Ocean underwater robot for coral reef ecological protection

Abstract

The invention provides a marine underwater robot for coral reef ecological protection, and relates to the technical field of underwater robots, a power propulsion system is mounted on a main body frame, a wheel-paddle vector hybrid power architecture is adopted, underwater full-degree-of-freedom maneuvering and hovering are realized through six propellers, and the underwater robot is mounted on a main body frame; moving and parking on the land or on the hard bottom surface are achieved through four-wheel independent driving. The mechanical arm operation system comprises a four-degree-of-freedom mechanical arm, the control of the four-degree-of-freedom mechanical arm by the hand action of an operator is realized through a data glove, and a sample collection task is realized; the sensing monitoring system synchronously acquires multi-dimensional sensing data through a multi-sensor array; the control system comprises a sealed cabin and a main control unit arranged in the sealed cabin, the sealed cabin is fixed to the upper portion of the main body frame, the main control unit is in communication connection with the power propelling system, the mechanical arm operation system and the sensing monitoring system, control over the power propelling system and the mechanical arm operation system is achieved according to the multi-dimensional sensing number, and automatic obstacle avoidance is conducted.

Description

Technical Field

[0001] This invention relates to the field of underwater robot technology, and specifically to a marine underwater robot for coral reef ecological protection. Background Technology

[0002] Coral reefs comprise only 0.1% of the ocean, yet they support over 25% of marine life, possessing immense ecological and economic value. However, due to multiple threats, their global coverage has sharply declined over the past three decades. Therefore, it is necessary to strengthen coral reef restoration and protection, establish a sound investigation, assessment, and early warning monitoring system, understand the distribution, ecological status, and changing trends of coral reef ecosystems, register and map them, and gradually build a comprehensive operational system for coral reef ecosystem investigation and assessment.

[0003] Against this backdrop, high-resolution, long-term coral reef monitoring data has become a crucial support for protection and restoration efforts. Currently, coral reef monitoring mainly relies on divers conducting on-site surveys, remote sensing image analysis, and underwater robots (ROVs). Among these, ROVs, due to their advantages such as remote operation and the ability to carry a variety of sensors, are gradually becoming an important tool for coral reef monitoring.

[0004] Currently, most mainstream underwater ROVs on the market focus on operational capabilities in a single underwater environment, such as deep-water exploration and underwater structure inspection. These ROVs typically employ a closed hull design and are equipped with multiple thrusters to achieve six degrees of freedom of movement underwater. Some high-end models also feature robotic arms for grasping or sampling operations. However, they can only operate in underwater environments and cannot adapt to the needs of alternating land and water terrains such as the intertidal zone. Furthermore, existing ROVs lack effective autonomous obstacle avoidance capabilities, relying on operator vision, resulting in a very high risk of collision in complex environments.

[0005] Based on the aforementioned background technology, existing ROVs have the following significant drawbacks when applied to coral reef ecological protection: Existing ROVs have limited functionality and poor task adaptability: observation-type ROVs can only perform video recording and lack sample collection capabilities, while operational ROVs equipped with robotic arms often lack robust local intelligent recognition functions. Furthermore, general-purpose ROVs on the market are not specifically designed for coral reef protection and lack necessary environmentally friendly design features.

[0006] Existing ROVs have a limited mobility and cannot overcome the barriers between water and land environments, resulting in significant monitoring blind spots. Current ROVs are all designed for underwater propulsion, relying heavily on buoyancy. In coral reef ecosystems, critical intertidal zones and shallow reef platforms can cause traditional ROVs to run aground at low tide, rendering them inoperable; conversely, at high tide, these areas are covered by currents, rendering land-based mobile platforms ineffective. This incompatibility between water and land results in a lack of continuous monitoring data in both time and space for core coral reef degradation areas, failing to meet the protection requirements of all-weather, all-terrain monitoring.

[0007] The robotic arms have poor control precision, high learning costs, and are unable to complete delicate tasks: Most existing ROV robotic arms are controlled by joysticks, keyboards or buttons, which are complex and counterintuitive to operate. They are difficult to complete delicate grasping, sample collection or sensor deployment under complex water flow interference.

