Waste sorting system for controlling robot arm using spatial information of target objects

Abstract

A waste sorting system for controlling a robot arm using spatial information of target objects according to the present invention comprises: a target object determination unit that analyses images of waste moving on a conveyor belt and determines pieces of the waste as being target objects to be sorted or non-target objects not to be sorted; a robot arm that sorts the target objects on the conveyor belt; a path planning unit that generates spatial information about the space occupied by the target objects on the conveyor belt and the space occupied by the non-target objects around the target objects, and uses the spatial information to derive a movement path for the robot arm; and an operation control unit that controls an avoidance operation and a picking operation of the robot arm on the basis of the movement path.

Description

A waste sorting system that controls a robot arm using spatial information of a target object

[0001] The present invention relates to a waste sorting system that controls a robot arm using spatial information of a target object, and more specifically, to a waste sorting system that controls a robot arm using spatial information of the space occupied by a target object on a conveyor belt and the space occupied by non-target objects surrounding the target object.

[0002] This patent application claims priority to Korean Patent Application No. 10-2024-0140279 filed with the Korean Intellectual Property Office on October 15, 2024, and the disclosures of said patent application are incorporated herein by reference.

[0003] A large amount of waste is generated daily from households, factories, restaurants, etc., and among general waste, recyclable waste, and food waste, many people separate and dispose of food waste.

[0004] However, recyclable waste is still being disposed of mixed with general waste, and manual labor is required to separate it.

[0005] Furthermore, companies using PET or glass bottles are currently importing recyclable waste from abroad because the waste is not properly separated and ends up being discarded as general waste.

[0006] In Korea as well, various inventions are being made to solve these problems, and in particular, methods, devices, and systems for separating PET bottles are being filed and registered as patents.

[0007] For example, Korean Published Patent No. 10-2023-0084017, "Recyclable Waste Classification Device, Method and Program Using Artificial Intelligence Model," relates to a recyclable waste classification device, wherein an image of waste on a conveyor is captured through at least one camera in a pre-set sorting area, and the type of each waste on the conveyor is determined by analyzing the captured image based on an artificial intelligence model, and the type of waste for each waste on the conveyor is determined by analyzing the captured image using the artificial intelligence model.

[0008] Various types of robotic arms are used as a means to sort waste moving on a conveyor belt. Robotic arms utilize pressure-based suction grippers or finger grippers as end-effectors to pick up waste.

[0009] However, a problem arises where the robot arm collides with the waste because other waste surrounds the waste to be sorted, or the waste picked up by the end effector collides with other waste, resulting in failure to pick up.

[0010] This problem can occur by determining the robot arm's movement path without considering the target objects to be classified and non-target objects around the target objects.

[0011] To solve the aforementioned problems, the present invention aims to provide a waste classification system that generates spatial information of a target object to be classified and its surroundings using an image of the waste, and controls a robot arm using the spatial information of the target object.

[0012] A waste sorting system for controlling a robot arm using spatial information of a target object according to an embodiment of the present invention may include: a target object determination unit that analyzes a waste image of waste moving on a conveyor belt and determines the waste as a target object to be classified and a non-target object not to be classified; a robot arm that classifies the target object on the conveyor belt; a path planning unit that generates spatial information of the space occupied by the target object on the conveyor belt and the space occupied by the non-target object around the target object, and derives a movement path of the robot arm using the spatial information; and a motion control unit that controls avoidance and gathering movements of the robot arm based on the movement path.

[0013] The above avoidance action may be the robot arm waiting at one of a plurality of preset waiting points, and the above grasping action may be the robot arm grasping the target object and moving along the movement path.

[0014] The above robot arm may include a gripper that uses air suction force to adsorb and move the target object.

[0015] The path planning unit may include a spatial information generation unit that generates first spatial information including at least one of the position coordinates, width, height, and point cloud of the target object and second spatial information including at least one of the position coordinates, width, height, and point cloud of the non-target object using the waste image; and a path generation unit that generates a first movement path in which the robot arm moves to pick up the target object and a second movement path in which the robot arm moves after picking up the target object.

