A mechanical arm digital twin system for deep space exploration

Abstract

The application discloses a kind of mechanical arm digital twin systems for deep space exploration, it is related to digital twin simulation technical field.The present application includes physical layer, data acquisition and transmission layer, digital twin model layer and application layer, the physical layer includes mechanical arm body and sensor network, wherein sensor network is used to collect mechanical arm body operating state information;The data acquisition and transmission layer includes data acquisition equipment and data transmission network, wherein data acquisition equipment is connected with sensor network, data transmission network respectively.The present application can guide actual mechanical arm body to carry out relevant action by virtual simulation data by completing the construction of mechanical arm physical layer, data acquisition and transmission layer, digital twin model layer, application layer, and can complete virtual-real body configuration comparison through visual interface, meet the digital twin needs of various spacecraft operations in future deep space exploration field.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of BB, and particularly relates to a mechanical arm digital twin system for deep space exploration. BACKGROUND

[0002] The digital twin technology is a technology of real-time synchronous interaction between a digital model and an actual system. Through the digital twin technology, a virtual model highly consistent with a real mechanical arm can be constructed, and simulation and optimization can be performed in a virtual environment. The technology can provide a highly controllable environment, so that engineers and researchers can safely, economically and efficiently test and optimize the mechanical arm.

[0003] The existing digital twin system is not accurate and intelligent in mechanical arm twin control, cannot realize high restoration of the motion state, is difficult to realize friendly human-computer interaction control, and does not have an intelligent motion control function based on arm historical data and running state. The application can overcome these defects, realize more accurate control, higher intelligent degree and better human-computer interaction effect, etc. based on the deep space exploration task scene by constructing a specific system architecture and using related algorithms.

[0004] Therefore, the application provides a mechanical arm digital twin system for deep space exploration. SUMMARY

[0005] The application aims to provide a mechanical arm digital twin system for deep space exploration to solve the problems in the background technology.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme: a mechanical arm digital twin system for deep space exploration, comprising a physical layer, a data acquisition and transmission layer, a digital twin model layer and an application layer, the physical layer comprising a mechanical arm body and a sensor network, wherein the sensor network is used for acquiring mechanical arm body running state information;

[0007] The data acquisition and transmission layer comprises a data acquisition device and a data transmission network, wherein the data acquisition device is connected with the sensor network and the data transmission network respectively, and the data acquisition device is used for converting the running state information collected by the sensor network into an electrical signal and transmitting the electrical signal to the data transmission network for storage;

[0008] The digital twin model layer comprises a geometric model construction module, a physical model construction module, and a model calibration and verification module. The geometric model construction module is used to create a high-precision geometric model of the robot arm body. The physical model construction module is used to construct a physical model of the robot arm body based on the geometric model, combined with the dynamics, kinematics principles, and material properties of the robot arm body. The model calibration and verification module is used to calibrate the precision of the constructed geometric model and physical model according to the actual running state data of the robot arm obtained by the data acquisition and transmission layer.

[0009] The application layer is specifically a visualization module. The visualization module is connected with the data acquisition and transmission layer and the digital twin model layer, and is used to receive the running state information of the robot arm body and display the motion trajectory, workspace range, and current task information of the robot arm body in real time.

[0010] Further, the sensor network is installed on the robot arm body, and the sensor network comprises joint position sensors, force sensors, vision sensors, and speed sensors.

[0011] Further, the data acquisition device is used to condition, sample, and digitize the analog signals collected by the sensor network to convert them into a digital format suitable for network transmission. The data acquisition device is specifically an integrated data acquisition card, and the data acquisition device is equipped with a redundant power module.

[0012] Further, the data transmission network uses a standard network transmission protocol to transmit the collected data from the physical layer to the digital twin model layer.

[0013] Further, the geometric model and the physical model are deeply connected through a multi-dimensional and multi-level data interaction link and an adaptive interface, and are used to realize digital twin control, specifically as follows:

[0014] The physical layer collects the current joint angle, angular velocity, force size and direction, and real-time spatial attitude running data of the physical model, and transmits them to the digital twin model layer where the geometric model is located via the data acquisition and transmission layer according to the defined data format and transmission protocol, so that the geometric model reproduces the shape and position of the physical model in space at the moment.

[0015] Further, the modeling deduction and twin simulation in the model calibration and verification module are implemented by using the Solidworks and matlab simulink platforms, specifically as follows:

[0016] (1) Construct a model assembly: a robot arm body assembly is designed by using Solidworks;

[0017] (2) Generate a geometric model URDF file;

[0018] (3) The specific method of building a twin model;

[0019] (4) Twin control interface construction.

[0020] Further, the mechanical arm body assembly is designed by Solidworks, specifically as follows:

[0021] (71) First, in the top-down design stage, the overall architecture and structural strength distribution of the mechanical arm are planned, and the top-level framework of the joint layout, link shape and end effector of the mechanical arm is designed;

[0022] (72) Then switch to the bottom-up design stage, create the basic components of the mechanical arm one by one, and use Solidworks to design the part structure, including joint shell, link structure, sensor structure;

[0023] (73) In the assembly stage, dynamically check the fitting accuracy and motion interference of each component, and real-time backtrack to correct the assembly relationship.