[0008] Current ROVs lack effective autonomous obstacle avoidance capabilities and rely heavily on operator vision, resulting in a high risk of collisions in complex environments. Most small and medium-sized ROVs currently rely heavily on operators manually judging and avoiding obstacles based on video feeds from front-end cameras. However, since cameras typically have a forward-facing view, there are blind spots on the left, right, and bottom sides of the ROV. In complex environments such as coral reefs, collisions are highly likely when moving laterally or approaching slowly. Furthermore, underwater visibility varies significantly; if the water is murky or lighting is insufficient, the video signal will fail, rendering the operator essentially operating blind and completely losing obstacle avoidance capabilities. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention proposes a marine underwater robot for coral reef ecological protection, comprising: a main frame, a power propulsion system, a robotic arm operation system, a sensing and monitoring system, and a control system. The main frame is composed of symmetrical side plate supports connected by horizontal plate supports and bottom supports. The propulsion system is installed on the main frame and adopts a wheel-propeller vector composite power architecture. It achieves underwater full-degree-of-freedom maneuvering and hovering through six thrusters, and achieves land or hard-bottom movement and parking through four-wheel independent drive. The robotic arm operation system includes a four-degree-of-freedom robotic arm, which is controlled by the operator's hand movements through a data glove to achieve the sample collection task. The sensing and monitoring system synchronously collects multi-dimensional sensing data through a multi-sensor array; The control system includes a sealed chamber and a main control unit located inside the sealed chamber. The sealed chamber is fixed to the upper part of the main frame. The main control unit is communicatively connected to the power propulsion system, the robotic arm operation system, and the sensing and monitoring system. It controls the power propulsion system and the robotic arm operation system based on the multi-dimensional sensor data and performs autonomous obstacle avoidance.

[0010] Furthermore, the sensing and monitoring system includes: an environmental perception sensor array, a positioning sensor, a vision sensor, and an ultrasonic ranging sensor.

[0011] Furthermore, the propulsion system includes: The vector thruster assembly consists of six thrusters, with two vertical thrusters symmetrically mounted on the side plate bracket and four horizontal thrusters installed at a 45° inward angle below the horizontal plate bracket. The wheel drive system consists of four independently driven waterproof motors and tires. The waterproof motors are respectively mounted at the four corners of the bottom of the main frame via motor mounting brackets.

[0012] Furthermore, the main control unit runs a deep learning-based target detection algorithm, performs adaptive anchor box calculation, image size standardization, and pixel value normalization preprocessing on the input video frames, extracts multi-scale features through the CSPDarknet backbone network, performs bidirectional feature fusion through the PANet neck network, and finally outputs through the decoupled head and performs non-maximum suppression processing to achieve real-time identification of underwater organisms.

[0013] Furthermore, the robotic arm operation system achieves motion control through a data glove. The data glove has a built-in MPU6050 sensor to collect three-axis acceleration and angular velocity data of the hand. The attitude is calculated by a complementary filtering algorithm, and the calculated hand attitude angle is linearly fitted and mapped to the PWM control signals of the servo motors of each joint of the robotic arm through the least squares method.

[0014] Furthermore, the complementary filtering algorithm performs the following steps: normalizing the accelerometer measurements; estimating the gravity vector using quaternions; calculating the cross product error between the measured gravity vector and the estimated gravity vector; correcting the gyroscope data using a PI controller; updating and normalizing the quaternions; and finally converting them to Euler angles to obtain the real-time hand posture.

[0015] Furthermore, the control system realizes autonomous obstacle avoidance function based on ultrasonic ranging sensor. The main control unit periodically reads ultrasonic ranging data, compares it with safe distance threshold after sliding average filtering, and triggers obstacle avoidance movement in the opposite direction when the measured value on either side is less than the threshold. The movement speed is negatively correlated with the measured distance within the safe threshold range. After two consecutive detections that the distance on both sides exceeds the threshold, the straight movement is resumed.

[0016] The present invention has the following technical effects: This invention upgrades the ROV from a single image acquisition tool into a comprehensive information platform capable of outputting specific environmental background and precise spatial location through the integration of multi-sensor arrays and data fusion, significantly enhancing its practical value in coral reef research and protection applications.

[0017] This invention pioneers a propeller-vector hybrid power architecture, employing a 6-propeller layout underwater to achieve omnidirectional precision maneuverability; and a 4-wheel independent drive system on land, breaking through the terrain limitations of tidal cycles, enabling switching from underwater to shore, and eliminating monitoring blind spots in the intertidal zone of coral reefs.

