Robot variable admittance compliant teleoperation control method based on human intention recognition

Abstract

The invention belongs to the technical field of robot compliant teleoperation, and particularly relates to a robot variable admittance compliant teleoperation control method based on human intention recognition, which comprises the following steps of: S1, accurately defining respective working spaces of master and slave equipment before matching the working spaces of the master and slave equipment, the working space of the master equipment is limited by the control range of an operator, and the working space of the slave equipment is limited by the geometric structure, kinematics characteristics and task requirements of the robot arm; s2, a new variable-scale workspace mapping method is provided for the master robot and the slave robot of the teleoperation system, in the speed mapping part, the mapping proportion between the master robot and the slave robot is adjusted in real time according to different requirements of tasks by introducing a speed scaling factor, and the slave equipment can accurately follow according to the operation speed of the master equipment; after human intention recognition, a variable admittance control algorithm is realized, so that admittance parameters are adjusted, and the robot is helped to move.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot compliant teleoperation, and in particular is a robot variable admittance compliant teleoperation control method based on human intention recognition. Background Art

[0002] With the rapid development of intelligent manufacturing and artificial intelligence technologies, the application of robots in industrial production, medical assistance, space exploration and other fields has gradually expanded. Compliant control technology, as an important method to achieve safe and efficient operation of robots, has become a research hotspot in academia and industry. The core of compliant control is to enable robots to perceive changes in external forces in dynamic environments and flexibly adjust their motion states, thereby meeting the high requirements of flexibility, precision and safety for complex tasks.

[0003] Chinese invention patent CN115556088B provides a control system for the contact force of a robotic arm, a robotic arm, and a robot. The control system includes: an I-PD force controller; an admittance controller, the admittance controller is used to output the posture deviation E of the robotic arm based on the output signal Fr of the I-PD force controller, and the I-PD force controller and the admittance controller are matched and designed; an interactive system, the interactive system is used to output the actual posture X of the robotic arm based on the reference posture Xr, and when the robotic arm contacts the environment, the interactive system outputs the actual contact force Fc of the robotic arm based on the environmental stiffness Ke, the actual posture X of the robotic arm, and the environmental position Xe. The contact force Fc is input to the input end of the interactive system through feedback. The control system has low control accuracy and high control difficulty.

[0004] Chinese invention patent CN116619393B provides a variable admittance control method, device and equipment for a robotic arm based on SVM. Through the OCSVM model, it identifies the moment when the robotic arm contacts the bone model and records the osteotomy area. Then, based on the real-time comparison of the tool position at the end of the robotic arm with this osteotomy position, it achieves the distinction between the robotic arm being freely dragged and the doctor dragging the robotic arm for osteotomy / bone grinding. This control method is difficult to operate and has low control efficiency.

[0005] However, in practical applications, robot compliance control still faces many challenges. For example, traditional fixed-parameter admittance control methods are difficult to adapt to dynamically changing contact forces or multi-stage task requirements. In addition, when robots need to complete precise tasks under the guidance of human operating intentions, existing control algorithms are unable to balance compliance and stability, which may lead to low task efficiency or safety hazards.

[0006] Robot teleoperation tasks involve a variety of force-related actions such as grasping, twisting, and connecting. During these actions, the end effector will inevitably come into contact or collide with the target object. When the contact force deviates from the normal range, it may cause potential damage to the target object or seriously affect the system balance of the robot itself. Summary of the Invention

[0007] In response to the above problems, the present invention provides a robot variable admittance compliant teleoperation control method based on human intention recognition.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a robot variable admittance compliant teleoperation control method based on human intention recognition, comprising the following steps:

[0009] Step S1: Before matching the workspaces of the master and slave devices, accurately define the workspaces of the master and slave devices;

[0010] Step S2: According to the different requirements of the task, the mapping ratio between the master and slave robots is adjusted in real time, and the slave device can accurately follow the operating speed of the master device;

[0011] Step S3: Human intention is divided into direct intention and indirect intention. Direct intention is the immediate response to the force or action signal applied by humans, that is, the robot adjusts its position or speed according to human input. Indirect intention reflects the long-term goal of humans, that is, moving along a specific trajectory in a straight line or curve. The role of the robot is divided into passive follower and obedient leader. In tasks where the trajectory is difficult to plan in advance or humans want to guide the robot to move freely, the robot behaves as a passive follower. When the robot passively follows the human's guidance, the human's intention is classified as acceleration or deceleration.

[0012] Step S4: Establish a discrete admittance control model, treating the robot end effector as an equivalent second-order spring-mass-damper system. When the robot interacts with the environment, the parameters of the spring-mass-damper system are adjusted to establish an ideal dynamic relationship between the external force and the robot position.

[0013] Step S5: The variable admittance control algorithm obtains force information from the environment and the operated object through sensors, and accurately perceives and controls the position of the robot end effector by updating the position. The results of human intention recognition are used to design the variable admittance algorithm. Variable admittance control is divided into two types: active variable admittance control and passive variable admittance control. According to the characteristics of the external force, it is divided into three scenarios: contact, non-contact, and emergency.

[0014] Preferably, in step S1: the initial maximum working space of the robot system with n degrees of freedom is defined as Γ0, and the working space after the limit is defined as Γ p, we can get Defined in Γ p The dimensions of the three directions of the modified workspace in the space are α, β, and γ, which are defined in Γ p The coordinate system in space is CS P , and define the origin of the coordinate system as CS0 = (cs x min cs y min cs z min ), the point farthest from the origin within the working range is defined as CS f =(cs x max cs y min cs z min ), then the corresponding geometric relationship is:

[0015]

[0016] Define the desired position as CS ep (t), then the expected position must satisfy {CS ep |CS0≤CS ep ≤CS f};

[0017] In Γ p In space, the current position point is defined as CS e (t), then the trajectory tracking error ξ(t) can be expressed as:

[0018] ξ(t)=CS ep (t)-CS e (t)

[0019] The derivative of the trajectory tracking error is:

[0020]

[0021] Define the end effector at Γ p The translation vector of space is CS t , then the movement speed satisfies:

[0022]

[0023] Where q is the joint angle and J(q) is the robot Jacobian matrix, we can get:

[0024]

[0025] Further simplify the differential equation and assume that in Γ p There are no singular positions in space, so we can get:

[0026]

[0027] Where K is a full-rank diagonal parameter matrix, and its error convergence is related to the eigenvalue of K, which can be obtained:

[0028]

[0029] Therefore, the larger the eigenvalue of K, the faster the convergence speed.

