Self-sensing multifunctional manipulator execution device based on motion platform

Abstract

The application relates to the technical field of intelligent robots and discloses a self-perception type multifunctional mechanical arm execution device based on a motion platform, which comprises bottom hardware, a perception module and a control module; the bottom hardware comprises a mobile platform, a mechanical arm, a self-calibration base, a dynamic counterweight mechanism and a magnetorheological compliant wrist carrying a work tool. After the work tool is replaced, the control module controls the mechanical arm to drive the tool to contact the self-calibration base, and the actual pose of the tool is solved and obtained; when overturning is prevented, the counterweight block is calculated and driven to translate based on the real-time center of mass position of the whole machine, so that the system dynamic stability is maintained; when contact work is carried out, the shear stress of the magnetorheological compliant wrist is closed-loop adjusted based on the contact force data of the perception module, so that hardware-level compliant control with extremely low delay is realized. The application effectively solves the technical problems that the robot mobile work is unstable, collision response is lagged and the pose is difficult to accurately reconstruct after the work tool is frequently replaced in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of intelligent robot technology, specifically to a self-sensing multifunctional robotic arm execution device based on a motion platform. Background Technology

[0002] With the development of automation technology, mobile robots consisting of mobile chassis and robotic arms are widely used in unstructured environments to perform tasks such as handling, assembly, and polishing.

[0003] In actual operation, the outward extension of the robotic arm or the grasping of variable-load heavy objects directly alters the spatial center of gravity distribution of the entire machine. Existing mobile robots often maintain physical stability by adding a fixed chassis weight or restricting the robotic arm's workspace. This approach increases the equipment's energy consumption and sacrifices its working range. Furthermore, when faced with unknown loads or terrain changes, there remains a risk that the system's center of gravity may deviate from the support area, leading to platform overturning.

[0004] When robotic arms perform contact operations such as grinding and assembly, there is direct physical interaction between the end effector and the target workpiece. Traditional robotic arm ends are typically rigidly connected, and the impact force generated at the moment of contact can easily damage the workpiece surface or the internal reducer. To achieve compliant contact, existing technologies usually employ purely software impedance control algorithms based on dynamic models. However, due to the time lag between signal acquisition from multi-dimensional force and torque sensors, calculation by the main control algorithm, and the underlying motor drive, the software control exhibits response lag when facing transient mechanical changes such as rigid collisions. It cannot provide immediate physical clearance space, easily causing high-frequency oscillations at the contact interface and making it difficult to maintain stable contact force.

[0005] Furthermore, multi-functional robots require frequent tool changes during on-site operations based on task requirements. After changing to tools of different sizes or shapes, the actual end-effector center point and spatial orientation of the new tool will shift relative to the robot arm's original coordinate system. Existing tool calibration methods primarily rely on externally deployed high-precision laser tracking equipment or manual point-by-point teaching. In unstructured environments with limited resources, the lack of external measuring equipment and cumbersome manual operation make it difficult for the system to autonomously acquire accurate pose data after tool changes. This leads to accumulated positioning errors, reducing the overall efficiency and accuracy of continuous operation. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a self-sensing multifunctional robotic arm execution device based on a motion platform, which solves the technical problems of mobile robots during operation, such as overall center of gravity instability and overturning, lag in response of pure software variable stiffness control which easily leads to collisions and oscillations, and difficulty in autonomously calibrating the end effector pose after tool changes.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: a self-sensing multifunctional robotic arm execution device based on a motion platform, comprising underlying hardware, a sensing module, and a control module;

[0008] The underlying hardware includes a mobile platform, a robotic arm, a self-calibrating base, a dynamic counterweight mechanism, and a magnetorheologically compliant wrist carrying the working tools; the dynamic counterweight mechanism includes a counterweight block and a counterweight drive mechanism.

[0009] The sensing module includes at least a visual sensor and a multi-dimensional force and torque sensor for data acquisition.

[0010] The control module is communicatively connected to the underlying hardware and the sensing module, and is configured to perform the following controls:

[0011] Calibration control: After changing the working tool, control the robotic arm to drive the working tool to contact the self-calibration base, and calculate and obtain the actual position and pose of the working tool;

[0012] Anti-tipping control: Based on the overall machine mass parameters and the calculated real-time center of gravity position of the whole machine, calculate the target position of the counterweight, and control the counterweight drive mechanism to drive the counterweight block to the target position of the counterweight;

[0013] Compliance control: During contact operation, an excitation current is generated based on the contact force data obtained by the multi-dimensional force and torque sensor and output to the magnetorheological compliant wrist to adjust the shear stress.

[0014] Furthermore, the control module is also configured to: locate the target workpiece by coordinate transformation based on the point cloud data acquired by the vision sensor and the current pose data of the robotic arm.

[0015] Furthermore, the control module calculates and obtains the actual pose of the working tool, specifically configured to: control the robotic arm to drive the working tool to make multi-point contact with the self-calibration base, and output a set bias excitation current to the magnetorheological compliant wrist during the contact process to keep it in a compliant state; when the contact force is detected to reach a preset calibration threshold, record the joint angle of the robotic arm; based on the joint angle and the known constraint position of the self-calibration base, solve the spatial deviation of the working tool relative to the magnetorheological compliant wrist through an iterative algorithm.

[0016] Furthermore, the control module calculates the real-time center of gravity position of the entire machine, specifically configured to: obtain the mass parameters and real-time center of gravity coordinates of the mobile platform, the robotic arm, the working tool, and the counterweight; perform weighted calculation on the mass parameters and real-time center of gravity coordinates to obtain the three-dimensional position of the overall center of gravity of the entire machine in the reference coordinate system of the mobile platform, which is used as the real-time center of gravity position of the entire machine.

