process of voxel division of the subspace region corresponding to any subtask based on voxel resolution in an embodiment of the present application and the process of obtaining feature information corresponding to each target voxel in any subspace region are described. Figure 2 A flowchart of another environment perception method provided in an embodiment of the present application.
[0059] like Figure 2 As shown, the environment perception method may include the following steps:
[0060] Step 201 , extracting features from visual data of a target space region to obtain feature information corresponding to a plurality of original voxels into which the target space region is divided.
[0061] It should be noted that in order to distinguish from the voxels divided according to the voxel resolution corresponding to the subtask in the embodiment of the present application, the embodiment of the present application refers to the voxels that have been divided before the voxels are divided according to the voxel resolution corresponding to the subtask as original voxels.
[0062] The visual data may include one or more of the following combinations: at least one two-dimensional image, at least one point cloud data, at least one depth map, etc. When the visual data includes multiple combinations, the visual data may also be referred to as multimodal data.
[0063] The feature information may include color, texture, material, edge and other features.
[0064] The size or voxel resolution of the original voxel can be set as needed. For example, in order to reduce the amount of calculation, it can be set to a lower voxel resolution, that is, a larger size. For example, the size of the original voxel is set to 1 cubic decimeter.
[0065] In some embodiments, taking the example of visual data including multiple two-dimensional images of the target space area from different perspectives, the process of performing feature extraction on the visual data of the target space area to obtain feature information corresponding to multiple original voxels into which the target space area is divided can be: extracting color, texture, edge and other features from multiple two-dimensional images of the target space area from different visual perspectives; reconstructing a three-dimensional point cloud or model using stereo vision or structure based on multiple two-dimensional images from different visual perspectives; converting the reconstructed three-dimensional model into a voxel grid, wherein each voxel in the voxel network is the original voxel into which the target space area is divided; mapping the features extracted from the two-dimensional images to each original voxel of the voxel grid to obtain feature information corresponding to each original voxel.
[0066] In some embodiments, taking the example of visual data including at least one two-dimensional image and at least one depth map of the target space area, the process of performing feature extraction on the visual data of the target space area to obtain feature information corresponding to multiple original voxels into which the target space area is divided can be: extracting color, texture, edge and other features from the two-dimensional image of the target space area; reconstructing a three-dimensional model based on the two-dimensional image and the depth map; converting the reconstructed three-dimensional model into a voxel grid, wherein each voxel in the voxel network is the original voxel into which the target space area is divided; mapping the features extracted from the two-dimensional image to each original voxel of the voxel grid to obtain feature information corresponding to each original voxel.
[0067] In some embodiments, taking the example that visual data includes at least one two-dimensional image of the target space area and at least one point cloud data, the process of performing feature extraction on the visual data of the target space area to obtain feature information corresponding to multiple original voxels into which the target space area is divided can be: extracting color, texture, edge and other features from the two-dimensional image of the target space area; extracting geometric features from the point cloud data; aligning the features in the two-dimensional image with the point cloud data to ensure that the two are aligned in three-dimensional space; converting the aligned point cloud data into a voxel grid, wherein each voxel in the voxel network is the original voxel into which the target space area is divided; mapping the features extracted from the two-dimensional image and the point cloud data to the corresponding original voxels of the voxel grid, and fusing the features extracted from the two-dimensional image and the point cloud data to obtain feature information corresponding to each original voxel.
[0068] In some embodiments, a neural network can be configured in an environmental perception device or an electronic device, and feature extraction can be performed on visual data through a feature extraction network of the neural network, such as a backbone network, to obtain feature information of multiple original voxels into which the target space area is divided. Among them, the feature extraction network can be any network that can realize feature extraction, and can be set as needed, and this application does not impose any restrictions on this. Among them, the base network is a basic network structure or backbone network for feature extraction, and its main function is to extract useful features from input data (such as images). These features can be low-level (such as edges, textures, etc.) or high-level (such as object shapes, categories, etc.).
[0069] In addition, before using the feature extraction network in the neural network to extract features, the neural network can be trained first, and then the feature extraction network in the trained neural network can be used to accurately extract features from the visual data of the target space area to obtain feature information of each original voxel.
[0070] Step 202: Divide the target task to be performed in the target space area into at least one subtask.
[0071] Step 203, for any subtask in at least one subtask, determine the subspace region required to be sensed in the target space region to perform any subtask, and determine the voxel resolution corresponding to any subtask based on the accuracy requirement of any subtask.