[0024] Further, the specific method for generating a geometric model URDF file is:

[0025] (81) After completing the construction of the mechanical arm assembly in Solidworks, use Solidworks and URDF conversion plug-in or conversion script to realize the innovative conversion from Solidworks assembly data to URDF format;

[0026] (82) During the conversion process, in addition to the conventional geometric shape, size, mass attribute information, the functional role, reliability requirement and environmental adaptability characteristics of each component in deep space exploration are also integrated into the URDF file;

[0027] (83) The flexible connection characteristics between the connecting components are described in the URDF file, including simulating the elastic deformation, damping effect and connection stiffness at the joint;

[0028] (84) Embed intelligent data tags in the URDF file to automatically update the state information of the corresponding components according to the real-time running data and simulation results of the mechanical arm;

[0029] (85) Use the method based on XML extension to add custom elements and attributes in the URDF file.

[0030] Further, the specific method of building a twin model;

[0031] (91) In the Matlab Simulink environment, based on the previously imported mechanical arm URDF file, the multi-physics modeling technology is deeply integrated, the mechanical arm rigid body dynamics model is built using the module library of Simulink, and the corresponding physical effect sub-model is constructed for the special environmental factors of deep space exploration. At the same time, the motion deviation of the dynamics model is compensated and corrected by using the microgravity compensation algorithm, and the adaptive control algorithm is used to adjust the control parameters in real time, as follows:

[0032] (91.1) The microgravity compensation algorithm is as follows:

[0033] The microgravity compensation algorithm is based on the dynamics modeling, and the dynamics model of the mechanical arm is described as follows:

[0034]

[0035] In the formula, τ is the joint torque, M(q) is the inertia matrix, is the Coriolis force and centrifugal force matrix, G(q) is the gravity compensation term, and F(q) is the friction torque;

[0036] In the microgravity environment, the influence of the gravity compensation term is significantly reduced, and the improved compensation term is expressed as:

[0037] G micro (q)=g micro ·J T (q)m

[0038] In the formula, g micro is the microgravity acceleration, J T is the Jacobian matrix, and m is the mass vector of each link of the mechanical arm;

[0039] (91.2) Adaptive control algorithm, the specific implementation steps are as follows:

[0040] Dynamic modeling: build a real-time dynamics model of the mechanical arm, describe the nonlinear relationship between joint torque and angle, speed, acceleration;

[0041] Parameter estimation: real-time estimation of model parameters using recursive least squares or gradient descent algorithm, including inertia, friction and joint stiffness;

[0042] Error analysis and feedback: calculate the deviation of the current output and the expected output through the error function, the error is defined as: e(t) = x d (t)-x a (t), x d is the desired trajectory, and x a is the actual trajectory;

[0043] Control law update: based on Lyapunov stability criterion, design control law: τa = τ n + K e e(t), K e is an error gain matrix, τ n is a nominal value;

[0044] (92) Build an intelligent control architecture based on model predictive control or adaptive control algorithm, tightly coupled with the physical model;

[0045] (93) With the help of Simulink 3D Animation toolbox, the fine geometric model built by SolidWorks is deeply bound with the physical model, and the virtual robot operation scene is rendered.

[0046] Further, the specific method for building a twin control interface is:

[0047] (101) In the design of the twin control interface, multi-modal information fusion display is introduced, including position, attitude, speed, force, audio, and tactile feedback information;

[0048] (102) Develop an intelligent assistant decision system integrated in the control interface;

[0049] (103) Build adaptive control functions for the interface to adapt to long-term tasks and high-delay communication in deep space exploration;

[0050] (104) Realize the distributed collaborative control of the interface, allowing multiple operators to access the control interface simultaneously through the network at different locations;

[0051] (105) Set up a multi-level security protection mechanism, including permission authentication and real-time risk assessment of operation commands;

[0052] (106) The system automatically records the operation history of the operator and the response of the robot, analyzes the pros and cons of the operation based on machine learning algorithm, and provides improvement suggestions for the operation skills of the operator.

[0053] The present application has at least the following advantages:

[0054] The present application can guide the actual robot body to perform related actions through virtual simulation data by completing the construction of the physical layer, data acquisition and transmission layer, digital twin model layer, and application layer of the robot, and can complete the comparison of virtual and actual body configurations through the visual interface, meeting the digital twin needs of various spacecraft operations in the future deep space exploration field.

[0055] Of course, implementing any product of the present application does not necessarily require all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The overall architecture schematic diagram of the digital twin system in the embodiment of the present application is shown in the figure.

[0057] Figure 2 The assembly schematic diagram of the construction robot body in the embodiment of the present application is shown in the figure.

