Compliant control method for massage robots with stiffness self-adaptation function

Abstract

This invention discloses a compliant control method for a massage robot with stiffness self-adaptation, comprising a six-degree-of-freedom robotic arm, an end effector, a six-dimensional force sensor, and a control host. By installing a six-dimensional force sensor at the end effector, the contact force between the robot and the human body surface can be collected in real time. Combined with end effector motion data obtained from a pose sensor, the system executes a compliant control closed loop to achieve adaptive and safe massage operation. The method can construct a force control closed loop in real time. Through the force sensor installed at the robot's end effector and combined with a compliant control algorithm based on displacement error, precise, safe, and compliant massage operation can be achieved without relying on a vision system. Simultaneously, it achieves stiffness detection and adaptive control of the massage area, effectively improving the user experience.

Description

Technical Field

[0001] This invention belongs to the field of robot control technology, and in particular relates to a compliant control method for a massage robot with stiffness adaptive function. Background Technology

[0002] With the widespread application of intelligent robots in the fields of medical care, rehabilitation, and daily life services, massage robots are gradually emerging as a promising assistive device. Most existing massage robots rely on visual sensing technology for body part recognition and path planning, resulting in high system costs, complex configurations, and uncertainties such as occlusion and changes in lighting, limiting the universality of the devices and the ease of user interaction. Furthermore, traditional control methods are mostly open-loop or semi-closed-loop control, lacking real-time force feedback information and the ability to respond in real-time to applied force and user reactions, which can easily lead to excessive or inaccurate massage intensity, causing injury to the user.

[0003] In addition, current massage robots lack the ability to detect and adaptively adjust the stiffness of the massaged area. They cannot adjust the intensity based on the user's muscle feedback, nor can they intelligently adjust the force according to the softness and hardness of different parts of the body, resulting in insufficient safety and comfort.

[0004] Therefore, it is necessary to propose a compliant control method that does not rely on a vision system, can establish a closed-loop control system through force sensing and displacement feedback, and can automatically detect and adapt to local stiffness, so as to improve the intelligence level and naturalness of human-computer interaction of massage robots. Summary of the Invention

[0005] To address the aforementioned difficulties and problems in existing methods, this invention innovatively proposes a compliant control method for massage robots with stiffness adaptive function. This method can construct a force control closed loop in real time. By using a force sensor installed at the end of the robot and combining it with a compliant control algorithm based on displacement error, it can achieve precise, safe, and compliant massage operations without relying on a vision system. At the same time, it can realize stiffness detection and adaptive control of the massage area, effectively improving the user experience.

[0006] The purpose of this invention is to propose a compliant control method for a massage robot with stiffness adaptive function, the specific solution of which is as follows:

[0007] Step 1: Force sensor installation and force data acquisition:

[0008] A six-dimensional force sensor is installed at the end of a six-degree-of-freedom robotic arm via a self-made flange and rigidly connected to the end effector through a self-made mechanism to ensure the accuracy of force signal transmission. During the robot's massage task, the sensor collects six-dimensional force data in real time when the robot's end is in contact with the human body, which serves as the basis for subsequent compliance control input.

[0009] Step 2: Acquiring pose information:

[0010] With a robotic arm having six degrees of freedom, the end effector consists of position and orientation. Using Euler angle modeling, a clear mapping relationship is established between orientation changes and angular velocity. To establish the mapping relationship, it is necessary to analyze the orientation rotation matrix of the end effector and introduce the Euler angular velocity transformation model. The orientation adopts the ZYX Euler angle modeling method, which is to first rotate around Z, then around Y, and finally around X.

[0011] Step 3: Construct a compliant control model:

[0012] By introducing a compliant control algorithm and combining impedance control with admittance control, the system collects the interactive force information generated when the robot comes into contact with the human body in real time through a six-dimensional force sensor installed at the end. Based on the error between the desired pose and the actual pose of the end, the system dynamically calculates the required response behavior, including displacement and attitude, and generates feedback control commands accordingly to achieve a compliant response effect that meets the needs of human-machine interaction and build a complete force-position closed-loop control system.

[0013] Step 4. Actuator response and closed-loop control implementation;

[0014] Based on the control target calculated in step 3, the controller generates the expected motion command at the end of the next cycle and controls the actuator to perform the motion through the joint actuator, thereby realizing complete force-pose closed-loop control. This closed loop does not rely on visual information and achieves compliant interaction entirely based on force and pose feedback.