[0072] The specific implementation process and principle of steps 202-203 may refer to the description of other embodiments and will not be repeated here.
[0073] Step 204 , when the voxel resolution corresponding to any subtask is the same as the voxel resolution of the original voxel, the original voxel in the subspace region corresponding to the any subtask is determined as the target voxel in the subspace region.
[0074] It can be understood that, for any subtask, when the voxel resolution determined according to the accuracy requirement of the arbitrary subtask is the same as the voxel resolution of the original voxel, the subspace area corresponding to the arbitrary subtask can be voxel-divided according to the voxel resolution of the original voxel.
[0075] Step 205 , when the voxel resolution corresponding to any subtask is different from the voxel resolution of the original voxel, the original voxel in the subspace region corresponding to the any subtask is reconstructed to obtain the target voxel in the subspace region.
[0076] It can be understood that, for any subtask, when the voxel resolution determined according to the accuracy requirement of the arbitrary subtask is different from the voxel resolution of the original voxels, the original voxels in the subspace area corresponding to the arbitrary subtask can be reconstructed to obtain the target voxels in the subspace area.
[0077] In some embodiments, reconstruction of the original voxels can be achieved through a neural network.
[0078] In some embodiments, when the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, for any original voxel in the subspace region corresponding to the arbitrary subtask, the arbitrary original voxel can be segmented based on the voxel resolution corresponding to the arbitrary subtask through a neural network to obtain multiple target voxels corresponding to the arbitrary original voxel. The neural network and the neural network including the feature extraction network can be the same neural network.
[0079] In some embodiments, when the voxel resolution corresponding to any subtask is smaller than the voxel resolution of the original voxel, a neural network can be used to integrate multiple adjacent original voxels in the subspace region corresponding to the arbitrary subtask based on the voxel resolution corresponding to the arbitrary subtask to obtain the target voxel corresponding to the multiple adjacent original voxels. The neural network and the neural network including the feature extraction network can be the same neural network.
[0080] For example, assuming that the voxel resolution of the original voxel corresponds to a voxel size of 1 cubic decimeter, and the target space region is a 20*20*1.6 cubic meter three-dimensional region, the target space region can be divided into 200*200*16 1 cubic decimeter original voxels. The first, second, and third values correspond to the x-axis, y-axis, and z-axis directions, respectively.
[0081] Assume that the target task to be performed in the target space region is divided into multiple subtasks. Among them, the subspace region corresponding to a subtask 1 is a 5*5*2 cubic decimeter region in the target space region, and the voxel resolution corresponding to subtask 1 corresponds to a voxel size of 1 cubic centimeter. Since the voxel resolution corresponding to subtask 1 is greater than the voxel resolution of the original voxel, each original voxel in the 5*5*2 cubic decimeter region can be further divided into 10*10*10 1 cubic centimeter target voxels through a neural network, so that the subspace region corresponding to subtask 1 can be divided into 50*50*20 1 cubic centimeter target voxels.
[0082] Assuming that the subspace area corresponding to a subtask 2 is a 20*20*10 cubic decimeter area in the target space area, and the voxel resolution corresponding to subtask 2 corresponds to a voxel size of 1 cubic meter, since the voxel resolution corresponding to subtask 2 is smaller than the voxel resolution of the original voxel, the neural network can be used to further integrate 10*10*10 adjacent original voxels in the 20*20*10 cubic decimeter area into a target voxel of 1 cubic meter, so that the subspace area corresponding to subtask 2 can be divided into 2*2*1 target voxels of 1 cubic meter.
[0083] Step 206 : for any target voxel into which the subspace region corresponding to any subtask is divided, feature information corresponding to the any target voxel is generated based on feature information of the original voxel corresponding to the any target voxel.
[0084] In some embodiments, when the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, for any target voxel into which the subspace area corresponding to the arbitrary subtask is divided, a neural network that has learned feature representation knowledge can generate feature information corresponding to the arbitrary target voxel based on the feature information of the original voxel corresponding to the arbitrary target voxel.
