Low-interference non-inductive grabbing method of floating target control mechanical arm based on artificial potential field

Abstract

The invention particularly relates to a floating target control mechanical arm low-interference non-inductive grabbing method based on an artificial potential field, and relates to the technical field of robot motion control and intelligent operation. The method comprises the following steps: planning an initial trajectory by adopting a quintic polynomial; constructing an artificial potential field comprising a radial potential field and a tangential potential field, generating a virtual force, introducing an impedance model to dynamically correct a track, and suppressing contact impact; constructing a super-torsion sliding mode controller based on symmetric preset performance to realize joint space trajectory tracking; introducing a reinforcement learning algorithm to carry out iterative optimization on the control parameters; and a grabbing detection and re-planning mechanism is set, and track re-planning is automatically triggered when grabbing fails. Low-contact-force grabbing can be achieved without depending on a force sensor, and the method has the advantages of being high in control precision, high in robustness, good in autonomy and the like and is suitable for mechanical arm grabbing tasks in complex environments such as floating targets.

Description

Technical Field

[0001] This invention relates to the field of robot motion control and intelligent operation technology, specifically to a method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field, belonging to the field of control technology that combines compliant robot control, motion planning, and learning optimization. Background Technology

[0002] With the widespread application of industrial robots in assembly, sorting, and grasping operations, achieving stable and reliable object grasping by robotic arms in unstructured or semi-structured environments has become a crucial research direction in the field of robot control. Especially in application scenarios where the target object's position is uncertain, the environment is disturbed, or high contact safety requirements exist, how to reduce unnecessary contact impacts and force interference while ensuring grasping success rates has been a key long-standing concern in this technological field.

[0003] Existing robotic arm grasping systems typically consist of a trajectory planning layer and a motion control layer. In the trajectory planning layer, methods such as polynomial trajectory planning and spline curve planning are commonly used to generate the desired motion trajectory of the end effector from its initial pose to its target pose. These planning methods primarily focus on the continuity of position, velocity, and acceleration, enabling the robot to move smoothly between given start and end poses. In the motion control layer, joint-space or Cartesian-space control methods are often used to achieve the robotic arm's tracking of the planned trajectory.

[0004] In tasks involving contact with target objects or the environment, impedance control and compliant control methods are widely adopted in robotic arm control systems to reduce contact impact and improve system compliance. These methods construct a dynamic relationship between force and displacement, enabling the robotic arm to exhibit characteristics similar to a spring-damped-mass system during contact, thereby reducing the impact risk associated with rigid control. Furthermore, the artificial potential field method, a classic motion planning and obstacle avoidance approach, guides the robotic arm towards the target and away from obstacles by constructing attractive and repulsive potential fields, and is widely used in robot path planning and local motion correction.

[0005] At the control strategy level, to address model uncertainties and external disturbances, some studies have introduced robust control methods such as sliding mode control and adaptive control to improve system stability and tracking accuracy. Meanwhile, with the improvement of computing power, data-driven methods such as reinforcement learning are gradually being applied to the field of robot control for parameter tuning, strategy optimization, or decision control in complex environments.

[0006] Although various methods for robotic arm grasping and contact control have been proposed in the existing technology, the following shortcomings still exist in terms of achieving zero-contact force grasping, control robustness, and parameter adaptation: (1) Existing grasping trajectory planning methods are mainly based on position planning and lack mechanical constraints on the contact process: Existing technologies widely employ methods such as polynomial trajectories and spline curves for trajectory planning at the end effector of robotic arms. These methods primarily focus on the continuity of position, velocity, and acceleration, assuming the target object's position is fixed and can be accurately reached. However, these methods do not explicitly constrain the force changes before and after contact when the end effector approaches the target object, easily leading to impact forces at the moment of contact. This makes it difficult to meet the grasping requirements of strictly limited or near-zero contact forces.

[0007] (2) Methods based on artificial potential fields are mostly used for obstacle avoidance or path guidance, and are rarely used directly for contact force control: Artificial potential field methods are mainly used for obstacle avoidance or path guidance in existing technologies, and their potential field forces are usually used as auxiliary corrections in the planning or motion layers. In existing methods, the potential field forces are not systematically introduced into the robot dynamics or impedance model, making it difficult to establish a clear physical correspondence between the potential field guidance effect and the contact force evolution process. Therefore, their application in delicate contact or low-force grasping tasks is limited.