[0015] Step 5. Stiffness testing and dynamic response adjustment:

[0016] Within each sampling period, the system calculates the stiffness of the current contact area based on the interaction force F and the pose change Δx, using the formula: k=F / Δx, where k is the stiffness of the human tissue. The stiffness result is used to adjust the controller stiffness and damping parameters in real time to adapt to the softness and hardness of the tissue and optimize the massage response characteristics.

[0017] Step 6. Introduce a model reference adaptive control mechanism:

[0018] A model reference adaptive control strategy is introduced to optimize and dynamically adjust the parameters in compliant control online. This method can adaptively adjust the control parameters according to the error between the target model and the actual system, ensuring that the robot can maintain the desired dynamic response characteristics when facing environments with different stiffness.

[0019] Step 7. Integrate voice control and intelligent interaction;

[0020] To improve the convenience and naturalness of human-computer interaction, enabling users to remotely control massage tasks or input high-level commands through natural language, this module is based on a trained voice recognition neural network model. It can semantically parse the commands issued by the user and convert them into a standard set of commands that can be executed by the robot control system. The system processes the voice signal through a voice big data model.

[0021] Step 8: Control process integration and overall scheduling;

[0022] All of the above control modules are integrated into the control host. Each hardware module communicates with the host via Ethernet. The entire control system can realize autonomous recognition of the massage area, adaptive adjustment of intensity, and natural language interaction, forming a complete closed-loop control process without the need for a vision system.

[0023] As a further improvement to the present invention, step 2, Euler angle modeling, is specifically as follows:

[0024] Euler angles ( , , corresponding angular velocity The relationship between it and its derivative can be written as:

[0025] , ;

[0026] , ;

[0027] This matrix represents the Euler angle derivative. Mapped to angular velocity in rigid body coordinates Angular velocity input used to control attitude error feedback;

[0028] To facilitate the control system in deriving the Euler angle derivative from the angular velocity acquired by the sensor, the inverse of this transformation matrix is ​​often used in numerical control:

[0029] = .

[0030] As a further improvement to the present invention, step 3 is specifically as follows:

[0031] 3.1 Impedance Model:

[0032] Simulates the dynamic behavior of a mass-spring-damped system to describe the target response state input when the end effector is subjected to an external force. The input is the desired pose and the output is the control force.

[0033] The general form of the impedance model is:

[0034] ;

[0035] in:

[0036] ;

[0037] ;

[0038] ;

[0039] The second derivative represents the error between the desired and actual acceleration at the end point of flight; the first derivative represents the velocity error; and the zeroth derivative term represents the static error between the end point position and attitude. The position representing the expected position;

[0040] : These represent the desired mass, damping, and stiffness matrices, respectively;

[0041] : The calculated force generated at the robot's end effector;

[0042] By adjusting impedance parameters, including stiffness This allows for the control of response speed and compliance, enabling the robot to have sufficient adaptability in different areas or environments.

[0043] 3.2 Admittance control:

[0044] Unlike impedance control, admittance control takes an external force as input and outputs the desired displacement change. Its mathematical form is:

[0045] ;

[0046] The equivalent mass matrix defined by the admittance model;

[0047] : Equivalent damping matrix, used to control the smoothing speed of the response;

[0048] : Equivalent stiffness matrix, used to set the displacement sensitivity of the end to external forces;

[0049] , , These represent terminal acceleration, velocity, and displacement, respectively.

[0050] : ;

[0051] 3.3. Dynamic Model Compensation:

[0052] The controller converts the end effector force into joint space torque using the Jacobian matrix, and compensates for it using the robot's dynamics model, which is as follows:

[0053] ;

[0054] Control torque in joint space

[0055] Joint angle;

[0056] : Joint angular velocity vector;

[0057] Joint angular acceleration vector;

[0058] The inertia matrix in joint space describes the effect of the robot's mass distribution on motion;

[0059] Coriolis force / centrifugal force matrix, describing the dynamic effects of velocity changes;

[0060] The gravity term represents the torque generated at each joint due to gravity.

[0061] Torque-based control implementation:

[0062] : The transpose of the Jacobian matrix from joint to Cartesian space, used to convert forces in Cartesian space into torques in joint space.

[0063] As a further improvement to the present invention, step 6 is specifically as follows:

[0064] 6.1. Model reference structure setting;

[0065] ;

[0066] in , , These represent the reference mass, damping, and stiffness matrices, respectively. Let X be the desired output trajectory. The actual system state is X, and the state error between the two is defined as:

[0067] -X;

[0068] 6.2. Construction and stability analysis of Lyapunov functions;

[0069] Construct a Lyapunov function of the following form to analyze the stability of the system and derive the parameter update law:

[0070] ;

[0071] in : This represents the error in the estimation of impedance stiffness;

[0072] >0: To adjust the gain, affecting the parameter convergence speed;

[0073] : is the desired system quality matrix;

[0074] By taking the derivative of V with respect to time and combining it with the system dynamics model, a stable adaptive rate can be obtained;

[0075] 6.3. Adaptive Rate Design;

[0076] Based on the actual external force detected With pose error The adaptive parameter adjustment rate is designed as follows;

[0077] Stiffness adaptive law:

[0078]

[0079] Damping adaptive rate:

[0080]

[0081] in , >0: Adjust the gain parameter;

[0082] f() is the velocity error response function;

[0083] All adjustment rates have physical constraint ranges. This prevents excessive stiffness or oscillation.