[0085] It is understandable that a neural network can be configured in an environmental perception device or an electronic device, and a feature extraction network in the neural network, such as a base network, can be used to extract features from the visual data of the target space area to obtain feature information of multiple original voxels into which the target space area is divided. Among them, the neural network can be trained first, and during the training process, the feature extraction network or other network layers of the neural network can learn feature representation knowledge, so that the feature information corresponding to a larger-sized voxel can be accurately obtained based on the feature information corresponding to the voxel, and the feature information corresponding to each small-sized voxel after the voxel is divided into multiple small-sized voxels. Then, in an embodiment of the present application, when the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, for any target voxel into which the subspace area corresponding to the arbitrary subtask is divided, the neural network that has learned the feature representation knowledge can be used to generate the feature information corresponding to the arbitrary target voxel based on the feature information of the original voxel corresponding to the arbitrary target voxel. For example, continuing the above example, the neural network that has learned the feature representation knowledge can be used to generate the feature information of each target voxel corresponding to the original voxel cut into the corresponding original voxel based on the feature information corresponding to each original voxel in the 5*5*2 cubic decimeter area corresponding to subtask 1. Among them, the training method of the neural network can be supervised training.
[0086] In some embodiments, when the voxel resolution corresponding to any subtask is smaller than the voxel resolution of the original voxel, for any target voxel into which the subspace area corresponding to the arbitrary subtask is divided, the feature information of the original voxel corresponding to the arbitrary target voxel can be fused to obtain the feature information corresponding to the arbitrary target voxel.
[0087] For example, continuing with the above example, the feature information corresponding to every 10*10*10 adjacent original voxels in the 20*20*10 cubic decimeter area corresponding to subtask 2 can be fused to obtain the feature information of the 1 cubic meter target voxel integrated from the 10*10*10 adjacent original voxels.
[0088] In some embodiments, when the voxel resolution corresponding to any subtask is equal to the voxel resolution of the original voxel, for any target voxel into which the subspace area corresponding to the arbitrary subtask is divided, the feature information of the original voxel corresponding to the arbitrary target voxel can be used as the feature information corresponding to the arbitrary target voxel.
[0089] It can be understood that when the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, for any target voxel into which the subspace area corresponding to the arbitrary subtask is divided, compared with directly performing feature extraction on the visual data of the target space area to obtain the feature information corresponding to the arbitrary target voxel, by first performing feature extraction on the visual data of the target space area to obtain the feature information of multiple original voxels into which the target space area is divided, and then generating the feature information corresponding to the arbitrary target voxel based on the feature information of the original voxel corresponding to the arbitrary target voxel, the computational amount of the feature information acquisition process can be reduced.
[0090] Step 207: Perform environmental perception on the target space region based on feature information corresponding to at least one target voxel into which each subspace region is divided.
[0091] The following example divides the target task performed in the target space area into n subtasks, and the voxel size of the original voxel is 1 cubic decimeter. Figure 3 , an example is given to illustrate the environment perception method provided in the embodiment of the present application. Wherein, n is an integer greater than or equal to 1.
[0092] refer to Figure 3 In the embodiment of the present application, feature extraction can be performed on the visual data of the target space area to obtain feature information corresponding to the multiple 1 cubic decimeter original voxels into which the target space area is divided. The process can be implemented, for example, by a feature extraction network in a trained neural network. The neural network can be configured in an environment perception device or an electronic device.
[0093] Furthermore, the target task to be performed in the target space region can be divided into n subtasks. For each subtask, the subspace region required to be sensed in the target space region to perform the subtask can be determined, thereby determining n subspace regions, and based on the accuracy requirements of each subtask, the voxel resolution corresponding to the subtask can be determined, and then each subspace region can be divided according to the corresponding voxel resolution to obtain the target voxel in each subspace region. Among them, assuming that the voxel resolution corresponding to the nth subtask is different from the voxel resolution of the original voxel, and the voxel size corresponding to the nth subtask is 1 cubic centimeter, then each original voxel in the nth subspace region can be divided to obtain multiple 1 cubic centimeter target voxels. Among them, this process can be implemented, for example, by the resolution adjustment network in the above-mentioned neural network.
[0094] Furthermore, for each target voxel in each subspace region, feature information corresponding to the target voxel can be generated based on feature information of the corresponding original voxel. This process can be implemented, for example, by a feature extraction network in the above neural network.
[0095] Furthermore, the target voxel can be classified based on the feature information corresponding to the target voxel to obtain the classification information of the target voxel. Thus, the classification information of each target voxel in the n subspace regions can be obtained. Then, based on the classification information of all target voxels in the space region, the environment perception of the target space region can be performed. Among them, the process of classifying the target voxel can be realized, for example, by the classification network in the above-mentioned neural network.
[0096] Among them, the training method of the neural network can be supervised training.
