An underwater garbage cleaning robot and a method for controlling the operation thereof

Abstract

The present application relates to a kind of underwater garbage cleaning robot and its operation control method, belong to the field of robot technology, to solve the underwater garbage cleaning robot of prior art There is difficult to adapt to complex underwater environment, poor mobility, short endurance time and low problem of grabbing effect to underwater plastic bottle and other sedimentary garbage.The present application includes mobile carrier module, buoyancy adjustment module, operation execution module, material collection and transmission module, visual sensor and hierarchical control system, including host computer computing unit and lower computer motion control unit;Wherein, mobile carrier module adopts track type drive structure, hierarchical control system is carried out robot underwater bottom operation by controlling buoyancy adjustment module and mobile carrier module.The present application is realized from automatic diving, depth cruise, target identification, stable grabbing to full load floating complete closed loop operation by mechatronics design, especially suitable for cleaning riverbed and shallow sea area of high-density sedimentary garbage.

Description

Technical Field

[0001] This invention belongs to the technical field of sewage garbage cleaning equipment, specifically relating to an underwater garbage cleaning robot and its operation control method. Background Technology

[0002] With the acceleration of global industrialization, marine plastic pollution has become an urgent environmental crisis. According to statistics from the United Nations Environment Programme, there are currently more than 150 million tons of plastic waste in the world's oceans, with an additional 8 to 12 million tons added each year. Studies show that approximately 94% of plastic waste eventually settles on the seabed, forming "benthic debris patches," which pose a serious threat to coral reef ecosystems, benthic communities, and fishery resources.

[0003] Currently, the main methods for cleaning up underwater debris rely on manual retrieval or simple mechanical operations, but these existing technologies have many limitations in practical applications: First, manual salvage methods are inefficient and risky. Divers operate in complex and ever-changing underwater environments, which not only limit their physical capabilities, resulting in short working hours, high labor intensity, and insufficient personal safety. Furthermore, due to the lack of specialized underwater identification and retrieval equipment, it is difficult for manual divers to accurately locate and effectively capture specific targets such as widely distributed and scattered plastic bottles.

[0004] Secondly, existing underwater cleaning robots lack sufficient environmental adaptability and grasping capabilities. Although underwater robot technology has advanced, existing equipment often suffers from limited functionality. Specifically: Poor adaptability of the end effector: Most robots' gripping mechanisms use rigid structures or simple opening and closing designs, lacking the ability to adapt to unstructured targets. When faced with plastic bottles of various shapes, smooth materials, and easy deformation, they cannot flexibly adjust the gripping force and posture according to the target characteristics, which can easily lead to the target slipping, breaking, or being missed, seriously affecting the cleaning effect.

[0005] Weak mobility and positioning / tracking capabilities: Existing underwater debris removal devices typically employ bulky mobile carriers, which suffer from high underwater fluid resistance and slow movement. When dealing with dynamic targets that float or move due to water currents, they struggle to achieve rapid response and accurate tracking. Furthermore, the complex hydrodynamic environment severely interferes with the stability of motion control, making it difficult for the robot to maintain precise positioning within the work area.

[0006] Secondly, the system's endurance and reliability are bottlenecks. Most underwater robots on the market use multi-degree-of-freedom thruster systems equipped with high-power brushless motors, resulting in huge energy consumption and making it difficult to support long-term continuous underwater operations. In addition, existing devices are often not specifically designed for the efficient transport and storage of plastic bottles, resulting in long dwell times and low transport efficiency during the collection process. Furthermore, some devices lack waterproof sealing and corrosion resistance design, making them prone to failure in harsh, high-salinity, and high-pressure underwater environments, leading to high maintenance costs. Therefore, an underwater garbage collection robot is needed to solve the problems of existing underwater garbage collection robots, such as difficulty adapting to complex underwater environments, poor maneuverability, short endurance, and low effectiveness in grasping underwater deposited garbage such as plastic bottles. Summary of the Invention

[0007] The purpose of this invention is to provide an underwater garbage cleaning robot to solve the technical problems of existing underwater cleaning equipment, such as poor mechanical adaptability in unstructured environments, insufficient body posture stability when performing grasping actions, and lack of a closed-loop operation of detection-grabbing-collection.

[0008] In addition, another objective of this invention is to provide an operation control method for an underwater garbage cleaning robot. By coordinating the buoyancy adjustment module and the tracked walking unit, and based on attitude compensation control using sensor feedback, the robot can achieve stable bottom-hugging cruising and precise anti-disturbance operation in complex hydrological environments, thereby improving the cleaning efficiency and success rate of deposited garbage such as plastic bottles.

[0009] The main design concept of the technical solution of this invention is as follows: To address the technical bottlenecks of existing underwater debris cleaning equipment in unstructured environments, such as poor mechanical adaptability, unstable operating posture, and low degree of autonomy, this invention adopts an integrated "mechanical-electrical-computer-control" system design strategy. The design concept of this invention is no longer limited to optimizing a single function, but rather constructs a complete autonomous cleaning closed-loop system through layered hardware integration and multi-physics-coordinated control logic. The specific design logic is as follows: 1. Design Strategy for a Mobility Platform Based on "Floating-Crawling" Dual Mode: In order to solve the problems of difficulty in hovering and inaccurate positioning of traditional remotely operated vehicles in strong current areas, this invention proposes a composite motion strategy that combines rigidity and flexibility.

[0010] Vertical dimension (Z-axis): Employs a variable volume aerodynamic buoyancy adjustment module. The design concept utilizes Archimedes' principle to achieve "submarine-like" vertical lifting and lowering by changing the displacement volume, enabling the robot to hover or sit on the bottom at a constant depth with low energy consumption, avoiding the high energy consumption caused by the continuous anti-gravity force of traditional propulsion.