[0058] Figure 3 The structure schematic diagram of the geometric model in the embodiment of the present application is shown in the figure.

[0059] Figure 4 The URDF file generation model schematic diagram in the embodiment of the present application is shown in the figure.

[0060] Figure 5 The coordinate system construction and model correction schematic diagram of the geometric model in the embodiment of the present application is shown in the figure.

[0061] Figure 6 The UR5 robot twin model generation schematic diagram in the embodiment of the present application is shown in the figure.

[0062] Figure 7 The UR5 robot twin model simulink simulation model schematic diagram in the embodiment of the present application is shown in the figure.

[0063] Figure 8 The control principle schematic diagram of the UR5 robot input controller in the embodiment of the present application is shown in the figure.

[0064] Figure 9 The control principle schematic diagram of the UR5 robot actuator in the embodiment of the present application is shown in the figure.

[0065] Figure 10 The control principle schematic diagram of the UR5 robot output controller in the embodiment of the present application is shown in the figure.

[0066] Figure 11 The UR5 robot twin control interface schematic diagram based on matlab in the embodiment of the present application is shown in the figure.

[0067] Figure 12 The UR5 robot digital twin comparison implementation schematic diagram in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present disclosure.

[0069] Please refer to Figures 1-12The application provides a technical scheme: a mechanical arm digital twin system for deep space exploration, comprising a physical layer, a data acquisition and transmission layer, a digital twin model layer, and an application layer, the physical layer comprising a mechanical arm body and a sensor network, wherein the sensor network is used to acquire mechanical arm body operating state information;

[0070] The data acquisition and transmission layer comprises data acquisition equipment and a data transmission network, wherein the data acquisition equipment is connected with the sensor network and the data transmission network respectively, the data acquisition equipment is used to convert the operating state information collected by the sensor network into an electrical signal and transmit it to the data transmission network for storage, and the data transmission network uses a wired network (such as an Ethernet network) to transmit the collected data from the physical layer to the digital twin model layer;

[0071] The digital twin model layer comprises a geometric model construction module, a physical model construction module, and a model calibration and verification module, the geometric model construction module is used to create a high-precision geometric model of the mechanical arm body, the physical model construction module is used to construct a physical model of the mechanical arm body based on the geometric model, combined with the dynamics, kinematics principles, and material properties of the mechanical arm body, and the model calibration and verification module is used to calibrate the precision of the constructed geometric model and physical model with the actual operating state data of the mechanical arm acquired by the data acquisition and transmission layer, to ensure that the model output results are highly consistent with the operating characteristics of the real physical system;

[0072] The application layer is specifically a visualization module, the visualization module is connected with the data acquisition and transmission layer and the digital twin model layer, the visualization module provides an intuitive and friendly human-computer interaction interface for users, the interface can display real-time collected physical parameters such as the position, speed, acceleration, and force of each joint of the mechanical arm, and synchronously update and display the motion trajectory, working space range, and current task information of the mechanical arm in the digital twin model, the running state of the mechanical arm is evaluated in real time, such as whether it is in a normal working range, whether there is abnormal vibration or overload, and the like, and alarm information is sent in a timely manner.

[0073] According to the technical scheme of the embodiment, the sensor network is installed on the mechanical arm body, and the sensor network includes a joint position sensor, a force sensor, a vision sensor and a speed sensor, which are used to collect physical parameters such as position, speed, acceleration and force of the mechanical arm body in the running state. The data acquisition device is used to process the analog signals collected by the sensor network, sample and digitize the analog signals to convert them into a digital format suitable for network transmission. The data acquisition device is specifically a multi-channel, high-precision integrated data acquisition card, which has a powerful signal conditioning function module. It can accurately remove noise interference and compensate for signal attenuation for different types of analog signals from the sensor network, such as weak voltage signals from pressure sensors, millivolt level fluctuation signals from acceleration sensors, and resistance change corresponding analog signals from strain gauges.

[0074] The digital processing unit is built-in with a high-speed and high-precision analog-to-digital converter (ADC) with a conversion accuracy of up to 24 bits or even higher, realizing seamless and low-error conversion of analog signals to digital signals. The converted data is packaged and encapsulated according to standard network transmission protocols, such as PROFINET and ETHERNET / IP protocols of industrial Ethernet or Wi-Fi 6 standard protocols of wireless transmission, to efficiently and stably transmit to the data transmission network, support real-time and large-capacity data transmission requirements, meet the rapid circulation of massive running state data of deep space exploration mechanical arms, and ensure the high-quality supply of the entire digital twin system data source.

[0075] At the same time, the data acquisition device is equipped with a redundant power module. Even in complex and variable deep space environments, it can still operate stably and continuously, preventing data loss, and has an intelligent self-checking program that periodically checks the functions of each acquisition channel and processing unit. Once a fault is found, it immediately sends an alarm message to the ground control center through an independent backup communication link, ensuring the reliability of the device itself and laying the foundation for the continuity of mechanical arm state monitoring.