[0097] The environmental perception method provided by the embodiment of the present application realizes the determination of suitable voxel resolutions for different subtasks according to the accuracy requirements of different subtasks of the target task executed in the target space area, and then based on the suitable voxel resolution, the subspace area corresponding to the different subtasks is voxel-divided and environmental perception is realized. For example, for subtasks that require high-precision perception, high voxel resolution is used to improve the perception accuracy, and for subtasks that do not require high-precision perception, low voxel resolution is used to reduce the amount of data calculation, so that the accuracy and amount of data calculation of the tasks that can be completed by the intelligent device can be optimally balanced. In addition, in the case where the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, for any target voxel divided into the subspace area corresponding to the arbitrary subtask, feature information of multiple original voxels divided into the target space area is obtained by first extracting features from the visual data of the target space area, and then generating feature information corresponding to the arbitrary target voxel based on the feature information of the original voxel corresponding to the arbitrary target voxel, which can reduce the amount of calculation in the feature information acquisition process.
[0098] In order to implement the above embodiment, the embodiment of the present application also proposes an environment sensing device.
[0099] Figure 4 A schematic diagram of the structure of an environment sensing device provided in an embodiment of the present application.
[0100] like Figure 4 As shown, the environment perception device 400 includes:
[0101] A first processing module 410 is used to divide the target task to be performed in the target space area into at least one subtask;
[0102] A determination module 420 is used to determine, for any subtask of the at least one subtask, a subspace region in the target space region that needs to be sensed to perform any subtask, and determine a voxel resolution corresponding to any subtask based on the accuracy requirement of any subtask;
[0103] A second processing module 430 is used to divide the subspace region corresponding to any subtask into at least one target voxel according to the corresponding voxel resolution;
[0104] The third processing module 440 is used to perform environmental perception on the target space region based on feature information corresponding to at least one target voxel into which each subspace region is divided.
[0105] It should be noted that the environmental perception device 400 provided in the embodiment of the present application can execute the environmental perception method provided in the embodiment of the present application. Among them, the environmental perception device can be an electronic device, or it can be configured in an electronic device so that the electronic device can perceive the environment, and the present application does not limit this. Among them, the electronic device can be any intelligent device with environmental perception capabilities, such as a general robot, a robot with specific functions, an autonomous driving vehicle, a controller of an autonomous driving vehicle, an unmanned aerial vehicle, a smart home device, a wearable device, etc., and the present application does not limit this.
[0106] In some embodiments, before the subspace region corresponding to any subtask is divided into at least one target voxel according to the corresponding voxel resolution, the target space region has been divided into a plurality of original voxels; accordingly, the second processing module 430 is used to:
[0107] When the voxel resolution corresponding to any subtask is the same as the voxel resolution of the original voxel, the original voxel in the subspace region corresponding to the any subtask is determined as the target voxel in the subspace region;
[0108] When the voxel resolution corresponding to any subtask is different from the voxel resolution of the original voxel, the original voxel in the subspace region corresponding to the any subtask is reconstructed to obtain the target voxel in the subspace region.
[0109] In some embodiments, the second processing module 430 is configured to:
[0110] In the case where the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, for any original voxel in the subspace region corresponding to any subtask, the original voxel is segmented based on the voxel resolution corresponding to any subtask through a neural network to obtain a plurality of target voxels corresponding to the original voxel;
[0111] When the voxel resolution corresponding to any subtask is smaller than the voxel resolution of the original voxel, a plurality of adjacent original voxels in the subspace region corresponding to any subtask are integrated based on the voxel resolution corresponding to the any subtask through a neural network to obtain target voxels corresponding to the plurality of adjacent original voxels.
[0112] In some embodiments, the environment sensing device 400 further includes:
[0113] A feature extraction module is used to extract features from the visual data of the target space area to obtain feature information corresponding to a plurality of original voxels into which the target space area is divided;
[0114] The generation module is used to generate feature information corresponding to any target voxel based on the feature information of the original voxel corresponding to the any target voxel, for any target voxel divided into the subspace area corresponding to any subtask.
[0115] In some embodiments, a generating module is used to:
[0116] When the voxel resolution corresponding to any subtask is greater than the voxel resolution of the original voxel, feature information corresponding to any target voxel is generated based on the feature information of the original voxel corresponding to any target voxel through a neural network that has learned feature representation knowledge;
[0117] When the voxel resolution corresponding to any subtask is smaller than the voxel resolution of the original voxel, the feature information of the original voxel corresponding to any target voxel is fused to obtain the feature information corresponding to any target voxel;
[0118] When the voxel resolution corresponding to any subtask is equal to the voxel resolution of the original voxel, the feature information of the original voxel corresponding to any target voxel is used as the feature information corresponding to the any target voxel.