[0011] Horizontal dimension (XY plane): A tracked differential drive chassis is adopted. The design concept is to utilize the high adhesion characteristics of the tracks so that the robot can switch to a "tank-like" bottom-hugging cruising mode after reaching the riverbed. This design uses the friction of the bottom to resist the impact of the water flow, providing a stable physical base for the robotic arm to grasp, fundamentally solving the "drifting" problem during suspended grasping.

[0012] 2. Layered hardware architecture design strategy of "separation of computing and control": In view of the contradiction between the limited computing power of underwater edge devices and the high real-time requirements of motion control, this invention establishes a decoupled architecture of "upper-layer perception and decision-making - lower-layer motion execution".

[0013] Perception Layer (Host Computer): This layer is responsible for processing high-throughput video stream data and running a lightweight version of the YOLOv5n algorithm, an improvement on the visual perception algorithm module. The design focuses on maximizing feature extraction efficiency with limited embedded computing resources by utilizing GhostNetV2 and Bidirectional Feature Pyramid Network (BiFPN) technology. It only outputs the target's coordinate deviation and category instructions, without participating in direct motor control, to prevent visual computation from blocking the control thread and ensure the system's real-time performance.

[0014] Execution layer (lower-level machine): Responsible for high-frequency pulse width modulation and sensor data fusion. The design focuses on millisecond-level response speed to ensure that the walking drive motor can be immediately activated for attitude compensation upon receiving instructions from the upper-level machine, thus guaranteeing the real-time performance of the system.

[0015] 3. Attitude disturbance rejection control strategy based on dual closed-loop PID: To address the problem of robot overturning caused by the forward shift of the center of gravity when the robot arm extends, this invention introduces an active attitude compensation mechanism.

[0016] The design concept is to model the robot's attitude control as a cascaded control system. The outer loop (angle loop) uses the pitch and roll angles locked in real time by the gyroscope as feedback, while the inner loop (speed loop) uses the motor speed fed back by the motor encoder.

[0017] When the robotic arm's movements cause a slight tilt of the robot body, the control algorithm does not rely on visual correction. Instead, it directly uses feedback from the attitude sensors to quickly adjust the differential speed output of the track motors, generating a counter-torque to counteract the overturning torque. This strategy ensures the robot's stability during dynamic grasping.

[0018] 4. Operation execution strategy of "flexible envelope + linear transmission": In view of the characteristics of smooth and easy slippage of target surfaces such as plastic bottles, this invention abandons the traditional rigid precision clamping approach and instead adopts a flexible envelope strategy with higher fault tolerance.

[0019] Gripping end: Employs a TPU soft gripper with a high coefficient of friction, which passively adapts to the geometric irregularities of the target by utilizing the elastic deformation of the material itself, thereby achieving an adaptive flexible envelope for unstructured targets.

[0020] Transmission end: To solve the problem of "nowhere to put" the garbage after it is grabbed, a linear transmission channel based on a synchronous belt and a lead screw was designed. The design concept is to transform the discrete grabbing action into continuous material transmission. With the flipping action of the end servo motor, the garbage can be seamlessly transferred from the "riverbed" to the "collection box", avoiding the time loss and center of gravity oscillation caused by the frequent large-amplitude rotation of the robotic arm.

[0021] 5. Visual-Guided Autonomous Servo Workflow Design: This invention constructs an automated state machine of "detection-proximity-lock-operation". The design concept utilizes visual feedback as the trigger condition for state transitions. Long-range phase: Utilize the wide-angle field of view of a wide-angle underwater camera to perform omnidirectional scanning; Mid-range phase: Visual servo correction is performed by driving the walking chassis to compensate for the center deviation of the detection frame of the wide-angle underwater camera. In the close-range phase: Depth information is perceived using a depth sensor to trigger the robotic arm's movements. This strategy eliminates reliance on expensive sensors such as sonar, which are noisy in shallow water, and achieves low-cost, high-precision autonomous operation entirely based on the fusion of monocular vision and kinematics.

[0022] To achieve the above objectives, the present invention provides the following technical solution: an underwater garbage cleaning robot, comprising: Mobile carrier module, used for underwater mobile cruising of robots; The buoyancy adjustment module, installed on the mobile carrier module, is used to adjust the robot's net buoyancy, enabling the robot to dive and surface for recovery. The operation execution module is installed at the front end of the mobile carrier module and is used to grab underwater debris; The material collection and transfer module is used to transport and store the grabbed waste into the collection bin; Visual sensors are used to acquire images and perform target recognition; The hierarchical control system includes a host computer computing unit and a slave computer motion control unit; the host computer computing unit is used to process visual sensor data and output target decision commands; the slave computer motion control unit is electrically connected to the mobile carrier module, buoyancy adjustment module and operation execution module, and is used to perform motion control. The mobile carrier module adopts a tracked drive structure, and the layered control system controls the buoyancy adjustment module and the mobile carrier module to perform underwater bottom-hugging operations for the robot.

[0023] Preferably, the mobile carrier module includes a chassis and two sets of tracked walking units; the tracked walking units are symmetrically installed on both sides of the chassis, and the chassis surface constitutes the mounting carrier for other module components; each set of tracked walking units includes two sub-tracked walking units, each sub-tracked walking unit including a walking frame, drive wheels, and tracks; the walking frame is fixedly installed on the chassis, and several drive wheels are rotatably installed on the walking frame at intervals, the tracks are sleeved on the outside of the drive wheels, and a drive motor is installed on the chassis at a position corresponding to one of the drive wheels, the power end of the drive motor being connected to the drive wheel.