[0017] Furthermore, the control module calculates the target position of the counterweight and controls the counterweight drive mechanism to drive the counterweight block. Specifically, it is configured to: set a safety support area to ensure the stability of the mobile platform; calculate the target position of the counterweight with the horizontal coordinate of the three-dimensional position located within the safety support area as a constraint, and send it to the counterweight drive mechanism.

[0018] Furthermore, the contact force data includes the actual measured contact force; the control module generates the excitation current, specifically configured to: calculate the difference between the actual measured contact force and the preset reference contact force to obtain the mechanical error; and based on the mechanical error, generate the excitation current under closed-loop control through a feedback control law.

[0019] Furthermore, the magnetorheological compliant wrist has an excitation coil inside and is filled with magnetorheological fluid; the excitation coil receives the excitation current to generate an external magnetic field, and the magnetic induction intensity of the external magnetic field is linearly corresponding to the excitation current, which is used to change the rheological state of the magnetorheological fluid to adjust the shear stress.

[0020] Furthermore, the control module generates the excitation current for closed-loop control through a feedback control law, specifically configured to: perform proportional gain calculation and time integral calculation on the mechanical error respectively to obtain a proportional component and an integral component; and add the proportional component and the integral component to obtain the excitation current.

[0021] Furthermore, the end of the magnetorheological compliant wrist is provided with a quick-change flange, through which different working tools can be switched.

[0022] Furthermore, after switching to a new working tool, when the control module calculates the actual pose of the new working tool, it simultaneously uses the mass and center of gravity parameters of the new working tool to update the mass parameters of the whole machine, thereby triggering the recalculation of the real-time center of gravity position of the whole machine and updating the counterweight target position, realizing the physical coordinated linkage of tool switching, perception calibration and chassis anti-tipping.

[0023] This invention provides a self-sensing multifunctional robotic arm execution device based on a motion platform. It has the following beneficial effects:

[0024] 1. This invention constructs a hardware-level variable stiffness mechanism based on underlying contact force feedback. By outputting continuously modulated excitation current to the magnetorheologically compliant end-effector, the system can directly change the yield shear stress of the internal fluid. This scheme utilizes the millisecond-level physical phase change of the magnetorheological fluid to absorb instantaneous impact kinetic energy, effectively overcoming the response lag problem caused by computational delay and dynamic model errors in traditional pure software impedance control. In the event of a rigid collision, the device can provide rapid physical elastic yielding, significantly reducing the risk of workpiece damage and high-frequency system oscillation under unstructured operations.

[0025] 2. This invention introduces a dynamic counterweight closed-loop intervention mechanism based on spatial dynamics, enhancing the force boundary of the mobile platform under extreme postures. The system calculates the overall center of mass of the entire machine, including each independent link of the robotic arm and the working load, in real time, driving the counterweight block within the mobile platform to actively translate. This dynamic transfer of spatial mass effectively and precisely counteracts the overturning moment caused by the robotic arm's significant outward extension during operation, stabilizing the overall center of gravity within the safe support boundary and providing reliable dynamic stability for the mobile platform.

[0026] 3. This invention designs a fully automatic self-calibration and collaborative mechanism, solving the technical challenge of autonomously reconstructing the spatial pose of a new tool during cross-task switching. After changing the working tool, the system controls the tool's end effector to make multi-point physical contact with the feature array of the self-calibration base, and uses contact geometric constraints and high-precision joint angles to solve for the actual translation and rotation deviations. This scheme reduces reliance on external high-precision spatial tracking instruments and eliminates the need for manual secondary teaching, enabling the rapid establishment of a high-precision physical reference system in complex work environments. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the system architecture and overall physical structure of the present invention;

[0028] Figure 2 This is a block diagram illustrating the internal physical structure of the magnetorheological compliant wrist and quick-change flange of the present invention.

[0029] Figure 3 This is a schematic diagram of the self-calibrating base structure and multi-point contact principle of the present invention;

[0030] Figure 4 This is a top view of the geometric projection of the mobile platform safety support area and dynamic counterweight of the present invention.

[0031] Figure 5 This is a block diagram of the hardware-level magnetorheological closed-loop control based on mechanical error feedback of the present invention.

[0032] Figure 6 This is a flowchart of the cross-task collaborative linkage and full-process control method of the present invention. Detailed Implementation

[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] It should be noted that the specific physical assembly and mechanical cross-sectional structure of the underlying hardware, such as the mobile platform, robotic arm, self-calibration base, and magnetorheological compliant wrist, detailed in this embodiment, are intended to provide a highly reliable physical execution carrier for the present invention, thereby achieving the optimal embodiment of the invention. The core innovative concept of this invention lies primarily in the multi-source data fusion and collaborative decision-making control logic operating within the control module. Through data interaction between the underlying hardware and the sensing module, the control module achieves a coordinated linkage of sensing calibration, dynamic anti-tipping, and hardware-level compliance via a collaborative algorithm. Those skilled in the art will understand that, without departing from the aforementioned control logic and collaborative mechanism of this invention, the specific mechanical transmission form and physical assembly method of the underlying hardware can be adaptively replaced according to actual operational needs, without deviating from the scope of protection of this invention.

[0035] Please see the appendix Figure 1 -Appendix Figure 6 This invention provides a self-sensing multifunctional robotic arm execution device based on a motion platform, comprising underlying hardware, a sensing module, and a control module. The underlying hardware includes: a mobile platform, a robotic arm mounted on the mobile platform, a self-calibrating base disposed on the top surface of the mobile platform, a dynamic counterweight mechanism disposed inside the mobile platform, and a magnetorheological compliant wrist connected to the end of the robotic arm. The dynamic counterweight mechanism includes a counterweight block and a counterweight drive mechanism. The sensing module includes a visual sensor and a multidimensional force and torque sensor. The control module is communicatively connected to the underlying hardware and the sensing module to achieve efficient physical coordination and closed-loop control of the underlying data between the hardware units.

