A deep sea robotic factory system

Abstract

The application provides a deep-sea robot factory system, comprising: a factory body subsystem for deep-sea landing and carrying; an environment sensing subsystem for acquiring deep-sea environment information; a decision unit subsystem taking the deep-sea environment information, preset task target and robot body state as input, outputting an assembly process of the robot, and inputting the deep-sea environment information, preset task target and assembled module robot, and outputting motion control to execute the task target; and a deep-sea in-situ robot assembly platform, comprising: a deep-sea robot module library comprising a plurality of functional modules for executing the task target; an assembly execution mechanism subsystem for selecting the functional modules according to the assembly process and assembling the functional modules, and the assembled module robot executing the task target according to the motion control; and a communication subsystem comprising a wireless communication module for communication between the functional modules. The system has the capabilities of environment sensing, decision making and automatic assembly.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of deep-sea equipment and intelligent manufacturing, in particular to a deep-sea robot factory system. BACKGROUND

[0002] Current deep-sea operating robots are mainly lowered through a water surface platform, and their structures and functions are integrated on the water surface or onshore factory. However, due to the highly uncertain deep-sea environment, such as complex terrain, strong ocean current, low visibility, and large temperature and pressure limits, current deep-sea exploration operations mainly rely on various pressure-resistant deep-sea robots, such as remote-operated vehicles (ROVs), autonomous underwater vehicles (AUVs), and manned submersibles (HOVs). These traditional robot systems usually adopt a single-body or limited reconfigurable design, but when dealing with the complexity of deep-sea extreme environments and the diversity of tasks, they face limitations such as easy damage to structure, low task flexibility, high communication dependence, low autonomy, and high cost.

[0003] A support table with adjustable support for underwater robot assembly is disclosed in Chinese Patent No. CN214081296U, which includes a box body, an assembly platform, a slide rail, two supports, two motors, a transmission box, a fixed arm, a movable arm, and a clamping plate. It can realize assembly operation and debugging of robots in different positions. However, its structure can only be used for ground operation and does not have the ability to operate in deep-sea extreme environments.

[0004] A manned submersible automatic release docking device and operation method are disclosed in Chinese Patent No. CN113830268B, which mainly includes a fixed base, a lifting steel ball, a movable base, an auxiliary mechanism, a lifting cylinder, and a magnet. It can realize automatic release and docking of the manned submersible during operation. However, the device must be carried on the manned submersible and cannot work independently. It can only release and recover the operation target and cannot adjust the structure of the operation target underwater.

[0005] Modular robotics technology, which enhances the flexibility and adaptability of deep-sea equipment, is an effective way to overcome the aforementioned challenges. Theoretically, it allows for dynamic combination and reconfiguration based on mission requirements. Patent CN120269588A provides a deep-sea self-reconfigurable robot system that can assemble robots into different configurations by assembling modules to meet different scenarios and tasks. However, existing modular robot technology is still in its early stages of development and faces several problems: modules rely on pre-assembly, requiring physical connection, system integration, and functional testing on the water surface. Once deployed to the deep sea, their configuration is fixed, making in-situ dynamic reconfiguration impossible based on the actual deep-sea environment. Furthermore, modular robots lack sufficient intelligence; assembly primarily relies on preset programs or remote operator control, lacking the ability to autonomously understand the environment, plan tasks, and make decisions based on real-time, multimodal perception (acoustic, optical, force, and tactile senses) of the deep-sea environment (terrain, obstacles, targets, ocean currents, and the robot's own state).

[0006] Currently, communication between deep-sea modules primarily relies on wired connections via physical connectors (such as wet-plug connectors). This method typically involves assembling all modules on land or on the mother ship's deck before the entire system is submerged to operate in the deep-sea environment. However, under high pressure, the connectors and their cables become weak points in the system's mechanical structure, easily damaged by pulling, squeezing, or collisions during deployment, retrieval, or in complex seabed environments. Furthermore, physical connectors offer poor maintenance and flexibility; adjustments to module functions or mechanisms often require redesigning the connector structure and location. To address these limitations, some deep-sea systems are attempting to replace physical connections with wireless communication technologies. Acoustic communication (ultrasound) is currently the most widely used deep-sea wireless communication method. However, its extremely low bandwidth, speed, and high latency severely restrict its application in high-speed, reliable inter-module communication. Therefore, a method or device is needed that eliminates physical connectors, adapts to extreme deep-sea environments, provides high-speed communication capabilities, and offers flexibility and maintainability for inter-module communication.

[0007] Therefore, there is an urgent need for a robot platform with deep-sea landing capability, environmental perception capability, modular assembly capability, autonomous task judgment capability, and in-situ assembly capability, as well as an intelligent system that enables efficient and reliable data transmission between modules, so as to realize the assembly and construction of robot modules on the seabed to meet dynamic task requirements. Summary of the Invention

[0008] The purpose of this invention is to provide a deep-sea robot factory system with environmental perception, decision-making and automatic assembly capabilities. By acquiring and analyzing deep-sea environmental information, it can make judgments on the combination of different functional modules and autonomously complete the assembly of deep-sea robots, thereby improving the flexibility and intelligence of deep-sea operations.

[0009] To achieve the above object, the present application provides the following technical solutions.

[0010] A deep-sea robot factory system, the system comprising:

[0011] A factory body subsystem for deep-sea landing and carrying an environment perception subsystem, a decision unit subsystem, a deep-sea in-situ robot assembly platform and a communication subsystem;

[0012] An environment perception subsystem for obtaining deep-sea environment information;

[0013] A decision unit subsystem taking the deep-sea environment information obtained by the environment perception subsystem, a preset task target and a robot body state as inputs, outputting an assembly process of a module robot; and taking the deep-sea environment information, the preset task target and an assembled module robot configuration as inputs, outputting motion control of the module robot to execute the task target;

[0014] A deep-sea in-situ robot assembly platform comprising a deep-sea robot module library and an assembly execution mechanism subsystem:

[0015] The deep-sea robot module library comprises a plurality of functional modules for executing a task target, the functional modules comprising a sensing module, a power module, a camera module and an operation module;

[0016] The assembly execution mechanism subsystem selects functional modules from the deep-sea robot module library according to the assembly process of the module robot decided by the decision unit subsystem and assembles the functional modules according to the assembly process to obtain an assembled module robot, the assembled module robot executing the task target according to the motion control;

[0017] The communication subsystem comprises a wireless communication module for communication between the functional modules.

[0018] The factory body subsystem provided by the present application is in a square cage structure, the outer frame is made of corrosion-resistant high-strength alloy or composite material, such as titanium alloy or high-strength carbon fiber reinforced composite material, and has corrosion resistance and pressure resistance.