[0024] Preferably, the buoyancy adjustment module includes: a depth sensor, a variable volume cavity mechanism, and a pneumatic drive unit; the variable volume cavity mechanism includes several variable volume water injection chambers; a slidable stopper is installed inside the water injection chamber to divide the water injection chamber into a variable volume water injection chamber and an air injection chamber, with the side of the water injection chamber away from the stopper being the open side, forming the water injection port; an air port is provided on the side of the air injection chamber away from the stopper; the pneumatic drive unit includes an air pump and several high-pressure air cylinders; the air port of the air pump is connected to the air port of the water injection chamber and the air port of the high-pressure air cylinders through connecting pipes. Preferably, the air pump of this invention is a dual-head air pump, which controls the sliding of the stopper in the water injection chamber by forward air extraction or reverse air injection, thereby changing the volume of the water injection chamber and thus changing the buoyancy generated by the air injection chamber.

[0025] Preferably, the operation execution module includes a flexible end effector and a multi-degree-of-freedom drive mechanism. The flexible end effector includes a robotic arm and a robotic gripper. The robotic gripper is made of TPU soft material and its surface is coated with friction pads. The multi-degree-of-freedom drive mechanism includes a first servo and a second servo. The first servo is a single-axis servo, with one end of the robotic arm connected to the power shaft of the first servo and rotatably mounted on the operation execution module in the vertical direction. The other end is connected to the robotic gripper via the second servo. The second servo is a dual-axis servo, with the robotic gripper connected to one of the power shafts and rotatably mounted on the robotic arm. The other power shaft is connected to the gripper's jaws to drive the gripper's clamping or releasing action.

[0026] Preferably, the material collection and transfer module includes: a linear lifting mechanism and a collection box. The linear lifting mechanism is located at the back of the mobile carrier module and includes a mounting frame, a stepper motor, a ball screw assembly, and a mounting plate. The mounting frame is fixedly installed on the mobile carrier module. The ball screw assembly's lead screw is radially rotatable and mounted on the mounting frame. The stepper motor is fixedly installed on the mounting frame, and its power end is connected to the lead screw via a synchronous belt. The nut end of the ball screw assembly is fixedly connected to the mounting plate. The operation execution module is mounted on the mounting plate. The surface of the lead screw is provided with an anti-corrosion and wear-resistant layer. The collection box is located at the end of the linear lifting mechanism. Its inlet is equipped with a diffuse reflection sensor and a guide plate. The sensing end of the diffuse reflection sensor faces the inside of the collection box to monitor the height of waste accumulation. The guide plate is installed at an inclined angle at the inlet of the collection box. A discharge door is installed at the rear of the collection box, and the discharge door is rotatably mounted on the collection box via a pin.

[0027] Preferably, the hierarchical control system adopts a computing and control separation architecture: it includes an upper-level computing unit and a lower-level motion control unit; the upper-level computing unit is equipped with a visual perception algorithm module, which is used to calculate the target's category and three-dimensional spatial coordinates. The visual perception algorithm module is connected to the lower-level motion control unit through a communication interface; the lower-level motion control unit integrates a dual closed-loop PID controller, which is connected to an attitude sensor and a motor encoder. The attitude sensor is a gyroscope, which is mounted on the chassis; the motor encoder is electrically connected to the drive motor, and is used to adjust the output torque of the drive motor of the moving carrier module in real time when the operation of the execution module causes a change in the system's center of gravity, so as to maintain the horizontal stability of the body attitude; the hierarchical control system has several waterproof chambers for installing the hierarchical control system components and the overall power supply. The control logic of the dual closed-loop PID controller of the present invention is as follows: the outer loop is the angle loop, which calculates the target differential speed signal based on the deviation between the fuselage pitch angle data fed back by the attitude sensor and the target angle; the inner loop is the speed loop, which performs incremental PID adjustment based on the real-time speed data of the drive motor fed back by the motor encoder, and drives the motor to adjust the electronic speed control based on the target differential speed signal of the outer loop.

[0028] As a preferred option, the visual sensor is a wide-angle underwater camera; the visual perception algorithm module running in the host computer computing unit adopts a lightweight convolutional neural network, which uses GhostNetV2 as the backbone network and integrates a bidirectional feature pyramid network and a receptive field convolutional block attention module.

[0029] The present invention also provides an operation control method for an underwater garbage cleaning robot, which includes the following steps: S1. Stable diving: The depth sensor detects the water depth at the target location. When it is determined to be a diving depth, the buoyancy adjustment module is controlled to reduce the drainage volume. At the same time, the attitude sensor monitors the pitch angle and adjusts the center of gravity to make the robot dive vertically to the predetermined depth. S2, Visual Servo Navigation: The robot uses a tracked walking unit to navigate underwater. When the visual sensor identifies the target, the visual perception algorithm module calculates the target deviation and controls the differential speed of the two tracks through a dual closed-loop PID controller to achieve a straight approach to the target. S3, Flexible Adaptive Grasping: After reaching the work position, based on the target posture feedback from the vision, the work execution module is controlled to adjust the pitch angle and open the mechanical claw of the flexible end effector to envelop and grasp the target; during this process, the lower-level motion control unit continuously controls the anti-overturning torque caused by the forward shift of the center of gravity through the PID controller; S4. Continuous transmission and storage: After the target is grasped, the target is sent into the collection box of the material collection and transmission module by controlling the lifting of the linear lifting mechanism and the flipping of the robotic arm of the operation execution module. S5, Full-load Float Recovery: When the diffuse reflection sensor of the collection box detects that the collection box is full, the buoyancy adjustment module is controlled to increase the drainage volume, so that the robot floats to the water surface.

