Intelligent household garbage identifying and grabbing device and control method

Abstract

The intelligent household garbage recognizing and grabbing device comprises a conveying belt, a 3D camera and a mechanical arm are arranged above the conveying belt, a metal suction cup and a pneumatic suction cup are installed at the bottom end of the mechanical arm, placing plates are evenly installed on the front face of the conveying belt, and storage boxes are slidably connected to the upper surfaces of the placing plates; side connecting plates are rotatably connected to the lower portions of the placing plates, buffer plates are evenly and rotatably connected to the upper surfaces of the side connecting plates, arc springs are arranged between the buffer plates and the side connecting plates, and the upper surfaces of the side connecting plates are attached to the lower surface of the storage box; side clamping plates are slidably connected to the upper surfaces of the placement plates, and the inner side surfaces of the side clamping plates are tightly attached to the outer surface of the storage box; fixing side plates are evenly arranged on the upper surfaces of the side connecting plates, the upper surfaces of the fixing side plates are attached to one side of the lower surface of the buffer plate, the 3D camera can recognize household garbage, a metal suction cup and a pneumatic suction cup of the mechanical arm can adapt to different types of garbage, and the grabbing adaptability and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of garbage disposal, and in particular to a domestic garbage intelligent identification and grabbing device and a control method. Background Art

[0002] Domestic waste is solid waste generated in daily life or from activities that provide services for daily life. Its management involves multiple aspects, including classification, treatment, environmental impact, and policy. Traditionally, domestic waste disposal involves burying it underground, rendering it harmless by covering it with an impermeable layer, draining leachate (wastewater produced by decaying waste), and collecting biogas (primarily methane, which can generate electricity). However, this method requires significant land, and long-term soil and groundwater contamination due to aging impermeable layers has led to a gradual upgrade to "sanitary landfill" or "ecological landfill." With growing environmental protection demands, the need for intelligent and efficient domestic waste disposal is becoming increasingly urgent.

[0003] In the prior art, the grabbing devices for domestic waste often lack targeted identification and adaptation structures, and the types of grabbing components are relatively single, which makes it difficult to adapt to domestic waste of different materials and forms, resulting in poor adaptability and low accuracy in grabbing different types of domestic waste. At the same time, the storage box for collecting the grabbed domestic waste lacks effective fixing means during the transportation process with the conveyor belt, and is prone to shaking or displacement due to the operation of the conveyor belt, and has poor stability. In addition, the buffer structure design during the placement and transportation of the storage box is unreasonable, which cannot effectively alleviate the impact of vibration, and the range of movement of the buffer component is lack of constraints, which can easily lead to unstable buffering effect and insufficient supporting performance, thereby affecting the efficiency and stability of the overall grabbing process, and it is difficult to ensure the stability and safety of the storage box during transportation. For this reason, an intelligent identification and grabbing device for domestic waste and a control method are proposed. Summary of the Invention

[0004] The present invention provides the following technical solution: a household garbage intelligent identification and grabbing device, comprising: A conveyor belt, a 3D camera and a robotic arm are installed above it, and the bottom of the robotic arm is respectively installed with a metal suction cup and a pneumatic suction cup. Placement plates are evenly installed on the front of the conveyor belt, and the upper surface of the placement plates is slidably connected to the storage box. The side connecting plate is rotatably connected to the lower part of the placement plate, and the upper surface of the side connecting plate is evenly rotatably connected to the buffer plate, and an arc spring is added between the buffer plate and the side connecting plate, and the upper surface of the side connecting plate is in contact with the lower surface of the storage box; The side clamping plate is slidably connected to the upper surface of the placement plate, and the inner side surface of the side clamping plate is tightly fitted with the outer surface of the storage box; The fixed side panels are evenly arranged on the upper surface of the side connecting panels, and the upper surface of the fixed side panels is in contact with one side of the lower surface of the buffer panel.

[0005] Start the conveyor belt, which drives the placement plate and the storage box on the placement plate to move. When domestic waste is transported into the recognition range of the 3D camera, the 3D camera identifies the domestic waste and determines the type and location information of the domestic waste. Based on the information recognized by the 3D camera, the robotic arm moves to the location of the domestic waste and selects the corresponding suction cup according to the type of domestic waste. If it is metal domestic waste, the robotic arm controls the metal suction cup to grab it; if it is other types of domestic waste, the robotic arm controls the pneumatic suction cup to grab it. After the robotic arm grabs the domestic waste, it moves to the top of the storage box and places the domestic waste into the storage box. When the storage box is placed on the placement plate, the lower surface of the storage box fits with the upper surface of the side connecting plate. As the storage box is placed, the buffer plate rotates under the pressure of the storage box, and the arc spring is compressed, which plays a buffering role for the storage box. At the same time, the upper surface of the fixed side plate fits with one side of the lower surface of the buffer plate, limiting the rotation range of the buffer plate. Slide the side clamping plate so that the inner side of the side clamping plate fits tightly against the outer surface of the storage box to fix the storage box and prevent the storage box from shaking during the movement of the conveyor belt.