[0008] (3) The control law parameters depend on manual setting, resulting in insufficient adaptability and robustness: While existing robotic arm control systems employ methods such as sliding mode control and super-torsional sliding mode control, which offer strong robustness, their control law parameters are typically obtained through manual experience or offline tuning. Under varying load conditions, different trajectory segments, or in the presence of modeling errors and external disturbances, fixed control parameters struggle to simultaneously ensure both tracking accuracy and control smoothness throughout the entire task, thus impacting the overall system performance.

[0009] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0010] This invention provides a method for low-disturbance, contactless grasping of floating targets by a robotic arm based on an artificial potential field. It aims to solve the problems of existing robotic arm grasping technology in floating target grasping scenarios, such as the lack of mechanical constraints in trajectory planning leading to contact impact, the ineffective use of artificial potential field method for contact force control, poor adaptability of manually set control parameters, and the lack of effective failure detection and replanning mechanisms, making it difficult to achieve low-disturbance, contactless, zero-contact force grasping.

[0011] Other features and advantages of the invention will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0012] According to a first aspect of the present invention, a method for low-disturbance, non-intrusive grasping of a floating target using a robotic arm based on an artificial potential field is provided, the method comprising: The initial pose of the robotic arm end effector and the desired grasping pose of the target object are obtained, and the initial planned trajectory of the end effector is generated by the fifth-order polynomial trajectory planning method. Based on the relative spatial relationship between the end effector and the target object, an artificial potential field is constructed, which includes a radial potential field and a tangential potential field, generating radial potential field force and tangential potential field force respectively, which together constitute the virtual force of the artificial potential field. The artificial potential field virtual force is introduced into the impedance control model to dynamically correct the initial planned trajectory and generate the reference trajectory after impedance correction. A super torsional sliding mode controller based on symmetric preset performance is constructed for trajectory tracking control of the robotic arm joint space, so that the robotic arm moves according to the reference trajectory; A reinforcement learning algorithm is introduced to iteratively optimize the control parameters in the super-torsional sliding mode controller; After each planned trajectory is executed, the relative positional relationship between the end effector and the target object is detected. If a grasping failure or target object position shift is detected, a replanning mechanism is triggered to regenerate the planned trajectory and repeat the above steps.

[0013] In some exemplary embodiments, the radial potential force is constructed in the following specific ways: Construct the radial potential function: ; in, , Let the radius be the radius of the block. The distance between the end effector of the robotic arm and the center of the object. , , , A constant defined by humans; Differentiation yields the radial potential force: .

[0014] In some exemplary embodiments, the tangential potential force is constructed as follows: Construct the tangential potential function: ; Differentiation yields the tangential potential force: ; in, Position of the robotic arm end effector relative to the spherical target radial error, A constant defined by humans.

[0015] In some exemplary embodiments, constructing the super-torsional sliding mode controller based on symmetric preset performance includes: A symmetric preset performance function is constructed to constrain the dynamic evolution of joint tracking error; Construct a super-torsional sliding mode control law to improve the system's robustness under modeling uncertainties and external disturbances; By combining the dynamic compensation term, a complete joint space control law is formed.

[0016] In some exemplary embodiments, the super-torsional sliding mode controller is represented as:

[0017]

[0018] in, For dynamic compensation terms, To control the maximum input amplitude, The inertial matrix of the robotic arm joint space. The Super-Twisting sliding mode control law is constructed.

[0019] In some exemplary embodiments, the reinforcement learning algorithm employs the Soft Actor-Critic algorithm, which constructs a reward function based on trajectory tracking error to optimize control parameters online or iteratively, thereby achieving adaptive parameter adjustment.

[0020] In some exemplary embodiments, the replanning mechanism includes: When the distance between the end effector and the target object exceeds a preset threshold, or the target object is pushed away from its original position or the grasping fails, trajectory replanning is automatically triggered. Based on the current terminal state and the updated target object position, the planned trajectory is regenerated, and the artificial potential field correction and control tracking process is re-executed.