[0030] Preferably, in S3, the flexible adaptive grasping includes the following steps: S30. Based on the identified target geometric features, dynamically adjust the rotation angle of the power shaft of the second servo motor to adjust the gripping stroke opening of the mechanical claw. S31. During the mechanical gripper's grasping process, the elastic deformation of the TPU soft material is used to adapt to the surface irregularities of the target, and the gripping friction force is provided by the friction pads on the surface. S32. During the grasping process, when the movement of the flexible end effector's robotic arm causes the robot to tilt slightly, the attitude sensor, i.e., the gyroscope, senses the shift in the center of gravity. The gyroscope feeds back the pitch and roll angle variables to the PID controller, which then controls the drive motor of the track walking unit through the motor encoder to generate a reverse torque to counteract the overturning torque.

[0031] As an alternative, when the S5 is fully loaded and floats to the surface for recovery, the following steps are also included: the robot uses its built-in GPS positioning module to obtain the water surface location information; after floating to the water surface, it sends out a recovery signal through an audible and visual alarm device; after the robot recovers the waste, the recovery personnel unlock the unloading door of the collection box to dump the waste and reset the system.

[0032] Beneficial Effects: 1. The underwater garbage cleaning robot of this invention greatly improves the posture stability and anti-disturbance capability during underwater operations. Existing underwater cleaning devices mostly adopt a floating ROV structure, which is prone to attitude drift when grabbing heavy objects or being impacted by water currents. This invention innovatively adopts a composite motion mode of "buoyancy adjustment + tracked propulsion," combined with a dual-closed-loop PID attitude control strategy. On the one hand, the tracked chassis achieves bottom-hugging cruising, utilizing the friction of the seabed to provide a stable working base; on the other hand, when the robotic arm extends and causes the system's center of gravity to shift forward, the PID controller can adjust the differential output torque of the motor in milliseconds to compensate. This design solves the problem of robot tipping and vibration caused by the robotic arm's movements, ensuring precise target alignment in complex flow fields.

[0033] 2. Building upon the foregoing, this invention employs a flexible end effector with a high-friction coefficient contact surface made of TPU soft material. This structure utilizes the material's inherent elastic deformation properties to adaptively conform to the target surface shape, achieving a flexible envelope. Combined with the variable angle control of a multi-degree-of-freedom servo motor, it maintains an extremely high grasping success rate even when the target position is deviated or partially buried, avoiding secondary pollution. This achieves efficient flexible envelope grasping of unstructured deposited waste, addressing the technical pain points of existing rigid robotic arms that easily slip and break when grasping smooth-surfaced, irregularly shaped plastic bottles.

[0034] 3. Building upon the foregoing, this invention establishes a hierarchical hardware topology of "upper-computer perception and decision-making - lower-computer motion execution." The upper-computer computing unit focuses on running an improved lightweight vision algorithm for target calculation, while the lower-computer motion control unit focuses on high-frequency motor control and sensor data fusion. This architecture achieves physical-level decoupling between computationally intensive tasks and real-time control tasks, ensuring the real-time response capability of the underlying motion control system to sudden attitude changes. A high-response hierarchical control architecture of "computation and control separation" is constructed to address the problem of underwater edge devices easily blocking motion control threads when handling high-load vision tasks.

[0035] 4. Building upon the foregoing, this invention integrates a unified linear transmission and automatic collection management system, breaking through the limitations of traditional equipment that suffers from "nowhere to store the collected waste" or "low efficiency per operation." Through the coordination of a synchronous belt screw transmission mechanism and a tilting servo motor, the discrete actions of grabbing are transformed into continuous material flow, achieving seamless transfer of waste from the riverbed to the collection bin. Combined with the constant-depth diving and full-load surfacing logic of the pneumatic buoyancy adjustment system, this invention can complete a full cleaning cycle without manual intervention, significantly improving cleaning efficiency and operational endurance per unit time. It achieves a fully autonomous closed-loop process from "searching" to "recycling." Attached Figure Description

[0036] Figure 1This is a schematic diagram of the underwater garbage cleaning robot in this embodiment; Figure 2 This is a schematic diagram of the mobile carrier module of the underwater garbage cleaning robot in this embodiment; Figure 3 This is a schematic diagram of the buoyancy adjustment module of the underwater garbage cleaning robot in this embodiment; Figure 4 This is a schematic diagram of the operation execution module and the material collection and transmission module of the underwater garbage cleaning robot in this embodiment; Figure 5 This is a schematic diagram of the collection box of the underwater garbage cleaning robot in this embodiment; Figure 6 This is a schematic diagram of the electrical connections of the components in the layered control system of the underwater garbage cleaning robot in this embodiment; Figure 7 This is the logic diagram of the dual closed-loop PID attitude control based on gyroscope feedback in the hierarchical control system of this embodiment; Figure 8 A flowchart illustrating the autonomous operation control method for the underwater garbage cleaning robot in this embodiment; Figure 9 This is a schematic diagram illustrating the target recognition effect of the host computer computing unit in this embodiment using a wide-angle underwater camera in a complex underwater environment. Figure 10 This is a flowchart illustrating the operational status of the underwater debris-collecting robot in this embodiment performing the continuous actions of "detection-grabbing-flipping-transfer". Figure 11 This is a schematic diagram of the overall network architecture of the visual perception algorithm module running in the host computer computing unit of this embodiment.