[0019] The environment perception subsystem comprises:

[0020] A camera unit for illumination and image acquisition and identifying seafloor topography, sediment state and ocean current information according to the acquired images;

[0021] A temperature-salinity-depth sensor for obtaining the temperature, salinity and water depth of the deep-sea surrounding environment in real time;

[0022] A chemical perception sensor for obtaining geochemical information in the local seafloor water / sediment.

[0023] Further, the camera unit is a deep-sea self-contained intelligent light camera unit, which provides directional lighting and image acquisition in dark seabed and can realize identification and classification of seabed current flow direction and sediment type; the geochemical information includes heavy metal, sulfide, pH, etc.

[0024] The decision unit subsystem includes a language model LLM, an encoder, an analyzer, an inferrer, a generator, and a planner: the encoder extracts features from deep-sea environment information to obtain a feature vector, the analyzer analyzes a preset task target to obtain semantic embedding of the task target, and then the feature vector, the semantic embedding of the task target, and the robot ontology state are spliced to generate an environment-task joint perception graph; the language model LLM and the inferrer evaluate the task performance of different functional module combinations in completing the preset task target in the current deep-sea environment according to the environment-task joint perception graph, and output an assembly process of the functional modules.

[0025] Further, the environment-task joint perception graph includes: a vector grid map, a local environment high-precision map centered on the platform, containing seabed topography elevation, obstacle position and contour, detected structure position, ocean current vector field, represented as attribute vectors within grid cells; semantic labels, identifying and labeling key environmental elements; task target: converting the preset task target into dense vector representation through the analyzer, and splicing or cross-attention fusion with the features of the vector grid map; robot ontology state, taking the pose state of the deep-sea in-situ robot assembly platform, the available functional modules in the deep-sea robot module library, and the initial or current robot configuration as state vectors input; the assembly process includes: a list of required functional modules, a connection topology relationship diagram between functional modules, and a functional module assembly sequence planning.

[0026] Wherein: the key environmental elements include hydrothermal vents, seabed target equipment, marine organisms, rocks and minerals, etc., as well as their confidence, position and key attributes; the preset task target can be sampling or maintenance or inspection or exploration, and specifically can be collecting samples near the hydrothermal vent, repairing equipment valves, and exploring regional topography; the module library state includes the types and quantities of available functional modules, as well as the remaining power of energy and control modules.

[0027] Further, after the module robot is assembled, the language model LLM and the inferrer perform motion planning and reasoning according to the module robot configuration and performance parameters, and output motion control; the motion control includes: global and local path planning, considering terrain, obstacles, and ocean currents, to plan a safe and efficient path from the current position to the work point; work posture planning, planning the trajectory of the module robot to reach the target pose for specific task targets and considering the kinematics and stability of the robot; module coordination strategy, generating coordination control instructions for multi-module collaborative work.

[0028] The performance parameters of the module robot include kinetic parameters, maximum speed and sensor field of view; the multi-module cooperative operation can coordinate the resistance of the current for multiple propellers, and the corresponding coordinated control instruction is thrust distribution.

[0029] Further, the energy and control module is divided into a micro energy module, a medium energy module and a large capacity energy module according to the battery capacity, such as lithium batteries with different battery capacities, and is used for communication and control of other functional modules; the power module includes several propellers; the camera module is divided into a positioning camera module and a general camera module according to functions; the operation module includes a sampling device (such as a mechanical gripper), a maintenance device or an exploration device; the sensing module includes a temperature-salinity-depth sensor and a chemical sensing sensor, which are used to sense the temperature, salinity, depth and concentration of various organic matters in the deep sea environment.

[0030] Further, the energy and control module and the functional module of the deep-sea robot factory system are encapsulated by high molecular materials, and the internal components are optimally arranged to ensure reliable operation in the deep-sea high-pressure environment.

[0031] The assembly execution mechanism subsystem provided by the application includes at least one linear guide rail module and a set of mechanical arm systems, which are used to grab, transport and assemble functional modules into module robots; the assembly execution mechanism subsystem has multiple assembly modes: the module can be assembled by a mechanical arm or a linear guide rail alone, or the module (functional module) can be grabbed, pushed out, positioned and spliced by the mechanical arm and the linear guide rail in a cooperative mode; the splicing mode adopts magnetic docking. Further, the assembly execution mechanism subsystem pushes the module to the target position accurately through the linear guide rail, and completes precise insertion, screwing, magnetic attraction or buckle assembly by the mechanical arm.

[0032] As an implementation manner, the assembly execution mechanism subsystem includes a screw rod propelling assembly module, which includes a screw rod propelling motor, a linear guide rail, a push plate, a lifting mechanism and a linear slider mounted on the linear guide rail; the limiting push plate fixed to the linear slider moves along the linear guide rail under the drive of the screw rod propelling motor, pushes out the functional module and splices it; the lifting mechanism vertically lifts the module robot after splicing to the starting position of the preset task target.

[0033] When the above-mentioned assembly execution mechanism subsystem is used, the deep-sea robot module library includes a plurality of module warehouses fixed at the tail of the screw linear guide rail, and the functional modules are placed on each module warehouse.

[0034] As another implementation manner, the assembly execution mechanism subsystem comprises a mechanical arm assembly module, the mechanical arm assembly module comprises a mechanical arm, a rotary table guide rail assembly and a rotary table rotating motor, the mechanical arm is used for grabbing the functional module, and the rotary table guide rail assembly is used for driving the functional module to rotate to complete multi-direction assembly; the rotary table guide rail assembly comprises a screw rod pushing motor, a linear slide, a rotary table and a rotary table rotating motor, the linear slide pushes the rotary table to move along the guide rail under the driving of the screw rod pushing motor, and the rotary table realizes multi-surface rotating positioning of the functional module under the driving of the rotating motor.

[0035] Further, after the current task target is executed, or the functional module fails, or the deep sea environment changes dramatically, the module robot returns to the deep sea in-situ robot assembly platform, and the decision unit subsystem is reconfigured according to the current deep sea environment information or / and a new task target and a current robot body state.

[0036] The deep sea in-situ robot assembly platform provided by the application can realize in-situ assembly and deployment of the robot module in the complex deep sea environment, and can select appropriate functional modules for combination based on the operation task and the environmental condition, and has high modularity, automation and environmental adaptability.

[0037] The wireless communication module comprises a communication module body arranged at an interface area between the functional modules of the deep sea equipment; the communication module body is embedded with a wireless communication circuit board comprising a micro control unit, a WIFI chip, a power regulating circuit, an antenna system and a battery.

[0038] Further, the antenna system of the communication module is a directional patch antenna, and the design frequency band is 2.4GHz or 5GHz.