[0016] The path planning unit may further include a collision prediction unit that calculates a collision probability in which the robot arm may collide with the target object or the non-target object based on the first spatial information, the second spatial information, the first movement path, and the second movement path.

[0017] The path generation unit may include an artificial intelligence model that generates a movement path of the robot arm through reinforcement learning.

[0018] The collision prediction unit may include an artificial intelligence model that receives the first spatial information, the second spatial information, the first movement path, and the second movement path, and outputs the collision probability.

[0019] The path generation unit may modify at least one of the first movement path and the second movement path if the collision probability obtained from the collision prediction unit is greater than or equal to a preset threshold value.

[0020] The path generation unit may modify at least one of the first movement path and the second movement path containing a curve to include only a straight line when the space occupied by the non-target object surrounds the space occupied by the target object and the collision probability is greater than or equal to the threshold value.

[0021] The above motion control unit can control the robot arm to avoid the target object or the non-target object when the collision probability of the robot arm moving along the above movement path colliding with the target object or the non-target object is greater than or equal to a preset threshold value.

[0022] If the above motion control unit starts moving before the robot arm is controlled to avoid the target object or the non-target object, the robot arm can be controlled to move back to the nearest waiting point.

[0023] It may further include an image collection unit for collecting waste images of waste moving on the conveyor belt.

[0024] The above image acquisition unit may include a vision sensor capable of acquiring 2D images and 3D images.

[0025] According to the present invention as described above, the following effects are achieved.

[0026] The present invention can derive a movement path that allows a robot arm to avoid collision with a target object.

[0027] The present invention can derive a movement path that allows a robot arm to avoid collisions with non-target objects around a target object.

[0028] The present invention can reduce gripping failures that may occur when a robot arm collides with a target object or a non-target object, and prevent damage to the robot arm.

[0029] In addition, other features and advantages of the present invention may be newly identified through the embodiments of the present invention.

[0030] FIG. 1 is a schematic diagram of a waste sorting device to which a waste sorting system that controls a robot arm using spatial information of a target object according to one embodiment of the present invention is applied.

[0031] FIG. 2 is a configuration diagram of a waste sorting system that controls a robot arm using spatial information of a target object according to one embodiment of the present invention.

[0032] FIG. 3 is a diagram illustrating the process of a target object identification unit analyzing a waste image according to an embodiment of the present invention.

[0033] FIG. 4 is a schematic diagram illustrating a robot arm according to one embodiment of the present invention.

[0034] FIG. 5 is a configuration diagram of a path planning unit according to one embodiment of the present invention.

[0035] FIG. 6 is a diagram illustrating the process of deriving a movement path of a robot arm according to an embodiment of the present invention.

[0036] FIG. 7 is a diagram illustrating the process of modifying the movement path of a robot arm according to an embodiment of the present invention.

[0037] It should be noted that in assigning reference numbers to the components of each drawing in this specification, identical components are assigned the same number whenever possible, even if they are shown in different drawings. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. As used in this specification, "comprises" and / or "comprising" do not exclude the presence or addition of one or more other components in addition to the components mentioned.

[0038] The above objects, other objects, features, and advantages of the present invention will be easily understood through the following preferred embodiments associated with the accompanying drawings. However, the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided to ensure that the disclosed content is thorough and complete and to ensure that the spirit of the invention is sufficiently conveyed to a person skilled in the art.

[0039]

[0040] Hereinafter, preferred embodiments of the present invention designed to solve the above problems will be described in detail with reference to the attached drawings. A waste classification system that controls a robot arm using spatial information of a target object according to one embodiment of the present invention will be described as abbreviated as "waste classification system."

[0041] FIG. 1 is a schematic diagram of a waste sorting device to which a waste sorting system that controls a robot arm using spatial information of a target object according to one embodiment of the present invention is applied.

[0042] Referring to FIG. 1, a waste sorting device according to one embodiment of the present invention can automatically sort waste moving on a conveyor belt (2000). A waste sorting system (1000) applied to the waste sorting device can collect waste images and analyze waste images to determine target objects (10) to be sorted and non-target objects (20) not to be sorted. The waste sorting system (1000) can sort waste by picking up and moving target objects (10) through a robot arm (300).