[0037] In the diagram: 1. Mobile carrier module; 10. Chassis; 11. Tracked walking unit; 12. Walking frame; 13. Drive wheel; 14. Track; 18. Drive motor; 2. Buoyancy adjustment module; 21. Water injection tank; 22. Dual-head air pump; 23. High-pressure air cylinder; 3. Operation execution module; 31. Mechanical claw; 32. Silicone friction plate; 33. First servo motor; 34. Second servo motor; 4. Material collection and transmission module; 40. Mounting frame; 41. Stepper motor; 42. Ball screw pair; 43. Mounting plate; 44. Guide plate; 45. Unloading gate; 46. Waterproof tank; 5. Layered control system; 51. Upper computer computing unit; 52. Lower computer motion control unit; 55. Attitude sensor; 6. Collection box; Detailed Implementation

[0038] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments.

[0039] Please see Figures 1-11 This embodiment provides an underwater garbage cleaning robot and its operation control method to solve the problems mentioned in the background art.

[0040] To achieve the above objectives, the present invention provides the following technical solution: an underwater garbage cleaning robot, comprising: Mobile carrier module 1 is used for underwater mobile cruising of the robot; Buoyancy adjustment module 2, installed on the mobile carrier module, is used to adjust the net buoyancy of the robot to enable the robot to dive and float back for recovery; Operation execution module 3 is installed at the front end of the mobile carrier module and is used to grab underwater debris; The material collection and transfer module 4 is used to transport and store the grabbed waste into the collection bin 6; The visual sensor, using a wide-angle underwater camera, is used to acquire images and perform target recognition; The hierarchical control system 5 includes an upper computer computing unit and a lower computer motion control unit; the upper computer computing unit is used to process visual sensor data and output target decision commands; the lower computer motion control unit is electrically connected to the mobile carrier module, the buoyancy adjustment module and the operation execution module, and is used to perform motion control; The mobile carrier module adopts a tracked drive structure, and the layered control system controls the buoyancy adjustment module and the mobile carrier module to perform underwater bottom-hugging operations for the robot.

[0041] The mobile carrier module 1 of this embodiment includes a chassis 10 and two sets of tracked walking units 11. Each set of tracked walking units includes two sub-tracked walking units. Tracked walking units are symmetrically installed on both sides of the chassis, and the chassis surface constitutes the mounting carrier for other module components. Each set of tracked walking units includes two sub-tracked walking units. The sub-tracked walking units include a walking frame 12, a drive wheel 13, and a track 14. The walking frame is fixedly installed on the chassis. Several drive wheels are rotatably installed on the walking frame at intervals. The track is sleeved on the outside of the drive wheels. A drive motor 18 is installed on the chassis at a position corresponding to one of the drive wheels. The power end of the drive motor is connected to the drive wheel.

[0042] The buoyancy adjustment module 2 of this embodiment includes: a depth sensor, a variable volume cavity mechanism, and a pneumatic drive unit; the variable volume cavity mechanism includes four variable volume water injection chambers 21; the four water injection chambers are installed at the four corners of the chassis, and the interior of the water injection chambers is slidably fitted with a partition to divide the water injection chamber into a variable volume water injection chamber and an air injection chamber, the side of the water injection chamber away from the partition is the opening side, forming a water injection port; the side of the air injection chamber away from the partition is provided with an air port; the pneumatic drive unit includes two double-headed air pumps 22 and four high-pressure air storage cylinders 23; the two double-headed air pumps and the four high-pressure air storage cylinders are symmetrically arranged in two groups on both sides of the chassis, and the air ports of the double-headed air pumps are connected to the air ports of the water injection chambers and the air ports of the high-pressure air storage cylinders through connecting pipes. In this embodiment, the water injection chamber adopts a medical-grade syringe structure, each with a capacity of 1.2L, and a total drainage adjustment capacity of 4.8L. The dual-head air pump is a UV-U3-24V dual-head air pump, which is connected to the air injection chamber and the built-in high-pressure air storage cylinder of the water injection chamber through connecting pipelines. It can establish a vacuum of up to 90% in a closed loop through series air pumping, thereby driving the movement of the diaphragm to change the buoyancy of the system. The depth sensor in this embodiment uses an acoustic ranging sensor to avoid the inaccuracy of using a conventional laser ranging sensor when the water quality is poor.

[0043] To improve target grasping efficiency and handle plastic bottles of different shapes, the operation execution module 3 of this embodiment includes a flexible end effector and a multi-degree-of-freedom drive mechanism. The flexible end effector includes a robotic arm and a robotic gripper 31. The robotic gripper is made of TPU soft material, and its surface is coated with a silicone friction pad 32 with a friction coefficient of μ=0.8. The multi-degree-of-freedom drive mechanism includes a first servo motor 33 and a second servo motor 34. The first servo motor is a single-axis servo motor. One end of the robotic arm is connected to the power shaft of the first servo motor and is rotatably mounted on the mounting plate of the operation execution module in the vertical direction. The other end is connected to the robotic gripper through the second servo motor. The second servo motor is a dual-axis servo motor. The robotic gripper is connected to one of the power shafts and is rotatably mounted on the robotic arm. The other power shaft is connected to the gripper of the robotic gripper to drive the gripper's gripping or releasing action. The mechanical gripper in this embodiment has two grippers arranged opposite each other. Each gripper is equipped with a driven gear. A main drive gear is installed at the end of the power shaft of the second servo motor. The main drive gear and the driven gear form a gear transmission, which enables the power shaft of the second servo motor to drive the two grippers to perform gripping or releasing actions.