[0076] Further, the data transmission network uses standard network transmission protocols to transmit the collected data from the physical layer to the digital twin model layer.

[0077] According to the technical solution of the embodiment, the geometric model is a simulation model constructed by using a three-dimensional modeling software (Solidworks or 3Dmax), and the physical model is a UR5 mechanical arm. The simulation model and the UR5 mechanical arm jointly perform virtual reality twin simulation. The output end posture and position data of the mechanical arm are compared. Specifically, the geometric model and the physical model are deeply connected through a multi-dimensional and multi-level data interaction link and an adaptive interface, so as to achieve precise digital twin control. At the data interaction level, a real-time bidirectional data channel is constructed. On the one hand, various sensors densely arranged on the physical model (the UR5 mechanical arm), such as high-precision encoders at joints, force sensors and attitude sensors on an end effector, continuously and constantly transmit the actual physical parameters of the mechanical arm at the moment, including joint angles, angular velocities, force sizes and directions, and quaternion information of real-time spatial attitudes, to the digital twin model layer where the geometric model is located, according to the strictly defined data format and transmission protocol, and are transmitted at a high speed. These actual operation data become the key driving force for the geometric model to synchronize the physical entity state in real time, ensuring that the geometric model can accurately reproduce the shape and position of the physical mechanical arm in space at the moment;

[0078] According to the technical solution of the embodiment, modeling deduction and twin simulation in the model calibration and verification module are implemented by using Solidworks and matlab simulink platforms, as follows:

[0079] (1) Construct a model assembly. The mechanical arm body assembly is designed by using Solidworks.

[0080] The construction of the model assembly specifically refers to the use of the powerful parametric design function of Solidworks to open an innovative top-down and bottom-up hybrid design mode, and to fully sculpt the mechanical arm body assembly.

[0081] First, in the top-down design stage, the overall architecture and key performance indicators of the mechanical arm are accurately planned according to the macro requirements of the deep space exploration task and the complex operation process. For example, a flexible working envelope space suitable for different celestial body sampling and complex equipment installation tasks is planned, as well as a structure strength distribution required to resist deep space radiation and micro meteorite impact. In this way, the top-level framework of the joint layout, link shape and end effector adaptation form of the mechanical arm is outlined, so as to ensure that each component is designed around the task core.

[0082] Then switch to the bottom-up design process, one by one to create the basic components of the robot arm; using Solidworks advanced surface modeling, solid modeling technology, not only strictly in accordance with the frontier of materials science design parts structure, such as the use of new lightweight high strength alloy to make joint shell, combined with the topological optimization algorithm to shape the internal hollow weight reduction and excellent mechanical properties of the connecting rod structure; also embedded in the parts of intelligent sensing microstructure, so that it can in situ sensing stress, temperature, strain and other parameters, for subsequent real-time monitoring and health management buried in the notes;

[0083] In the assembly link, abandon the traditional static assembly concept, introduce the dynamic assembly verification mechanism based on kinematics, dynamics virtual prototype technology; simulate the whole life cycle motion process of the robot arm under the deep space of a variety of typical working conditions, from the launch stage of the severe vibration impact, to the microgravity attitude adjustment in orbit, to the work stress on different celestial bodies after landing, dynamic check the matching precision and motion interference of each component, real-time backtracking correction of assembly relationship, to ensure that the assembly body can withstand the test in the virtual environment, and the final geometric model is like a precise projection of the real robot arm in the digital world, laying a solid foundation for the deep application of the twin model, and realizing the innovative construction paradigm that is deeply consistent with the stringent requirements of deep space exploration from the design source;

[0084] (2) generate a geometric model URDF file, the specific method is:

[0085] First, after completing the fine construction of the robot arm assembly in Solidworks, use Solidworks and URDF conversion plug-in or self-developed customized conversion script to realize the innovative conversion from Solidworks assembly data to URDF format; the conversion process is not a simple data format conversion, but a deep optimization and information expansion of the data considering the unique needs of deep space exploration scene;

[0086] In the conversion process, for each part and connecting component of the robot arm, in addition to the regular geometric shape, size, mass properties and other information, the functional role, reliability requirements and environmental adaptability characteristics of each component in deep space exploration are also integrated into the URDF file; for example, for the key joint components, not only the regular information such as rotation range and moment of inertia is recorded, but also special information such as lubrication characteristics in high vacuum environment, radiation damage resistance index, and performance attenuation coefficient of materials at extreme temperature is added; at the same time, each component is assigned a unique deep space exploration task label, which clearly defines its use state and importance level in different task stages (such as sampling, exploration, maintenance, etc.);

[0087] In addition, the connection relationship of the manipulator is innovatively and optimally described; multi-physical field coupling information is introduced to accurately depict the flexible connection characteristics between the connecting components in the URDF file, such as simulating the elastic deformation and damping effect at the joint, so as to better reflect the connection behavior under the complex force field in space; and according to different task scenarios, the stiffness and damping coefficient of the connection are adaptively adjusted, so that the URDF file can dynamically reflect the change of the connection performance of the manipulator in different deep space environments;