[0039] The wireless communication module provided by the application has high structural integration and perfect functions, further comprises a WIFI chip, a radio frequency matching circuit, a micro control unit and a power supply battery module. The WIFI chip is used for transmitting and receiving high-frequency wireless signals, the radio frequency matching circuit optimizes the signal transmission efficiency and stability, the micro control unit is responsible for logical control functions such as data protocol analysis, node management and link scheduling, and the battery module provides independent and reliable energy guarantee, so that the wireless communication module has certain autonomous operation capability.

[0040] Further, the wireless communication module as a whole is packaged with a high polymer potting material with low dielectric loss to form a whole structure resistant to deep-sea high pressure; the high polymer potting material is a high polymer material with low dielectric loss. The selected potting material needs to have the following characteristics: stable dielectric constant, small loss, ensuring signal penetration without distortion; excellent mechanical strength, resistant to at least 11000 meters of water depth (about 110 MPa) pressure environment; corrosion and seawater environment aging resistant, ensuring long-term stable operation of the communication module; the communication module body is integrally formed by a mold casting process, suitable for deep-sea environments below 11000 meters; the functional modules provided by the application are connected through wireless communication technologies such as wifi or lora, and electromagnetic waves are transmitted between the functional modules formed by the high polymer potting material, realizing efficient cooperation.

[0041] The wireless communication module provided by the application supports three forms of face-to-face wifi communication, wired high-speed bus, underwater acoustic communication and laser communication, and has automatic link switching capability. The communication device supports one-to-one, one-to-many and many-to-many communication topologies, and has channel switching, adaptive frequency band adjustment and multi-module synchronous control capabilities.

[0042] One-to-one master-slave communication mode; one-to-many broadcast control mode; many-to-many networking cooperation mode, all communications can be managed and interfered through scheduling protocols, compatible with different underwater task requirements, greatly improving the overall communication flexibility and efficiency of the system. The communication module supports automatic channel adaptation and power adjustment, limits the signal leakage range based on near-field coupling design, and effectively suppresses the influence of water absorption and interference sources. The thickness and material of the packaging shell of the communication parties are optimized through EM simulation, so that the loss of radio frequency signals is minimized when penetrating the potting material.

[0043] The wireless communication module provided by the application has the following advantages: reliability improvement: eliminating mechanical connectors and their failure points, improving communication stability in deep-sea environment; good pressure resistance and sealing performance: the overall potting structure can withstand high water pressure and high corrosion environment for a long time, avoiding communication failure and equipment water ingress; high communication freedom: supporting multiple communication topologies, breaking through the point-to-point limitation of traditional wired methods; compact structure, easy to integrate: the communication module is small and compact, suitable for different deep-sea equipment interface requirements.

[0044] The wireless communication module realizes a deep-sea functional module communication method, including the following steps: after each module is powered on, the communication module performs identity recognition and frequency matching; a face-to-face short-distance WIFI communication link is established; instruction and data transmission is performed; according to the channel monitoring result, the communication frequency band and transmission power are automatically adjusted.

[0045] The decision unit subsystem and the assembled modular robot can interact information through underwater acoustic communication or optical communication, so that the decision unit subsystem can realize the effect of commanding the modular deep-sea robot.

[0046] The system provided by the application adopts a high-pressure self-contained packaging process: the internal core electronic devices of the module are filled with resin and arranged in an optimized structure, effectively resisting deep-sea high pressure, and ensuring reliable operation of each unit.

[0047] The application further provides an assembling and control method of the modular robot in a deep-sea in-situ environment, comprising the following steps:

[0048] Step 1: The environment perception subsystem obtains deep-sea environment information, including seabed topography, sediment state, ocean current information, temperature, salinity, water depth, geochemical information in water / sediment, and obtains the robot body state and task target, wherein the task target is selected from sampling, maintenance, inspection or exploration;

[0049] Step 2: The deep-sea environment information, the preset task target and the robot body state are input into the decision unit subsystem, and the decision unit subsystem outputs the assembly process of the functional module, wherein the assembly process comprises: a required functional module list, a functional module connection topological relationship diagram and a functional module assembly sequence planning;

[0050] Step 3: The assembly execution mechanism subsystem selects the corresponding functional module from the deep-sea robot module library according to the assembly process output by the decision unit subsystem, and assembles according to the assembly process, to obtain the assembled modular robot;

[0051] Step 4: The decision unit subsystem outputs the motion control according to the deep-sea environment information, the task target and the configuration of the assembled modular robot, wherein the motion control comprises: global and local path planning, considering the terrain, obstacles, ocean currents, planning a safe and efficient path from the original position to the work point; work posture planning, planning the trajectory of the modular robot to reach the target pose according to the specific task target and considering the kinematics and stability of the modular robot; and module coordination strategy, generating coordinated control instructions for multi-module cooperative work;

[0052] Step 5: The modular robot is separated from the deep-sea in-situ robot assembly platform after receiving the motion control of the decision unit subsystem, and executes the corresponding task target.

[0053] The deep-sea robot factory system continuously perceives the changes of the environment and the task state, and if the reconstruction condition (such as task change, module failure, environmental mutation, etc.) is triggered, the system automatically returns to the assembly platform and starts dynamic re-planning and re-assembly according to step 2.

[0054] The specific structure and implementation method of the subsystem involved in each step are the same as those of the deep-sea robot factory system, and will not be described here.

[0055] Compared with the prior art, the present application has the following excellent effects:

[0056] The present application automatically generates the most suitable robot configuration for the current environment and task by combining the environmental perception subsystem with the decision unit subsystem and the deep-sea in-situ robot assembly platform, realizing the true "environment is the task, and the task is the assembly". Specifically, the decision unit subsystem provided by the present application combines the environmental perception subsystem to output the assembly process and motion control planning, and the deep-sea in-situ robot assembly platform realizes the closed-loop autonomous assembly of the robot in the deep-sea in-situ environment according to the assembly process, and supports adaptive reconfiguration of the module robot body, realizing autonomous dynamic assembly and reconfiguration of the module robot in the deep-sea extreme environment without water surface pre-assembly or manual intervention, completely solving the defects of fixed configuration of traditional systems and inability to respond to environmental changes, and significantly improving the operation safety.

[0057] The energy and control module and the functional module provided by the present application have high pressure resistance, high sealing, and high integration, and support hot plug connection, improving the flexibility and reliability of on-site construction; the deep-sea in-situ robot assembly platform provided by the present application supports mechanical arm assembly, linear guide assembly, and various assembly forms of mechanical arm and linear guide cooperative matching, and can flexibly select the assembly path according to the module type and space layout.

[0058] The present application reduces the dependence on water surface communication through the wireless communication module, significantly improves the operation safety and continuity, and enables the communication closed loop between the subsystems of the deep-sea robot factory.