[0044] Furthermore, the material collection and transfer module 4 of this embodiment includes: a linear lifting mechanism and a collection box. The linear lifting mechanism is located on the back of the mobile carrier module and includes a mounting frame 40, a stepper motor 41, two ball screw pairs 42, and a mounting plate 43. The mounting frame is fixedly installed on the mobile carrier module. The screws of the two ball screw pairs are radially rotatably mounted on the mounting frame. The stepper motor is fixedly installed on the mounting frame, and its power end is connected to the screw via a synchronous belt (not shown). The nut ends of the two ball screw pairs are fixedly connected to the mounting plate. The operation execution module is installed on the mounting plate. The surface of the screw is provided with an anti-corrosion and wear-resistant layer. The collection box is located at the end of the linear lifting mechanism. A diffuse reflection sensor and a guide plate 44 are provided at its inlet. The sensing end of the diffuse reflection sensor faces the inside of the collection box and is used to monitor the height of the garbage accumulation. The guide plate is installed at an inclined angle at the inlet of the collection box. A discharge door 45 is installed at the rear of the collection box. The discharge door is rotatably mounted on the collection box via a hinge and locked by a mechanical pin.

[0045] The hierarchical control system in this embodiment adopts a computing and control separation architecture: it includes an upper computer computing unit 51 and a lower computer motion control unit 52; the upper computer computing unit is equipped with a visual perception algorithm module, which is used to calculate the target category and three-dimensional spatial coordinates. The visual perception algorithm module is connected to the lower computer motion control unit through a communication interface; the lower computer motion control unit integrates a dual closed-loop PID controller. The PID controller is connected to an attitude sensor, a motor encoder and a dual-head air pump. The attitude sensor 55 adopts an HWT9073-485 high-precision gyroscope with a static measurement accuracy better than 0.001° and has an anti-magnetic field interference algorithm, which is used to provide real-time feedback on the body attitude. It is installed on the chassis; the motor encoder is electrically connected to the drive motor, which is used to adjust the output torque of the drive motor of the moving carrier module in real time when the operation of the execution module causes the system center of gravity to change, so as to maintain the horizontal stability of the body attitude; the hierarchical control system has three waterproof chambers 46 for installing the hierarchical control system components and the power supply of the whole machine. In addition, the host computer computing unit 51 in this embodiment runs a visual perception algorithm module. This module is used to process images acquired by a wide-angle underwater camera in real time to identify the type and location of underwater debris targets. Specifically, the visual perception algorithm module adopts a lightweight target detection network based on YOLOv5n and uses GhostNetV2 as the backbone network to reduce the computational load. In the feature fusion part, a bidirectional feature pyramid network BiFPN is introduced to enhance the multi-scale feature extraction capability for debris targets of different sizes. The receptive field convolutional block attention module RFCBAMConv is integrated in the detection head to improve the target representation capability in turbid, low-contrast underwater environments and reduce the interference of suspended particle noise on the recognition results.

[0046] During operation, underwater images captured by the wide-angle underwater camera are first input to the host computer computing unit 51. The visual perception algorithm module detects debris targets in the images and outputs the target category label, the target detection box position, and the target recognition confidence score. Figure 9 As can be seen in the figure, "plastic_bottle" indicates that the target category identified by the visual perception algorithm is plastic bottle waste. Subsequent values ​​such as 0.71, 0.93, 0.41, and 0.61 represent the confidence level of the target belonging to the plastic bottle category. The rectangular box represents the position range of the corresponding target in the image. The host computer computing unit 51 further calculates the relative orientation information between the robot and the target based on the positional deviation of the target detection box center relative to the image center, and sends this deviation information to the lower-level motion control unit 52 to control the robot to complete target approach and grasping.

[0047] This embodiment also provides an operation control method for an underwater garbage cleaning robot, which includes the following steps: S1. Stable Descent: After the robot enters the water, the depth sensor detects the water depth at the target location. When the depth is determined to be suitable for descent, the layered control system initiates the descent program. The lower-level motion control unit controls the dual-head air pumps of the buoyancy adjustment module to operate. By pumping air, the diaphragm in the water injection chamber is drawn towards the air injection chamber, drawing air from the air injection chamber into the high-pressure air storage cylinder. As the diaphragm operates under negative pressure, the buoyancy generated in the air injection chamber decreases, and the robot begins to descend. After descending to a certain depth, water enters from the water injection port of the water injection chamber, filling it and reducing the drainage volume, allowing the robot to descend further. During this process, the robot's gyroscope monitors the body's pitch angle in real time. If tilting is detected, the lower-level motion control unit fine-tunes the dual-head air pumps on both sides of the chassis to ensure the robot descends vertically to the predetermined depth. When the robot sinks to the bottom, the dual-head air pumps stop working, and the robot is in a state of slight negative buoyancy or zero buoyancy.

[0048] S2. Visual Servo Navigation: The robot navigates underwater using a tracked walking unit, while the host computer acquires images via a wide-angle camera. Upon target detection, the robot calculates the target's deviation in the image coordinate system and sends the result to the lower-level motion control unit. The lower-level motion control unit executes a dual-closed-loop PID control algorithm. Outer ring (position ring / angle ring): Calculates the target heading angle based on visual bias; Inner loop (speed loop): Adjusts the speed difference between the left and right track drive motors based on feedback from the motor encoder. The robot maintains a straight line as it approaches the target. When it gets close to the work area, it automatically decelerates and stops, waiting to be grabbed. S3. Flexible Adaptive Grasping: After the robot arrives at the work position and stops and stabilizes, it controls the work execution module to adjust the pitch angle and open the mechanical claw of the flexible end effector to grasp the target based on the target posture feedback from vision. Attitude adjustment: The first servo motor adjusts the pitch angle of the flexible end effector based on the target depth information from visual feedback, so that it is aligned with the target.