[0088] Intelligent data tags can also be embedded in the URDF file, which can automatically update the state information of the corresponding components according to the real-time running data and simulation results of the manipulator; for example, when the manipulator is executing a high-load sampling task, the stress at the joint exceeds the set threshold, and the corresponding joint part in the URDF file will automatically update its load-bearing capacity tag, so that the subsequent digital twin model and analysis software can quickly perceive and make corresponding decisions, such as triggering fault warning and adjusting operation strategy;

[0089] Furthermore, based on the XML extension method, custom elements and attributes are added to the URDF file to describe the non-traditional structure and function of the manipulator; for example, for the new detection equipment that may be carried, custom elements are used in the URDF file to describe its detection range, accuracy, energy demand and cooperation with other components of the manipulator, and external scripts are allowed to dynamically update these custom elements, ensuring that the URDF file can flexibly adapt to the development of deep space exploration technology and task requirements;

[0090] The URDF file generated in this way will become a comprehensive, dynamic and rich deep space exploration information description file, providing more accurate and forward-looking information support for subsequent simulation and control operations on platforms such as Matlab, greatly improving the performance and adaptability of the digital twin system;

[0091] (3) The specific method for building the twin model is:

[0092] In the Matlab Simulink environment, based on the previously imported URDF file of the manipulator, the multi-physics modeling technology is deeply integrated; using the rich module library of Simulink, not only the rigid body dynamics model of the manipulator is built, but also the joint kinematics relationship and the force transmission path are accurately simulated, and the corresponding physical effect sub-models are constructed for special environmental factors of deep space exploration, such as microgravity field, high and low temperature radiation heat exchange, material property change in vacuum environment, etc. For example, the microgravity compensation algorithm is written through the self-defined S-Function module and embedded into the core loop of the dynamics model to correct the motion deviation caused by the absence of gravity in real time. The heat conduction finite element module is used to simulate the heat distribution and heat dissipation process in the manipulator structure, and the closed-loop thermal control strategy modeling is realized combined with the temperature sensor feedback to ensure that the model can accurately reflect the physical performance evolution of the manipulator in the deep space extreme environment, as follows:

[0093] It should be noted that the microgravity compensation algorithm is based on the dynamics modeling, mainly for adjusting the weakening of the gravity compensation term in the microgravity environment, and the dynamics model of the manipulator is described as follows:

[0094]

[0095] In the formula, τ is the joint torque, M(q) is the inertia matrix, is the Coriolis force and centrifugal force matrix, G(q) is the gravity compensation term, and F(q) is the friction torque.

[0096] In the microgravity environment, the influence of the gravity compensation term is significantly reduced, and the improved compensation term can be expressed as:

[0097] G micro (q)=g micro ·J T (q)m

[0098] In the formula, g micro is the microgravity acceleration, J T is the Jacobian matrix, and m is the mass vector of each link of the manipulator.

[0099] The algorithm implementation includes:

[0100] State acquisition: real-time acquisition of the angle, angular velocity, acceleration of the joints of the manipulator, and external force information;

[0101] Parameter adjustment: setting g micro according to the microgravity conditions of the deep space environment;

[0102] Compensation calculation: calculating the compensation torque of the joint based on the Jacobian matrix;

[0103] Dynamics correction: adjusting the dynamics model, removing the original gravity compensation term, and adding the microgravity correction term;

[0104] Real-time control: Combined with the prediction module, dynamically adjust the control input to ensure the accuracy of the robot trajectory.

[0105] At the same time, an intelligent control architecture based on model predictive control (MPC) or adaptive control algorithm is built, which is tightly coupled with the physical model. This control architecture generates control sequences in real time according to task objectives, continuously compares model prediction output with physical model actual response, and dynamically adjusts control parameters, giving the twin model the ability to adapt to complex working conditions and uncertain disturbances, realizing the leap from simple mechanical motion reproduction to complex task intelligent decision execution, and greatly improving the performance and reliability of the robot in the whole life cycle of deep space exploration tasks.

[0106] It should be noted that the adaptive control algorithm adjusts the control parameters according to the task objectives and environmental changes by monitoring the dynamic behavior of the robot in real time, and solves the problem of uncertainty in complex working conditions. The specific implementation steps are as follows:

[0107] Dynamic modeling: Construct a real-time dynamics model of the robot, which describes the nonlinear relationship between joint torque and angle, velocity, and acceleration.

[0108] Parameter estimation: Use recursive least squares (RLS) or gradient descent algorithm to estimate model parameters in real time, including inertia, friction, and joint stiffness, etc.

[0109] Error analysis and feedback: Calculate the deviation between the current output and the expected output through the error function, and the error is defined as: e(t) = x d (t) - x a (t), x d is the expected trajectory, and x a is the actual trajectory.