[0030] Preferably, in step S2: after the working spaces of the master and slave robots are restricted, the motion spaces of the endpoints are Γ m =x am ×x bm ×x cm and Γ s =x as ×x bs ×x cs , the moving range of the end effector of the master and slave robots is defined as η mi ∈[η mi min ,η mi max ] i＝{x,y,z} and η si ∈[η si min ,η si max ] i＝{x,y,z} , so we can know that x am =|η mx max -η mx min |, x bm =|η my max -η my min |,x cm =|η mz max -η mz min |, x as =|η sx max -η sx min |,x bs =|η sy max -η sy min |, x cs =|η sz max -η sz min |;

[0031] In order to make the position of the end of the slave robot relative to the confined space correspond to the position of the end of the master robot relative to the confined space, you can set:

[0032]

[0033] It can be deduced that:

[0034]

[0035] where θ x =x am / x as ,θ y =xbm / x bs ,θ z =x cm / x cs , the parameters represent the spatial mapping scale coefficients of the three directions of the master end and the slave end, which can be written as θ=[θ x θ y θ z ].

[0036] Preferably, in step S2: after the working spaces of the master and slave robots are restricted, the motion spaces of the endpoints are Γ m =x am ×x bm ×x cm and Γ s =x as ×x bs ×x cs , the moving range of the end effector of the master and slave robots is defined as η mi ∈[η mi min ,η mi max ] i＝{x,y,z} and η si ∈[η si min ,η si max ] i＝{x,y,z} , so we can know that x am =|η mx max -η mx min |, x bm =|η my max -η my min |, x cm =|η mz max -η mz min |, x as =|η sx max -η sx min |, x bs =|η sy max -η sy min |, x cs =|η sz max -η sz min |;

[0037] In order to make the position of the end of the slave robot relative to the confined space correspond to the position of the end of the master robot relative to the confined space, you can set:

[0038]

[0039] It can be deduced that:

[0040]

[0041] where θ x =x am / x as ,θ y =x bm / x bs ,θ z =x cm / x cs , the parameters represent the spatial mapping scale coefficients of the three directions of the master end and the slave end, which can be written as θ=[θ x θ y θ z ].

[0042] Preferably, in step S2: if the proportional coefficient matrix is ​​given a fixed value, the master-slave proportional mapping coefficient is completely determined by the running speed of the master robot. When the master robot is too fast, the proportional mapping coefficient will become larger, resulting in excessive displacement of the slave robot. On the contrary, when the master robot is too slow, the proportional mapping coefficient will become smaller, resulting in too little movement of the slave robot, or even almost no movement. The value range of the speed proportional function is limited to v i ∈[v i min ,v i max ] i＝{x,y,z} , for the velocity scaling function Ψ(V m ), gives:

[0043]

[0044] The proportional mapping coefficient is expressed as

[0045]

[0046] where θ i are the elements of the coefficient matrix θ, w1, w2, w3 are the gain coefficients of different working modes used when performing normal, fast and fine tasks.

[0047] Preferably, in step S3: during the deceleration process, the following two scenarios are considered: (a) the human gradually reduces the applied force; (b) the human suddenly changes the direction of the applied force, and the dot product of the velocity vector and the force vector is introduced to evaluate whether the applied force has reversed direction; the compliant leader can be further subdivided into semi-active and active types. The robot's response can adapt when humans intervene for safety reasons, and quickly resume the task after the intervention is eliminated. This situation is called semi-active behavior. Such applications need to provide users with a soft interactive experience, that is, reduce the admittance parameters; assuming that the task scenario is walking or sit-stand conversion assistance, when humans need strong support, the robot should become rigid enough. In this case, the robot is defined as an active compliant leader.

[0048] Preferably, in step S3: the stiffness should increase as the force applied by the human increases to ensure accurate tracking of the predetermined trajectory under the premise of safety. The human-machine interaction at this stage is then further subdivided into normal interaction or abnormal interaction: on the robot side, abnormal human behavior can be further subdivided into two situations: supportable and dangerous. If the interaction is abnormal but still supportable, the robot will take appropriate measures, that is, increase the stiffness to offset the additional force applied; on the contrary, if the abnormal interaction is judged to be dangerous, the robot will activate the emergency stop mechanism to prevent further damage. Once a safe state is restored, the user can redirect the robot to resume the task. At this time, compliance is achieved by switching the robot role to a passive follower.

[0049] Preferably, in step S4: by combining the position, velocity, acceleration and force information of the robot end effector, a second-order differential equation model, namely, an impedance control model, is constructed:

[0050]

[0051] Among them, M, B, and K represent the inertia, damping, and stiffness coefficient matrices of the system respectively; X c Respectively represent the adjusted position, velocity, and acceleration of the robot; X d They represent the initial reference position, velocity, and acceleration of the robot, respectively. F is the actual force in three-dimensional space when the robot contacts the environment. F d For the power of expectation;

[0052] The following parts are described and processed using a one-dimensional model:

[0053]

[0054] When the robot end effector is not in contact with the object, it is assumed that the robot can reach the desired position, that is, x d =x c , because there are no obstacles hindering its movement during the movement, in order to ensure that the robot end effector stops moving completely after reaching the target position, it is necessary to ensure and Therefore, the admittance model can be reformulated as:

[0055]

[0056] Where Δf=ff d , using the zero-order holder technique (ZOH), the discrete form of the admittance model is given, where the time step is denoted as k and the sampling time is denoted as T:

[0057]

[0058] Writing the acceleration term in differential form gives:

[0059]

[0060] Finally, we get:

[0061]

[0062] When the external force remains constant, the velocity of the robot end will converge to a certain value, which means Therefore, the maximum speed can be expressed as:

[0063]

[0064] Preferably, in step S5: in step S5: active variable admittance adjusts the admittance parameters in real time according to the displacement of the robot end or the contact force between the robot and the environment; passive variable admittance control adjusts the admittance parameters when the robot end accidentally hits an obstacle, ends the current motion state to play a protective role, and divides the variable admittance control into two stages: normal working state and abnormal working state; normal working state: the robot is in a rapid motion process or the robot end maintains constant force output after contacting the environment; abnormal working state: the current target trajectory needs to be adjusted so that it is corrected in the direction of reducing external force.