[0049] Enveloping gripping: The second servo motor drives the mechanical claw to open, envelop the target, and then close. The deformation of the flexible TPU claw and the high friction of the silicone friction pad work together to ensure tight gripping of waste of different shapes (such as flat and cylindrical).

[0050] Steady-state control: During the movement of the robotic arm, the lower-level motion control unit continuously monitors the change of the center of gravity and controls the track drive motor to output reverse torque through PID algorithm to prevent the body from tilting forward; S4. Continuous Transfer and Storage: After the target is grasped, the transfer program is executed. By controlling the lifting of the linear lifting mechanism and the flipping of the robotic arm in the work execution module, the target is sent into the collection box of the material collection and transfer module; specifically: Linear conveying: The stepper motor starts and rotates through the synchronous belt pulley and screw, driving the mounting plate to lift the garbage to the inlet side height of the collection box; Tilting and feeding: The first servo motor drives the robotic arm to rotate towards the collection box side, which in turn moves the robotic claw and the garbage towards the inlet side of the collection box. When it reaches the guide plate of the collection box, the second servo motor drives the flexible end effector to rotate as a whole, so that the garbage is aligned with the inlet of the guide plate and rotated. Then the robotic claw is released to put the garbage into the inlet of the collection box through the guide plate. Monitoring and counting: After the diffuse reflection sensor located at the entrance of the collection bin detects the passage of garbage, the system count is incremented by one, and the robotic arm is reset to prepare for the next round of grabbing.

[0051] S5. Full-load buoyancy recovery: When the diffuse reflection sensor of the collection box detects that the collection box is full, the buoyancy adjustment module is controlled to increase the drainage volume. Specifically, the lower-level motion control unit starts the dual-head air pump to inject air into the air injection chamber of the water injection tank, so as to drive the diaphragm in the water injection tank to move towards the water injection chamber side, and drain the water in the water injection chamber. At the same time, air is continuously injected into the air injection chamber to increase buoyancy and make the robot float to the water surface.

[0052] After the robot emerges from the water, it sends a location signal using its GPS module and triggers a audible and visual alarm to prompt for retrieval. During the manual unloading phase: once the robot is retrieved to the shore or a vessel, the operator manually unlocks the mechanical latch located on the back of the collection bin. The unloading door opens around the hinge axis, quickly dumping the waste. After cleaning, the bin door is locked, and the robot resets to prepare for the next task.

[0053] Working Principle: The underwater debris-collecting robot of this invention sinks to the bottom by controlling the drainage volume of the buoyancy adjustment module. It utilizes a mobile carrier module, a layered control system, and visual sensors for linear movement and angle adjustment, achieving precise positioning. A flexible mechanical claw made of TPU soft material, combined with servo motor adjustments, enables the grasping of objects of different shapes. A stepper motor drives a synchronous pulley to transport debris to a collection bin, and a diffuse reflection sensor detects the debris collection status. This design improves the efficiency and accuracy of debris collection, reduces the risk of debris leakage and secondary pollution, and simultaneously reduces the workload and safety risks for operators. This invention is of great significance for improving the efficiency and thoroughness of underwater debris collection and achieving efficient seabed environmental purification.

[0054] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. An underwater garbage cleaning robot, characterized in that, include: Mobile carrier module, used for underwater mobile cruising of robots; A buoyancy adjustment module is installed on the mobile carrier module to adjust the net buoyancy of the robot, enabling the robot to dive and surface for recovery. An operation execution module, installed at the front end of the mobile carrier module, is used to grab underwater debris; The material collection and transfer module is used to transport and store the grabbed waste into the collection bin; Visual sensors are used to acquire images and perform target recognition; The hierarchical control system includes a host computer computing unit and a slave computer motion control unit; the host computer computing unit is used to process visual sensor data and output target decision commands; the slave computer motion control unit is electrically connected to the mobile carrier module, the buoyancy adjustment module and the operation execution module, and is used to perform motion control. The mobile carrier module adopts a tracked drive structure, and the layered control system controls the buoyancy adjustment module and the mobile carrier module to perform the robot's underwater bottom-hugging operation.

2. The underwater garbage cleaning robot according to claim 1, characterized in that, The mobile carrier module includes a chassis and two sets of tracked walking units. The tracked walking units are symmetrically installed on both sides of the chassis, and the chassis surface constitutes the mounting carrier for other module components. Each set of tracked walking units includes two sub-tracked walking units, each sub-tracked walking unit including a walking frame, drive wheels, and tracks. The walking frame is fixedly installed on the chassis, and several drive wheels are rotatably installed on the walking frame at intervals. The tracks are sleeved on the outside of the drive wheels. A drive motor is installed on the chassis at a position corresponding to one of the drive wheels, and the power end of the drive motor is connected to the drive wheel.

3. The underwater garbage cleaning robot according to claim 1, characterized in that, The buoyancy adjustment module includes: a depth sensor, a variable volume cavity mechanism, and a pneumatic drive unit; the variable volume cavity mechanism includes several variable volume water injection chambers; a partition is slidably installed inside the water injection chamber to divide the water injection chamber into a variable volume water injection chamber and an air injection chamber, with the side of the water injection chamber away from the partition being the open side, forming a water injection port; an air port is provided on the side of the air injection chamber away from the partition; the pneumatic drive unit includes an air pump and several high-pressure air cylinders; the air port of the air pump is connected to the air port of the water injection chamber and the air port of the high-pressure air cylinders through connecting pipes.