[0021] According to a second aspect of the present invention, a low-disturbance, non-intrusive grasping system for a floating target manipulation robotic arm based on an artificial potential field is provided, the system comprising: The trajectory planning module is used to generate an initial planned trajectory based on the initial pose and the target pose. The artificial potential field construction module is used to generate radial and tangential potential forces based on the relative spatial relationship between the end point and the target, thus forming a virtual artificial potential field force. The impedance trajectory correction module is used to introduce the virtual force of the artificial potential field into the impedance model to dynamically correct the planned trajectory. The joint space control module is used to build a super torsional sliding mode controller based on symmetrical preset performance to realize the trajectory tracking control of the robotic arm; The control parameter learning module is used to iteratively optimize the control parameters using reinforcement learning algorithms. The crawling detection and replanning module is used to detect crawling results and determine whether a replanning mechanism has been triggered.

[0022] The low-disturbance, contactless grasping method for floating target manipulation robotic arms based on artificial potential fields provided by embodiments of the present invention can effectively control the contact process of the robotic arm without relying on real contact force sensing. Simultaneously, by optimizing the joint space control law parameters through reinforcement learning, the robustness and adaptability of the system are improved, and the reliability of the grasping task is enhanced through failure detection and reprogramming mechanisms. Compared with existing technologies, it has the following beneficial effects: (1) This invention introduces a virtual force generated by an artificial potential field into impedance control to dynamically correct the planned trajectory of the end effector. This makes contact control no longer solely dependent on the smoothness of the position trajectory, but rather constrains the approach process through the force-displacement relationship. Since the radial potential field force can actively limit the approach speed of the end effector towards the target object, and the tangential potential field force can guide the end effector to adjust its attitude and position, the end effector gradually enters a compliant state before contact occurs, effectively suppressing the impact force at the moment of contact. This enables a grasping effect with low contact force or even near-zero contact force, improving the safety and stability of the grasping process.

[0023] (2) The artificial potential field force used in this invention is a virtual force calculated based on geometric relationships. It can control the contact process without relying on force sensors or additional hardware detection devices. This avoids the problems of increased system cost, structural complexity and signal noise interference caused by the introduction of force sensors, making the overall system structure simpler and more feasible for engineering implementation. It is suitable for application scenarios with high requirements for cost and reliability.

[0024] (3) This invention employs a super-torsional sliding mode controller based on symmetric preset performance for trajectory tracking in joint space, and optimizes the control law parameters online or iteratively through reinforcement learning algorithms, enabling the control parameters to adaptively adjust according to different trajectory segments, robotic arm configurations, and external disturbance conditions. Compared with fixed parameter control methods that rely on manual experience tuning, this method improves trajectory tracking accuracy and control smoothness while ensuring system robustness, and enhances the system's adaptability under complex working conditions.

[0025] (4) The present invention introduces a grasping result detection and replanning mechanism during the grasping process. When the target object deviates, the end clips through the clipping or the grasping fails, the trajectory replanning can be automatically triggered and the grasping process can be re-executed, reducing the need for manual intervention and improving the system's autonomy and grasping success rate in unstructured environments.

[0026] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0027] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0028] Figure 1 This is a schematic diagram of the low-disturbance, non-intrusive grasping method of the floating target control robotic arm based on artificial potential field according to the present invention. Figure 2 This is a schematic diagram of the planned trajectory and the actual trajectory of the robotic arm end effector after impedance control correction, which is an exemplary embodiment of the present invention. Figure 3 This is a schematic diagram illustrating a successful grasping action by the robotic arm in an exemplary embodiment of the present invention. Detailed Implementation

[0029] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the invention will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0030] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0031] To address the shortcomings and deficiencies of existing technologies, this example embodiment provides a low-disturbance, non-intrusive grasping method for a floating target manipulation robotic arm based on an artificial potential field. By introducing a virtual force of an artificial potential field during the trajectory planning stage and combining it with impedance control, contact impact suppression is achieved during the end effector's approach to the target object. Simultaneously, the joint space control law parameters are iteratively optimized using a reinforcement learning algorithm, improving the robotic arm's trajectory tracking performance and robustness under conditions of modeling errors and disturbances. Furthermore, in the event of grasping failure or target position deviation, the system can automatically replan the trajectory based on position detection results, thereby enhancing the grasping success rate and system autonomy.