[0006] Preferably, a top plate is installed transversely on the upper surface of the conveyor belt, and the 3D camera and the robotic arm are evenly arranged on the lower surface of the top plate.

[0007] A top plate is installed horizontally on the upper surface of the conveyor belt, and then the 3D camera and robotic arm are evenly arranged on the lower surface of the top plate; when the conveyor belt is running, the 3D camera and robotic arm are supported by the top plate and carry out corresponding operations above the conveyor belt.

[0008] Preferably, the upper surface of the placement plate is evenly provided with limiting side grooves, and the inner cavities of the limiting side grooves are all slidably connected to the limiting side blocks.

[0009] Limiting side grooves are evenly opened on the upper surface of the placement plate, and then the limiting side blocks are slidably connected to the inner cavity of the limiting side grooves so that the limiting side blocks can slide along the inner cavity of the limiting side grooves.

[0010] Preferably, an extrusion spring is provided between the inner wall of the limiting side groove and the surface of the limiting side block, and the lower surface of the side clamping plate is connected to the upper surface of the limiting side block.

[0011] An extrusion spring is installed between the inner wall of the limiting side groove and the surface of the limiting side block, and then the lower surface of the side clamp is connected to the upper surface of the limiting side block; when the limiting side block slides along the limiting side groove, the extrusion spring is deformed and generates elastic force, driving the limiting side block and the side clamp to move synchronously.

[0012] Preferably, the upper surface of the side connecting plate is evenly rotatably connected to a rotating rod, and the buffer plate is connected to the rotating rod, and the fixed side plate is arranged on the outer side of the rotating rod.

[0013] The connecting rod is evenly rotated on the upper surface of the side connecting plate to connect the buffer plate and the rotating rod, and then a fixed side plate is set on the outside of the rotating rod; when the buffer plate is subjected to external force, the rotating rod rotates accordingly, and the fixed side plate limits the rotation range of the rotating rod.

[0014] Preferably, the outer side of the placement plate is rotatably connected to a rotating shaft, and a drive motor is coaxially connected to the end of the rotating shaft, and the side connecting plate is connected to the rotating shaft.

[0015] The rotating shaft is rotated and connected on the outer side of the placement plate, the driving motor is coaxially connected to the end of the rotating shaft, and then the side connecting plate is connected to the rotating shaft; the driving motor is started, the driving motor drives the rotating shaft to rotate, and the rotating shaft then drives the side connecting plate to rotate synchronously.

[0016] A method for intelligently identifying and grasping household waste is provided, based on the above-mentioned intelligent device for identifying and grasping household waste, and comprises: S1. Multimodal data acquisition and preprocessing: The 3D camera collects three-dimensional point cloud data and surface texture information of the garbage on the conveyor belt in real time, and simultaneously obtains the material properties of the garbage through the material sensor to form a multimodal data set. The collected data is subjected to noise reduction processing, and edge features and geometric features are extracted to generate standardized input data. S2. Intelligent identification and classification decision-making: The HybridNet model analyzes preprocessed data, generates bounding boxes and classification results for the garbage, and identifies material types such as metal, plastic, and glass. The depth sensor is used to calculate the number of stacked layers of garbage, prioritizing surface objects as grab targets and matching suction cup types based on material properties. S3, dynamic path planning and crawling execution: The DQN deep reinforcement learning algorithm is used to optimize the robotic arm's grasping path, avoiding interference from adjacent objects and ensuring the suction cup accurately contacts the target trash. The robotic arm controls the selected suction cup to absorb the trash, and dynamically adjusts the suction cup's suction force based on the trash's weight to complete the grasping action. S4, automatic cabinet management and buffer protection: When the storage box is full of garbage, the driving motor drives the side connecting plate to rotate and separate from the placement plate through the rotating shaft, and the storage box descends along the inner wall of the placement plate; during the descent, the bottom surface of the storage box contacts the buffer plate, and the arc spring absorbs the impact force. At the same time, the side clamping plate releases the storage box under the action of the extrusion spring, completing the automatic replacement process.

[0017] A 3D camera captures the household waste transported on the conveyor belt in real time, collecting 3D point cloud data and surface texture information. Simultaneously, the device's material sensors acquire the waste's material properties. These 3D point cloud data, surface texture information, and material properties together form a multimodal dataset. This multimodal dataset undergoes noise reduction to remove interference, then extracts the waste's edge and geometric features. This processed feature information is integrated to generate standardized input data, providing the foundation for subsequent identification and classification.

[0018] The standardized input data generated in S1 is input into the HybridNet model, which analyzes and processes the data to generate the corresponding bounding boxes and classification results of the garbage, accurately identifying the material types of the garbage, such as metal, plastic, and glass. At the same time, the depth sensor in the device is used to calculate the number of stacked layers of garbage on the current conveyor belt, and preferentially mark the surface objects of the stack as targets to be grasped. In addition, according to the identified material properties of the garbage, the suction cup type in the device is matched to the material to prepare for subsequent grasping actions.