4. The underwater garbage cleaning robot according to claim 1, characterized in that, The operation execution module includes a flexible end effector and a multi-degree-of-freedom drive mechanism. The flexible end effector includes a robotic arm and a robotic gripper. The robotic gripper is made of TPU soft material and its surface is coated with friction pads. The multi-degree-of-freedom drive mechanism includes a first servo motor and a second servo motor. The first servo motor is a single-axis servo motor. One end of the robotic arm is connected to the power shaft of the first servo motor and is rotatably mounted on the operation execution module in the vertical direction. The other end is connected to the robotic gripper through the second servo motor. The second servo motor is a dual-axis servo motor. The robotic gripper is connected to one of the power shafts and is rotatably mounted on the robotic arm. The other power shaft is connected to the gripper's jaws to drive the gripper's clamping or releasing action.

5. The underwater garbage cleaning robot according to claim 1, characterized in that, The material collection and transfer module includes a linear lifting mechanism and a collection box. The linear lifting mechanism is located on the back of the mobile carrier module and includes a mounting frame, a stepper motor, a ball screw assembly, and a mounting plate. The mounting frame is fixedly installed on the mobile carrier module. The ball screw assembly's screw is radially rotatably mounted on the mounting frame. The stepper motor is fixedly installed on the mounting frame, and its power end is connected to the screw via a synchronous belt. The nut end of the ball screw assembly is fixedly connected to the mounting plate. The operation execution module is mounted on the mounting plate. The surface of the screw is provided with an anti-corrosion and wear-resistant layer. The collection box is located at the end of the linear lifting mechanism. Its inlet is equipped with a diffuse reflection sensor and a guide plate. The sensing end of the diffuse reflection sensor faces the inside of the collection box to monitor the height of waste accumulation. The guide plate is installed at an inclined angle at the inlet of the collection box. A discharge door is installed at the rear of the collection box, and the discharge door is rotatably mounted on the collection box via a pin.

6. The underwater garbage cleaning robot according to claim 2, characterized in that, The hierarchical control system adopts a computing and control separation architecture, comprising an upper-level computing unit and a lower-level motion control unit. The upper-level computing unit is equipped with a visual perception algorithm module for calculating the target's category and three-dimensional spatial coordinates. The visual perception algorithm module is connected to the lower-level motion control unit via a communication interface. The lower-level motion control unit integrates a dual-closed-loop PID controller, which is connected to an attitude sensor and a motor encoder. The attitude sensor is a gyroscope mounted on the chassis. The motor encoder is electrically connected to the drive motor and is used to adjust the output torque of the drive motor of the moving carrier module in real time when the system's center of gravity changes due to the operation of the execution module, so as to maintain the horizontal stability of the machine's attitude. The hierarchical control system has several waterproof chambers for installing the hierarchical control system components and the overall power supply.

7. An underwater garbage cleaning robot according to claim 6, characterized in that, The visual sensor is a wide-angle underwater camera; the visual perception algorithm module running in the host computer computing unit adopts a lightweight convolutional neural network, which uses GhostNetV2 as the backbone network and integrates a bidirectional feature pyramid network and a receptive field convolutional block attention module.

8. The operation control method for an underwater garbage cleaning robot according to any one of claims 1-7, characterized in that, It includes the following steps: S1. Stable diving: The depth sensor detects the water depth at the target location. When it is determined to be a diving depth, the buoyancy adjustment module is controlled to reduce the drainage volume. At the same time, the attitude sensor monitors the pitch angle and adjusts the center of gravity to make the robot dive vertically to the predetermined depth. S2, Visual Servo Navigation: The robot uses a tracked walking unit to navigate underwater. When the visual sensor identifies the target, the visual perception algorithm module calculates the target deviation and controls the differential speed of the two tracks through a dual closed-loop PID controller to achieve a straight approach to the target. S3, Flexible Adaptive Grasping: After reaching the work position, based on the target posture feedback from the vision, the work execution module is controlled to adjust the pitch angle and open the mechanical claw of the flexible end effector to envelop and grasp the target; during this process, the lower-level motion control unit continuously controls the anti-overturning torque caused by the forward shift of the center of gravity through the PID controller; S4. Continuous transmission and storage: After the target is grasped, the target is sent into the collection box of the material collection and transmission module by controlling the lifting of the linear lifting mechanism and the flipping of the robotic arm of the operation execution module. S5, Full-load Float Recovery: When the diffuse reflection sensor of the collection box detects that the collection box is full, the buoyancy adjustment module is controlled to increase the drainage volume, so that the robot floats to the water surface.

9. The operation control method for an underwater garbage cleaning robot according to claim 8, characterized in that, In S3, flexible adaptive grasping includes the following steps: S30. Based on the identified target geometric features, dynamically adjust the rotation angle of the power shaft of the second servo motor to adjust the gripping stroke opening of the mechanical claw. S31. During the mechanical gripper's grasping process, the elastic deformation of the TPU soft material is used to adapt to the surface irregularities of the target, and the gripping friction force is provided by the friction pads on the surface. S32. During the grasping process, when the movement of the flexible end effector's robotic arm causes the robot to tilt slightly, the attitude sensor, i.e., the gyroscope, senses the shift in the center of gravity. The gyroscope feeds back the pitch and roll angle variables to the PID controller, which then controls the drive motor of the track walking unit through the motor encoder to generate a reverse torque to counteract the overturning torque.

10. The operation control method for an underwater garbage cleaning robot according to claim 8, characterized in that, When the S5 is fully loaded and floats to the surface for recovery, the following steps are also taken: the robot uses its built-in positioning module to obtain the water surface position information; after floating to the water surface, it sends out a recovery signal through an audible and visual alarm device; after the robot recovers the waste, the recovery personnel unlock the unloading door of the collection box to dump the waste and reset the system.

Images