[0032] refer to Figure 1 As shown, the specific steps may include: Step S1: Obtain the initial pose of the robotic arm end effector and the desired grasping pose of the target object, and generate the initial planned trajectory of the end effector using a fifth-order polynomial trajectory planning method. Step S2: Based on the relative spatial relationship between the end effector and the target object, an artificial potential field is constructed. The artificial potential field includes a radial potential field and a tangential potential field, which generate radial potential field force and tangential potential field force respectively, and together constitute the virtual force of the artificial potential field. Step S3: Introduce the artificial potential field virtual force into the impedance control model to dynamically correct the initial planned trajectory and generate the reference trajectory after impedance correction. Step S4: Construct a super torsional sliding mode controller based on symmetric preset performance for trajectory tracking control of the robotic arm joint space, so that the robotic arm moves according to the reference trajectory; Step S5: Introduce a reinforcement learning algorithm to iteratively optimize the control parameters in the super torsional sliding mode controller; Step S6: After each planned trajectory is executed, the relative positional relationship between the end effector and the target object is detected. If a grasping failure or target object position shift is detected, the replanning mechanism is triggered to regenerate the planned trajectory and repeat the above steps.

[0033] The steps in this exemplary embodiment will now be described in more detail with reference to the accompanying drawings and embodiments.

[0034] In step S1, at the start of the grasping task, based on the initial pose of the robotic arm's end effector and the desired grasping pose of the target object, a fifth-order polynomial trajectory planning method is used to generate the desired trajectory of the end effector. The fifth-order polynomial trajectory satisfies the continuity constraints of position, velocity, and acceleration within the planning period to ensure trajectory smoothness. The generated trajectory serves as the initial planned trajectory for the end effector, primarily providing a reference trajectory for subsequent virtual force control and impedance correction, rather than being directly used to perform contact actions.

[0035] Specifically, a two-dimensional, two-degree-of-freedom robotic arm is established to generate a two-dimensional spherical block with a radius of... Distance between the end of the robotic arm and the center of the block Define the position of the robotic arm's end effector in Cartesian space. robotic arm joint angle Joint angular velocity Joint angular acceleration ,time Planning time Target location The initial planning time is 5 seconds, the replanning time is 2 seconds, and the initial position of the robotic arm gripper is... initial position of the block spherical target position ; Design a fifth-order polynomial , Require: ; ; ; Solve Parameters are used to obtain the expected tracking trajectory in Cartesian space. The expected tracking speed in Cartesian space The expected tracking acceleration in Cartesian space ; Will , , Substituting the inverse kinematics of the robotic arm, the desired motion trajectory in joint space can be obtained. Desired joint angular velocity Desired joint angular acceleration ; Solving for trajectory feedforward acceleration , Define joint position error , Joint speed error .

[0036] In step S2, an artificial potential field is constructed to generate a virtual force based on the relative spatial relationship between the end effector and the target object. This virtual force is used to regulate the motion behavior of the end effector as it approaches the target object. The artificial potential field includes a radial potential field and a tangential potential field. A radial potential function is constructed based on the relative distance between the end effector and the target object, and its gradient is used to generate a radial potential force to limit the approach speed of the end effector in the radial direction, thereby suppressing excessive impact at the moment of contact. A tangential potential function is constructed based on the tangential direction of the end effector relative to the surface of the target object, and its gradient is used to generate a tangential potential force to guide the end effector to adjust along the target surface direction, improving the reachability of the end effector's attitude and contact position. The radial and tangential potential forces together constitute the virtual force of the artificial potential field acting on the end effector. This virtual force does not rely on real contact force sensing signals but is calculated based on geometric relationships and potential functions.

[0037] Specifically, construct the radial potential function: ; in, , , , , A constant defined by humans; Differentiation yields the radial potential force: ; Construct the tangential potential function: ; Differentiation yields the tangential potential force: ; in, Position of the robotic arm end effector relative to the spherical target radial error, A constant defined by humans; Total potential force .

[0038] In step S3, the artificial potential field virtual force is introduced into the impedance model to dynamically correct the planned trajectory, so that the end effector exhibits controlled compliant behavior as it approaches the target object. This allows the contact process to be controlled before actual contact occurs, achieving low or near-zero contact force contact.

[0039] Specifically, the total potential field force With quintic polynomial programming trajectory Substituting into the second-order linear impedance control model in Cartesian space:

[0040] Obtain the planned trajectory corrected by impedance control .in , , These are the expected virtual inertia matrix, expected virtual damping matrix, and expected virtual stiffness matrix in Cartesian space at the end of the robotic arm, respectively.