[0019] The DQN deep reinforcement learning algorithm is used to optimize the robotic arm's grasping path, ensuring that the robotic arm's motion trajectory can avoid adjacent objects on the conveyor belt to prevent interference. After determining the optimal path, the robotic arm is controlled to move along the planned path, driving the matching suction cup in S2 to accurately contact the marked target garbage. During the adsorption process, the suction force of the suction cup is dynamically adjusted according to the weight of the target garbage to ensure that the garbage is stably adsorbed and ultimately complete the grasping action.

[0020] When the garbage in the storage box accumulates to a full state, the device controls the drive motor to start, and the drive motor drives the side connecting plate to rotate through the rotating shaft, so that the side connecting plate is separated from the placement plate; at this time, the storage box descends along the inner wall of the placement plate; during the descent, the bottom surface of the storage box contacts the buffer plate, and the elastic deformation of the arc spring absorbs the impact force generated during the descent process, thereby achieving buffering protection; at the same time, the side clamping plate releases its clamping of the storage box under the resetting action of the extrusion spring, completing the automatic replacement process of the storage box.

[0021] Preferably, in the above S2, intelligent identification and classification decision, the HybridNet model generates a comprehensive identification result including the location, size and material category of the garbage space by fusing the three-dimensional point cloud and material information.

[0022] During the intelligent recognition and classification decision-making stage, the HybridNet model starts running. The model receives 3D point cloud information and material information, fuses the two types of information, and generates a comprehensive recognition result after processing. This result includes the spatial location, size, and material category of the garbage.

[0023] Preferably, in the S3, dynamic path planning and grasping execution, the DQN deep reinforcement learning algorithm generates optimal control instructions including joint angles, movement speeds and suction cup switching timing by simulating the motion trajectory of the robotic arm in different stacking scenarios.

[0024] During the dynamic path planning and grasping execution phase, the DQN deep reinforcement learning algorithm begins to work. The algorithm simulates the motion trajectory of the robot arm in different stacking scenarios and generates optimal control instructions based on the simulation results. The instructions include the robot arm's joint angles, movement speed, and suction cup switching timing.

[0025] Preferably, in the said S4, automatic management and buffer protection of the box, during the descent of the storage box, the system calculates the descent speed of the storage box in real time by monitoring the current changes of the drive motor, and dynamically adjusts the output torque of the drive motor, so that the storage box completes the descent process in a variable acceleration manner, realizing impact-free switching.

[0026] During the automatic management and buffer protection stage of the box, when the storage box enters the descent process, the system monitors the current changes of the drive motor in real time, calculates the descent speed of the storage box in real time based on the current changes, and then dynamically adjusts the output torque of the drive motor based on the calculated descent speed, so that the storage box completes the descent process with variable acceleration and realizes impact-free switching.

[0027] In summary, compared with the prior art, the present invention provides a household waste intelligent identification and grasping device and control method, which has the following beneficial effects: The 3D camera above the conveyor belt of the present invention can identify domestic waste, and the metal suction cup and pneumatic suction cup at the bottom of the robotic arm can adapt to different types of domestic waste, which helps to improve the adaptability and accuracy of grabbing different types of domestic waste; The storage box on the placement plate can be used to collect the domestic waste after being grabbed. The sliding side clamps can fix the storage box by closely fitting with the outer surface of the storage box, preventing the storage box from shaking or deflecting during transportation by the conveyor belt, and enhancing the stability of the storage box placement; On the side connecting plate that is rotatably connected at the lower part of the placement plate, the arc spring between the buffer plate and the side connecting plate can play a buffering role during the placement or transportation of the storage box, reducing the vibration of the storage box, and the fixed side plate is in contact with one side of the lower surface of the buffer plate to limit the rotation range of the buffer plate and ensure the stability of the buffering effect. The upper surface of the side connecting plate is in contact with the lower surface of the storage box, which can cooperate with the buffer structure to further enhance the support and buffering effect of the storage box. The synergistic effect of the overall structure helps to improve the efficiency and stability of domestic waste identification and grasping, while ensuring the stability and safety of the storage box during transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a structural schematic diagram of the present invention.

[0029] Figure 2 It is a schematic diagram of the cross-sectional structure of the top plate of the present invention.

[0030] Figure 3 It is a schematic diagram of the storage box and its connection structure of the present invention.

[0031] Figure 4 It is a schematic diagram of the placement plate and its connection structure of the present invention.

[0032] Figure 5 It is a schematic diagram of the side splint and its connection structure of the present invention.

[0033] Figure 6 It is a right side view of the side connecting plate of the present invention and a schematic diagram of its connection structure.

[0034] Figure 7 It is a schematic block diagram of the method of the present invention.