[0059] The assembly and control method provided by the present application can realize intelligent autonomous assembly task and motion control planning of the module robot in the deep-sea in-situ environment. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 A structure diagram of a deep-sea robot factory system provided for embodiment 1;

[0061] Figure 2 A structure diagram of an assembly execution mechanism subsystem provided for embodiment 1;

[0062] Figure 3 A structure diagram of a wireless communication module provided for embodiment 1;

[0063] Figure 4 A structure diagram of a wireless communication module supporting one-to-many communication capability provided for embodiment 1;

[0064] Figure 5A flow chart of an assembly and control method of a module robot in a deep-sea in-situ environment for embodiment 1 is provided.

[0065] Figure 6 A structural diagram of another deep-sea robot factory system for embodiment 2 is provided.

[0066] Figure 7 A structural diagram of another assembly actuator subsystem for embodiment 2 is provided.

[0067] Figure 8 A structural diagram of another deep-sea robot factory system for embodiment 3 is provided.

[0068] Figure 9 A structural diagram of a wireless communication module supporting a many-to-many communication structure for embodiment 6 is provided.

[0069] Figure 10 A structural diagram of a wireless communication module supporting another many-to-many communication structure for embodiment 7 is provided.

[0070] Figure 11 A structural diagram of a wireless communication module of a directional package one-to-one communication structure for embodiment 8 is provided.

[0071] Wherein: 1, factory frame; 2, camera unit; 3, temperature-salinity-depth sensor; 4, module robot warehouse; 5, mechanical arm; 6, robot splicing platform; 7, chemical sensing sensor; 8, decision unit subsystem; 9, turntable assembly; 10, energy and control module; 11, sensing module; 12, power module; 13, camera module; 14, operation module; 15, lead screw linear guide; 16, lifting module; 17, linear slide; 18, limit push plate; 19, module warehouse; 20, module warehouse door; 21, lead screw drive motor; 22, module robot; 23, turntable drive motor; 24, turntable. DETAILED DESCRIPTION

[0072] The deep-sea robot factory system and the intelligent assembly and control method of a module robot in a deep-sea in-situ environment based on embodied intelligence provided by the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0073] Embodiment 1

[0074] As shown in Figure 1 , the present embodiment provides a deep-sea robot factory system comprising:

[0075] a factory body subsystem for a deep-sea landing and carrying environment sensing subsystem, a decision unit subsystem, a deep-sea in-situ robot assembly platform, and a communication subsystem;

[0076] an environment sensing subsystem for obtaining deep-sea environment information;

[0077] The decision-making unit subsystem takes the deep-sea environment information, preset task objectives, and robot body state acquired by the environmental perception subsystem as input, and outputs the assembly process of the modular robot; and takes the deep-sea environment information, preset task objectives, and the assembled modular robot configuration as input, and outputs the motion control of the modular robot to execute the task objectives.

[0078] Deep-sea in-situ robot assembly platform, including a deep-sea robot module library and an assembly actuator subsystem:

[0079] The deep-sea robot module library includes several functional modules for performing mission objectives, including a sensing module, a power module, a camera module, and an operation module.

[0080] The assembly execution mechanism subsystem selects functional modules from the deep-sea robot module library and assembles them according to the assembly process of the modular robot determined by the decision unit subsystem, thereby obtaining the assembled modular robot. The assembled modular robot executes the task objective according to motion control.

[0081] The communication subsystem includes a wireless communication module for communication between functional modules.

[0082] In this embodiment, as Figure 1 As shown, the plant body subsystem includes a cage-shaped plant frame 1, a camera unit 2 and a temperature, salinity and depth sensor 3 mounted on the two side frames of the plant frame 1, and a chemical sensing sensor 1 and a decision subsystem 8 mounted on the top of the plant frame 1.

[0083] In this embodiment, the decision-making unit subsystem 8 includes a large language model reasoning unit and a control instruction generation module, which are used to generate assembly schemes based on environmental information and task requirements: it embeds a large language model (LLM) core computing unit; it automatically plans module combination according to perception data and preset task descriptions; it outputs assembly instructions to the assembly execution mechanism subsystem to realize the automatic construction of task-oriented robots; the large language model reasoning unit has a built-in knowledge graph and semantic strategy for environment-structure-task mapping, which can generate the optimal robot configuration scheme according to task semantics and environmental parameters; the module scheduling order during the assembly process is generated by the large language model reasoning module to generate a task tree structure, and the optimal path is planned according to the module library surplus and spatial layout.

[0084] In this embodiment, the deep-sea robot factory system has a state monitoring and feedback mechanism, which can continuously monitor the changes in the surrounding environment through the environmental perception module and optimize the assembly strategy in real time; the module splicing adopts standardized multi-point docking ports, including power on-off interfaces, data bus interfaces and mechanical positioning guide structures, supporting hot plugging; the decision unit subsystem 8 can dynamically reconstruct the triggering mechanism, that is, the module robot can independently realize intelligent closed-loop operation in the case of task change, module failure or environmental change, etc. The decision unit subsystem 8 adopts a three-layer architecture of multi-modal perception-physical rule embedding-task chain cooperation, and can realize the embodiment computing power deployment on the deep-sea edge device through lightweight strategy.

[0085] In this embodiment, as shown in Figure 1 and Figure 2 , the deep-sea in-situ robot assembly platform includes a module robot warehouse 4, a mechanical arm 5 and a rotary table assembly 9, and the rotary table assembly 9 includes a lead screw linear guide 15, a linear slide 17, a lead screw propulsion motor 21, a rotary table drive motor 23 and a rotary table 24.

[0086] Before work, the functional modules (including the energy and control module 10, the sensing module 11, the power module 12, the camera module 13 and the operation module 14) are placed in the grooves of the module robot warehouse 4, and the functional modules are fixed with the module robot warehouse 4 by magnetic attraction. When starting work, the lead screw propulsion motor 21 pushes the linear slide 17 to make the energy and control module 10 as the center module move to the assembly position with the rotary table 24, the mechanical arm carries a functional module needed to be used after receiving the instruction transmitted by the upper computer, and assembles it with the energy and control module 10 on the rotary table 24, and then takes another functional module on the module robot warehouse 4 for assembly. At this time, the lead screw propulsion motor 1 pushes the linear slide 17, the rotary table rotation motor 23 rotates the rotary table 24, and the mechanical arm 5 cooperates with each other to make the next installation surface of the energy and control module 10 face the mechanical arm 5, and leave a suitable assembly space to start the assembly work, and so on until the module robot is assembled according to the assembly process output by the decision unit subsystem 8 to meet the task requirements. Before the module robot starts to work, the lead screw propulsion motor 21 pushes the linear slide 17 to move the combined module robot to the end away from the mechanical arm 5 to prevent interference.