[0119] In some embodiments, the third processing module 440 is configured to:
[0120] For any target voxel divided into any subspace region, based on the feature information corresponding to the any target voxel, the any target voxel is classified to obtain classification information of the any target voxel, wherein the classification information includes: whether the any target voxel is occupied, and if the any target voxel is occupied, the type of the object occupying the any target voxel belongs to;
[0121] Based on the classification information of each target voxel divided into each subspace area, environmental perception is performed on the target space area.
[0122] It should be noted that the aforementioned explanation of the embodiment of the environment perception method is also applicable to the environment perception device of this embodiment and will not be repeated here.
[0123] Figure 5A schematic diagram of the structure of an electronic device provided for an embodiment of the present application. Among them, the electronic device 500 in this embodiment is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.
[0124] like Figure 5 As shown, the electronic device 500 includes:
[0125] The memory 501 and the processor 502, a bus 503 connecting different components (including the memory 501 and the processor 502), the memory 501 stores a computer program, and when the processor 502 executes the program, the environment perception method of the embodiment of the present application is implemented.
[0126] Bus 503 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnection (PCI) bus.
[0127] The electronic device 500 typically includes a variety of electronic device readable media. These media can be any available media that can be accessed by the electronic device 500, including volatile and non-volatile media, removable and non-removable media.
[0128] The memory 501 may also include a computer system readable medium in the form of a volatile memory, such as a random access memory (RAM) 504 and / or a cache memory 505. The electronic device 500 may further include other removable / non-removable, volatile / non-volatile computer system storage media. For example only, the storage system 506 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, usually called a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the bus 503 via one or more data medium interfaces. The memory 501 may include at least one program product having a set (e.g., at least one) of program modules, which are configured to perform the functions of the various embodiments of the present application.
[0129] A program / utility 508 having a set (at least one) of program modules 507 may be stored, for example, in the memory 501, such program modules 507 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 507 generally perform the functions and / or methods of the embodiments described herein.
[0130] The electronic device 500 may also communicate with one or more external devices 509 (e.g., keyboard, pointing device, display 511, etc.), one or more devices that enable a user to interact with the electronic device 500, and / or any device that enables the electronic device 500 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed through an input / output (I / O) interface 512. Furthermore, the electronic device 500 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 513. Figure 5As shown, the network adapter 513 communicates with other modules of the electronic device 500 via the bus 503. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0131] The processor 502 executes various functional applications and data processing by running the programs stored in the memory 501 .
[0132] It should be noted that the implementation process and technical principles of the electronic device of this embodiment refer to the aforementioned explanation of the environmental perception method of the embodiment of the present application, and will not be repeated here.
[0133] In order to implement the above embodiments, the present application also proposes 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 methods provided by the above embodiments.
[0134] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.
[0135] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.
[0136] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.
[0137] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.
[0138] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0139] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0140] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.
[0141] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.
[0142] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0143] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0144] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0145] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A method for environmental perception, characterized in that: The following steps are involved: Dividing a target task to be performed in a target space region into at least one subtask, wherein the target task is any task that can be performed by a smart device with environmental perception capability; For any subtask among the at least one subtask, determining a subspace region in the target space region that needs to be sensed to perform the arbitrary subtask, and determining a voxel resolution corresponding to the arbitrary subtask based on the accuracy requirement of the arbitrary subtask, wherein a corresponding relationship between the degree of the accuracy requirement of the subtask and the voxel resolution is preset; Dividing the subspace region corresponding to the arbitrary subtask into at least one target voxel according to the corresponding voxel resolution; Based on the feature information corresponding to the at least one target voxel into which each of the subspace regions is divided, environmental perception is performed on the target space region.
2. The method according to claim 1, characterized in that Before dividing the subspace region corresponding to the arbitrary subtask into at least one target voxel according to the corresponding voxel resolution, the target space region has been divided into a plurality of original voxels; dividing the subspace region corresponding to the arbitrary subtask into at least one target voxel according to the corresponding voxel resolution includes: In a case where the voxel resolution corresponding to the arbitrary subtask is the same as the voxel resolution of the original voxel, determining the original voxel in the subspace region corresponding to the arbitrary subtask as the target voxel in the subspace region; In the case that the voxel resolution corresponding to the arbitrary subtask is different from the voxel resolution of the original voxel, the original voxel in the subspace region corresponding to the arbitrary subtask is reconstructed to obtain the target voxel in the subspace region.