[0035] Description of reference numerals: 1. Conveyor belt; 2. Top plate; 3. 3D camera; 4. Robotic arm; 5. Placement plate; 6. Storage box; 7. Limiting side groove; 8. Limiting side block; 9. Extrusion spring; 10. Side clamping plate; 11. Rotating shaft; 12. Drive motor; 13. Side connecting plate; 14. Rotating rod; 15. Fixed side plate; 16. Buffer plate; 17. Arc spring. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] See also Figures 1 to 6 The present invention provides a technical solution, a domestic waste intelligent identification and grabbing device, comprising: Conveyor belt 1, and 3D camera 3 and robot arm 4 are installed above, and the bottom end of robot arm 4 is respectively installed with metal suction cup and pneumatic suction cup, and placement plates 5 are evenly installed on the front of conveyor belt 1, and the upper surface of placement plates 5 is slidably connected to storage box 6; The side connecting plate 13 is rotatably connected to the lower portion of the placement plate 5, and the upper surface of the side connecting plate 13 is evenly rotatably connected to the buffer plate 16, and an arc spring 17 is added between the buffer plate 16 and the side connecting plate 13, and the upper surface of the side connecting plate 13 is in contact with the lower surface of the storage box 6; The side clamping plate 10 is slidably connected to the upper surface of the placement plate 5, and the inner surface of the side clamping plate 10 is tightly fitted with the outer surface of the storage box 6; The fixed side plates 15 are evenly arranged on the upper surface of the side connecting plate 13 , and the upper surface of the fixed side plates 15 is in contact with one side of the lower surface of the buffer plate 16 .

[0038] The conveyor belt 1 is started, and the conveyor belt 1 drives the placement plate 5 and the storage box 6 on the placement plate 5 to move. When the domestic waste is transported into the recognition range of the 3D camera 3 , the 3D camera 3 recognizes the domestic waste and determines the type and location information of the domestic waste. According to the information recognized by the 3D camera 3, the robotic arm 4 moves to the location of the domestic waste and selects the corresponding suction cup according to the type of domestic waste. If it is metal domestic waste, the robotic arm 4 controls the metal suction cup to grab it; if it is other types of domestic waste, the robotic arm 4 controls the pneumatic suction cup to grab it. After grabbing the domestic garbage, the robotic arm 4 moves to the top of the storage box 6 and places the domestic garbage into the storage box 6. When the storage box 6 is placed on the placement plate 5, the lower surface of the storage box 6 is in contact with the upper surface of the side connecting plate 13. As the storage box 6 is placed, the buffer plate 16 rotates under the pressure of the storage box 6, and the arc spring 17 is compressed, which plays a buffering role on the storage box 6. At the same time, the upper surface of the fixed side plate 15 is in contact with one side of the lower surface of the buffer plate 16, limiting the rotation range of the buffer plate 16. Slide the side clamping plate 10 so that the inner side surface of the side clamping plate 10 is tightly fitted with the outer surface of the storage box 6 to fix the storage box 6 to prevent the storage box 6 from shaking during the movement of the conveyor belt 1; The 3D camera 3 identifies domestic waste, and the robotic arm 4 selects the corresponding metal suction cup or pneumatic suction cup to grab it based on the identification result, realizing intelligent identification and classified grabbing of different types of domestic waste, and improving the accuracy and automation of classified grabbing of domestic waste; The storage box 6 is slidably connected to the upper surface of the placement plate 5, which makes it easy to take out and replace the storage box 6, thereby improving the operational convenience of the device.

[0039] See also Figure 2 A top plate 2 is horizontally installed on the upper surface of the conveyor belt 1, and a 3D camera 3 and a robotic arm 4 are evenly arranged on the lower surface of the top plate 2.

[0040] A top plate 2 is installed horizontally on the upper surface of the conveyor belt 1, and then a 3D camera 3 and a robotic arm 4 are evenly arranged on the lower surface of the top plate 2. When the conveyor belt 1 is running, the 3D camera 3 and the robotic arm 4 are supported by the top plate 2 and perform corresponding operations above the conveyor belt 1. The top plate 2 provides a stable installation carrier for the 3D camera 3 and the robotic arm 4, ensuring that the installation positions of the 3D camera 3 and the robotic arm 4 are fixed; the 3D camera 3 and the robotic arm 4 are evenly distributed on the lower surface of the top plate 2, which can ensure that they can perform comprehensive and uniform inspection or operation on the items transported on the conveyor belt 1, thereby improving the operation coverage.

[0041] See also Figure 3 and Figure 4 The upper surface of the placement plate 5 is evenly provided with limited side grooves 7 , and the inner cavities of the limited side grooves 7 are slidably connected to the limited side blocks 8 .