[0043] When waste is discharged onto a conveyor belt, it moves in the direction (X-axis) in which the conveyor belt moves. The waste may move onto the conveyor belt without being aligned or positioned. The target object (10) and the non-target object (20) move while mixed randomly, and the non-target object (20) may surround the target object (10), or the non-target object (20) and the target object (10) may overlap.

[0044] In such a case, the robot arm (300) may collide with a non-target object (20) while performing movement and picking operations to pick up the target object (10). Alternatively, even if the robot arm (300) succeeds in picking up the target object (10), it may lose the target object (10) by colliding with a non-target object (20).

[0045] In one embodiment of the present invention, another waste classification system (1000) can efficiently perform waste classification by controlling a robot arm while considering not only the target object (10) but also the non-target object (20).

[0046] Meanwhile, if waste exceeding the driving radius of the robot arm (300) is fed into the waste sorting device, there is a risk that the robot arm (300) and the waste will collide and the robot arm (300) will be damaged.

[0047] In such a case, a waste sorting system (1000) according to one embodiment of the present invention can predict the probability of collision with waste in advance and control the robot arm (300) to avoid the waste.

[0048] FIG. 2 is a configuration diagram of a waste classification system that controls a robot arm using spatial information of a target object according to an embodiment of the present invention, and FIG. 3 is a diagram explaining the process of a target object identification unit analyzing a waste image according to an embodiment of the present invention.

[0049] FIG. 4 is a schematic diagram for explaining a robot arm according to an embodiment of the present invention, and FIG. 5 is a configuration diagram of a path planning unit according to an embodiment of the present invention.

[0050] Referring to FIG. 2, a waste classification system (1000) according to one embodiment of the present invention may include an image collection unit (100), a target object identification unit (200), a robot arm (300), a path planning unit (400), and a motion control unit (500).

[0051] The image collection unit (100) can collect images of waste moving on a conveyor belt. The image collection unit (100) can collect images of waste and then preprocess the image data.

[0052] The waste images collected by the image collection unit (100) include 2D images and 3D images, and may include all images according to the type and function of the sensor.

[0053] The image acquisition unit (100) may include a vision sensor capable of acquiring 2D images and 3D images. For example, the image acquisition unit (100) may include a vision camera capable of capturing 2D images, and may include a Time of Flight (TOF) camera or a LiDAR camera capable of capturing 3D images. The type and function of the vision sensor are not limited and can be appropriately selected.

[0054] The image collection unit (100) may be installed at various locations according to the embodiment. For example, the image collection unit (100) may include a first image collection unit (110) installed at the beginning of the waste moving on the conveyor belt (2000) and a second image collection unit (120) installed by being coupled to the robot arm (2000).

[0055] The first image collection unit (110) can collect images of waste moving on a conveyor belt, and the second image collection unit (120) can collect images of target objects that can be classified within the activity radius of the robot arm (300) that classifies target objects and images of non-target objects that are not subject to classification within the activity radius of the robot arm (300).

[0056] Referring to FIG. 3, the target object identification unit (200) can analyze a waste image of waste moving on a conveyor belt. The target object identification unit (200) acquires a waste image collected by the image collection unit (100) and can determine whether the acquired waste image is an image of a target object to be classified using an artificial intelligence model.

[0057] The artificial intelligence model used by the target object discrimination unit (200) may be a deep learning or neural network. The neural network models may include various types of models such as Convolutional Neural Network (CNN), Region with Convolutional Neural Network (R-CNN), Region Proposal Network (RPN), Recurrent Neural Network (RNN), Stacking-based Deep Neural Network (S-DNN), State-Space Dynamic Neural Network (S-SDNN), Deconvolution Network, Deep Belief Network (DBN), Restricted Boltzmann Machine (RBM), Fully Convolutional Network, Long Short-Term Memory Network (LSTM), Classification Network, etc., such as GoogleNet, AlexNet, VGG Network, etc., but are not limited thereto.