[0018] This invention uses data gloves to collect operator hand movements, enabling intuitive control of the 4-DOF robotic arm. It transforms robotic arm control from a cumbersome joystick-keyboard mode to an intuitive hand-mimicking action, enabling grasping actions such as sample collection. This significantly avoids secondary damage to coral reefs when sampling bleached areas.

[0019] This invention utilizes bidirectional ultrasonic ranging sensors to monitor distances to the sides and bottom blind spots in real time, and automatically triggers obstacle avoidance actions via a lower-level computer. It overcomes the limitations of underwater visibility and operator blind spots, achieving fully autonomous and rapid-response collision avoidance. This significantly improves the safety and reliability of the equipment in complex reef environments, protecting the ROV itself and, more importantly, the coral reefs. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0021] Figure 1 This is a side view of the ROV frame of the present invention; Figure 2 This is a front view of the ROV frame of the present invention; Figure 3 This is a bottom view of the ROV frame of the present invention; Figure 4 This is a schematic diagram of the tire motor device of the present invention; Figure 5 This is a complete front view of the ROV of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] (1) The high integration and intelligence of the multi-sensor array of underwater robots.

[0024] In this embodiment, the underwater robot is equipped with a high-definition camera system to collect underwater video streams. The low-latency video is transmitted back to the host computer via wired communication from the Raspberry Pi, and the host computer runs the YOLOv5 model, a deep learning-based object detection algorithm.

[0025] The YOLOv5 target detection algorithm adopts an end-to-end detection architecture. First, in the input stage, adaptive anchor box calculation is performed on video frames. K-means clustering is used to generate the optimal initial anchor box size to adapt to the underwater target morphology. Next, image size normalization is performed, scaling the input frames to 640×640 pixels proportionally using letterboxing. Simultaneously, pixel value normalization is performed to convert the intensity range to the [0,1] interval. During the training stage, Mosaic data augmentation technology is used to improve the model's robustness. The preprocessed data enters the backbone network CSPDarknet, which consists of convolutional layers, batch normalization layers, and SiLU activation functions. Through CSPNE... The system optimizes gradient flow and reduces computational complexity by combining a spatial pyramid pooling module with multi-scale features to effectively capture multi-granular features ranging from coral texture to overall morphology. The feature maps then enter the neck network PANet, where a top-down path fuses deep semantic information to enhance small target recognition capabilities. A bottom-up path then transmits precise localization information, forming a bidirectional feature pyramid structure. Finally, a decoupled head design is used in the head network to predict bounding box coordinates, target confidence, and class probability. The sigmoid function constrains the offset and supports multi-label classification. The final result is output using a non-maximum suppression algorithm, enabling real-time identification and analysis of underwater organisms such as corals and fish. The system can also acquire and save images and videos.

[0026] The underwater robot in this embodiment is equipped with an array of sensors, including GPS, depth sensors, temperature sensors, and water quality sensors, to continuously and synchronously collect data. GPS provides global location information, which can mark the location of image acquisition or sampling, enabling the ROV to perform horizontal analysis and comparison of the growth status of the same coral over time; the depth sensor calibrates the underwater depth; and the temperature and water quality sensors are used to record the micro-ecological environment of the coral reef.

[0027] When each frame of an image is acquired, the corresponding depth, temperature, and water quality data at that location are also recorded and saved, ensuring that each frame of an image is recorded in an accurate environmental background, providing highly reliable, multi-dimensional data for the observation and protection of coral reefs.

[0028] (2) The propeller-vector hybrid power architecture of underwater robots.

[0029] The underwater robot of this invention adopts a propeller-vector hybrid propulsion architecture, which combines the characteristics of standardized functional modules, modular installation, structural rigidity, and excellent hydrodynamic performance. The overall structural framework is as follows: Figures 1-5 As shown.