[0042] The limiting side grooves 7 are evenly opened on the upper surface of the placement plate 5, and then the limiting side blocks 8 are slidably connected to the inner cavity of the limiting side grooves 7 so that the limiting side blocks 8 can slide along the inner cavity of the limiting side grooves 7; The limiting side groove 7 provides a sliding guide for the limiting side block 8, limits the sliding direction of the limiting side block 8, and ensures the stable movement of the limiting side block 8; the sliding connection method enables the position of the limiting side block 8 to be adjusted according to actual needs, thereby improving the adaptability of the structure.

[0043] See also Figure 5 An extrusion spring 9 is provided between the inner wall of the limiting side groove 7 and the surface of the limiting side block 8, and the lower surface of the side clamping plate 10 is connected to the upper surface of the limiting side block 8.

[0044] An extrusion spring 9 is installed between the inner wall of the limiting side groove 7 and the surface of the limiting side block 8, and the lower surface of the side clamping plate 10 is connected to the upper surface of the limiting side block 8; when the limiting side block 8 slides along the limiting side groove 7, the extrusion spring 9 is deformed and generates an elastic force, driving the limiting side block 8 and the side clamping plate 10 to move synchronously; The extrusion spring 9 can provide continuous elastic extrusion force, so that the side splint 10 can automatically adjust the clamping force according to the size of the object, thereby achieving stable clamping of objects of different specifications; the connection method between the side splint 10 and the limiting side block 8 ensures the synchronization of the movement of the side splint 10 with the limiting side block 8, thereby improving the clamping reliability.

[0045] See also Figure 6 The upper surface of the side connecting plate 13 is evenly rotated and connected to the rotating rod 14, and the buffer plate 16 is connected to the rotating rod 14, and the fixed side plate 15 is arranged on the outer side of the rotating rod 14.

[0046] The rotating rod 14 is evenly rotated on the upper surface of the side connecting plate 13 to connect the buffer plate 16 to the rotating rod 14. Then, a fixed side plate 15 is set on the outer side of the rotating rod 14. When the buffer plate 16 is subjected to an external force, the rotating rod 14 rotates accordingly, and the fixed side plate 15 limits the rotation range of the rotating rod 14. The rotation characteristics of the rotating rod 14 enable the buffer plate 16 to adjust its angle according to the action of external force, effectively buffering external impact; the fixed side plate 15 can prevent the rotating rod 14 from excessive rotation, ensure the stable working position of the buffer plate 16, and improve the impact resistance and working stability of the structure.

[0047] The outer side of the placement plate 5 is rotatably connected to a rotating shaft 11 , and a driving motor 12 is coaxially connected to the end of the rotating shaft 11 , and a side connecting plate 13 is connected to the rotating shaft 11 .

[0048] The rotating shaft 11 is rotated and connected to the outer side of the placement plate 5, and the drive motor 12 is coaxially connected to the end of the rotating shaft 11, and then the side connecting plate 13 is connected to the rotating shaft 11; the drive motor 12 is started, and the drive motor 12 drives the rotating shaft 11 to rotate, and the rotating shaft 11 then drives the side connecting plate 13 to rotate synchronously; The drive motor 12 provides power for the rotation of the rotating shaft 11, ensuring the stability and controllability of the rotation of the rotating shaft 11; the connection method between the rotating shaft 11 and the side connecting plate 13 enables the side connecting plate 13 to be adjusted in angle along with the rotating shaft 11 to adapt to different operating requirements and enhance the flexibility of the structure.

[0049] See also Figure 7 A method for intelligently identifying and grasping domestic waste is provided, based on the above-mentioned intelligent device for intelligently identifying and grasping domestic waste, comprising: S1. Multimodal data acquisition and preprocessing: 3D camera 3 collects 3D point cloud data and surface texture information of the garbage on conveyor belt 1 in real time, and simultaneously obtains the material properties of the garbage through material sensors to form a multimodal data set; it performs noise reduction on the collected data, extracts edge features and geometric features, and generates standardized input data; S2. Intelligent identification and classification decision-making: The HybridNet model analyzes preprocessed data, generates bounding boxes and classification results for the garbage, and identifies material types such as metal, plastic, and glass. The depth sensor is used to calculate the number of stacked layers of garbage, prioritizing surface objects as grab targets and matching suction cup types based on material properties. S3, dynamic path planning and crawling execution: The DQN deep reinforcement learning algorithm is used to optimize the grasping path of Robot Arm 4, avoiding interference from adjacent objects and ensuring that the suction cups precisely contact the target trash. Robot Arm 4 is controlled to drive the selected suction cups to absorb the trash, and the suction force of the suction cups is dynamically adjusted according to the weight of the trash to complete the grasping action. S4, automatic cabinet management and buffer protection: When the storage box 6 is full of garbage, the driving motor 12 drives the side connecting plate 13 to rotate and separate from the placement plate 5 through the rotating shaft 11, and the storage box 6 descends along the inner wall of the placement plate 5; during the descent, the bottom surface of the storage box 6 contacts the buffer plate 16, and the impact force is absorbed by the arc spring 17. At the same time, the side clamping plate 10 releases the storage box 6 under the action of the extrusion spring 9, completing the automatic replacement process.