[0041] In step S4, after obtaining the reference trajectory corrected by the impedance model, the robotic arm is driven to execute the trajectory using a joint space control method. Specifically, a super-torsional sliding mode controller (SPP-ST-SM) based on symmetric preset performance is constructed for trajectory tracking control in joint space. The controller constrains the dynamic evolution of joint tracking error by introducing a symmetric preset performance function, causing the error to converge to a given range within a predetermined time. At the same time, the super-torsional sliding mode structure is used to improve the robustness of the system under conditions of modeling uncertainty and external disturbances.

[0042] Specifically, construct a preset performance function. , ; right Taking the derivative, we get , ; in, , , A constant defined by humans; Constructing asymmetric mapping functions

[0043] in, , A constant defined by humans and satisfying ; definition Differentiating, we get , achievable

[0044] Constructing sliding surfaces , in, , A constant defined by humans; Constructing Super-Twisting sliding mode control laws , in, , A constant defined by humans; Constructing state-dependent time-varying coefficient functions , , in,

[0045] ; Differentiation yields, ; Constructing dynamic compensation terms ; in, ; Within the planned timeframe, based on the tracking trajectory error, a super-torsional sliding mode controller (SPP-ST-SM) with symmetrical preset performance is constructed. The controller form is as follows: ; ; in, To control the maximum input amplitude, The inertial matrix of the robotic arm joint space. The Super-Twisting sliding mode control law is constructed.

[0046] Based on the dynamic model of the robotic arm's gripper end:

[0047] in, The coupling matrix of Coriolis force and centrifugal force in the joint space of the robotic arm. The gravitational torque compensation vector in the joint space of the robotic arm.

[0048] Solving for joint angular acceleration The joint angular velocity is obtained by integration. and the joint angle of the robotic arm ; Robotic arm joint angle Substitute the forward kinematics of the robotic arm into the equations to solve for the position of the robotic arm's end effector in Cartesian space. .

[0049] In step S5, to overcome the problem that traditional sliding mode controller parameters rely on manual experience for tuning, this invention introduces a reinforcement learning algorithm to iteratively optimize the joint space control law parameters. The reinforcement learning algorithm employs the Soft Actor-Critic (SAC) algorithm, which interacts with the robotic arm control system to construct a reward function based on the trajectory tracking error. While ensuring system stability, the reinforcement learning module adjusts the key control parameters in the SPP-ST-SM controller, enabling the controller parameters to adapt to different trajectory segments, robotic arm configurations, and disturbance conditions, thereby improving overall tracking performance and control stability.

[0050] Specifically, a black-box parameter optimization framework based on Soft Actor-Critic (SAC) is established to optimize control parameters. , , , , , , , To optimize the problem, the multidimensional parameter tuning problem of the SPP-ST-SM controller is modeled as a black-box optimization problem in a continuous action space.

[0051] Each step in the reinforcement learning environment corresponds to a complete robotic arm simulation, with the state set as a constant placeholder [0], thus formalizing the problem into a stateless continuous parameter optimization problem based on policy search.

[0052] The SAC algorithm is used to generate continuous parameter vectors, with the action space focused on the ±30% range around the empirical initial values, and combined with an automatic entropy adjustment mechanism to balance exploration and convergence.

[0053] Before training, empirical parameters are evaluated by calling real simulation functions and injected into the empirical replay pool as valid samples to achieve a warm start and alleviate the performance degradation caused by cold start.

[0054] By using a custom callback function to periodically perform independent simulation evaluations using the deterministic output of the current policy, the system is only used to update and save historical optimal parameters and models, and does not participate in policy training, thereby ensuring the stability and reliability of the optimal solution record.

[0055] In step S6, after each planned trajectory segment is executed, the relative positional relationship between the end effector and the target object is detected. If the detection result indicates that the target object has been pushed away from its original position or the grasping has failed, a replanning mechanism is triggered. The replanning mechanism regenerates the planned trajectory based on the current end effector state and the updated target object position, and repeats the above-mentioned artificial potential field construction, impedance trajectory correction, and control execution process, thereby achieving automatic recovery of the grasping task and improving the system's autonomy and success rate in complex environments.

[0056] Specifically, Position detection is performed at any time. if , Program ended else if && , Replanning else if && This refers to clipping, where in the actual physical environment, the object will collide with objects at a distance. Block position updated to At this point, the new target location is obtained. Re-planning; Among them, the acceleration of the block Simplify the block into an elastic system. Its elastic coefficient.