[0030] like Figure 1 , Figure 2 and Figure 3 The core load-bearing structure of the entire framework consists of symmetrical side plate supports on both sides, secured with screws to four horizontal plate supports at the front and rear, and a bottom support, forming a robust three-dimensional rectangular load-bearing structure. The side plate supports are constructed from a flat sheet of corrosion-resistant PE material with multiple standard mounting holes. The sheet has multiple standard equipment mounting areas and various threaded mounting hole arrays of M3, M4, and M6 sizes. These holes follow standard spacing, allowing users to directly install various third-party devices, such as robotic arms, sonar, tracers, and USBL beacons, without the need for custom adapters. This open design facilitates equipment installation and commissioning from the bottom of the ROV, greatly improving maintenance and configuration convenience. The horizontal plate supports and bottom support are 3D printed from high-fill PETG material, with reinforced joints to ensure extremely high torsional stiffness. The four horizontal plate supports not only serve a connecting function but also have pre-drilled standard mounting holes on their surfaces for mounting additional equipment, such as thrusters or other small sensors. The horizontal plate support has four vertically threaded holes at its four corners, which are securely fixed to the float cover with small screws. This allows for the loading of counterweight plates and buoyancy blocks to achieve precise buoyancy balance and center of gravity adjustment for the ROV. Two semi-circular brackets are fastened to the horizontal plate support with double bolts, facilitating the stable and horizontal placement of the sealed chamber. The sealed chamber is made of transparent, high-strength injection-molded material, allowing for quick assembly and disassembly, and is pressure-resistant up to 100 meters in water depth.

[0031] like Figure 3Six brushless DC thrusters are securely fixed in a vector layout using four M3 screws. Two vertical thrusters are mounted on two side plate supports, providing vertical lift and pitch torque. Four horizontal thrusters are mounted below the four horizontal plate supports 3 at a 45° inward angle, providing forward / reverse propulsion and yaw torque. This six-thruster layout gives the ROV multi-degree-of-freedom maneuverability: forward / reverse, lateral movement, heave, roll, and yaw; it enables precise vertical movement and hovering at any depth in the water; it allows for precise control of the ROV's roll and yaw angles, ensuring stable attitude in complex water currents; and it works in conjunction with the wheeled system to assist horizontal propulsion, providing additional horizontal thrust, such as for assisting propulsion on soft substrates or achieving rapid lateral translation.

[0032] like Figure 4 The tire-motor unit consists of a motor mounting bracket, a waterproof motor, a coupling, and a tire. Braking, speed regulation, forward and reverse rotation functions, as well as overcurrent, overvoltage, and overtemperature protection are achieved through signal feedback from a Hall encoder. The waterproof motor features a fully enclosed structure to prevent water from entering the motor, ensuring a long lifespan, strong torque for powerful operation, and reliable performance. The tire is connected to the coupling with hexagonal screws to secure it to the motor shaft. The motor has its own threaded holes for fastening to the motor bracket with screws. The motor bracket is secured to the four pre-drilled threaded holes at the corners of the side plate bracket 1 with high-strength screws, forming a stable mobile platform for land or hard water surfaces. They are primarily responsible for precise movement on the horizontal plane, including: forward and backward straight movement and speed adjustment: providing stable and efficient forward and backward movement capabilities, especially suitable for long-distance inspections on flat seabeds, underwater platform surfaces, or tank walls. Compared to relying solely on propellers, they consume less energy and provide more precise positioning; braking and parking: achieving rapid and stable braking through motor counter-torque, enabling the ROV to firmly adhere to the work surface in strong water currents or complex environments, providing an extremely stable working platform for tools such as robotic arms; steering: achieving flexible on-the-spot steering or cornering by controlling the different speeds of the wheels on both sides through differential speed control.

[0033] The advanced nature of this underwater robot based on a propeller-vector hybrid propulsion architecture lies in its ability to flexibly switch to the optimal propulsion mode according to complex mission scenarios and hydrological environments. When patrolling open waters, the system relies entirely on vector thrusters to achieve full-degree-of-freedom maneuverability. In intertidal zones or shallow reef areas, especially when the water depth is insufficient after low tide, the underwater robot can lower its chassis and switch to a pure wheeled drive mode, thereby achieving efficient and stable horizontal movement in extremely shallow waters. The wheeled structure not only provides a highly energy-efficient mode of movement, but its physical contact with the base also provides a stable platform for the underwater robot, reducing body sway under water currents or robotic arm operations, creating favorable conditions for tasks such as high-definition observation and precise sampling. In summary, this system uses a central controller to perform real-time calculation of motion commands and manage motion status, realizing on-demand calling and functional complementarity between wheeled drive and vector propulsion: the wheel system dominates efficient translation and operational stability on shallow water and hard bottom surfaces; the propeller system is fully responsible for floating, attitude adjustment, and current-resistant hovering. These two elements combine to create a seamless connection between the two major operational domains of benthic and suspended water, ultimately enabling underwater robots to possess the three core advantages of high adaptability, high stability, and high energy efficiency in performing tasks in complex underwater environments.