[0065] Preferably, in step S5: design of a non-contact active variable admittance control algorithm: adjusting the desired force according to the distance between the robot end and the target position. This method can stop at the target position faster and is conducive to rapid movement over a wide range:

[0066]

[0067] Among them, δ represents the distance between the current and target positions, δ t represents the distance from the starting position to the target position, b max is the maximum value of damping;

[0068] In the contact-type passive emergency stop variable admittance control algorithm, when the robot end effector is at point x1 and encounters an unexpected obstacle, causing the contact force to suddenly increase significantly and exceed the normal threshold, the passive variable admittance control algorithm is triggered to quickly respond to this situation and quickly stop the robot motion. By passively adjusting the admittance parameters, the robot end effector accelerates the motion from x1 to x2 and accelerates the reverse acceleration from x2 to x3. The change in admittance parameters can be expressed as:

[0069]

[0070] in b f Is the default damping value, bmax and b min are the default maximum damping value and minimum damping value respectively, gd() function is the Gudermannian function, Indicates the default maximum acceleration of the robot end.

[0071] Contact active variable admittance control algorithm, the admittance model can be expressed as:

[0072]

[0073] in ξ=x d -x c , k(t) can be adjusted according to the contact force. Tracking force error ξ when the robot end interacts with the environment f Expressed as ξ f =f d -f. The following formula can be derived:

[0074]

[0075] An active variable admittance controller for contact is designed:

[0076]

[0077] where k p and k' d is a variable parameter. Combining the above two equations, we get:

[0078]

[0079] After sorting, we get

[0080]

[0081] In order to prove the stability of the proposed active variable admittance controller, the Lyapunov function is used:

[0082]

[0083] in The proposed Lyapunov function can be regarded as the sum of kinetic energy and potential energy for motion control and force tracking control, so we have

[0084]

[0085] According to the above, f = k e (xx e ), we can get

[0086] ξ f =fd -f=f d -k e (xx e )

[0087] According to this formula, So the derivative of the Lyapunov function is

[0088]

[0089] Therefore, combining the above formulas we get

[0090]

[0091] Therefore, the stability of the contact-type active variable admittance controller is demonstrated.

[0092] Compared with the prior art, the present invention has the following beneficial effects:

[0093] 1. The present invention can dynamically adjust the mapping ratio of master and slave devices. This method optimizes the utilization of workspace and improves the operating accuracy and efficiency of robots in remote operation tasks.

[0094] 2. The present invention dynamically adjusts the robot's admittance parameters and operating mode by real-time recognition of human operating intentions. This method enhances the collaborative ability of robots and humans in a multi-tasking environment and improves the accuracy and flexibility of task completion.

[0095] 3. The present invention combines active variable admittance control with passive variable admittance control. This structure can dynamically adjust the admittance parameters according to different motion stages and external force conditions, thereby achieving rapid motion, collision impact relief, and stability of terminal force output.

[0096] 4. The present invention addresses the different requirements of the non-contact stage, contact stage, and trajectory tracking stage, achieving rapid response, impact relief, and stable force output, significantly improving the robot's flexibility and adaptability in complex tasks.

[0097] 5 The effectiveness of the control method of the present invention has been verified through actual experiments in remote operation scenarios such as micro-satellite assembly, and the advantages of the present invention in flexibility, accuracy and mission efficiency have been demonstrated through comparative analysis with existing algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 The space restriction control structure of the present invention;

[0099] Figure 2 A human intention recognition table of the present invention;

[0100] Figure 3 This is a block diagram of the variable admittance control of the present invention;

[0101] Figure 4 For the emergency stop process of the present invention;

[0102] Figure 5 The comparison experiment results of the non-contact active variable admittance of the present invention are as follows;

[0103] Figure 6 The comparative experimental results of the contact-type passive emergency stop variable admittance control algorithm of the present invention are shown;

[0104] Figure 7 The robot terminal trajectory diagram using the control method of the present application is shown in the figure;

[0105] Figure 8 This is a contact force trajectory diagram of the robot end using the control method of the present application;

[0106] Figure 9 The robot terminal trajectory diagram using the traditional control method of the present invention;

[0107] Figure 10 This is the robot terminal trajectory diagram using the traditional control method of the present invention. DETAILED DESCRIPTION

[0108] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0109] First embodiment

[0110] Robot teleoperation tasks involve a variety of force-related actions, such as grasping, twisting, and connecting. During these actions, the end effector will inevitably come into contact or collide with the target object. When the contact force deviates from the normal range, it may cause potential damage to the target object or seriously affect the robot's own system balance. Therefore, strict requirements are placed on the robot's contact force in teleoperation tasks. In challenging operating environments, robots need to have compliant control capabilities to effectively adapt to changing working conditions. Therefore, this application designs an active-passive combined variable admittance compliant algorithm based on human intention recognition for teleoperation systems, which makes rigid joints have better compliant characteristics, thereby achieving the purpose of compliant control. The specific steps include:

[0111] Step S1: The present invention first proposes an effective solution to the problem of workspace matching between master and slave devices. In teleoperation systems, the workspaces of master and slave devices are often irregular in shape, and the robots may have multiple pose solutions within their reach, or even unique solutions for certain positions. Therefore, before matching the workspaces of the master and slave devices, their respective workspaces must be accurately defined.

[0112] The workspace of the master device is limited by the operator's control range, while the workspace of the slave device is constrained by the robot arm's geometry, kinematics, and task requirements. This paper clearly defines the workspaces of the master and slave devices by comprehensively considering factors such as device degrees of freedom, joint limitations, and range of motion. It also specifically addresses the multi-solution nature of the robot arm's pose solutions. This method ensures that all possible pose solutions are accounted for during workspace matching, providing a clear and accurate framework for subsequent workspace mapping and ensuring efficient control of the teleoperation system.

[0113] Specifically, the initial maximum working space of the robot system with n degrees of freedom is defined as Γ0, and the limited working space is defined as Γ p , we can get Defined in Γ p The dimensions of the three directions of the modified workspace in the space are α, β, and γ, which are defined in Γ p The coordinate system in space is CS P , and define the origin of the coordinate system as CS0 = (cs x min cs y min cs z min ), the point farthest from the origin within the working range is defined as CS f =(cs x max cs y min cs z min ). Then the corresponding geometric relationship is:

[0114]

[0115] Define the desired position as CS ep (t), then the expected position must satisfy {CS ep |CS0≤CS ep ≤CS f}.