[0057] in, This is the tolerance threshold for the grasping accuracy of the robotic arm's end effector. This refers to the duration of the collision between the end effector of the robotic arm and the object.

[0058] The above method is implemented in a computer, using dual robotic arms to grasp the object in two-dimensional space. After the program finishes, an image is generated, and the planned trajectory of the robotic arm's end effector, corrected by impedance control, and the actual trajectory of the end effector are plotted. Figure 2 As shown, an animation of the robotic arm grasping the block is generated, and a diagram illustrating a successful grasp is drawn, as follows. Figure 3 As shown.

[0059] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the claims.

[0060] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is defined only by the appended claims.

Claims

1. A method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field, characterized in that, The method includes: The initial pose of the robotic arm end effector and the desired grasping pose of the target object are obtained, and the initial planned trajectory of the end effector is generated by the fifth-order polynomial trajectory planning method. Based on the relative spatial relationship between the end effector and the target object, an artificial potential field is constructed, which includes a radial potential field and a tangential potential field, generating radial potential field force and tangential potential field force respectively, which together constitute the virtual force of the artificial potential field. The artificial potential field virtual force is introduced into the impedance control model to dynamically correct the initial planned trajectory and generate the reference trajectory after impedance correction. A super torsional sliding mode controller based on symmetric preset performance is constructed for trajectory tracking control of the robotic arm joint space, so that the robotic arm moves according to the reference trajectory; A reinforcement learning algorithm is introduced to iteratively optimize the control parameters in the super-torsional sliding mode controller; After each planned trajectory is executed, the relative positional relationship between the end effector and the target object is detected. If a grasping failure or target object position shift is detected, a replanning mechanism is triggered to regenerate the planned trajectory and repeat the above steps.

2. The method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field according to claim 1, characterized in that, The radial potential force is constructed as follows: Construct the radial potential function: ； in, , Let the radius be the radius of the block. The distance between the end effector of the robotic arm and the center of the object. , , , A constant defined by humans; Differentiation yields the radial potential force: 。 3. The method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field according to claim 1, characterized in that, The tangential potential force is constructed as follows: Construct the tangential potential function: ； Differentiation yields the tangential potential force: ； in, Position of the robotic arm end effector relative to the spherical target radial error, A constant defined by humans.

4. The method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field according to claim 1, characterized in that, The construction of the super-torsional sliding mode controller based on symmetrical preset performance includes: A symmetric preset performance function is constructed to constrain the dynamic evolution of joint tracking error; Construct a super-torsional sliding mode control law to improve the system's robustness under modeling uncertainties and external disturbances; By combining the dynamic compensation term, a complete joint space control law is formed.

5. The method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field according to claim 4, characterized in that, The super-torsional sliding mode controller is represented as follows: in, For dynamic compensation terms, To control the maximum input amplitude, The inertial matrix of the robotic arm joint space. The Super-Twisting sliding mode control law is constructed.

6. The method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field according to claim 1, characterized in that, The reinforcement learning algorithm employs the Soft Actor-Critic algorithm, which constructs a reward function based on trajectory tracking error to optimize control parameters online or iteratively, thereby achieving adaptive parameter adjustment.

7. The method for low-disturbance, non-intrusive grasping of a floating target by a robotic arm based on an artificial potential field according to claim 1, characterized in that, The replanning mechanism includes: When the distance between the end effector and the target object exceeds a preset threshold, or the target object is pushed away from its original position or the grasping fails, trajectory replanning is automatically triggered. Based on the current terminal state and the updated target object position, the planned trajectory is regenerated, and the artificial potential field correction and control tracking process is re-executed.

8. A low-disturbance, non-intrusive grasping system for a floating target manipulation robotic arm based on an artificial potential field, characterized in that, The system includes: The trajectory planning module is used to generate an initial planned trajectory based on the initial pose and the target pose. The artificial potential field construction module is used to generate radial and tangential potential forces based on the relative spatial relationship between the end point and the target, thus forming a virtual artificial potential field force. The impedance trajectory correction module is used to introduce the virtual force of the artificial potential field into the impedance model to dynamically correct the planned trajectory. The joint space control module is used to build a super torsional sliding mode controller based on symmetrical preset performance to realize the trajectory tracking control of the robotic arm; The control parameter learning module is used to iteratively optimize the control parameters using reinforcement learning algorithms. The crawling detection and replanning module is used to detect crawling results and determine whether a replanning mechanism has been triggered.

Images