[0034] (3) Motion control of the robotic arm of an underwater robot The intuitive control function of the underwater robot's robotic arm in this invention is mainly achieved through the following technical solutions. In terms of hardware composition, the system includes a data glove and a 4-DOF robotic arm. The data glove serves as the main hand controller, with key components being an STM32F103C8T6 microcontroller and an MPU6050 sensor, integrating a three-axis gyroscope and a three-axis accelerometer. It communicates with a host computer via a serial port, and the host computer then sends the parsed instructions to the underwater robot's slave device. Each joint of the robotic arm is driven by a servo motor. The core of motion control lies in mapping the operator's hand posture to the movement of the robotic arm joints. The specific implementation steps are as follows: The MPU6050 sensor collects raw data of the operator's hand movements in real time using a high-frequency sampling rate of 100Hz, including triaxial acceleration and triaxial angular velocity. After the collected data is preprocessed by moving average filtering to suppress noise interference, a complementary filtering algorithm is used to calculate the attitude. Specifically, the real-time attitude angles of the hand in space are obtained through the following steps: roll angle (Roll), pitch angle (Pitch), and yaw angle (Yaw).

[0035] First, the accelerometer measurement value (a) x , a y , a z Normalization is performed to obtain the gravity vector in the body coordinate system. , , ),Right now Then, the gravity vector (v) in the body coordinate system is estimated using the current quaternions (q0, q1, q2, q3). x , v y , v z Quaternions are a mathematical tool used to describe the rotation of an object in three-dimensional space. They consist of four parameters (q0, q1, q2, q3) and can more effectively avoid gimbal lock problems than Euler angles. q0 represents the scalar part (related to the rotation angle), and q1, q2, q3 represent the vector part (related to the rotation axis).

[0036] Then calculate the cross product error (e) between the measured gravity vector and the estimated gravity vector. x , e y , e z ),Right now PI control is used to measure the gyroscope values ​​(g) x , g y , g z The angular velocity is corrected to obtain the corrected angular velocity. , , ) Where K p For proportional gain, K i For integral gain; update the quaternion based on the corrected angular velocity, i.e. Where T is the sampling period, and the quaternions are normalized; finally, the quaternions are converted into Euler angles, where the roll angle is... Pitch angle The calculation formula is as follows: Yaw angle Since there is no accelerometer reference, there will be accumulated error, which is updated by integrating the gyroscope z-axis data.

[0037] Finally, based on the least squares method and linear fitting calculation, the proportionality coefficient K is obtained, establishing a linear mapping relationship between the hand posture angle and the motion of each joint of the robotic arm. That is, the pitch angle change corresponds to the pitch motion of the upper arm joint of the robotic arm, the roll angle change corresponds to the rotation motion of the lower arm joint of the robotic arm, and the yaw angle change corresponds to the opening and closing of the robotic gripper. in, , Let represent the attitude angle measured in the i-th measurement and the PWM duty cycle of the corresponding servo rotation, respectively, and n be the number of measurements; This represents the average value of each attitude angle and the corresponding PWM duty cycle of the servo rotation after n measurements.

[0038] Based on this mapping relationship, the attitude angle data is converted into the target control quantity of each servo motor of the robotic arm, which is ultimately reflected in the duty cycle of the PWM signal. The conversion formula can be simplified to: PWM duty cycle = K * current angle + neutral position duty cycle, where the current angle is the calculated real-time attitude angle; the neutral position duty cycle is the PWM signal duty cycle required to keep the servo motor in the center of its range of motion.

[0039] The lower-level microcontroller outputs the calculated PWM duty cycle signal to each servo motor of the robotic arm, driving it to rotate precisely to the target angle, thereby realizing the correspondence between hand movements and robotic arm posture, and replicating the operator's hand movements.