[0058] The artificial intelligence model used by the target object identification unit (200) can learn waste images labeled according to a pre-set waste classification standard table.

[0059] Referring to FIG. 4, the robot arm (300) can classify target objects on a conveyor belt. The robot arm (300) can pick up and move target objects.

[0060] According to one embodiment of the present invention, the robot arm (300) may include a gripper (320) that adsorbs and moves a target object using air suction force. For example, the gripper may be a suction gripper.

[0061] A gripper (320) can be mounted on the end of a robot arm. The robot arm (300) may include an exhaust nozzle (2100) that transmits air suction force to the gripper (320).

[0062] The gripper (320) may be equipped with a spring to allow for positional adjustment to effectively adsorb the target object, and may be equipped with a rubber suction cup.

[0063] According to another embodiment of the present invention, the gripper (320) of the robot arm (300) can be formed in various materials and shapes in addition to the suction type, taking into account the type of waste and the classification environment, and is not limited thereto.

[0064] Referring to FIG. 5, the path planning unit (400) may include a spatial information generation unit (410), a path generation unit (420), and a collision prediction unit (430).

[0065] The path planning unit (400) can generate spatial information of the space occupied by the target object on the conveyor belt and the space occupied by non-target objects around the target object, and use the spatial information to derive the movement path of the robot arm (300).

[0066] Spatial information may include location coordinates, width, and height of the space occupied by target objects and non-target objects, and may be generated from 2D waste images. Spatial information may include a point cloud representing the space occupied by target objects and non-target objects, and may be generated from 3D waste images.

[0067] Spatial information may include additional data other than location coordinates, area, height, and point clouds, and the types of data that can be included are not limited.

[0068] Specifically, the spatial information generation unit (410) can generate first spatial information including at least one of the location coordinates, width, height, and point cloud of a target object using a waste image. For example, the space occupied by the target object may be defined as the bounding box of the target object, and the spatial information generation unit (410) can generate the center location coordinates of the bounding box, the width, and the height of the bounding box as the first spatial information.

[0069] The spatial information generation unit (410) can generate second spatial information including at least one of the location coordinates, width, height, and point cloud of a non-target object using a waste image. For example, the space occupied by the non-target object can be defined as the bounding box of the target object, and the spatial information generation unit (410) can generate the center location coordinates of the bounding box, the width, and the height of the bounding box as second spatial information.

[0070] The first spatial information and the second spatial information may indicate the state in which waste is placed on a conveyor belt. For example, the first spatial information and the second spatial information may indicate whether the target object and the non-target object overlap, or whether the non-target object surrounds the target object. Additionally, the height, shape, size, etc., of the non-target object surrounding the target object may be indicated.

[0071] The first spatial information and the second spatial information generated by the spatial information generation unit (410) can be transmitted to the path generation unit (420), and the path generation unit (420) can generate a movement path of the robot arm (300) based on the first spatial information and the second spatial information.

[0072] Specifically, the path generation unit (420) can generate a first movement path for the robot arm (300) to move to pick up a target object.

[0073] For example, the path generation unit (420) can identify a state in which a non-target object, which is taller than the target object, surrounds the target object based on the first spatial information and the second spatial information. In such a case, when the robot arm (300) moves toward the target object to pick it up, the robot arm (300) may collide with the non-target object surrounding the target object. To prevent collision, the path generation unit (420) can generate a path that moves toward the target object while positioning the end effector of the robot arm (300) at a height higher than that of the non-target object. By moving along this path, the robot arm (300) can avoid collision with the non-target object and approach the target object.

[0074] As another example, the path generation unit (420) can identify a state in which a non-target object, which is taller than the target object, surrounds only a part of the target object according to the first spatial information and the second spatial information. In such a case, to prevent collision, the path generation unit (420) can generate a path that moves toward the target object while positioning the end effector of the robot arm (300) at a height higher than that of the non-target object, or generate a path that moves in a direction in which the non-target object does not surround the target object.

[0075] The path generation unit (420) can generate a second movement path for the robot arm (300) to move after picking up a target object.