[0116] In Γ p In space, the current position point is defined as CS e (t), then the trajectory tracking error ξ(t) can be expressed as:

[0117] ξ(t)=CS ep (t)-CSe (t)

[0118] The derivative of the trajectory tracking error is:

[0119]

[0120] Define the end effector at Γ p The translation vector of space is CS t , then it is obvious that the movement speed satisfies:

[0121]

[0122] Where q is the joint angle and J(q) is the robot Jacobian matrix:

[0123]

[0124] The relationship between the end error and the joint velocity vector. Further simplify the differential equation and assume that in Γ p There is no singular position in space, so we can get

[0125]

[0126] Where K is a full-rank diagonal parameter matrix, and its error convergence is related to the eigenvalue of K, which can be obtained:

[0127]

[0128] It is obvious that the larger the eigenvalue of K, the faster the convergence speed.

[0129] Therefore, the space-limited control structure can be designed as follows Figure 1 As shown in this control structure, Π is a position calculation unit of the end effector. In a robot system with only rotation joints, it is only related to the joint angle. In addition, J -1 (q) is used to compensate J(q). Furthermore, in the process of solving the corresponding Jacobian matrix inverse kinematics, it is necessary to integrate the velocity of each joint to obtain the real-time joint angle, which is then substituted into the nonlinear elements of the feedback branch to form a closed loop.

[0130] Step S2: A new variable-scale workspace mapping method is proposed for the master and slave robots of the teleoperation system. In the teleoperation system, this method takes into account the spatial matching problem between the master and slave robots and the dynamic adjustment of speed requirements. The core of the variable-scale workspace mapping method is to be able to flexibly adjust the proportional relationship between the two according to the motion state of the master device, the actual needs of the workspace, and the motion limitations of the slave device. This adjustment mechanism enables the master and slave robots to achieve accurate spatial matching in a larger range, thereby maximizing the utilization efficiency of the workspace. At the same time, the speed mapping part introduces a speed scaling factor to adjust the mapping ratio between the master and slave robots in real time according to different task requirements (such as precise mode, high-speed mode, etc.), so that the slave device can accurately follow the operating speed of the master device, thereby improving the accuracy and safety of task execution.

[0131] The advantage of this approach lies in that it not only optimizes workspace utilization by adjusting spatial proportions but also incorporates a speed-adaptive mechanism, enabling flexible adaptation to dynamic changes in complex operational scenarios. This approach ensures coordination and control precision between the master and slave robots, both during high-speed motion and delicate operations. This improves the overall performance and efficiency of the teleoperation system, particularly during multitasking and complex scene switching, significantly enhancing operational stability and responsiveness.

[0132] Specifically, assuming that the working space of the master and slave robots is restricted, the motion spaces of the endpoints are Γ m =x am ×x bm ×x cm and Γ s =x as ×x bs ×x cs The moving range of the end effector of the master and slave robots is defined as η mi ∈[η mi min ,η mi max ] i＝{x,y,z} and η si ∈[η si min ,η si max ] i＝{x,y,z} , so we can know that x am =|η mx max -η mx min |,x bm =|η my max -η my min |,x cm =|η mz max -η mz min |,x as =|η sx max -η sx min |,x bs =|η sy max-η sy min |,x cs =|η sz max -η sz min |.

[0133] In order to make the position of the end of the slave robot relative to the confined space correspond to the position of the end of the master robot relative to the confined space, you can set:

[0134]

[0135] It can be deduced that:

[0136]

[0137] where θ x =x am / x as ,θ y =x bm / x bs ,θ z =x cm / x cs , the parameters represent the spatial mapping scale coefficients of the three directions of the master end and the slave end, which can be written as θ=[θ x θ y θ z ].

[0138] Normal proportional space mapping can achieve good master-slave tracking capabilities. However, due to the limited workspace, the slave robot loses a large portion of its workspace. Therefore, incremental space mapping has been widely used in previous research. Incremental space mapping, also known as rate mapping, is a method that sets a motion origin for the master and slave robots. The slave robot uses the master's displacement per unit time as a reference at the end. By manipulating the master robot, the slave robot's displacement per unit time is mapped. In addition, a certain proportional coefficient matrix is ​​added to achieve the master-slave robot displacement per unit time mapping.

[0139] Set d dpu is the expected motion displacement per unit time, σ m is the displacement of the master robot relative to the origin of the world coordinate system, σ m0 is the displacement of the origin of motion relative to the origin of the world coordinate system. Then we stipulate:

[0140] d dpu =θ(σ m -σ m0 )

[0141] After each unit time of movement, σ mcan be considered as the origin of the next unit motion time, so (10) can be rewritten as:

[0142] ΔP s =θ·ΔP m

[0143] where ΔP s and ΔP m is the motion displacement vector of the master and slave robots per unit time, and then we can infer:

[0144] P s =P s0 +θ·ΔP m

[0145] Taking its derivative we can get:

[0146]

[0147] Where V m (t)={v mx (t)v my (t)v mz (t)}.

[0148] Incremental spatial mapping does not limit the master / slave workspace, and also allows the use of a wide range of slave end space. However, this mapping method also has the problem that the slave device cannot change direction quickly. Taking into account the limitations of traditional proportional mapping and incremental mapping methods, this application incorporates the master operation speed as a proportional factor into the mapping algorithm, and improves these two methods. The method proposed in this application can not only make more extensive use of the end effector's workspace, but also divide the operation tasks into multiple modes such as normal, high speed and precision according to the speed of the master robot, realizing a speed-based variable-scale spatial mapping method.

[0149] Specifically, the master-slave space mapping based on the master robot's running speed is to transform the scale function Ψ(V m ) is added to the incremental mapping model:

[0150] d dpu =θ·Ψ(V m )·(σ m -σ m0 )

[0151] The instantaneous expression of the slave robot position is:

[0152] P s =P s0 +θ·Ψ(V m )·ΔP m

[0153] From the above formula, we can see that as long as the speed scaling function Ψ(V m ) and the proportional coefficient matrix θ, the mapping relationship between the master and slave robot end positions can be realized.

[0154] If the proportional coefficient matrix is ​​given a fixed value, the master-slave proportional mapping coefficient is completely determined by the running speed of the master robot. When the master robot is too fast, the proportional mapping coefficient will become larger, resulting in excessive displacement of the slave robot. On the contrary, when the master robot is too slow, the proportional mapping coefficient will become smaller, resulting in too little movement of the slave robot, or even almost no movement.

[0155] In order to solve the above problem, this application will limit the value range of the speed proportional function, and limit the value range of the speed proportional function to v i ∈[v i min ,v i max ] i＝{x,y,z} , for the velocity scaling function Ψ(V m ), this application provides:

[0156]

[0157] The proportional mapping coefficient is expressed as

[0158]

[0159] where θ i are the elements of the coefficient matrix θ, w1, w2, w3 are the gain coefficients of different working modes used when performing normal, fast and fine tasks.