[0036] The mobile platform serves as the load-bearing and power base for the entire machine, providing stable motion support and omnidirectional movement capability for the working modules above it. The base of the robotic arm is fixedly mounted on the load-bearing top surface of the mobile platform, providing a wide range of three-dimensional working space through the rotation and coordination of its joints.

[0037] A self-calibration base is fixedly installed in the calibration area on the top surface of the mobile platform. The surface of this self-calibration base is machined with a feature array exhibiting a known spatial geometric arrangement. This feature array includes at least four non-collinearly arranged conical positioning recesses. The three-dimensional coordinate data of the vertices of each recess in the mobile platform's reference coordinate system are pre-stored in the registers of the control module. After the robotic arm changes its working tool, a standard spherical calibration probe can be attached to the end of the current working tool. The control module drives the spherical probe to make multi-point contact with the conical positioning recesses on the self-calibration base. For working tools with regular shapes, their own known features such as vertices and end faces can also be used as contact references. The automatic centering characteristic of the conical surface or the geometric constraints of known features ensure the repeatability and accuracy of the contact points, thereby providing an absolute physical reference system for obtaining the actual spatial pose of the current tool's end.

[0038] To counteract the overturning moment caused by the extension of the robotic arm and the grasping of different mass loads at the end effector, a dynamic counterweight mechanism is installed inside the mobile platform. This dynamic counterweight mechanism specifically includes a counterweight drive mechanism and a counterweight block; the counterweight drive mechanism is a multi-axis guide rail driver arranged orthogonally along the longitudinal and transverse directions inside the mobile platform, with the counterweight block slidably mounted on the multi-axis guide rail driver. After receiving the target position command from the control module, the servo transmission component inside the counterweight drive mechanism forces the counterweight block to translate within the horizontal plane inside the platform chassis, intervening in real time and adjusting the overall physical center of gravity distribution of the machine.

[0039] Instead of a traditional rigid connection between the robotic arm's end effector and the working tool, a magnetorheological compliant wrist is connected in series via a flange. This compliant wrist employs a coaxial disc shearing structure, comprising an input flange, an output flange, a sealed shearing chamber, an excitation coil, and a magnetically conductive outer shell. The input flange is fixedly connected to the robotic arm's end effector. The annular sealed shearing chamber is positioned within the gap between the input and output discs, and is uniformly filled with magnetorheological fluid. The excitation coil is embedded within the magnetically conductive outer shell, surrounding the sealed shearing chamber, with its terminals extending through the shell to electrically connect to the output of the control module. To ensure the physical reliability of the electrical and communication links during continuous rotation of the robotic arm's end effector, a hollow isolation tube is installed along the central axis of the compliant wrist, containing multi-channel conductive slip rings.

[0040] The output end of the magnetorheological compliant wrist is equipped with an opto-mechatronic quick-change flange. This quick-change flange consists of a male and a female end. The female end is fixed to the outside of the output end of the magnetorheological compliant wrist, while the male end is fixedly connected to the working tool. To provide multi-degree-of-freedom deformation allowance, the input and output flanges of the compliant wrist are coaxially connected via an annular elastic support. The flange integrates a steel ball mechanical locking structure, a cylindrical end face positioning reference, an annular electrical power supply contact, and a fiber optic communication interface. During docking, radial precision positioning is achieved through the end face, and the locking mechanism drives the steel ball to engage, completing the mechanical locking. Simultaneously, the contacts synchronously connect the power and communication links. The wiring is connected to the internal wiring harness of the robotic arm via a multi-channel conductive slip ring within the hollow isolation tube, allowing the device to switch between different types of working tools (grabbing, grinding, assembly, etc.) in a plug-and-play manner at the work site, while simultaneously connecting the physical status monitoring and power transmission links of the new tool.

[0041] When the output end is subjected to relative torsion by an external force, the magnetorheological fluid in the disk gap undergoes circumferential shear deformation. When the output end is subjected to radial force causing deflection, the disk gap is compressed on one side and stretched on the other under the action of the elastic support, resulting in uneven radial shear deformation of the magnetorheological fluid in the edge region. Under the combined effect of these two deformations, when the excitation control current is input, the excitation coil generates a variable external magnetic field perpendicular to the shear plane inside the chamber. By adjusting the yield shear stress of the magnetorheological fluid, the equivalent stiffness of the compliant wrist can be continuously adjusted in multiple directions of torsion and deflection. The magnetic induction intensity of the external magnetic field... With the input excitation control current They exhibit a linear positive correlation, and their basic electromagnetic mechanical expression is as follows:

[0042] ;

[0043] in, The permeability of the magnetorheological fluid is given by the value of . This refers to the effective number of turns of the excitation coil. This is the effective closed length of the magnetic circuit.

[0044] When magnetorheological fluids are subjected to an external magnetic field, the micron-sized magnetic particles suspended within them overcome thermal motion and rapidly align along the magnetic field lines to form chain-like or columnar structures. Macroscopically, this manifests as a rapid transformation of the liquid's rheological state from a Newtonian fluid to a solid-like state within an extremely short time (milliseconds). This process reduces the shear stress of the fluid within the magnetorheological fluid, thus improving its flexibility. It obeys the Bingham plastic fluid constitutive model, and its mechanical constitutive equation is expressed as:

[0045] ;

[0046] in, The dynamic yield shear stress is related to the magnetic flux density. This represents the zero-field apparent dynamic viscosity of the magnetorheological fluid in the absence of a magnetic field. The shear strain rate of the magnetorheological fluid when the flexible wrist is deformed by force.