[0110] Control law update: Based on Lyapunov stability criterion, design control law: τ a = τ n + K e e(t), K e is the error gain matrix, and τ n is the nominal value.

[0111] In addition, to realize high-fidelity visual twin display, use the Simulink 3D Animation toolbox to deeply bind the fine geometric model built by SolidWorks with the physical model, and render a realistic virtual robot running scene, supporting immersive interactive inspection. The operator can observe the robot structure stress and strain, joint drive details from any angle, providing intuitive basis for fault troubleshooting and performance tuning, and making the model building results serve the deep space exploration complex operation process control in all directions.

[0112] (4) Twin control interface construction, the specific method is:

[0113] In the design of the twin control interface, multi-modal information fusion display is introduced; in addition to the traditional graphical interface display of the state information of the robot arm (such as position, attitude, speed, force, etc.), audio, haptic feedback and other information are also fused; for example, when the robot arm approaches the limit of the operating range or a potential fault occurs, the operator is prompted through specific sound signals and haptic vibrations, so that he can perceive the running state of the robot arm through hearing and touch in addition to vision, enhancing the situational awareness capability;

[0114] At the same time, an intelligent auxiliary decision-making system is developed and integrated into the control interface; this system can automatically recommend the best operation strategy and control parameters according to the current task target and real-time state of the robot arm; for example, when the robot arm needs to grasp an irregular deep space object, the system will automatically plan the optimal motion trajectory and grasping posture according to the shape, position of the object and the current position of the robot arm, and visualize it on the interface, and the operator can choose to accept the suggestion or adjust it according to his own experience;

[0115] In order to adapt to the characteristics of long-term tasks and high-delay communication of deep space exploration, the adaptive control function of the interface is constructed; in the case of high communication delay with the actual robot arm, the interface can predict the motion trajectory and state of the robot arm during the delay period according to the current control command and known robot dynamics model, so that the operator can make adjustments in advance to avoid operation errors caused by communication delay;

[0116] In addition, the distributed collaborative control of the interface is realized; multiple operators can simultaneously access the control interface through the network at different locations to jointly complete complex deep space exploration tasks; each operator is responsible for different operating parts, the interface can clearly divide their operating permissions and task ranges, and real-time synchronize the operating information between different operators to ensure the smoothness and safety of collaborative operation;

[0117] In terms of the safety of the control interface, a multi-level security protection mechanism is set up; in addition to the regular permission authentication, real-time risk assessment of the operation command is also performed, and for high-risk operations that may cause damage to the robot arm or task failure, a warning window will pop up and the operator will be forced to confirm or provide higher level of permission authentication to ensure the safety and reliability of the operation;

[0118] Finally, an operation record and learning function driven by machine learning is added to the interface; the system automatically records the operation history of the operator and the response of the robot arm, analyzes the pros and cons of the operation through a machine learning algorithm, provides improvement suggestions for the operator's operation skills, and continuously optimizes its own operation recommendation algorithm, realizes the continuous improvement of human-machine cooperation, makes the operator and the twin control interface grow together, and better serves the deep space exploration mission;

[0119] According to the technical scheme of the embodiment, the system automatically records the operation history of the operator and the response of the robot arm, analyzes the pros and cons of the operation by using an algorithm architecture combining deep reinforcement learning and supervised learning, and provides improvement suggestions for the operator's operation skills. Reinforcement learning continuously optimizes the operation strategy through interaction with the environment, and supervised learning uses historical data to provide clear feedback and improvement direction for the operation. Compared with a single reinforcement learning method, this combined strategy can converge faster, and in the case of less data, the supervised learning can make up for the shortcomings of the early model.

[0120] The specific steps of the algorithm include:

[0121] Firstly, the system collects the running state of the robot arm (such as joint position, speed, force, etc.) and the control input of the operator in real time, and performs denoising and normalization processing on these data, extracts key features such as task completion degree and response time, etc.; then, through the supervised learning model, the historical operation data is analyzed to find out the rules of successful and failed operations, and to provide basic support for operation skill improvement; at the same time, in the reinforcement learning stage, the operation strategy is optimized through the deep Q network (DQN), and the operation success rate and efficiency are taken as the reward signal to continuously improve the control strategy; finally, the system combines the historical experience of supervised learning and the real-time optimization of reinforcement learning through a hybrid decision mechanism to generate targeted operation suggestions; the feedback results are presented in an intuitive way, including operation skill improvement direction, invalid operation reduction suggestions, and force input optimization, etc.