[0050] 3D camera 3 captures the household waste transported on conveyor belt 1 in real time, collecting 3D point cloud data and surface texture information. Simultaneously, the device's material sensors acquire the waste's material properties. These 3D point cloud data, surface texture information, and material properties together constitute a multimodal dataset. This multimodal dataset undergoes noise reduction to remove interference. The waste's edge and geometric features are then extracted and integrated to generate standardized input data, providing the foundation for subsequent identification and classification.

[0051] The standardized input data generated in S1 is input into the HybridNet model, which analyzes and processes the data to generate the corresponding bounding boxes and classification results of the garbage, accurately identifying the material types such as metal, plastic, and glass to which the garbage belongs. At the same time, the depth sensor in the device is used to calculate the number of stacked layers of garbage on the current conveyor belt 1, and preferentially mark the surface objects of the stack as targets to be grasped. In addition, according to the identified garbage material properties, the suction cup type in the device is matched to the material to prepare for subsequent grasping actions.

[0052] The DQN deep reinforcement learning algorithm is used to optimize the grasping path of the robot arm 4 to ensure that the movement trajectory of the robot arm 4 can avoid adjacent objects on the conveyor belt 1 to avoid interference. After determining the optimal path, the robot arm 4 is controlled to move according to the planned path, driving the matching suction cup in S2 to accurately contact the marked target garbage. During the adsorption process, the suction force of the suction cup is dynamically adjusted according to the weight of the target garbage to ensure that the garbage is stably adsorbed and finally complete the grasping action.

[0053] When the garbage in the storage box 6 accumulates to a full state, the device controls the drive motor 12 to start, and the drive motor 12 drives the side connecting plate 13 to rotate through the rotating shaft 11, so that the side connecting plate 13 is separated from the placement plate 5; at this time, the storage box 6 descends along the inner wall of the placement plate 5; during the descent, the bottom surface of the storage box 6 contacts the buffer plate 16, and the elastic deformation of the arc spring 17 absorbs the impact force generated during the descent, thereby achieving buffer protection; at the same time, the side clamping plate 10 releases the clamping of the storage box 6 under the reset action of the extrusion spring 9, completing the automatic replacement process of the storage box 6; The 3D camera 3 collects 3D point cloud data and surface texture information, and combines this with material sensors to obtain material properties to form a multimodal dataset. Compared to single data collection methods, this can more comprehensively reflect the physical characteristics and material properties of garbage, providing richer basic data for subsequent intelligent identification and helping to improve the accuracy of garbage identification. The HybridNet model analyzes preprocessed data to accurately generate bounding boxes and classification results, clarifying the type of garbage material. It also uses the number of stacked layers to mark surface targets and match suction cups, effectively avoiding misgrabbing of lower-level garbage, improving target accuracy, reducing ineffective grasping actions, and increasing grasping efficiency. The DQN deep reinforcement learning algorithm is used to optimize the gripping path of the robotic arm 4, effectively avoiding interference from adjacent objects and ensuring precise contact between the suction cup and the target trash. Dynamic adjustment of the suction force based on the trash weight prevents trash from falling off due to insufficient suction force or damage caused by excessive suction force, thus enhancing the stability of the gripping process. When the storage box 6 is full, the storage box 6 is automatically separated and lowered by the driving motor 12, the rotating shaft 11, the side connecting plate 13 and other structures without manual intervention, thereby improving the degree of automation of garbage disposal; during the descent, the buffering effect of the buffer plate 16 and the arc spring 17 can reduce the impact between the storage box 6 and the device, and the cooperation of the side clamping plate 10 and the extrusion spring 9 ensures that the storage box 6 is smoothly separated, which not only protects the device structure, but also avoids the scattering of garbage due to impact, thereby ensuring the safety of the storage and replacement process.

[0054] In S2, intelligent identification and classification decision-making, the HybridNet model generates comprehensive identification results including the location, size, and material category of garbage space by fusing 3D point cloud and material information.

[0055] During the intelligent identification and classification decision-making phase, the HybridNet model is activated. It receives 3D point cloud information and material information, fuses the two, and generates a comprehensive identification result that includes the spatial location, size, and material category of the garbage. By fusing 3D point clouds with material information using the HybridNet model, we can comprehensively capture the spatial, dimensional, and material characteristics of garbage, making identification results more complete and accurate, and providing reliable basic data support for subsequent garbage disposal processes.

[0056] During dynamic path planning and grasping execution in S3, the DQN deep reinforcement learning algorithm simulates the motion trajectory of the robot arm 4 in different stacking scenarios and generates optimal control instructions including joint angles, movement speeds, and suction cup switching timing.