[0076] For example, the path generation unit (420) can identify a state in which a non-target object, which is taller than the target object, surrounds the target object according to the first spatial information and the second spatial information. In such a case, when the robot arm (300) moves after picking up the target object, the robot arm (300) may collide with the non-target object surrounding the target object, or the target object attached to the robot arm (300) may collide with the non-target object. To prevent collisions, the path generation unit (420) can generate a path in which the target object attached to the robot arm (300) moves while positioning it higher than the height of the non-target object. By moving along this path, the robot arm (300) can avoid collisions with the non-target object and perform classification of the target object.

[0077] As another example, the path generation unit (420) can identify a state in which a non-target object, which is taller than the target object, surrounds only a part of the target object according to the first spatial information and the second spatial information. In such a case, to prevent collision, the path generation unit (420) can generate a path in which the target object attached to the robot arm (300) moves while positioned higher than the height of the non-target object, or generate a path in which the non-target object moves in a direction in which it does not surround the target object.

[0078] The path generation unit (420) may include an artificial intelligence model that generates a movement path of the robot arm (300). The artificial intelligence model used by the path generation unit (420) may be a deep learning or neural network. The neural network models may include various types of models such as a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), a Deep Neural Network (DNN), a Fully Convolutional Network, a Long Short-Term Memory (LSTM) Network, and Reinforcement Learning Models, but are not limited thereto.

[0079] The collision prediction unit (430) can calculate the collision probability that the robot arm (300) may collide with a target object or a non-target object based on the first spatial information, the second spatial information, the first movement path, and the second movement path.

[0080] More specifically, the collision prediction unit (430) may include an artificial intelligence model that receives first spatial information, second spatial information, first movement path and second movement path and outputs a collision probability, and can calculate the collision probability using the same.

[0081] The collision prediction unit (430) may include an artificial intelligence model that generates a movement path of the robot arm (300). The artificial intelligence model used by the collision prediction unit (430) may be a deep learning or neural network.

[0082] For example, the artificial intelligence models used by the collision prediction unit (430) may include various types of models such as Convolutional Neural Networks (CNN), Region with Convolutional Neural Networks (R-CNN), Region Proposal Network (RPN), Recurrent Neural Network (RNN), Fully Convolutional Network, Long Short-Term Memory Network (LSTM), and Reinforcement Learning Models, such as GoogleNet, AlexNet, and VGG Network, but are not limited thereto.

[0083] A motion control unit (500) according to one embodiment of the present invention can control the avoidance motion and the grasping motion of the robot arm based on the movement path generated by the path planning unit (400).

[0084] The avoidance action may be the robot arm (300) waiting at one of a plurality of preset waiting points. The picking action may be the robot arm (300) picking up a target object and moving along a movement path.

[0085] FIG. 6 is a diagram illustrating the process of deriving a movement path of a robot arm according to one embodiment of the present invention, and FIG. 7 is a diagram illustrating the process of modifying a movement path of a robot arm according to one embodiment of the present invention.

[0086] Referring to Fig. 6, the process of deriving and modifying the movement path of the robot arm is explained in more detail.

[0087] An image collection unit (100) according to one embodiment of the present invention can collect an image of waste on a conveyor belt (S10). A target object identification unit (200) can analyze the waste image and identify the waste on the conveyor belt as a target object and a non-target object (S20).

[0088] The spatial information generation unit (410) can generate first spatial information of a target object based on a waste image (S30). The spatial information generation unit (420) can generate second spatial information of a non-target object based on a waste image (S40).

[0089] The path generation unit (420) can generate a first movement path, which is a movement path in which the robot arm (300) moves to pick up a target object, based on the first spatial information and the second spatial information (S50). The path generation unit (420) can generate a second movement path, which is a movement path in which the robot arm (300) moves after picking up the target object (S60).

[0090] The collision prediction unit (430) can calculate the probability of collision between the robot arm (300) and the target object, and between the robot arm (300) and the non-target object based on the first spatial information, the second spatial information, the first movement path, and the second movement path (S70).