[0047] Furthermore, dynamic yield shear stress With magnetic induction intensity The coupling response relationship can be expressed by exponential fitting as follows:

[0048] ;

[0049] in, and The fitting coefficients related to the intrinsic properties of magnetorheological fluid materials are exponential. denoted as the magnetization saturation constant of the material.

[0050] Through the aforementioned force-electromagnetic coupling mechanism, the control module transforms the contact mechanics data acquired by the multi-dimensional self-sensing module into a specific excitation control current, thereby precisely altering the dynamic yield shear stress of the magnetorheological fluid. This change in shear stress directly maps to variations in the damping and stiffness of the compliant wrist in resisting external impacts. Ultimately, this achieves continuously adjustable hardware-level variable stiffness control at the end of the robotic arm, avoiding the response lag and high-frequency jitter problems commonly found in pure software impedance control.

[0051] In this embodiment, the multidimensional self-sensing module serves as the core data acquisition source for the entire robotic arm execution device to physically and informationally interact with the external unstructured environment, and is responsible for capturing multidimensional field parameters in real time and at high frequency.

[0052] The vision sensor employs an integrated depth vision device, which is fixedly mounted to the end link or end flange side of the robotic arm via a high-strength rigid bracket, thereby constructing an end-effector-driven vision measurement and control topology. The multi-dimensional force and torque sensors are spatially connected in series and coaxially between the magnetorheological compliant wrist and the opto-mechatronic quick-change flange, used to measure in real-time and accurately all linear contact force components and angular torque components generated in three-dimensional space when different tools interact with external objects.

[0053] During the preparation phase of operations in unstructured environments, a vision sensor continuously scans and acquires a color two-dimensional image matrix of the target work area and the corresponding pixel depth information. The control module extracts the depth channel data from the image frames and fuses it with the color pixels through point-to-point registration to generate a high-density point cloud dataset containing rich geometric shapes and three-dimensional spatial topology of the target surface.

[0054] To overcome the inherent limitations of localized viewpoints in vision sensors and enable the acquired spatial data to be directly used for the coordinated planning of chassis movement and robotic arm motion, the control module performs rigorous spatial coordinate transformation calculations to construct a globally unified coordinate system throughout the entire machine. This globally unified coordinate system is directly bound and aligned with the aforementioned mobile platform reference coordinate system.

[0055] Let the homogeneous coordinates of any feature point on the surface of the target workpiece in the camera-independent coordinate system of the vision sensor be expressed as: .

[0056] Let the external spatial pose relationship between the camera's independent coordinate system and the robot arm's end flange coordinate system be given by the homogeneous transformation matrix. Description. This matrix relies on rigid constraints imposed by the hardware installation and contains two independent components: three-dimensional rotation and three-dimensional translation. During the initial system initialization, the control module drives the robotic arm end effector to observe the standard calibration board in multiple poses, and uses the Tsai-Lenz hand-eye calibration algorithm to solve for and solidify the homogeneous transformation matrix. This eliminates systematic errors introduced by machining and physical installation.

[0057] At any point during system operation, the control module reads the current actual joint angle vectors of each joint of the robotic arm via the internal high-speed bus, and substitutes these vectors into a preset standard model of the robotic arm's forward kinematics (e.g., the DH parameter model) to calculate in real time the homogeneous transformation matrix of the end flange coordinate system relative to the robotic arm base coordinate system. .

[0058] Meanwhile, given that the robotic arm's base is rigidly mounted on the mobile platform, the rigid static transformation matrix of the robotic arm's base coordinate system relative to the mobile platform's reference coordinate system (i.e., the globally unified coordinate system) is defined as follows: .

[0059] Based on the continuous multiplication transitivity of homogeneous coordinate transformation space, the absolute homogeneous coordinates of feature points on the surface of the target workpiece in a globally unified coordinate system. The mapping derivation equation is established as follows:

[0060] ;

[0061] Among them, any of the above adjacent associated coordinate systems arrive Homogeneous transformation matrix between They all have a unified standard algebraic form:

[0062] ;

[0063] In the above standard algebraic form, It is a 3×3 direction cosine orthogonal rotation matrix used to describe the coordinate system. Relative to coordinate system The three-dimensional spatial attitude deflection relationship; It is a 3×1 spatial translation column vector, representing the geometric distance vector between the origins of two coordinate systems; A one-dimensional zero vector to ensure dimensional compatibility in matrix multiplication.

[0064] The control module first performs pass-through filtering, voxel downsampling, and statistical outlier removal preprocessing on the raw point cloud data acquired by the vision sensor. Then, it uses a random sampling consensus algorithm to segment and separate the target workpiece point cloud from the complex background point cloud. For the extracted target workpiece point cloud, the system calculates its surface edges, corners, and normal vectors, and performs feature matching with a pre-set standard 3D model of the workpiece in memory, thereby initially calculating the target workpiece's pose in the camera-independent coordinate system.

[0065] Next, the control module substitutes the verified target workpiece surface feature points, one by one or in batches through parallel matrix operations, into the aforementioned continuous mapping transfer equation. Through this series of rigid body spatial transformations, the scattered local camera viewpoint point clouds are accurately reconstructed and mapped into the global unified coordinate system of the entire machine, forming a globally consistent three-dimensional spatial environment topology model. Based on this model, the control module uses a feature matching extraction algorithm to accurately output the global spatial absolute position and attitude positioning data of the target workpiece relative to the moving platform. This data conversion eliminates the cumulative positioning error caused by multi-stage serial mechanisms, providing unbiased initial spatial input conditions for subsequent triggering of dynamic counterweight planning and end-effector variable stiffness contact operations of the moving platform.