[0057] During the dynamic path planning and grasping execution phase, the DQN deep reinforcement learning algorithm begins to work. This algorithm simulates the motion trajectory of Robot Arm 4 in different stacking scenarios and generates optimal control instructions based on the simulation results. These instructions include Robot Arm 4's joint angles, movement speed, and suction cup switching timing. By using the DQN deep reinforcement learning algorithm to simulate the motion trajectory of Robot Arm 4 in different stacking scenarios, the generated control instructions can be more adapted to the actual scenario requirements, improving the rationality of path planning and the accuracy of grasping execution, ensuring that Robot Arm 4 can effectively complete grasping operations in various stacking scenarios.

[0058] S4. In the automatic management and buffer protection of the box, during the descent of the storage box 6, the system calculates the descent speed of the storage box 6 in real time by monitoring the current changes of the drive motor 12, and dynamically adjusts the output torque of the drive motor 12, so that the storage box 6 completes the descent process in a variable acceleration manner, realizing impact-free switching.

[0059] During the automatic management and buffer protection phase of the storage box 6, when the storage box 6 enters the descent process, the system monitors the current changes of the drive motor 12 in real time, calculates the descent speed of the storage box 6 in real time based on the current changes, and then dynamically adjusts the output torque of the drive motor 12 based on the calculated descent speed, so that the storage box 6 completes the descent process in a variable acceleration manner, achieving shock-free switching; By monitoring the current changes of the drive motor 12 to adjust the output torque, the storage box 6 is made to descend with a variable acceleration, which can effectively avoid the impact during the descent process, achieve impact-free switching, protect the storage box 6 and the items inside, and improve the stability and safety of the automatic management process of the storage box 6.

[0060] This solution starts the conveyor belt 1, which drives the placement plate 5 and the storage box 6 on the placement plate 5 to move. The 3D camera 3 takes real-time photos of the domestic waste transported on the conveyor belt 1, collecting the 3D point cloud data and surface texture information of the waste. At the same time, the material sensor synchronously obtains the material properties of the waste to form a multimodal dataset. The collected multimodal dataset is subjected to noise reduction processing, and the edge features and geometric features of the waste are extracted to generate standardized input data. The standardized input data is fed into the HybridNet model, which analyzes and processes the data, generates bounding boxes and classification results for the garbage, identifies the material type of the garbage, and uses the depth sensor to calculate the number of layers of garbage on conveyor belt 1. Objects on the surface of the stack are prioritized as targets to be grasped, and the corresponding suction cup type is matched based on the identified garbage material properties. The DQN deep reinforcement learning algorithm is used to optimize the grasping path of the robotic arm 4. The motion trajectory of the robotic arm 4 in different stacking scenarios is simulated, and optimal control instructions including joint angles, movement speeds, and suction cup switching timing are generated. The robotic arm 4 is controlled to move according to the planned path, driving the matching suction cups to accurately contact the target garbage. The suction force of the suction cups is dynamically adjusted according to the weight of the garbage. If the garbage is metal, the robotic arm 4 controls the metal suction cup to grasp it. If it is other types of garbage, the robotic arm 4 controls the pneumatic suction cup to grasp it. After grasping the garbage, the robotic arm 4 moves to the top of the storage box 6 and places the garbage into the storage box 6. When the storage box 6 is placed on the placement plate 5, the lower surface of the storage box 6 is in contact with the upper surface of the side connecting plate 13. As the storage box 6 is placed, the buffer plate 16 rotates under the pressure of the storage box 6, and the rotating rod 14 rotates accordingly. The arc spring 17 is compressed, which plays a buffering role for the storage box 6. At the same time, the upper surface of the fixed side plate 15 is in contact with one side of the lower surface of the buffer plate 16, limiting the rotation range of the buffer plate 16. The side clamping plate 10 is slid so that the inner side surface of the side clamping plate 10 is in close contact with the outer surface of the storage box 6, thereby fixing the storage box 6 and preventing the storage box 6 from shaking during the movement of the conveyor belt 1. When the limiting side block 8 slides along the limiting side groove 7, the extrusion spring 9 is deformed and generates an elastic force, driving the limiting side block 8 and the side clamping plate 10 to move synchronously. When the storage box 6 is full of garbage, the drive motor 12 is started, and the drive motor 12 drives the rotating shaft 11 to rotate, and the rotating shaft 11 then drives the side connecting plate 13 to rotate and separate from the placement plate 5, and the storage box 6 descends along the inner wall of the placement plate 5; During the descent process, the bottom surface of the storage box 6 contacts the buffer plate 16, and the arc spring 17 absorbs the impact force, completing the replacement process.