[0160] It can be seen that when the speed exceeds the limit range, the proportional mapping coefficient is independent of the speed, which solves the problem of excessive or insufficient speed in the speed mapping mentioned above.

[0161] Step S3: Classification of human intentions. In human-machine physical interaction, intention recognition is the core issue for achieving efficient collaboration and precise control. However, the definition of human intention has a broad meaning, and different studies have different understandings and applications of it. From the summary of previous studies, human intentions can be divided into two categories: direct intentions and indirect intentions. Direct intentions are immediate responses to the force or action signals applied by humans, such as robots quickly adjusting their position or speed based on human input; while indirect intentions reflect humans’ long-term goals, such as straight or curved motion along a specific trajectory. Distinguishing and effectively utilizing these two types of intentions is of great significance for improving the performance of robots in dynamic interaction scenarios. This application proposes to classify human intentions using the following method:

[0162] A robot's understanding of human intentions depends on the task the robot is performing, and the type of task determines whether the human or the robot takes the lead in control.

[0163] To this end, it is necessary to clearly define the role of robots in human-computer interaction. This application divides the role of robots into passive followers and obedient leaders, such as Figure 2 As shown:

[0164] (1) Passive follower: In tasks where the trajectory is difficult to plan in advance or humans want to guide the robot to move freely, the robot behaves as a passive follower. When the robot passively follows the human's guidance, this application classifies the human's intention into two situations: acceleration or deceleration. During the deceleration process, the following two scenarios are considered: (a) the human gradually reduces the applied force; (b) the human suddenly changes the direction of the applied force. By introducing the dot product of the velocity vector and the force vector, it is used to evaluate whether the applied force has reversed direction.

[0165] (2) The compliant leader can be further divided into semi-active and active types. For example, when the main goal of the robot is to complete the task independently (such as grasping and placing objects), the robot becomes a semi-active compliant leader. In this case, the robot's response can adapt when humans intervene for safety reasons, and quickly resume the task after the intervention is eliminated. This application refers to this situation as semi-active behavior. Such applications need to provide users with a soft interactive experience, that is, reduce the admittance parameters. On the contrary, assuming that the task scenario is walking or sit-stand conversion assistance, when humans need strong support (such as falling or losing balance), the robot should become sufficiently rigid.

[0166] In this case, the robot is defined as an active compliant leader, where the stiffness should increase with the increase of the force applied by the human to ensure accurate tracking of the predetermined trajectory under the premise of safety. Subsequently, this application further subdivides the human-computer interaction at this stage into normal interaction or abnormal interaction. On the robot side, abnormal human behavior (such as loss of balance or falling) can be further subdivided into two situations: supportable and dangerous. If the interaction is abnormal but still supportable, the robot will take corresponding measures, that is, increase the stiffness to offset the additional force applied. On the contrary, if the abnormal interaction is judged to be dangerous, the robot will activate the emergency stop mechanism to prevent further damage. Once the safety state is restored, the user can redirect the robot to resume the task. At this time, compliance is achieved by switching the robot role to a passive follower.

[0167] Human intent is variable and specific across different tasks, and different human preferences make it user-oriented. The general human intent summarized in this application is not sufficient to represent all scenarios, but rather provides a basic framework for designing the parameters of the variable admittance control algorithm described later.

[0168] Step S4: Establish a discrete admittance control model. In order to establish a contact model between the robot and its environment, this application regards the robot end effector as an equivalent second-order spring-mass-damper system.

[0169] When the robot interacts with the environment, an ideal dynamic relationship between the external force and the robot's position can be established by adjusting the parameters of the spring-mass-damper system. Therefore, by combining the position, velocity, acceleration, and force information of the robot's end effector, a second-order differential equation model, namely the impedance control model, is constructed:

[0170]

[0171] Among them, M, B, and K represent the inertia, damping, and stiffness coefficient matrices of the system respectively; X c Respectively represent the adjusted position, velocity, and acceleration of the robot; X d They represent the initial reference position, velocity and acceleration of the robot respectively.

[0172] In addition, F is the actual force in three-dimensional space when the robot contacts the environment. d For the expected force, in order to facilitate analysis and solution, the following part uses a one-dimensional model for description and processing:

[0173]

[0174] When the robot end effector is not in contact with the object, it can be assumed that the robot can reach the desired position, that is, x d =x c , because there are no obstacles hindering its movement during the movement. In addition, in order to ensure that the robot end effector stops moving completely after reaching the target position, thereby effectively preventing collisions that may be caused by further movement, it is necessary to ensure that and Therefore, the admittance model can be reformulated as:

[0175]

[0176] Where Δf=ff d Using the zero-order holder technique (ZOH), this application presents a discrete form of the admittance model, where the time step is denoted as k and the sampling time is denoted as T:

[0177]

[0178] Writing the acceleration term in differential form gives:

[0179]

[0180] Finally, we get the following results given by this application:

[0181]

[0182] Obviously, when the external force remains constant, the velocity of the robot end will converge to a certain value, which means

[0183] Therefore, the maximum speed can be expressed as:

[0184]

[0185] Step S5: Design of the variable admittance control structure. Among the many compliant control algorithms, variable admittance control (VAC) has attracted much attention due to its outstanding features, including safety, high efficiency, and strong robustness. VAC obtains force information from the environment and the manipulated object through sensors and achieves precise sensing and force control by updating the position of the robot's end effector. This approach ensures the robot's safety and significantly improves its accuracy and compliance.

[0186] Variable admittance control can be divided into two types: active variable admittance control and passive variable admittance control. Active variable admittance adjusts the admittance parameters in real time according to the displacement of the robot end or the contact force between the robot and the environment, thereby improving the robot's compliance to meet work requirements. Passive variable admittance control can quickly adjust the admittance parameters and quickly end the current motion state to provide protection when the robot end accidentally encounters an obstacle.

[0187] Based on the above recognition of human intention, this application proposes a variable admittance control method based on active-passive hybrid human intention recognition. This method dynamically adjusts the physical parameters of the robot to enable it to adapt to the requirements of different operating tasks, especially showing significant advantages when dealing with unexpected or uncertain scenarios.

[0188] By evaluating the motion state in real time, the method can accurately determine and set the appropriate level of compliance, thereby optimizing the robot's performance in a changing environment and ensuring the efficient completion of mission objectives. Based on the recognition of human intention mentioned above, the control method of this application divides variable admittance control into two stages: normal working state and abnormal working state.