[0087] During the assembly process, the rotation of the rotary table 24 and the propulsion of the linear slide 17 ensure the multi-surface contactability of the energy and control module 10, which improves the assembly efficiency and accuracy. This assembly method is suitable for complex three-dimensional robot assembly, and has high flexibility and freedom.

[0088] In this embodiment, as shown in Figure 3As shown, including WIFI chip, radio frequency matching circuit, micro control unit and battery. Wireless communication module and functional module adopt "face to face" type assembly, the surface of the communication module is provided with magnetic positioning structure, which is used for improving the close fitting degree and signal coupling efficiency between modules. The specific implementation is: each functional module reserves a communication window area in the structure design, installs an embedded wireless communication module, and stably transmits information under the condition of close contact between modules by controlling the direction and distance of the signal, forms a "pore-free" wireless connection with structural integration, and replaces the module communication mode which depends on the connector with "face to face" type short distance WIFI wireless transmission; the whole shell is covered with a high-molecular encapsulation material with low dielectric loss to form a high-strength pressure-resistant package; the surface can be provided with alignment marks and positioning magnetic attraction structure to facilitate the module face to face cooperation.

[0089] As shown in Figure 4 The embodiment provides a deep sea inter-module wireless communication system structure supporting one-to-many communication capability. The central functional module is provided with a master communication module group, and the peripheral functional modules are respectively provided with slave communication module groups. The whole is assembled in a "face to face" manner to form a high-sealing and close communication interface network.

[0090] Each wireless communication module group is packaged in a six-sided unit structure, which is integrally encapsulated and formed by a high-molecular material with low dielectric loss, so as to ensure that the structure does not deform in a deep sea pressure environment of 110 MPa or more, and has excellent electromagnetic wave transmission performance.

[0091] The master communication module is located at the center position and serves as a communication scheduling control node. The internal units include a micro controller unit (MCU), a WIFI radio frequency chip, a directional face-to-face antenna array and a power regulation module. The peripheral slave communication modules serve as controlled nodes to complete signal receiving and response.

[0092] In the working process of the system, the master module simultaneously broadcasts signals and issues instructions to multiple slave modules through the face-to-face directional antenna, realizing a "one-to-many" parallel information transmission structure. The green arrow direction is shown in the figure, representing the data interaction path between the master module and the peripheral slave modules.

[0093] The one-to-many communication structure has the following advantages: clear communication topology and compact physical structure: adaptive to cellular or modular installation structure, suitable for deep sea equipment integrated arrangement; high channel utilization rate: the master module has multiple channel scheduling capabilities and can dynamically allocate receiving time slots for each slave module; high reliability: face-to-face arrangement effectively reduces signal leakage and bit error rate; strong adaptability: the number of peripheral slave modules can be flexibly adjusted to support the module access requirements in different task scenarios.

[0094] As shown in Figure 5As shown, the deep-sea robot factory system provided by the embodiment provides a method for assembling and controlling a modular robot in a deep-sea in-situ environment, which comprises the following steps:

[0095] (1) After the overall modular robot assembly platform lands on the seabed, the environment perception subsystem observes the environment around the platform and obtains relevant multi-modal environment data, for example, obtains deep-sea scene image information through a camera combined with a sonar, obtains three-dimensional flow velocity and flow direction information through a Doppler velocity log (DVL), obtains water depth, temperature and salinity and calculates seawater density information through a conductivity-temperature-depth (CTD) measurement system, and obtains platform landing attitude information through an inertial measurement unit.

[0096] (2) Data preprocessing is performed on the information obtained in step (1): time and space alignment, time stamp interpolation, sliding window buffering, dynamic time warping, and other methods are used to synchronize asynchronous sensor data of different sampling rates (for example, camera information at a frequency of 30 Hz and IMU information at a frequency of 100 Hz), and sensor data of different perspectives or positions are mapped to a unified coordinate system (sonar point cloud, camera perspective, and DVL vector are uniformly projected to a three-dimensional coordinate system with the platform base as the center); filtering to remove obvious noise (such as image noise, IMU drift, and obviously erroneous data), and enhancing some data (such as contrast and visual effect enhancement of camera images and sonar images); optional compression and dimensionality reduction processing, especially for high-dimensional data such as images and point clouds, to reduce computational load and optimize memory; standardization processing to map data of different dimensions and ranges to a unified scale; so as to ensure that the data input to the decision subsystem is clean and accurate, has low redundancy, high consistency, and deep-sea task adaptability.

[0097] (3) The decision unit subsystem 8 adopts a three-layer architecture of multi-modal perception-physical rule embedding-task chain cooperation: the embodied intelligent large model is deployed in the intelligent decision unit, Nvidia Jetson is used as the hardware deployment platform and communicates with the sensing unit to accept the input of multi-modal sensing data, multi-modal perception encoder is used to extract features from the sensing data (such as PointNet++ for extracting three-dimensional obstacle maps from sonar point clouds and labeling passable areas; LSTM network for predicting flow rate time series changes to extract flow field vector features, and ResNet for extracting seabed terrain, structure outline and biological distribution features), all feature vectors and semantic embedding of task targets are spliced to generate an environment-task joint perception graph, which is a structured representation containing: vector grid map, local environment high-precision map centered on the platform, containing seabed terrain elevation, obstacle position and outline, detected structure position, current vector field, etc., represented as attribute vectors within grid cells; semantic labels, identified and labeled key environmental elements such as hydrothermal vents, seabed target equipment, marine organisms, and rock minerals, as well as their confidence, position and key attributes; task target: the pre-set task target (such as collecting samples near the hydrothermal vent, repairing the A device valve, and exploring the B area terrain) is converted into a dense vector representation through an analyzer and spliced or cross-attention fused with the features of the environment graph; robot body state, the pose state of the assembled platform itself, the state of the available module library (type, quantity, remaining power), the initial or current robot configuration, etc. are input as state vectors.