[0122] Next, the digital twin simulation verification of the robot arm is carried out:

[0123] The digital twin simulation verification of the robot arm is mainly realized by using the matlab simulink platform, which mainly includes four parts of building model assembly, generating model URDF file, twin model building, and twin control interface. Finally, the simulation model will be combined with the actual UR5 robot arm to carry out virtual reality twin simulation, and the output of the robot arm end posture and related position data, etc. are compared, as follows:

[0124] (1) The construction model assembly is specifically to build the assembly of the robot arm body by using Solidworks, and the assembly is assembled into a geometric model, such as Figure 2 andFigure 3 as shown in

[0125] (2) Generate a geometric model URDF file, as shown in Figure 4 and Figure 5 as shown in

[0126] (3) The specific method of building a twin model is:

[0127] As shown in Figures 6-10 , open matlab, install matlab R2021b installation robot toolbox (MatlabToolbox), add the path of ur5.urdf file to matlab path, input instruction: smimport(′ur5_urdf.urdf′) import system;

[0128] (4) As shown in Figure 11 , the twin control interface sends the relevant parameters required by the UR5 robot arm through the setting of relevant parameters, and displays and supports the relevant parameters through the relevant parameters fed back by the UR5 robot arm to the control box of the robot arm through Ethernet, and completes the control of the robot arm.

[0129] Demonstration and verification of robot digital twin:

[0130] Table 1 UR5 robot twin simulation parameter information

[0131]

[0132] Figure 12 The UR5 robot digital twin control implementation diagram is shown. By completing the construction of the physical layer, data acquisition and transmission layer, digital twin model layer and application layer of the robot, the actual robot body can be guided to perform related actions through virtual simulation data, and the virtual and real body configuration comparison can be completed through the visual interface, which can meet the digital twin needs of various spacecraft operations in the future deep space exploration field.

[0133] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0134] Those of ordinary skill in the art, with the benefit of this disclosure, would understand the specific meaning of the above terms in the context of the present application. When an element is referred to as being "on", "connected to", "mounted to", or "disposed to" another element, it can be directly on, connected to, mounted to, or disposed to the other element or intervening elements can be present. When an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements can be present. As used herein the terms "vertical", "horizontal", "up", "down", "left", "right", and the like as well as adjectival and adverbial derivatives thereof (e.g., "vertically", "horizontally", "upwardly", etc.) are intended to describe relative positions on the apparatus as the apparatus is oriented in the particular drawing figure to which reference is made.

[0135] While embodiments of the application have been shown and described, it is to be understood that various modifications, substitutions, combinations, and variations can be made in the application by those of ordinary skill in the art without departing from the spirit and scope of the present application, which is defined by the following claims and their equivalents.

[0136] In the description of the specification, reference to "one embodiment", "an example", "a specific example", or the like means that a particular feature, structure, material, or characteristic being referred to is included in at least one embodiment or example of the disclosure. The appearances of the above expressions in various places in the specification are not necessarily referring to the same embodiment or example. Moreover, the particular features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

Claims

1. A digital twin system for a robotic arm designed for deep space exploration, comprising a physical layer, a data acquisition and transmission layer, a digital twin model layer, and an application layer, characterized in that, The physical layer includes the robotic arm body and a sensor network, wherein the sensor network is used to collect the operating status information of the robotic arm body; The data acquisition and transmission layer includes a data acquisition device and a data transmission network. The data acquisition device is connected to the sensor network and the data transmission network respectively. The data acquisition device is used to convert the operating status information acquired by the sensor network into electrical signals and transmit them to the data transmission network for storage. The digital twin model layer includes a geometric model construction module, a physical model construction module, and a model calibration and verification module. The geometric model construction module is used to create a high-precision geometric model of the robotic arm body. The physical model construction module is used to construct a physical model of the robotic arm body based on the geometric model, combined with the dynamics, kinematics, and material properties of the robotic arm body. The model calibration and verification module is used to perform precision calibration on the constructed geometric model and physical model based on the actual operating status data of the robotic arm obtained by the data acquisition and transmission layer. The application layer is specifically a visualization module, which is connected to the data acquisition and transmission layer and the digital twin model layer. It is used to receive the operating status information of the robotic arm body and display the motion trajectory, workspace range and currently executed task information of the robotic arm body in real time. Specific methods for building a twin model: (91) In the Matlab Simulink environment, based on the previously imported URDF file of the robotic arm, multiphysics modeling technology is deeply integrated. Using the Simulink module library, a rigid body dynamic model of the robotic arm is built, and a corresponding physical effect sub-model is constructed for the special environmental factors of deep space exploration. At the same time, the motion deviation of the dynamic model is compensated and corrected by microgravity compensation algorithm, and the control parameters are adjusted in real time by adaptive control algorithm, as follows: (91.1) The microgravity compensation algorithm is as follows: The microgravity compensation algorithm is based on dynamic modeling. The dynamic model of the robotic arm is described as follows: In the formula, For joint torque, The inertia matrix, The matrix represents the Coriolis force and the centrifugal force. For gravity compensation, This is the frictional torque; In a microgravity environment, the influence of the gravity compensation term is significantly reduced, and the improved compensation term is expressed as: In the formula, Microgravity acceleration, For Jacobian matrices, Let be the mass vector of each link in the robotic arm; (91.2) Adaptive control algorithm, the specific implementation steps are as follows: Dynamic modeling: Construct a real-time dynamic model of the robotic arm to describe the nonlinear relationship between joint torque and angle, velocity, and acceleration; Parameter estimation: Model parameters, including inertia, friction, and joint stiffness, are estimated in real time using recursive least squares or gradient descent algorithms. Error Analysis and Feedback: The deviation between the current output and the expected output is calculated using an error function, where the error is defined as: , For the desired trajectory, This is the actual trajectory; Control Law Update: Based on the Lyapunov stability criterion, design the control law: , Here is the error gain matrix. It is the nominal value; (92) Build an intelligent control architecture based on model predictive control or adaptive control algorithms, which is tightly coupled with the physical model; (93) Using the Simulink 3D Animation toolbox, the fine geometric model built by SolidWorks is deeply bound to the physical model to render the virtual robotic arm operation scene.