[0189] like Figure 3As shown in the figure, the normal working state is when the robot is in rapid motion or maintains constant force output after the end of the robot contacts the environment. This usually occurs in large-scale motion scenarios under load conditions such as object handling. In this stage, it is necessary to optimize the admittance parameters to improve the robot's motion performance and give it a certain degree of flexibility while ensuring motion efficiency, so as to adapt to the requirements of the operation task.

[0190] Abnormal operating conditions: Due to collisions, the robot may experience sudden and unpredictable increases in external forces. To mitigate the impact of such collisions, the current target trajectory must be quickly adjusted to reduce external forces while minimizing the collision duration to protect the robot system and ensure smooth mission execution.

[0191] The structure of the control algorithm is as follows Figure 3 As shown:

[0192] According to the characteristics of the external force, it can be divided into three scenarios: contact, non-contact and emergency. The state of the robot is divided into three stages, and each stage adopts a different variable admittance control strategy. The first stage is the large-scale movement of the robot in a non-contact state. In this stage, it is necessary to improve the flexibility of the robot so that it can reach the target position faster. The corresponding non-contact active variable admittance control strategy is used; the second stage is used to deal with the situation where the robot has an accidental collision. At this time, a passive variable admittance control strategy is adopted to quickly change the current motion state to avoid collisions and damage to the robot and the operated object; in the third stage, the robot end moves while maintaining a specific output force. This stage adopts a contact-type active variable admittance control strategy.

[0193] Step S6: Design of a non-contact active variable admittance control algorithm: To ensure that the robot end's velocity drops to zero when it reaches the target position, this application adjusts the desired force based on the distance between the robot end and the target position. This method enables faster stopping at the target position and facilitates rapid movement over a wide range.

[0194]

[0195] Among them, δ represents the distance between the current and target positions, δ t represents the distance from the starting position to the target position, b max is the maximum value of damping.

[0196] Step S7: Design of variable admittance control algorithm for contact-type passive emergency stop: When the robot is in motion, unexpected collisions may cause irreversible damage to the robot and its interactive objects, and these collisions will have a profound impact on the successful completion of the task.

[0197] To address this problem, this paper designs a passive variable admittance control algorithm for emergency stop, which enables the robot end effector to quickly change its direction of motion when facing unexpected obstacles.

[0198] like Figure 4 As shown in the figure, when the robot's end effector is at point x1 and encounters an unexpected obstacle, the contact force suddenly increases significantly and exceeds the normal threshold. In this case, the passive variable admittance control algorithm is triggered to quickly respond to this situation and quickly stop the robot's motion. By passively adjusting the admittance parameters, the algorithm accelerates the robot's end effector's motion from x1 to x2 and accelerates the reverse acceleration from x2 to x3. By reducing the contact time, this method effectively reduces the impact of the collision and prevents potential damage.

[0199] Therefore, the change of the admittance parameter can be expressed as:

[0200]

[0201] in b f Is the default damping value, b max and b min are the default maximum damping value and minimum damping value respectively, gd() function is the Gudermannian function, Indicates the default maximum acceleration of the robot end.

[0202] Step S8: Design of the contact active variable admittance control algorithm: In this stage, the robot typically tracks the desired trajectory while maintaining a constant output force, mimicking the state of a human hand contacting the environment. To apply greater force, the robot arm becomes more rigid (increasing stiffness); when the applied force is relatively small, the arm becomes more flexible (reducing stiffness). By adjusting the stiffness, different contact forces can be achieved.

[0203] In order to achieve a control effect similar to that of a human arm, this application designs a contact-type active variable admittance control algorithm to adjust the stiffness parameters when the robot contacts the environment. The admittance model of the robot can be expressed as:

[0204]

[0205] in ξ=x d -x c , k(t) can be adjusted according to the contact force, and the tracking force error ξ when the robot end interacts with the environment f Expressed as ξ f =f d -f. The following formula can be derived:

[0206]

[0207] This application designs an active variable admittance controller during contact:

[0208]

[0209] where k p and k' d is a variable parameter. Combining the above two equations, we get:

[0210]

[0211] After sorting, we get

[0212]

[0213] In order to prove the stability of the proposed active variable admittance controller, we choose a Lyapunov function:

[0214]

[0215] in

[0216] The proposed Lyapunov function can be regarded as the sum of kinetic energy and potential energy for motion control and force tracking control, so we have

[0217]

[0218] According to the above, f = k e (xx e ), we can get

[0219] ξ f =f d -f=f d -k e (xx e )

[0220] According to this formula, So the derivative of the Lyapunov function is

[0221]

[0222] Therefore, combining the above formulas we get

[0223]

[0224] Therefore, the stability of the contact-type active variable admittance controller proposed in this application is proven.

[0225] Second embodiment

[0226] In order to verify the performance of the proposed control algorithm, this application designed a series of physical experiments and built a remote operation experimental platform, including a UR3 robot, a PHANToM Omni robot device, a six-dimensional force sensor, a mechanical gripper, and a micro-satellite model for assembly. The ground verification experiment of the micro-satellite model assembly includes multiple steps. First, the assembly object is identified and grasped, then it is transferred to the assembly position, the specified hole position is searched using the force sensor, and finally the assembly object is firmly installed in the corresponding assembly hole of the satellite model. The corresponding assembly process verifies the advantages of the algorithm proposed in this application compared with other algorithms.

[0227] 1. No-contact active variable admittance control algorithm verification experiment.

[0228] During the movement phase of the slave robot, the non-contact active variable admittance control algorithm proposed in this application is used to enable the robot to quickly reach the target position while maintaining flexibility. In order to evaluate its performance, we focused on the robot's unidirectional movement and conducted comparative experiments using the traditional admittance control algorithm and the non-contact active variable admittance control algorithm.

[0229] The experimental results are as follows Figure 5 As shown in the figure, the red curve corresponds to the traditional admittance control algorithm, and the blue curve represents the algorithm proposed in this paper. The acceleration process of the traditional algorithm is relatively slow, and its maximum speed is determined by the desired force, with a peak value of 0.014m / s. It is worth noting that the deceleration phase accounts for nearly half of the entire process and lasts for about 3 seconds. In contrast, the proposed algorithm shows a significant improvement in speed, reaching a speed of 0.065m / s in just 11 seconds and quickly reducing it to 0 within 1 second. By comparing the roughly equal areas under the red and blue curves, it can be clearly seen that in the two experiments, the robot traveled the same distance, however, the proposed algorithm enabled the robot to complete the movement nearly 5 seconds earlier than expected. This greatly improves the robot's operational efficiency, thereby achieving an improvement in overall performance.