[0098] (4) The physical engine inference layer embeds a lightweight differentiable fluid simulator, a configuration-performance mapping neural network and a module performance knowledge base, which converts deep sea dynamics constraints into a computation graph in real time to help large model inference; the task chain cooperation layer decomposes the task into assembly and motion control stages through a dynamic decomposer:

[0099] (4-1) The decision unit subsystem 8 reasons out the optimal module robot configuration according to the graph input, the reasoning process uses a pre-trained policy network based on reinforcement learning and a predictor based on graph neural network GNN to evaluate the expected performance (success rate, efficiency, energy consumption, stability) of different module combinations (such as main frame type, number and vector of thrusters, sensor suite) in completing the current task under the current environment. The output of the execution decision includes: the list of required functional modules, the connection topology graph between functional modules, and the assembly sequence planning of functional modules;

[0100] (4-2) Based on the decision-making assembly scheme, the decision unit subsystem 8 further generates detailed executable assembly instruction sequence. Including: which module is grabbed by the mechanical arm of the assembly platform, which precise docking position is moved to under the 6DOF pose, which connection operation is performed (such as electromagnetic locking, hydraulic bolt), sequential dependency, obstacle avoidance path in the process (considering the assembly platform mechanical arm workspace and environmental obstacles). The output format is the robot operating system (ROS) compatible action sequence or task level instruction, which is parsed and executed by the assembly execution platform controller;

[0101] (4-3) After waiting for the assembly platform to complete the module robot splicing, the decision unit subsystem 8 is activated again, and the precise configuration and performance model (such as dynamic parameters, maximum speed, sensor field of view) of the currently assembled module robot are input. The decision unit subsystem 8 performs motion planning reasoning and outputs the optimal control instruction sequence, including: global and local path planning, considering terrain, obstacles, currents, planning a safe and efficient path from the current position to the work point; work posture planning, planning the trajectory of the module robot to reach the target pose for specific tasks (such as maintenance, sampling) and considering the kinematics and stability of the robot; module coordination strategy, for multi-module cooperative work (such as multiple thrusters coordinating against currents), generate coordinated control instructions (such as thrust distribution);

[0102] (4-4) After receiving the control instruction sequence, the module robot starts the motion state of the robot and detaches from the assembly platform to perform related work tasks according to the preset content of the control instruction. When the module robot needs to change tasks after executing the current task, or individual module failure occurs, and the environment changes dramatically (such as sudden strong currents, earthquakes causing terrain changes), etc. The module robot automatically returns to the assembly platform to perform module robot reconstruction according to the preset instructions, and feeds back the current state (environmental information, module robot state, new task target) to the decision unit subsystem 8. The decision unit subsystem 8 re-reasoning generates a new assembly scheme and motion control instruction, executes the re-assembly process, and then the reconstructed robot receives and executes the motion control instruction generated by the large model to carry out the work task.

[0103] In addition, the deployment and optimization of the decision subsystem in the present application are as follows: the hardware platform is preferably deployed on a high-performance embedded AI computing platform such as NVIDIA Jetson AGX Orin, which meets the requirements of power consumption, volume and computing power in the deep-sea pressure-resistant cabin. In terms of software, it is based on deep learning frameworks such as PyTorch and TensorFlow, and integrates ROS 2 (Robot Operating System) for inter-module communication (with sensors, actuators) and task scheduling. In terms of model acceleration, techniques such as model pruning, quantization, and knowledge distillation are used to optimize the pre-trained embodied intelligent large model for lightweight, ensuring that the inference speed meets the requirements of practical applications (such as assembly decision-making within seconds and motion planning within seconds).

[0104] The present embodiment is a patrol type deep-sea robot assembly application: a deep-sea robot factory is deployed around a certain submarine hydrothermal vent. The environment perception subsystem collects information on strong corrosive ocean currents and high temperature gradients, and the decision unit subsystem 8 generates an assembly process: the "low-power high-agility patrol robot" configuration consists of a miniature energy and control module, a small propulsion unit (small thruster), and a camera module; the operation mode is to sequentially push each functional module from the storage position to the splicing position by the assembly execution mechanism subsystem, and complete the functional module docking assembly; the functional effect is that after the assembly is completed, the module robot carries out crack patrol and thermal flow image collection tasks near the vent according to the motion control through its own propulsion unit.

[0105] Embodiment 2

[0106] The difference between this embodiment and embodiment 1 is that this embodiment uses a lead screw propulsion type assembly module.

[0107] As shown in Figure 6 and Figure 7 , the lead screw propulsion type assembly module includes a center module warehouse 19, a small module warehouse, a large module warehouse, a lead screw propulsion motor 21, a lead screw linear guide 15, a warehouse door 20, a center module, a functional module, a lifting mechanism 16, a linear slide 17, and a limit push plate 18.

[0108] Working principle of the above assembly mechanism:

[0109] Sixth, the linear guide rail 15, the center module warehouse 19, four small module warehouse, large module warehouse, lifting mechanism 16 are fixed with the bottom plate of the platform frame, the screw rod propulsion motor 21 is fixed at the tail of the linear guide rail 15 for advancing the limiting push plate 18, the limiting push plate 18 is fixed on the linear slide 17 of the linear guide rail 15, the warehouse door 20 is fixed on the center module warehouse 19, small module warehouse and large module warehouse for protecting functional modules: wherein, the energy and control module 10 is arranged in the center module warehouse 19, the sensor module 11, power module 12, camera module 13 and operation module 14 are placed in the small module warehouse and large module warehouse according to size.

[0110] When working, all warehouses close the warehouse door 20, after receiving the instruction transmitted by the upper computer, the in-situ self-assembly is carried out according to the assembly process output by the decision unit subsystem 8: the screw rod propulsion motor 21 starts to drive the limiting push plate 18 on the linear guide rail 15 to move linearly, first, the center module (energy and control module 10) is pushed out from the center module warehouse 19 to the center position through the limiting push plate 18, during the process, the warehouse door 20 is opened by the functional module and remains open state; then, the appropriate functional module is selected from the small module warehouse or / and the large module warehouse for linear propulsion, the dispersed functional modules are pushed onto the lifting mechanism 16; the screw rod propulsion motor 21 stops working, the module robot automatically completes the overall assembly; after the assembly is completed, the screw rod propulsion motor 21 starts to reverse, the limiting push plate 18 retreats into the warehouse of each functional module, at this time, the module robot has completed the assembly and will not move with the limiting push plate 18, so the assembled robot is separated from the limiting push plate 18, after the limiting push plate 18 is completely retreated, the screw rod propulsion motor 21 stops working; the lifting mechanism 16 starts to work, lifting the assembled module robot to a certain height; the assembled module robot starts to execute the task target according to the motion control output by the decision unit subsystem 1.

[0111] Embodiment 3

[0112] The difference from embodiment 1 is that this embodiment simultaneously uses mechanical arm assembly assembly platform and screw rod propulsion type assembly module, as shown in Figure 8 .

[0113] By combining the screw rod propulsion type assembly and the mechanical arm assembly, the assembly precision and the functional module installation freedom are ensured, the complex robot assembly can be realized, and the system reliability is improved.

[0114] Embodiment 4

[0115] The difference from embodiment 1 is that this embodiment is a sampling type deep sea robot assembly application.