2. The digital twin system for a robotic arm for deep space exploration according to claim 1, characterized in that: The sensor network is mounted on the robotic arm body and includes joint position sensors, force sensors, vision sensors, and speed sensors.

3. The digital twin system for a robotic arm for deep space exploration according to claim 2, characterized in that: The data acquisition device is used to condition, sample, and digitize the analog signals acquired by the sensor network to convert them into a digital format suitable for network transmission. Specifically, the data acquisition device is configured as an integrated data acquisition card and is equipped with a redundant power supply module.

4. The digital twin system for a robotic arm for deep space exploration according to claim 3, characterized in that: The data transmission network uses a standard network transmission protocol to transmit the collected data from the physical layer to the digital twin model layer.

5. A digital twin system for a robotic arm for deep space exploration according to claim 4, characterized in that: The geometric model and the physical model are deeply connected through multi-dimensional, multi-layered data interaction links and adaptation interfaces to achieve digital twin control, as detailed below: The physical layer collects the current joint rotation angles, angular velocities, force magnitudes and directions, and real-time spatial attitude data of the physical model. Through the data acquisition and transmission layer, the data is transmitted to the digital twin model layer where the geometric model is located, according to the defined data format and transmission protocol, so that the geometric model can reproduce the current shape and position of the physical model in space.

6. A digital twin system for a robotic arm for deep space exploration according to claim 5, characterized in that: The model calibration and verification module utilizes Solidworks and MATLAB Simulink platforms for modeling, deduction, and twin simulation, as detailed below: (1) Constructing the model assembly: The robotic arm body assembly was designed using Solidworks; (2) Generate the geometric model URDF file; (3) Specific methods for constructing twin models; (4) Construction of twin control interface.

7. A digital twin system for a robotic arm for deep space exploration according to claim 6, characterized in that, The robotic arm body assembly was designed using Solidworks, as follows: (71) First, in the top-down design stage, plan the overall architecture and structural strength distribution of the robotic arm, and design the top-level framework of the joint layout, linkage shape and end effector of the robotic arm. (72) Then switch to the bottom-up design stage, create the basic components of the robotic arm one by one, and use Solidworks to design the part structure, including joint shell, linkage structure and sensor structure; (73) During the assembly stage, the fit accuracy and motion interference of each component are dynamically verified, and the assembly relationship is corrected in real time.

8. A digital twin system for a robotic arm for deep space exploration according to claim 6, characterized in that, The specific method for generating the geometric model URDF file is as follows: (81) After completing the construction of the robotic arm assembly in Solidworks, the Solidworks and URDF conversion plugin or conversion script is used to realize the innovative conversion from Solidworks assembly data to URDF format; (82) During the conversion process, for each part and connecting component of the robotic arm, in addition to the conventional geometric shape, size and mass attribute information, the functional role, reliability requirements and environmental adaptability characteristics of each component in deep space exploration are integrated into the URDF file; (83) Characterize the flexible connection characteristics between connecting components in the URDF file, including simulating elastic deformation, damping effect, and connection stiffness at the joint; (84) Embed smart data tags in the URDF file to automatically update the status information of the corresponding components based on the real-time operation data of the robotic arm and the simulation results; (85) Use an XML-based extension method to add custom elements and attributes to the URDF file.

9. A digital twin system for a robotic arm for deep space exploration according to claim 8, characterized in that, The specific method for constructing a twin control interface is as follows: (101) In the design of the twin control interface, multimodal information fusion display is introduced, specifically including position, posture, speed, force, audio, and tactile feedback information; (102) Develop an intelligent auxiliary decision-making system and integrate it into the control interface; (103) Construct an adaptive control function for the interface to adapt to long-duration missions and high-latency communication in deep space exploration; (104) Implement distributed collaborative control of the interface, allowing multiple operators to access the control interface simultaneously via the network from different locations; (105) Set up a multi-layered security protection mechanism, including access authentication and real-time risk assessment of operation commands; (106) The system automatically records the operator's operation history and the response of the robotic arm, analyzes the advantages and disadvantages of the operation based on machine learning algorithms, and provides operators with suggestions for improving their operation skills.

Images