3. The method according to claim 2, characterized in that The reconstructing the original voxels in the subspace region corresponding to the arbitrary subtask to obtain the target voxels in the subspace region includes: In the case where the voxel resolution corresponding to the arbitrary subtask is greater than the voxel resolution of the original voxel, for any original voxel in the subspace region corresponding to the arbitrary subtask, the arbitrary original voxel is segmented based on the voxel resolution corresponding to the arbitrary subtask through a neural network to obtain a plurality of target voxels corresponding to the arbitrary original voxel; When the voxel resolution corresponding to the arbitrary subtask is smaller than the voxel resolution of the original voxel, the neural network is used to integrate multiple adjacent original voxels in the subspace area corresponding to the arbitrary subtask based on the voxel resolution corresponding to the arbitrary subtask to obtain target voxels corresponding to the multiple adjacent original voxels.
4. The method according to claim 3, characterized in that Before performing environmental perception on the target space region based on the feature information corresponding to the at least one target voxel divided into each of the subspace regions, the method further includes: Performing feature extraction on the visual data of the target space region to obtain feature information corresponding to the plurality of original voxels into which the target space region is divided; For any target voxel into which the subspace region corresponding to the arbitrary subtask is divided, feature information corresponding to the arbitrary target voxel is generated based on feature information of the original voxel corresponding to the arbitrary target voxel.
5. The method according to claim 4, characterized in that The generating the feature information corresponding to the arbitrary target voxel based on the feature information of the original voxel corresponding to the arbitrary target voxel comprises: In a case where the voxel resolution corresponding to the arbitrary subtask is greater than the voxel resolution of the original voxel, generating feature information corresponding to the arbitrary target voxel based on feature information of the original voxel corresponding to the arbitrary target voxel by means of the neural network that has learned feature representation knowledge; In the case where the voxel resolution corresponding to the arbitrary subtask is smaller than the voxel resolution of the original voxel, the feature information of the original voxel corresponding to the arbitrary target voxel is fused to obtain the feature information corresponding to the arbitrary target voxel; When the voxel resolution corresponding to the arbitrary subtask is equal to the voxel resolution of the original voxel, the feature information of the original voxel corresponding to the arbitrary target voxel is used as the feature information corresponding to the arbitrary target voxel.
6. The method according to any one of claims 1 to 5, characterized in that The performing environmental perception on the target space region based on the feature information corresponding to the at least one target voxel into which each of the subspace regions is divided includes: For any target voxel divided into any subspace region, classify the arbitrary target voxel based on feature information corresponding to the arbitrary target voxel to obtain classification information of the arbitrary target voxel, wherein the classification information includes: whether the arbitrary target voxel is occupied, and if the arbitrary target voxel is occupied, the type of the object occupying the arbitrary target voxel belongs to; Based on the classification information of each target voxel into which each subspace region is divided, environmental perception is performed on the target space region.
7. An environment sensing device, characterized in that: include: A first processing module, configured to divide a target task to be performed in a target space region into at least one subtask, wherein the target task is any task that can be performed by a smart device having an environment perception capability; a determination module, configured to determine, for any subtask among the at least one subtask, a subspace region in the target space region that needs to be sensed for executing the arbitrary subtask, and determine a voxel resolution corresponding to the arbitrary subtask based on the accuracy requirement of the arbitrary subtask, wherein a corresponding relationship between a high degree of the accuracy requirement of the subtask and the voxel resolution is preset; A second processing module, used for dividing the subspace region corresponding to the arbitrary subtask into at least one target voxel according to the corresponding voxel resolution; The third processing module is used to perform environmental perception on the target space area based on feature information corresponding to the at least one target voxel into which each of the subspace areas is divided.
8. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.
9. 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 6 when executed by a processor.
10. 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 6 when being executed by a processor.
Citation Information
Patent Citations
Task-oriented space-time optimal path evaluation method and system
CN114254808A
Image rendering method and device
CN117218273A
Local feature fusion-based end-to-end real tunnel point cloud three-dimensional reconstruction method
CN117635867A
Environment sensing method and device and unmanned vehicle
CN118279873A
Systems and methods for distribution-aware goal prediction for modular autonomous vehicle control
US20240109557A1