[0116] In this embodiment, the deep-sea robot factory is deployed in a sea floor sediment belt rich in metal nodules; the chemical sensor detects an abnormally high concentration of manganese and iron elements; the system task requirement is to collect sediments and rocks; the functional modules are: a medium-sized energy and control module + a high-torque propulsion unit (high-torque thruster) + an operating unit module (with a mechanical gripper) + a camera module; the assembly process: the assembly mechanism execution subsystem successively completes the grabbing, positioning and splicing assembly of each functional module to obtain a module robot; the splicing result: the module robot can perform point sampling, grab sediment blocks and transmit images to the water surface support platform.

[0117] Embodiment 5

[0118] The difference from Embodiment 1 is that this embodiment is a deep-sea robot assembly application for cable maintenance.

[0119] In this embodiment, the deep-sea robot factory is deployed in a certain deep-sea cable laying area, and the task target is to perform routine inspection work on the cable path, and obtain cable layout state and environmental image information; the functional modules are: a large-capacity energy and control module + a high-precision positioning camera module + a double thruster module; the assembly path: each functional module is pushed out to the predetermined position in turn by the lead screw linear guide 15, and the mechanical arm 5 completes grabbing, positioning and precise splicing under the instruction of the decision unit subsystem 8, realizing the collaborative assembly of the lead screw linear guide 15 and the mechanical arm 5; after assembly is completed, the module robot carries out deep-sea cable patrol tasks, and records image information in real time, and can report relevant data to the decision unit subsystem 8 if an abnormal state is found.

[0120] Embodiment 6

[0121] The difference from Embodiment 1 is that this embodiment is a many-to-many communication structure.

[0122] As shown in Figure 9 , this embodiment provides a many-to-many communication structure suitable for deep-sea equipment modules. Each functional module in the system is configured with a wireless communication module having the same structure, and each wireless communication module has bidirectional communication capability, can dynamically establish a communication link between different functional modules, and realizes collaborative transmission and control of data.

[0123] The many-to-many communication system has the following structural characteristics: all communication modules are embedded with radio frequency antennas, microcontrollers (MCU), WIFI or LoRa communication chips and power regulation circuits; each module adopts a "face-to-face" installation layout to ensure the signal coupling strength of the short-distance wireless link; the green arrow indicates the multi-directional communication path between modules, and any module can act as a transmitting end or a receiving end, and has equivalent interconnection capability.

[0124] During the operation of the system, each wireless communication module can dynamically establish a point-to-point, broadcast, or relay link according to a task. Each wireless communication module has the functions of frequency band selection, transmission power adjustment, and channel collision detection, and is suitable for a multi-node communication scenario in a complex underwater operation.

[0125] The technical advantages of the many-to-many communication mode include: highly distributed, centerless structure: avoids global failure caused by master node failure, improves system robustness; communication link reconfiguration: suitable for flexible combination of modules or fault isolation in operation; strong parallel transmission capability: supports concurrent multiple communication sessions, improves overall communication bandwidth utilization; good task adaptability: suitable for underwater multi-machine cooperative tasks such as data acquisition cooperation, synchronous control, etc.

[0126] Embodiment 7

[0127] The difference between this embodiment and Embodiment 6 is that this embodiment is another many-to-many communication structure.

[0128] As shown in Figure 10 , this embodiment provides a vertically serially arranged many-to-many communication structure, which is suitable for wireless communication scenarios in a long and narrow deep sea equipment system (such as a cable accessory module chain, a columnar monitoring platform, etc.).

[0129] Each communication module is assembled in the vertical direction in a “face-to-face” manner, and the physical structure is compact. Each module has bidirectional communication capability and can establish a communication link with the upper and lower modules, or even non-adjacent modules, to realize a chain and cross-layer hybrid multi-point connection. As shown in Figure 10 , the green arrows mark the multi-directional communication paths between the modules, including: adjacent point-to-point data transmission; cross-node long-distance communication (relay or direct communication); upper and lower control and state feedback.

[0130] The vertical chain communication structure has the following characteristics: high module decoupling: any module can be started and stopped independently without affecting other nodes in the link; strong link reconfiguration capability: supports relay mode communication and automatically bypasses failed modules; good spatial adaptability: suitable for deep sea systems with limited vertical arrangement device space; anti-interference and directionality optimization: the internal antenna array of the communication module can be enhanced in the up and down direction to improve the efficiency and stability of the chain communication.

[0131] When the system is running, the communication protocol supports dynamic link state detection and re-routing mechanism, which can ensure that the overall communication network has high availability and low delay characteristics in complex deep sea environments.

[0132] Embodiment 8

[0133] The difference between this embodiment and Embodiment 1 is that this embodiment is a directional packaging one-to-one communication structure.

[0134] AsFigure 11 As shown, the embodiment provides a surface-to-surface wireless communication mode suitable for different structural arrangement requirements, and particularly for the installation of irregularly spaced functional modules in deep-sea equipment, by changing the form of the polymer encapsulation body, the directional configuration and long-distance fitting of the signal path are realized.

[0135] The structure includes a functional module unit provided with a wireless communication module group at both ends, and an intermediate encapsulation channel connecting the two. The channel is made of a low dielectric loss polymer material, and has good electromagnetic wave transmission ability and deep-sea pressure resistance.

[0136] Specifically, the communication module group is located at both ends of the encapsulation body and directly faces each other through surface-to-surface fitting; the intermediate connecting section can be designed as a straight line, a bend, an oblique direction or any other shape according to the actual installation space; the overall structure is integrally formed by mold injection or extrusion process, avoiding any mechanical joints or sealing points to ensure water tightness; the signal is stably coupled and propagated through the intermediate polymer material conducting area.

[0137] The green arrow in the figure represents the transmission path of the WIFI or LoRa signal in the encapsulation material, and can realize one-to-one high-speed and low-attenuation data communication within a range of tens of centimeters to 1 meter.

[0138] The technical advantages of the structure design include: strong form adaptability; the communication structure can be flexibly customized according to the module layout, and can adapt to complex structures such as vertical, oblique and embedded; avoid rigid coupling restrictions; compared with traditional connectors, the wireless communication path is not constrained by mechanical position; high encapsulation integrity; the encapsulation layer plays a role in structure, sealing and electromagnetic shielding, and improves the integration level of the module; low-loss and high-reliability communication; due to the low dielectric loss of the encapsulation material, the signal transmission attenuation is small and the communication quality is stable.

[0139] The present application can independently configure functional modules according to task requirements, and has high reliability in module identification, precise assembly and task deployment in deep-sea environment, and is suitable for deep-sea exploration, operation and long-term operation in various complex marine environment tasks. The present application can significantly improve the flexibility, autonomy and intelligent manufacturing capability of deep-sea operation.