[0066] In this embodiment, the technical problem of unknown displacement of the tool center point and spatial rotation posture of the new tool relative to the end of the robotic arm after the multi-functional robotic arm performs opto-mechatronics quick-change flange switching operation is explained in detail, along with the specific implementation logic of vehicle-mounted self-calibration and the underlying mechanism of compliant anti-collision.

[0067] After confirming that the physical locking connection of the new working tool is complete, the control module automatically triggers the on-board self-calibration process, controlling the robotic arm to drive the end of the current tool to perform a probing spatial movement in the direction of the feature array on the self-calibration base.

[0068] Throughout the entire motion approach process, from the tool tip to its final physical contact with the rigid feature array, the control module synchronously outputs a constant-amplitude, low-stiffness bias excitation current to the excitation coil of the magnetorheologically compliant wrist. In this embodiment, the bias current is set to 10% to 20% of the rated maximum excitation current, ensuring the wrist is stably locked in a preset low-stiffness, high-damping rheological state. This purely hardware-level compliance setting can directly absorb and dissipate the impact kinetic energy generated when the tool touches the rigid base by relying on fluid shear deformation, effectively avoiding the risk of rigidity failure due to position closed-loop control errors.

[0069] During multi-point physical contact, a series of multi-dimensional force and torque sensors sample contact mechanics data in three-dimensional space at high frequency in real time and extract the resultant force vector. The control module continuously calculates the magnitude of the resultant force vector and rigorously compares it with a pre-set calibration threshold.

[0070] When the magnitude of the resultant force vector accurately reaches the preset calibration threshold, the system determines that the tool end and the self-calibration base feature array form a stable physical constraint connection. The preset calibration threshold is numerically set to be strictly greater than the background noise amplitude of the multidimensional force and torque sensor, and less than the maximum contact force threshold that the magnetorheological fluid can withstand under bias current. In this embodiment, the preset calibration threshold can be 5N~15N, and can be dynamically adjusted within this range according to the tool weight and sensor range to ensure that the contact action is effectively captured while remaining within the fluid elastic clearance protection range. At this time, the control system immediately triggers a high-speed interrupt response, capturing and recording the high-precision joint angle vector of the robotic arm at the moment of contact. ,in This represents the total number of motion axes of the robotic arm.

[0071] In order to fully calculate the spatial pose of the current tool, the robotic arm is controlled to change its posture and complete at least four non-collinear contact operations with different preset points of the feature array, thereby obtaining multiple sets of matching joint angle vectors.

[0072] For the Upon successful contact, the control module will record the joint angle vector. Substituting the theoretical forward kinematics model of the robotic arm, the basic pose matrix of the magnetorheologically compliant wrist end in the reference coordinate system of the moving platform at that instant was calculated. .

[0073] Let the spatial translational and rotational deviation parameter vector of the current tool tip relative to the magnetorheologically compliant wrist tip be . Based on the kinematic transmission chain, in the first... Under the initial contact condition, the calculated theoretical spatial position vector of the tool tip in the reference coordinate system This can be expressed as:

[0074] ;

[0075] in, It is a nonlinear spatial mapping function that includes homogeneous matrix multiplication and coordinate projection.

[0076] At the same time, the pre-stored data on the self-calibration base in the control module is retrieved. The actual absolute position vector of each physical contact point in the reference coordinate system is denoted as... .

[0077] Under ideal, absolutely error-free physical contact, the theoretical spatial position vector should be strictly equal to the actual absolute position vector. Based on this constraint, a nonlinear spatial distance residual objective function is constructed for all probe contact points. :

[0078] ;

[0079] in, To perform the total number of multi-point physical contacts, and To satisfy the rank condition for solving the six-degree-of-freedom spatial parameters.

[0080] To efficiently solve the aforementioned nonlinear least squares optimization problem, a Gauss-Newton iterative optimization algorithm is embedded in the control module. During the iterative calculation, the residual vector is first defined... And calculate the residual vector with respect to the unknown parameter vector. The first-order partial derivative Jacobian matrix JJ:

[0081] ;

[0082] Subsequently, based on the error gradient descent direction, the parameter update step size for each iteration is calculated using the following matrix equation. :

[0083] ;

[0084] Get update step size Subsequently, the control module performs nonlinear pose updates on the preceding parameter vectors: for spatial translation components, it performs vector addition and accumulation; for spatial rotation components, it performs multiplication and right multiplication updates by converting them into rotation matrices or quaternions, until the objective function is reached. The calculated value converged and decreased to less than Within the extremely small allowable tolerance range of mm, the iteration terminates, and the final output convergence parameters are the precise spatial translation and rotation deviation data of the new tool. This process not only filters out random errors caused by system mechanical backlash through algorithm iteration, but also achieves fully automatic and precise reconstruction of the actual pose of the tool end without the need for any external laser tracking equipment.

[0085] This embodiment details the spatial dynamics modeling and dynamic counterweight intervention mechanism that maintains the absolute dynamic stability of the machine's center of gravity when the system performs large-scale extension or load change operations in unstructured scenarios.

[0086] The overall physical mass distribution of the system is a highly dynamic, multivariate, time-varying parameter. To implement precise anti-tipping control, the control module first initializes and constructs a global static physical parameter array in memory, containing the moving platform, the independent links of the robotic arm, the current working tool, and the counterweight. This array stores detailed calibration mass data for each of the aforementioned independent components, denoted as set. ,in For the stationary mass of the mobile platform, to The following are the inherent masses of each link in the robotic arm, in order. For the quality of the currently mounted tools, The mass of the dynamic counterweight.