[0230] 2. Verification experiment of contact passive emergency stop variable admittance control algorithm.

[0231] During the movement of a teleoperated robot, unexpected rigid collisions may have a significant impact on the robot system and may even cause mission failure. To evaluate the effectiveness of the algorithm of the present invention in mitigating the impact of such collisions, in the second phase of the experiment, obstacles were deliberately placed in the trajectory of the slave robot, causing the end of the slave robot to have a rigid collision with the obstacle. In addition, as a comparative analysis, the present invention also conducted a rigid collision experiment using a traditional admittance control algorithm.

[0232] like Figure 6As shown, when a collision event occurs, the present invention uses an emergency stop passive variable admittance control algorithm to enable the robot end to achieve rapid deceleration in the deceleration phase and rapid acceleration in the reverse acceleration phase. Experimental results show that the present invention significantly shortens the duration of the collision event and effectively reduces the impact of the collision on the robot system. The present invention further improves the robot's compliance and operational stability in complex environments, thereby achieving efficient protection of the equipment.

[0233] 3. Contact active variable admittance control experiment.

[0234] During the remote-operated assembly process of a microsatellite, a hole-finding operation method for locating a target hole is involved, which includes maintaining a preset contact force between the robot end effector and the assembly plane while performing an Archimedean spiral motion path along the assembly plane to search until the target hole is located. The operation method requires the robot end effector to have high flexibility so that it can still output a stable and consistent contact force when the motion direction changes rapidly, thereby ensuring the smoothness and accuracy of the robot's movement.

[0235] To verify the effectiveness of the above operation method, an experiment was conducted to test the robot's performance during the hole-finding phase. The experiment compared the performance of the proposed algorithm with that of a traditional admittance control algorithm.

[0236] The experimental results are as follows Figure 7-10 As shown, it helps to verify the technical advantages of the method in achieving precise hole positioning and stable force output.

[0237] like Figure 7 As shown in the figure, during the hole-finding process along the Archimedean spiral, the robot end maintains stable operation in the hole-finding plane, and only small fluctuations are observed in the direction perpendicular to the plane. Figure 8 From the axial force diagram in , it can be seen that when there is no contact with the plane, the force value fluctuates slightly around 0N; after contact, the contact force is always stable in the range of 10N to 15N.

[0238] In contrast, Figure 9 As shown in the figure, the experimental results of the traditional admittance control algorithm show that the fluctuation amplitude of the robot end when it contacts the plane is slightly larger than that of the proposed algorithm. In addition, Figure 10 It shows that when there is no contact, the contact force fluctuates greatly around 0N; after contact, the contact force changes significantly and is difficult to control within the desired range of 10N to 15N.

[0239] Therefore, it can be inferred that compared with the traditional admittance control algorithm, the proposed algorithm has higher stability on the compliant trajectory and can maintain more consistent output force values. These characteristics make it more suitable for practical application scenarios.

[0240] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0241] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A robot variable admittance compliant teleoperation control method based on human intention recognition, characterized in that: The following steps are involved: Step S1: Before matching the workspaces of the master and slave devices, accurately define the workspaces of the master and slave devices; Step S2: According to the different requirements of the task, the mapping ratio between the master and slave robots is adjusted in real time, and the slave device can accurately follow the operating speed of the master device; Step S3: Human intention is divided into direct intention and indirect intention. Direct intention is the immediate response to the force or action signal applied by humans, that is, the robot adjusts its position or speed according to human input. Indirect intention reflects the long-term goal of humans, that is, moving along a specific trajectory in a straight line or curve. The role of the robot is divided into passive follower and obedient leader. In tasks where the trajectory is difficult to plan in advance or humans want to guide the robot to move freely, the robot behaves as a passive follower. When the robot passively follows the human's guidance, the human's intention is classified as acceleration or deceleration. Step S4: Establish a discrete admittance control model, treating the robot end effector as an equivalent second-order spring-mass-damper system. When the robot interacts with the environment, the parameters of the spring-mass-damper system are adjusted to establish an ideal dynamic relationship between the external force and the robot position. Step S5: The variable admittance control algorithm obtains force information from the environment and the operated object through sensors, and accurately perceives and controls the position of the robot end effector by updating the position. The results of human intention recognition are used to design the variable admittance algorithm. Variable admittance control is divided into two types: active variable admittance control and passive variable admittance control. According to the characteristics of the external force, it is divided into three scenarios: contact, non-contact, and emergency.

2. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 1 is characterized in that: In step S1: the initial maximum workspace of the robot system with n degrees of freedom is defined as Γ0, and the limited workspace is defined as Γ p ,get Defined in Γ p The dimensions of the three directions of the modified workspace in the space are α, β, and γ, which are defined in Γ p The coordinate system in space is CS P , define the origin of the coordinate system as CS0 = (cs xmin cs ymin cs zmin ), the point farthest from the origin within the working range is defined as CS f =(cs xmax cs ymin cs zmin ), the corresponding geometric relationship is: Define the desired position as CS ep (t), the expected position must satisfy {CS ep |CS0≤CS ep ≤CS f }; In Γ p In space, the current position point is defined as CS e (t), the trajectory tracking error ξ(t) is expressed as: ξ(t)=CS ep (t)-CS e (t) The derivative of the trajectory tracking error is: Define the end effector at Γ p The translation vector of space is CS t , the movement speed satisfies: Where q is the joint angle and J(q) is the robot Jacobian matrix, we get: Further simplify the differential equation and assume that in Γ p There are no singular positions in the space, so we get: Where K is a full-rank diagonal parameter matrix, and its error convergence is related to the eigenvalue of K, which is: The larger the eigenvalue of K, the faster the convergence speed. In the time-constrained control structure, Π is a position calculation unit of the end effector. In a robot system with only rotating joints, it is only related to the joint angle. In addition, J -1 (q) is used to compensate J(q). In the process of solving the corresponding Jacobian matrix inverse kinematics, it is necessary to integrate the velocity of each joint to obtain the real-time joint angle, which is then substituted into the nonlinear elements of the feedback branch to form a closed loop.

3. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 1 is characterized in that: In step S2: After the working space of the master and slave robots is restricted, the motion spaces of the endpoints are Γ m =x am ×x bm ×x cm and Γ s =x as ×x bs ×x cs , the moving range of the end effector of the master and slave robots is defined as η mi ∈[η mimin ,η mimax ] i＝{x,y,z} and η si ∈[η simin ,η simax ] i＝{x,y,z} , we know that x am =|η mxmax -η mxmin |, x bm =|η mymax -η mymin |, x cm =|η mzmax -η mzmin |, x as =|η sxmax -η sxmin |, x bs =|η symax -η symin |, x cs =|η szmax -η szmin |; set up: Learned: where θ x =x am / x as ,θ y =x bm / x bs ,θ z =x cm / x cs , the parameters represent the spatial mapping scale coefficients of the three directions of the master end and the slave end, written as θ=[θ x θ y θ z ].

4. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 1 is characterized in that: In step S2: Incremental space mapping is called rate mapping. It sets a motion origin for the master and slave robots in advance. The end of the slave robot takes the unit time displacement of the master robot as a reference. By operating the master robot, the displacement of the slave robot in unit time is mapped out. The proportional coefficient matrix is ​​added to realize the unit time displacement mapping of the master and slave robots: Set d dpu is the expected motion displacement per unit time, σ m is the displacement of the master robot relative to the origin of the world coordinate system, σ m0 is the displacement of the origin of motion relative to the origin of the world coordinate system, set: d dpu =θ(σ m -s m0 ) σ m is considered to be the origin of the next unit motion time, rewritten as: ΔP s =θ·ΔP m where ΔP s and ΔP m is the motion displacement vector of the master and slave robots per unit time, and it can be inferred that: P s =P s0 +θ·ΔP m get: where V m (t) = {v mx (t)v my (t)v mz (t)}; The master-slave space mapping based on the master robot's running speed is to convert the master robot's running speed scale function Ψ(V m ) is added to the incremental mapping model: d dpu =θ·Ψ(V m )·(s m -s m0 ) The instantaneous expression of the slave robot position is: P s =P s0 +θ·Ψ(V m )·ΔP m 。 5. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 4 is characterized in that: In step S2: the proportional coefficient matrix is ​​given a fixed value. The master-slave proportional mapping coefficient is determined by the running speed of the master robot. When the master robot is too fast, the proportional mapping coefficient becomes larger, resulting in excessive displacement of the slave robot. When the master robot is too slow, the proportional mapping coefficient becomes smaller, resulting in excessive movement of the slave robot. The value range of the speed proportional function is limited to v i ∈[v imin ,v imax ] i＝{x,y,z} , for the velocity scaling function Ψ(V m ), gives: The proportional mapping coefficient is expressed as where θ i are the elements of the coefficient matrix θ, w1, w2, w3 are the gain coefficients of different working modes used when performing normal, fast and fine tasks.

6. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 1 is characterized in that: In step S3: During the deceleration process, the following two scenarios are considered: (a) the human gradually reduces the applied force; (b) the human suddenly changes the direction of the applied force; the dot product of the velocity vector and the force vector is introduced to evaluate whether the applied force has reversed direction; the compliant leader is further subdivided into semi-active and active types. The robot's response can adapt when humans intervene for safety reasons and resume the task after the intervention is eliminated, which is called semi-active behavior; when humans need strong support, the robot should become sufficiently rigid, which is called an active compliant leader.

7. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 6 is characterized in that: In step S3: the stiffness increases with the increase of the force applied by the human, ensuring accurate tracking of the predetermined trajectory under the premise of safety. The human-computer interaction at this stage is further subdivided into normal interaction or abnormal interaction: on the robot side, abnormal human behavior is subdivided into two situations: supportable and dangerous. If the interaction is abnormal but still supportable, the stiffness is increased to offset the additional force applied; if the abnormal interaction is judged to be dangerous, the robot will activate the emergency stop mechanism; compliance is achieved by switching the robot role to a passive follower.

8. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 1 is characterized in that: In step S4: By combining the position, velocity, acceleration and force information of the robot end effector, a second-order differential equation model, the impedance control model, is constructed: Among them, M, B, and K represent the inertia, damping, and stiffness coefficient matrices of the system respectively; X c Respectively represent the adjusted position, velocity, and acceleration of the robot; X d They represent the initial reference position, velocity, and acceleration of the robot, respectively. F is the actual force in three-dimensional space when the robot contacts the environment. F d For the power of expectation; Use one-dimensional model to describe and process: The admittance model can be reformulated as: Where Δf=ff d , using the zero-order holder technique, the discrete form of the admittance model is given, where the time step is denoted as k and the sampling time is denoted as T: Writing the acceleration term in differential form gives: get: The maximum speed is expressed as:

9. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 1 is characterized in that: In step S5: the active variable admittance adjusts the admittance parameters in real time according to the displacement of the robot end or the contact force between the robot and the environment; the passive variable admittance control adjusts the admittance parameters when the robot end accidentally hits an obstacle, ending the current motion state to play a protective role. The variable admittance control is divided into two stages: normal working state and abnormal working state; normal working state: the robot is in a rapid motion process or maintains constant force output after the robot end contacts the environment; abnormal working state: the current target trajectory needs to be adjusted so that it is corrected in the direction of reducing external force.

10. The robot variable admittance compliant teleoperation control method based on human intention recognition according to claim 9, characterized in that: In step S5: Design of non-contact active variable admittance control algorithm: Adjust the desired force according to the distance between the robot end and the target position: Among them, δ represents the distance between the current and target positions, δ t represents the distance from the starting position to the target position, b max is the maximum value of damping; In the contact-type passive emergency stop variable admittance control algorithm, when the robot end effector is at point x1 and encounters an unexpected obstacle, causing the contact force to suddenly increase significantly and exceed the normal threshold, the passive variable admittance control algorithm is triggered to quickly respond to this situation and quickly stop the robot motion. By passively adjusting the admittance parameters, the robot end effector accelerates the motion from x1 to x2 and accelerates the reverse acceleration from x2 to x3. The change in admittance parameters is expressed as: in b f Is the default damping value, b max and b min are the default maximum damping value and minimum damping value respectively, gd() function is the Goodman function, Indicates the default maximum acceleration of the robot end; Contact type active variable admittance control algorithm, the admittance model is expressed as: in ξ=x d -x c , k(t) can be adjusted according to the contact force, and the tracking force error ξ when the robot end interacts with the environment f Expressed as ξ f =f d -f; It turns out that: Active variable admittance controller during contact: where k p and k' d is a variable parameter, we get: After sorting, we get Lyapunov function: in The proposed Lyapunov function is the sum of kinetic energy and potential energy for motion control and force tracking control, which is obtained as: f = k e (xx e ),get: ξ f =f d -f=f d -k e (x-x e ) get The derivative of the Lyapunov function is: get:

Images