Claims

1. A deep-sea robotic factory system, characterized in that, The system includes: The factory body subsystem is used for deep-sea landing and carrying of the environmental perception subsystem, decision-making unit subsystem, deep-sea in-situ robot assembly platform and communication subsystem; The environmental sensing subsystem is used to acquire deep-sea environmental information; The decision-making unit subsystem takes the deep-sea environment information, preset task objectives, and robot body state acquired by the environmental perception subsystem as input, and outputs the assembly process of the modular robot; and takes the deep-sea environment information, preset task objectives, and the assembled modular robot configuration as input, and outputs the motion control of the modular robot to execute the task objectives. Deep-sea in-situ robot assembly platform, including a deep-sea robot module library and an assembly actuator subsystem: The deep-sea robot module library includes several functional modules for performing mission objectives, including an energy and control module, a sensing module, a power module, a camera module, and an operation module. The assembly execution mechanism subsystem selects functional modules from the deep-sea robot module library and assembles them according to the assembly process of the modular robot determined by the decision unit subsystem, thereby obtaining the assembled modular robot. The assembled modular robot executes the task objective according to motion control. The communication subsystem includes a wireless communication module for communication between functional modules.

2. The deep-sea robotic factory system according to claim 1, characterized in that, The environmental perception subsystem includes: The camera unit is used for illumination and image acquisition, and identifies seabed topography, sediment conditions and ocean current information based on the acquired images; Temperature, salinity, and depth sensors are used to acquire real-time data on the temperature, salinity, and water depth of the deep-sea environment. Chemosensing sensors are used to acquire geochemical information in local seabed waters / sediments.

3. The deep-sea robotic factory system according to claim 1, characterized in that, The decision-making unit subsystem includes a language model (LLM), an encoder, an analyzer, an inferencer, a generator, and a planner. The encoder extracts features from deep-sea environmental information to obtain feature vectors. The analyzer analyzes the preset task objectives to obtain the semantic embedding of the task objectives. Then, the feature vectors, the semantic embedding of the task objectives, and the robot's body state are concatenated to generate an environment-task joint perception map. The language model (LLM) and the inferencer evaluate the task performance of different functional module combinations in completing the preset task objectives in the current deep-sea environment based on the environment-task joint perception map, and output the assembly process of the functional modules.

4. The deep-sea robotic factory system according to claim 3, characterized in that, The environment-task joint perception map includes: a vector grid map, a high-precision local environment map centered on the platform, containing seabed topography elevation, obstacle location and outline, detected structure location, ocean current vector field, represented as attribute vectors within grid cells; semantic tags, identifying and labeling key environmental elements; task objectives: the preset task objectives are converted into dense vector representations by an analyzer and then stitched together or fused with cross-attention features of the vector grid map; robot body state, taking the pose state of the deep-sea in-situ robot assembly platform itself, the functional modules available in the deep-sea robot module library, and the initial or current robot configuration as state vector inputs; the assembly process includes: a list of required functional modules, a topology diagram of connections between functional modules, and a planning of the functional module assembly sequence.

5. The deep-sea robotic factory system according to claim 3, characterized in that, After the modular robot is assembled, the Language Model (LLM) and inference engine perform motion planning and inference based on the modular robot's configuration and performance parameters, and output motion control. The motion control includes: global and local path planning, taking into account terrain, obstacles, and ocean currents, to plan a safe and efficient path from the current position to the work point; work posture planning, for specific task objectives and considering the robot's kinematics and stability, to plan the trajectory of the modular robot to reach the target posture; and module cooperation strategy, for multi-module collaborative operations, to generate coordinated control commands.

6. The deep-sea robotic factory system according to claim 1, characterized in that, The energy and control module is divided into micro energy and control module, medium energy and control module and large capacity energy and control module according to the battery capacity, and is also used for communication and control function modules; the power module includes several thrusters; the camera module is divided into positioning camera module and ordinary camera module according to function; the operation module includes sampling device, maintenance device or exploration device; the sensing module includes temperature, salinity and depth sensor and chemical sensing sensor.

7. The deep-sea robotic factory system according to claim 1, characterized in that, The assembly execution mechanism subsystem includes a screw-driven assembly module, which includes a screw-driven motor, a linear guide rail, a push plate, a lifting mechanism, and a linear slider mounted on the linear guide rail. The limiting push plate, fixed to the linear slider, moves along the linear guide rail under the drive of the screw-driven motor to push out and assemble the functional modules. The lifting mechanism vertically lifts the assembled module robot to the preset starting position of the task target.

8. The deep-sea robotic factory system according to claim 1, characterized in that, The assembly execution mechanism subsystem includes a robotic arm assembly module, which comprises a robotic arm, a turntable guide rail assembly, and a turntable rotary motor. The robotic arm is used to grasp functional modules, and the turntable guide rail assembly is used to drive the functional modules to rotate to complete multi-directional assembly. The turntable guide rail assembly includes a lead screw drive motor, a linear slider, a turntable, and a turntable rotary motor. The linear slider pushes the turntable to move along the guide rail under the drive of the lead screw drive motor, and the turntable realizes multi-faceted rotational positioning of the functional modules under the drive of the rotary motor.

9. The deep-sea robotic factory system according to claim 1, characterized in that, The wireless communication module includes a communication module body, which is located in the interface area of ​​each energy and control module and functional module; the communication module body has an embedded wireless communication circuit board, including a microcontroller unit, a WIFI chip, a power regulation circuit and an antenna system and a battery.

10. A method for assembling and controlling a modular robot in a deep-sea in-situ environment, comprising the following steps: Step 1: The environmental perception subsystem acquires deep-sea environmental information, including seabed topography, sediment status, ocean current information, temperature, salinity, water depth, geochemical information in water / sediments, as well as the robot's status and task objectives, wherein the task objectives are selected from sampling, maintenance, inspection or exploration. Step 2: Input the deep-sea environment information, the preset mission objectives and the robot's body status into the decision unit subsystem. The decision unit subsystem outputs the assembly process of the functional modules. The assembly process includes: a list of required functional modules, a topology diagram of the connection between functional modules, and a planning of the assembly sequence of functional modules. Step 3: The assembly execution mechanism subsystem selects the corresponding functional modules from the deep-sea robot module library according to the assembly process output by the decision unit subsystem, and assembles them according to the assembly process to obtain the assembled modular robot; Step 4: The decision-making unit subsystem outputs motion control based on deep-sea environmental information, mission objectives, and the assembled modular robot configuration. Motion control includes: global and local path planning, considering terrain, obstacles, and ocean currents, to plan a safe and efficient path from the original position to the work site; work posture planning, considering the kinematics and stability of the modular robot and planning the trajectory of the modular robot to reach the target posture for specific mission objectives; and module cooperation strategy, generating coordinated control commands for multi-module collaborative operations. Step 5: After receiving motion control from the decision-making unit subsystem, the modular robot detaches from the deep-sea in-situ robot assembly platform and executes the corresponding task objectives.

Images