[0040] (4) Autonomous obstacle avoidance based on ultrasonic ranging The autonomous obstacle avoidance function of the underwater robot in this embodiment is mainly achieved through the following technical solutions. In terms of hardware composition, ultrasonic ranging modules are symmetrically installed on the left and right sides of the underwater robot body. This invention uses the DYP-L08*-V2.0 ultrasonic underwater obstacle avoidance sensor. This ranging module establishes a data connection with the core control unit STM32F407ZGT6 microcontroller via serial communication protocol. The microcontroller drives the ROV's thruster motor through an electronic speed controller. In terms of control logic and process, obstacle avoidance control is a closed-loop feedback process. Its specific implementation steps are as follows: The microcontroller reads the data from the left and right ultrasonic sensors at 200ms intervals, and obtains D after moving average filtering (removing outliers and taking the average of the most recent 5 valid data). left and D right Then the microcontroller compares the real-time distance data with the preset safe distance threshold (denoted as D). safe (Compare)

[0041] When the measured value on one side is less than the safety threshold, obstacle avoidance is triggered, and the vehicle moves in the opposite direction. Its speed is negatively correlated with the measured distance between 0 and the safety threshold. After obstacle avoidance, the vehicle maintains a two-cycle detection period before resampling. If the distance on both sides exceeds D consecutively, the obstacle avoidance is triggered. safe At that time, exit obstacle avoidance and resume straight driving.

[0042] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A marine underwater robot for coral reef ecological protection, characterized in that, include: The main frame, power propulsion system, robotic arm operation system, sensor monitoring system, and control system are all included. The main frame is composed of symmetrical side plate supports connected by horizontal plate supports and bottom supports. The propulsion system is installed on the main frame and adopts a wheel-propeller vector composite power architecture. It achieves underwater full-degree-of-freedom maneuvering and hovering through six thrusters, and achieves land or hard-bottom movement and parking through four-wheel independent drive. The robotic arm operation system includes a four-degree-of-freedom robotic arm, which is controlled by the operator's hand movements through a data glove to achieve the sample collection task. The sensing and monitoring system synchronously collects multi-dimensional sensing data through a multi-sensor array; The control system includes a sealed chamber and a main control unit located inside the sealed chamber. The sealed chamber is fixed to the upper part of the main frame. The main control unit is communicatively connected to the power propulsion system, the robotic arm operation system, and the sensing and monitoring system. It controls the power propulsion system and the robotic arm operation system based on the multi-dimensional sensor data and performs autonomous obstacle avoidance.

2. The marine underwater robot for coral reef ecological protection according to claim 1, characterized in that, The sensing and monitoring system includes: an environmental perception sensor array, a positioning sensor, a vision sensor, and an ultrasonic ranging sensor.

3. The marine underwater robot for coral reef ecological protection according to claim 1, characterized in that, The propulsion system includes: The vector thruster assembly consists of six thrusters, with two vertical thrusters symmetrically mounted on the side plate support and four horizontal thrusters arranged in a specific configuration. It is installed at an inward tilt below the horizontal plate support; The wheel drive system consists of four independently driven waterproof motors and tires. The waterproof motors are respectively mounted at the four corners of the bottom of the main frame via motor mounting brackets.

4. The marine underwater robot for coral reef ecological protection according to claim 1, characterized in that, The main control unit runs a deep learning-based target detection algorithm, performs adaptive anchor box calculation, image size standardization and pixel value normalization preprocessing on the input video frames, extracts multi-scale features through the CSPDarknet backbone network, performs bidirectional feature fusion through the PANet neck network, and finally outputs through the decoupled head and performs non-maximum suppression processing to achieve real-time identification of underwater organisms.

5. The marine underwater robot for coral reef ecological protection according to claim 1, characterized in that, The robotic arm operation system achieves motion control through a data glove. The data glove has a built-in MPU6050 sensor to collect three-axis acceleration and angular velocity data of the hand. The attitude is calculated by a complementary filtering algorithm, and the calculated hand attitude angle is linearly fitted and mapped to the PWM control signals of the servo motors of each joint of the robotic arm through the least squares method.

6. The marine underwater robot for coral reef ecological protection according to claim 5, characterized in that, The complementary filtering algorithm performs the following steps: normalizing the accelerometer measurements; estimating the gravity vector using quaternions; calculating the cross product error between the measured gravity vector and the estimated gravity vector; and correcting the gyroscope data using a PI controller. Update and normalize the quaternion; finally convert it to Euler angles to obtain the real-time hand pose.

7. The marine underwater robot for coral reef ecological protection according to claim 2, characterized in that, The control system achieves autonomous obstacle avoidance based on ultrasonic ranging sensors. The main control unit periodically reads ultrasonic ranging data, filters it by moving average, and compares it with a safe distance threshold. When the measured value on either side is less than the threshold, it triggers obstacle avoidance movement in the opposite direction. The movement speed is negatively correlated with the measured distance within the safe threshold range. After two consecutive detections where the distance on both sides exceeds the threshold, it resumes straight movement.

Images