[0061] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0062] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A household garbage intelligent identification and grabbing device, characterized in that: include: A conveyor belt (1), and a 3D camera (3) and a robotic arm (4) arranged above the conveyor belt, wherein a metal suction cup and a pneumatic suction cup are respectively installed at the bottom end of the robotic arm (4), and a placement plate (5) is evenly installed on the front of the conveyor belt (1), and a storage box (6) is slidably connected to the upper surface of the placement plate (5); The side connecting plate (13) is rotatably connected to the lower portion of the placement plate (5), and the upper surface of the side connecting plate (13) is evenly rotatably connected to the buffer plate (16), and an arc spring (17) is provided between the buffer plate (16) and the side connecting plate (13), and the upper surface of the side connecting plate (13) is in contact with the lower surface of the storage box (6); A side clamping plate (10) is slidably connected to the upper surface of the placement plate (5), and the inner side surface of the side clamping plate (10) is tightly fitted to the outer surface of the storage box (6); The fixed side plates (15) are evenly arranged on the upper surface of the side connecting plate (13), and the upper surface of the fixed side plates (15) is in contact with one side of the lower surface of the buffer plate (16).

2. The intelligent household waste identification and grabbing device according to claim 1, characterized in that: A top plate (2) is laterally mounted on the upper surface of the conveyor belt (1), and a 3D camera (3) and a robotic arm (4) are evenly arranged on the lower surface of the top plate (2).

3. The intelligent household waste identification and grabbing device according to claim 1, characterized in that: The upper surface of the placement plate (5) is evenly provided with limited side grooves (7), and the inner cavities of the limited side grooves (7) are all slidably connected to limited side blocks (8).

4. The intelligent household waste identification and grabbing device according to claim 3, characterized in that: An extrusion spring (9) is provided between the inner wall of the limiting side groove (7) and the surface of the limiting side block (8), and the lower surface of the side clamping plate (10) is connected to the upper surface of the limiting side block (8).

5. The intelligent household waste identification and grabbing device according to claim 1, characterized in that: The upper surface of the side connecting plate (13) is evenly rotatably connected to a rotating rod (14), and the buffer plate (16) is connected to the rotating rod (14). The fixed side plate (15) is arranged on the outside of the rotating rod (14).

6. The intelligent household waste identification and grabbing device according to claim 1, characterized in that: The outer side of the placement plate (5) is rotatably connected to a rotating shaft (11), and a driving motor (12) is coaxially connected to the end of the rotating shaft (11), and a side connecting plate (13) is connected to the rotating shaft (11).

7. A method for intelligently identifying and grasping domestic waste, based on a device for intelligently identifying and grasping domestic waste according to any one of claims 1 to 6, characterized in that: include: S1. Multimodal data acquisition and preprocessing: The 3D camera (3) collects the three-dimensional point cloud data and surface texture information of the garbage on the conveyor belt (1) in real time, and simultaneously obtains the material properties of the garbage through the material sensor to form a multimodal data set; the collected data is subjected to noise reduction processing, and edge features and geometric features are extracted to generate standardized input data; S2. Intelligent identification and classification decision-making: The HybridNet model analyzes the preprocessed data, generates bounding boxes and classification results for the garbage, and identifies the material types of metal, plastic, and glass. The depth sensor is used to calculate the number of stacked layers of garbage, prioritizing surface objects as grab targets and matching the corresponding suction cup type based on material properties. S3, dynamic path planning and crawling execution: The DQN deep reinforcement learning algorithm is used to optimize the grasping path of the robotic arm (4), avoid interference from adjacent objects, and ensure that the suction cup accurately contacts the target garbage; the robotic arm (4) is controlled to drive the selected suction cup to absorb the garbage, and the suction force of the suction cup is dynamically adjusted according to the weight of the garbage to complete the grasping action; S4, automatic cabinet management and buffer protection: When the storage box (6) is filled with garbage, the driving motor (12) drives the side connecting plate (13) to rotate and separate from the placement plate (5) through the rotating shaft (11), and the storage box (6) descends along the inner wall of the placement plate (5); during the descent, the bottom surface of the storage box (6) contacts the buffer plate (16), and the arc spring (17) absorbs the impact force. At the same time, the side clamping plate (10) releases the storage box (6) under the action of the extrusion spring (9), completing the automatic replacement process.

8. The method for intelligently identifying and controlling household waste according to claim 7, characterized in that: In S2, intelligent identification and classification decision-making, the HybridNet model generates a comprehensive identification result including the location, size and material category of garbage space by fusing 3D point cloud and material information.

9. The method for intelligently identifying and controlling household waste according to claim 7, characterized in that: In the above-mentioned S3, dynamic path planning and grasping execution, the DQN deep reinforcement learning algorithm generates optimal control instructions including joint angles, moving speeds and suction cup switching timing by simulating the motion trajectory of the robot arm (4) in different stacking scenarios.

10. The method for intelligently identifying and controlling household waste according to claim 7, characterized in that: In the above-mentioned S4, automatic management and buffer protection of the box, during the descent of the storage box (6), the system calculates the descent speed of the storage box (6) in real time by monitoring the current change of the drive motor (12), and dynamically adjusts the output torque of the drive motor (12), so that the storage box (6) completes the descent process in a variable acceleration manner, thereby achieving shock-free switching.

Images