[0087] During operation, the system reads encoder feedback data from each joint in real time via the underlying high-speed bus. Through forward kinematics calculations, it calculates the instantaneous three-dimensional centroid coordinate vector of each independent link in the current pose, within the reference coordinate system of the moving platform. The instantaneous spatial displacement information of all components is denoted as a set. .in, For the fixed centroid coordinate vector of the mobile platform, to All are time-varying coordinate vectors that change in real time with the joint motion state. This is the feedback coordinate vector of the current counterweight block.

[0088] Based on the aforementioned mass array and instantaneous coordinate vector set, the control module calculates the three-dimensional position vector of the system's overall centroid in the mobile platform's reference coordinate system using a spatially weighted geometric projection equation. Its mathematical solution formula is defined as follows:

[0089] ;

[0090] in, The total mass scalar of all hardware configurations in the current system is expressed as follows: The three-dimensional position vector of the system's overall centroid in the mobile platform's reference coordinate system, calculated using the above formula. This serves as the real-time center of gravity position of the entire machine, which is used as the basis for subsequent anti-tipping control calculations.

[0091] The mobile platform is supported on the ground by an array of wheels. Connecting the vertical support points of each wheel to the ground on a horizontal plane forms a convex polygon. Based on the internal geometry of this polygon, the control module reduces a predetermined safety margin boundary inwards, defining a safe support area to prevent the mobile platform from overturning. Specifically, the reduction in the safety margin boundary is based on the zero-moment point stability margin theory, and is set at 10% to 15% of the overall contour feature size of each side of the convex polygon, reduced inwards towards the geometric center, to reserve dynamic anti-overturning redundancy for the system in response to unexpected ground undulations.

[0092] To counteract the overturning moment generated when the robotic arm extends outward, the vertical projection of the overall center of gravity of the machine must always fall strictly within the aforementioned safety support area. (Control module extraction) horizontal components Establish the boundary equations for the inequality region:

[0093] ;

[0094] Under the constraint of this inequality region, the target centroid coordinates corresponding to the robot arm's posture at the next expected moment of operation are substituted into... In the formula, the system will use the coordinates of the center of mass of the counterweight. As the sole independent adjustment variable, the gradient projection method or convex quadratic programming algorithm is used for optimization to calculate the target position vector of the counterweight block that enables the horizontal centroid component to return to the geometric center of the safety support area. .

[0095] After the calculation is completed, the control module will contain the target position vector. The digital drive commands are sent to the counterweight drive mechanism (multi-axis guide rail driver) inside the mobile platform via the underlying controller area network bus. The counterweight drive mechanism responds immediately, using servo motors and precision lead screw modules to physically drive the counterweight block to smoothly migrate to the target coordinate position. This process is executed iteratively throughout the entire cycle of the robotic arm's movement, significantly reducing the risk of mechanical instability of the multi-functional robotic arm under extreme working conditions from the underlying hardware structure.

[0096] In this embodiment, a hardware-level magnetorheological variable stiffness closed-loop control mechanism based on mechanical error feedback is described in detail to address the technical problem that the end effector of a multi-functional robotic arm is prone to workpiece damage and system oscillation due to rigid collision when performing contact operations such as grinding and assembly.

[0097] Throughout the entire lifecycle of the contact operation, multi-dimensional force and torque sensors installed at the end effector acquire real-time three-dimensional contact mechanics data during the robot's interaction with the external environment at a set high-frequency sampling rate. This contact mechanics data, after signal conditioning and filtering, is extracted into the actually measured contact force vector, denoted as... .

[0098] To maintain constant and safe physical contact, the control module pre-sets a reference contact force vector that characterizes the ideal task profile, denoted as . The system obtains the mechanical error vector at the current operation moment by performing a vector subtraction operation on the two sets of spatial vectors mentioned above. :

[0099] ;

[0100] After obtaining the mechanical error vector, the control module calls the underlying adaptive feedback control law to generate the excitation control current for closed-loop drive. The adaptive feedback control law specifically employs a proportional-integral control architecture that combines spatial decoupling.

[0101] In the process of adaptive feedback control law calculation, the mechanical error vector is first multiplied by the proportional gain matrix to perform proportional gain calculation, generating a proportional control component for rapid response to transient shocks. Simultaneously, the mechanical error vector is continuously integrated over time and multiplied by the integral gain matrix to generate an integral control component used to eliminate steady-state contact error. Its mathematical expression is:

[0102] ;

[0103] ;

[0104] in, For the set proportional gain matrix, For the given integral gain matrix, The continuous integration time variable; in this embodiment, the range of the proportional gain is set to... The integral gain range is set to In engineering applications, parameter tuning can be accomplished using the critical proportionality method or the Ziegler-Nichols method.

[0105] Subsequently, the proportional control component and the integral control component calculated above are algebraically superimposed to finally output the excitation control current that is dynamically adjusted in real time. :

[0106] ;

[0107] The excitation control current is directly injected into the excitation coil inside the magnetorheological compliant wrist. Accompanied by the high-frequency pulsation of the excitation control current, the excitation coil generates an external magnetic field of alternating strengths and weaknesses within the sealed cavity. According to the aforementioned constitutive model of the magnetorheological fluid material, the instantaneous change in the magnetic induction intensity of the external magnetic field directly and linearly regulates the rheological state and dynamic yield shear stress of the magnetorheological fluid.