[0091] The path generation unit (420) can obtain a collision probability from the collision prediction unit and compare the collision probability with a preset threshold value (S80). For example, the threshold value may be 95%.

[0092] The path generation unit (420) can modify at least one of the first path and the second path when the collision probability is greater than or equal to a threshold value (S90).

[0093] For example, when the robot arm (300) moves along a first movement path to pick up a target object, the probability of colliding with the target object or a non-target object is 25%, and when it moves along a second movement path after picking up the target object, the probability of colliding with a non-target object is 96%. In such a case, the path generation unit (420) can maintain the first movement path and modify only the second movement path.

[0094] Referring to FIG. 7, the path generation unit (420) can modify at least one of the first movement path and the second movement path containing curves to include only straight lines when the space occupied by the non-target object surrounds the space occupied by the target object and the collision probability is greater than or equal to the threshold value.

[0095] For example, a non-target object surrounds a target object, so the collision probability calculated by the collision prediction unit based on the first movement path and the second movement path may be greater than or equal to a threshold value.

[0096] Generally, the first and second movement paths may include curves for efficient movement. The movement of the robot arm (300) along a movement path that includes curves is called J movement (Joint movement). The joints included in the robot arm (300) can be configured to move the robot arm along the curves.

[0097] As shown in FIG. 7 (a), if the robot arm (300) moves along a path that includes a curve, the robot arm (300) may collide with non-target objects surrounding the target object and fail to pick up.

[0098] The path generation unit (420) can modify the movement path to include only straight lines to avoid collisions. The movement of the robot arm (300) moving along a movement path that includes only straight lines is called L movement (Linear movement).

[0099] As shown in Fig. 7(b), when the robot arm (300) moves along a path that includes only straight lines, the robot arm (300) can avoid collisions with non-target objects surrounding the target object.

[0100] The motion control unit (500) can control the picking motion of the robot arm (300) according to the modified first movement path and second movement path (S100).

[0101] Even if the path generation unit (420) modifies the movement path, the probability of collision may be greater than a threshold value. In such cases, it may be difficult to avoid a collision even if the movement of the robot arm (300) is controlled along the modified movement path. In situations where it is difficult to avoid a collision between the robot arm (300) and waste, the movement control unit (500) can control the movement of the robot arm (300) to avoid the target object or non-target object.

[0102] Specifically, the motion control unit (500) can control the robot arm (300) to avoid the target object or non-target object when the probability of the robot arm (300) moving along the path colliding with the target object or non-target object is greater than or equal to a preset threshold value.

[0103] For example, the target object may be a large waste object that exceeds the driving radius of the robot arm (300). In such a case, since it is difficult to avoid a collision between the robot arm (300) and the target object, the motion control unit (500) can control the robot arm (300) to avoid the target object.

[0104] The avoidance action may be that the robot arm (300) does not move to pick up the target object, or that the robot arm (300) waits at a designated waiting point until the target object passes by. The avoidance action may be configured in various ways depending on the type of waste sorting device, the conditions of the location where the waste sorting device is installed, etc., and is not limited.

[0105] The motion control unit (500) can control the robot arm (300) to move back to the nearest waiting point if it starts moving before being controlled to avoid a target object or a non-target object.

[0106] For example, if waste is introduced with a collision probability exceeding a threshold value, the motion control unit (500) must control the robot arm (300) to avoid the waste, but the robot arm (300) may start moving before the control command is transmitted. In such a case, if the robot arm (300) attempts to stop moving and return to its original position, there is a risk of collision with the waste.

[0107] Therefore, the motion control unit (500) can control the robot arm (300) to move back to the nearest waiting point, rather than controlling the robot arm (300) to stop moving and return to its original position to avoid a collision.

[0108] The nearest waiting point can be appropriately selected considering the installation status of the waste sorting device, the type of waste being fed in, and the movement speed of the conveyor belt, and its location is not limited. In addition, the number of waiting points can also be appropriately selected.