[0108] In terms of macroscopic mechanical transmission, this change in shear stress directly maps to a physical change in the equivalent mechanical stiffness of the compliant wrist. When the actual contact force vector suddenly increases (i.e., encountering a rigid impact), the mechanical error vector surges positively, and the feedback control law instantly reduces the output amplitude of the excitation control current. At this time, the internal magnetic field weakens, the yield stress of the magnetorheological fluid drops sharply, and the wrist exhibits an extremely low stiffness and high damping state in the contact direction, absorbing the impact kinetic energy through a slight physical clearance at the end.

[0109] Conversely, when the actual contact force is insufficient, the feedback control law increases the excitation control current, causing the magnetorheological fluid to rapidly coalesce into a high-strength, chain-like solid-like structure. The wrist stiffness increases accordingly, thereby applying a stable and effective working contact force to the target workpiece. This control closed loop deeply couples electrical signal processing with the physical phase change of the smart fluid material, overcoming the computational power loss and response lag caused by inaccurate dynamic models in pure software impedance control, and achieving hardware-level variable stiffness operation with extremely low response latency.

[0110] This embodiment addresses the issue of abrupt changes in the overall system state caused by frequent tool changes in unstructured operations. It details a multi-source global physical collaborative workflow triggered by an opto-mechatronics quick-change flange. This workflow deeply integrates sensing and acquisition, motion control, and underlying physical hardware to construct a closed-loop adaptive operation system.

[0111] When the operational requirements change, the end effector of the robotic arm unloads the current work tool via the opto-mechatronics quick-change flange and completes the physical locking and electrical communication connection of the new work tool. At the instant the locking mechanism is fully closed, the internal communication interface of the opto-mechatronics quick-change flange sends a hardware-level high-level interrupt trigger signal to the control module. Upon receiving this trigger signal, the control module immediately reads the device-specific physical parameters stored in the read-only memory (ROM) of the new work tool via the communication bus, thereby obtaining the inherent calibration quality of the new work tool. and the theoretical centroid distribution vector in the tool's own local coordinate system. .

[0112] While acquiring the built-in static parameters of the aforementioned tools, the control module simultaneously initiates the vehicle-mounted self-calibration process described earlier; at this time, system parameter updates and dynamic counterweight intervention are strictly divided into two time-series execution phases:

[0113] Phase 1 (Rapid Pre-weighting): The control module first utilizes the read inherent calibration mass of the new working tool. With the theoretical centroid distribution vector The system performs a preliminary and rapid overwrite of the overall machine mass and center of mass parameter array, and uses this to calculate and drive the counterweight to perform initial translation. This stage utilizes theoretical parameters to achieve a zero-delay response, quickly offsetting the major abrupt change in the center of mass caused by the instantaneous replacement of the tool body.

[0114] The second stage (high-precision correction): The system forces the robotic arm to make multi-point physical contact between the new working tool and the feature array on the self-calibration base. An iterative algorithm is used to solve for the precise spatial pose transformation matrix of the end effector of the new working tool relative to the magnetorheologically compliant wrist, denoted as... After self-calibration is complete, the control module uses the precisely projected actual pose parameters to perform a final correction on the overall parameter array, driving the counterweight to fine-tune to the final absolute equilibrium position. Let the original working tool mass before unloading be... The original centroid distribution vector is The original overall weight of the system before the update was: The original three-dimensional position vector of the composite centroid is .

[0115] The control module executes the following global centroid state update equation to recalculate the three-dimensional position vector of the overall centroid of the system after the new operating tool is installed. :

[0116] ;

[0117] In the above equation, This represents the spatial vector obtained by accurately projecting the local centroid of the new working tool onto the global coordinate system using the pose matrix obtained through self-calibration. The update equation algebraically fuses the parameter jumps caused by tool switching with the actual spatial pose obtained through on-board self-calibration.

[0118] Based on the newly calculated integrated centroid three-dimensional position vector Based on the preset safety support area constraints, the control module recalculates and outputs the latest target position vector for the counterweight. This position vector, as a low-level drive command, is directly sent to the counterweight drive mechanism inside the mobile platform, driving the counterweight to move to the new physical position and completing the pre-balancing of the additional overturning moment introduced by the new working tool load.

[0119] After completing the self-calibration detection and mobile platform parameter reconstruction, the robotic arm, carrying the new tool, enters the actual unstructured contact operation mode. At this time, the multi-dimensional self-sensing module, jointly constructed by vision sensors and multi-dimensional force and torque sensors, continuously monitors environmental feedback. When rigid contact occurs, the system calls a mechanical error feedback algorithm containing a proportional-integral control law to output continuously modulated excitation control current to the magnetorheological compliant wrist in real time, adjusting the shear stress of the internal fluid to change the equivalent physical stiffness of the wrist.

[0120] The entire execution sequence is triggered by a single opto-electro-mechanical connection action, seamlessly linking the capture of new environmental features, precise geometric pose recalibration, transfer of the machine's dynamic spatial counterweight, and electromagnetic-physical-rheological control during the operation sequence. This collaborative workflow enables the multi-functional robotic arm to autonomously maintain the dynamic balance of the underlying physical structure and ensure safe and smooth contact during complex task switching, without requiring manual secondary configuration or external equipment intervention.

[0121] The following section, using the actual working conditions of heavy-duty grinding operations and switching gripping operations, further illustrates the complete full-process control logic of the coordinated linkage of various modules of the device of the present invention:

[0122] Global positioning and entry: The mobile platform moves to the work area, the vision sensor scans the environment and collects the workpiece point cloud. Through feature extraction, matching and coordinate continuous spatial mapping transformation, the target workpiece is positioned with high precision in a global unified coordinate system.

[0123] Dynamic stability and anti-tipping: The system reads the parameters of the currently mounted grinding tools, calculates the position of the center of gravity of the whole machine in the extended posture of the robotic arm in real time, and drives the counterweight block inside the mobile platform to translate to the target position through inequality constraint solution to counteract the overturning torque and ensure the absolute stability of the platform.

[0124] Hardware-level constant-force grinding: A robotic arm moves the grinding tool closer to and makes physical contact with the workpiece surface. Multi-dimensional force and torque sensors collect the contact force in real time and calculate the mechanical error. An adaptive feedback control law dynamically adjusts the excitation control current flowing into the magnetorheological compliant wrist based on this, achieving zero-delay stiffness adjustment and constant-force compliant grinding by changing the fluid shear stress.

[0125] Cross-task global physical collaboration: If a gripping tool needs to be changed after grinding, the system unloads the original tool and locks the new tool using an opto-mechatronics quick-change flange. The hardware locking signal of the flange instantly triggers a collaborative interruption: the system automatically reads the mass of the new tool, synchronously starts a self-calibration process to solve for the absolute pose of the new tool, immediately overwrites the overall mass and center of mass parameter array using the new parameters, and synchronously drives the counterweight mechanism to transfer the counterweight to the new equilibrium point. The entire collaborative linkage process is completed in a closed loop without any manual intervention, and then the robotic arm automatically moves into the next compliant gripping operation.

[0126] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended technical solutions and their equivalents.

Claims

1. A self-sensing multifunctional robotic arm execution device based on a motion platform, characterized in that, This includes the underlying hardware, sensing module, and control module; The underlying hardware includes a mobile platform, a robotic arm, a self-calibrating base, a dynamic counterweight mechanism, and a magnetorheologically compliant wrist carrying the working tools; the dynamic counterweight mechanism includes a counterweight block and a counterweight drive mechanism. The sensing module includes at least a visual sensor and a multi-dimensional force and torque sensor for data acquisition. The control module is communicatively connected to the underlying hardware and the sensing module, and is configured to perform the following controls: Calibration control: After changing the working tool, control the robotic arm to drive the working tool to contact the self-calibration base, and calculate and obtain the actual position and pose of the working tool; Anti-tipping control: Based on the overall machine mass parameters and the calculated real-time center of gravity position of the whole machine, calculate the target position of the counterweight, and control the counterweight drive mechanism to drive the counterweight block to the target position of the counterweight; Compliance control: During contact operation, an excitation current is generated based on the contact force data obtained by the multi-dimensional force and torque sensor and output to the magnetorheological compliant wrist to adjust the shear stress.

2. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 1, characterized in that, The control module is also configured to: locate the target workpiece by coordinate transformation based on the point cloud data acquired by the vision sensor and the current pose data of the robotic arm.

3. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 1, characterized in that, The control module calculates and obtains the actual pose of the working tool, and is specifically configured as follows: The robotic arm is controlled to drive the working tool to make multi-point contact with the self-calibration base, and during the contact process, a set bias excitation current is output to the magnetorheological compliant wrist to keep it in a compliant state. When the contact force reaches a preset calibration threshold, the joint angle of the robotic arm is recorded; Based on the joint angle and the known constraint position of the self-calibration base, the spatial deviation of the working tool relative to the magnetorheologically compliant wrist is solved by an iterative algorithm.

4. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 1, characterized in that, The control module calculates the real-time centroid position of the entire machine and is specifically configured as follows: Obtain the mass parameters and real-time centroid coordinates of the mobile platform, the robotic arm, the working tool, and the counterweight; The mass parameters and their real-time centroid coordinates are weighted and calculated to obtain the three-dimensional position of the overall centroid of the machine in the reference coordinate system of the mobile platform, which is used as the real-time centroid position of the machine.

5. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 4, characterized in that, The control module calculates the target position of the counterweight and controls the counterweight drive mechanism to drive the counterweight block, specifically configured as follows: Establish a secure support area to ensure the stability of the mobile platform; Using the constraint that the horizontal coordinates of the three-dimensional position are located within the safety support area, the target position of the counterweight is calculated and sent to the counterweight drive mechanism.

6. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 1, characterized in that... The contact force data includes the actual measured contact force; the control module generates the excitation current, specifically configured as follows: The mechanical error is obtained by calculating the difference between the actual measured contact force and the preset reference contact force. Based on the aforementioned mechanical error, the excitation current is generated through a feedback control law to achieve closed-loop control.

7. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 6, characterized in that, The magnetorheologically compliant wrist has an excitation coil inside and is filled with magnetorheological fluid; The excitation coil receives the excitation current to generate an external magnetic field. The magnetic induction intensity of the external magnetic field is linearly related to the excitation current, and is used to change the rheological state of the magnetorheological fluid to adjust the shear stress.

8. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 6, characterized in that, The control module generates the excitation current for closed-loop control through a feedback control law, specifically configured as follows: The mechanical error is subjected to proportional gain calculation and time integration calculation respectively to obtain proportional component and integral component; the proportional component and the integral component are added to obtain the excitation current.

9. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 1, characterized in that... The magnetorheological compliant wrist is equipped with a quick-change flange at its end, through which different working tools can be switched.

10. The self-sensing multifunctional robotic arm execution device based on a motion platform according to claim 9, characterized in that, After switching to a new working tool, the control module simultaneously uses the mass and center of gravity parameters of the new working tool to update the overall machine mass parameters when calculating the actual position of the new working tool. This triggers the recalculation of the real-time center of gravity position of the entire machine and updates the counterweight target position, thereby achieving physical coordination and linkage between tool switching, perception calibration, and chassis anti-tipping.

Images