[0109]

[0110] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or components such as the described structures and devices are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0111] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

[0112] A waste sorting system according to an embodiment of the present invention analyzes images of waste on a conveyor belt to distinguish between target objects and non-target objects, and generates spatial information regarding the space occupied by not only target objects but also non-target objects. By utilizing this information, the system can derive an optimal movement path that allows a robot arm to avoid collisions. By precisely controlling the avoidance and picking movements of the robot arm according to the movement path, the picking failure rate can be reduced and damage to the robot arm prevented, thereby improving the efficiency and safety of the waste sorting operation.

Claims

1. A target object determination unit that analyzes an image of waste moving on a conveyor belt and determines the waste as a target object to be classified and a non-target object not to be classified; A robot arm that classifies the target object on the conveyor belt; A path planning unit that generates spatial information of the space occupied by the target object on the conveyor belt and the space occupied by non-target objects around the target object, and derives the movement path of the robot arm using the spatial information; and A waste sorting system that controls a robot arm using spatial information of a target object, comprising a motion control unit that controls avoidance and picking movements of the robot arm based on the above movement path.

2. In Paragraph 1, The above avoidance action is for the robot arm to wait at one of a plurality of preset waiting points, and A waste sorting system that controls a robot arm using spatial information of a target object, wherein the above-mentioned picking operation is the robot arm picking up the target object and moving along the above-mentioned movement path.

3. In Paragraph 1, A waste sorting system that controls the robot arm using spatial information of a target object, wherein the robot arm includes a gripper that adsorbs and moves the target object using air suction force.

4. In Paragraph 1, The above path planning department A spatial information generation unit that generates first spatial information including at least one of the location coordinates, width, height, and point cloud of the target object and second spatial information including at least one of the location coordinates, width, height, and point cloud of the non-target object using the waste image; and A waste sorting system that controls a robot arm using spatial information of a target object, comprising a path generation unit that generates a first movement path for the robot arm to move to pick up the target object and a second movement path for the robot arm to move after picking up the target object.

5. In Paragraph 4, The above path planning department A waste sorting system that controls a robot arm using spatial information of a target object, further comprising a collision prediction unit that calculates a collision probability in which the robot arm may collide with the target object or the non-target object based on the first spatial information, the second spatial information, the first movement path, and the second movement path.

6. In Paragraph 4, A waste classification system that controls a robot arm using spatial information of a target object, wherein the path generation unit includes an artificial intelligence model that generates a movement path of the robot arm through reinforcement learning.

7. In Paragraph 5, A waste sorting system that controls a robot arm using spatial information of a target object, comprising an artificial intelligence model that receives the first spatial information, the second spatial information, the first movement path, and the second movement path, and outputs the collision probability.

8. In Paragraph 5, A waste sorting system that controls a robot arm using spatial information of a target object, wherein the path generation unit modifies at least one of the first movement path and the second movement path when the collision probability obtained from the collision prediction unit is greater than or equal to a preset threshold value.

9. In Paragraph 8, A waste sorting system that controls a robot arm using spatial information of a target object, wherein the path generation unit modifies at least one of a first movement path and a second movement path containing a curve to include only a straight line when the space occupied by the non-target object surrounds the space occupied by the target object and the collision probability is greater than or equal to the threshold value.

10. In Paragraph 1, A waste sorting system that controls a robot arm using spatial information of a target object, wherein the above-described motion control unit controls the robot arm to avoid the target object or the non-target object when the collision probability of the robot arm moving along the above-described path colliding with the target object or the non-target object is greater than or equal to a preset threshold value.

11. In Paragraph 10, A waste sorting system that controls a robot arm using spatial information of a target object, wherein the motion control unit controls the robot arm to move back to the nearest waiting point if the robot arm starts moving before being controlled to avoid the target object or the non-target object.

12. In Paragraph 1, A waste sorting system that controls a robot arm using spatial information of a target object, further comprising an image collection unit that collects waste images of waste moving on the conveyor belt.

13. In Paragraph 12, A waste sorting system that controls a robot arm using spatial information of a target object, wherein the image acquisition unit includes a vision sensor capable of acquiring 2D images and 3D images.

Images