[0084] As a further improvement to the present invention, step 7 is specifically as follows:

[0085] 7.1 Speech Acquisition and Feature Extraction;

[0086] Users input natural language commands through a microphone, the system collects voice data, and extracts Mel spectrograms, MFCCs, or other high-dimensional feature vectors as model input;

[0087] 7.2 Command Recognition and Intent Parsing;

[0088] The system uses a trained speech recognition model to map speech information into a standard semantic structure, and then uses the intent recognition module to convert it into predefined operation labels.

[0089] 7.3 Control command generation and integration;

[0090] The voice commands mentioned above are converted into control commands and encapsulated into standard control functions, which are then embedded into the current compliant control flow. Specifically, the controller receives the control function command, updates the function parameters, passes the command to the actuator, the actuator executes it, and then receives feedback until the user says "okay" or stops and ends the process.

[0091] 7.4 Control process integration;

[0092] After the voice-converted control commands are encapsulated, they are embedded into the compliant control process and work in conjunction with the end force feedback, stiffness detection, and adaptive adjustment modules to ensure that the system behavior remains dynamically stable and interactively compliant under the influence of language input.

[0093] Compared with existing technologies, this invention achieves significant innovations in structural design and control strategy. Firstly, it constructs a complete force control closed-loop system based on compliant control. Utilizing a six-dimensional force sensor and a posture error feedback mechanism, the massage robot can perceive and respond to external interactive forces in real time, thereby achieving safe and precise force control operation. Secondly, by introducing a stiffness detection mechanism and a model reference adaptive control method, the system can automatically adjust control parameters based on the hardness of the contact area, making the robot's massage intensity more adaptable to the actual physiological characteristics of the human body, effectively improving user comfort and personalized experience. Furthermore, the system completely eliminates the traditional reliance on visual recognition modules, avoiding recognition errors caused by occlusion, lighting, and other issues. This not only significantly reduces system cost and structural complexity but also simplifies the use of the device. Further, this invention integrates a large-model control interface based on speech recognition and semantic understanding, allowing users to remotely input commands or interact in real time via natural language. This effectively enhances the naturalness, intelligence, and convenience of human-computer interaction, comprehensively improving the practical value and technological competitiveness of the massage robot in medical, rehabilitation, and home application scenarios. Attached Figure Description

[0094] Figure 1 Schematic diagram of the overall structure of the massage robot;

[0095] Figure 2 : Block diagram of a compliant control system;

[0096] Figure 3 Adaptive control flowchart;

[0097] Figure 4 User interaction and response logic flowchart. Detailed Implementation

[0098] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0099] The massage robot system of the present invention, such as Figure 1 As shown, the system consists of a six-degree-of-freedom robotic arm, an end effector, a six-dimensional force sensor, and a control unit. By installing a six-dimensional force sensor at the end effector, the contact force between the robot and the human body surface can be collected in real time. Combined with the end effector motion data obtained from the posture sensor, the system executes a compliant control closed loop to achieve adaptive and safe massage operation.

[0100] After the robot enters task mode, the control system, as follows: Figure 2 As shown, pose adjustment and force regulation are performed based on an impedance-admittance control structure. This control system utilizes the end-effector force information F and displacement error. Construct a compliant control model and obtain the target dynamic behavior through the following expression:

[0101]

[0102]

[0103]

[0104] in : These represent the desired mass, damping, and stiffness matrices, respectively. The equivalent mass matrix defined for the admittance model. This is the equivalent damping matrix, used to control the smoothing speed of the response. This is the equivalent stiffness matrix, used to set the displacement sensitivity of the end effector to external forces. , , These represent the terminal acceleration, velocity, and displacement, respectively, with the subscript 'd' indicating the desired position. Based on the above control objectives, the system generates the desired terminal position for the next cycle. The instructions are transmitted to the lower-level machine to drive the actuator and achieve precise movement.

[0105] During the continuous contact between the robot and the human body, the system uses the interaction force F and the end displacement error acquired in real time during each sampling cycle. Calculate the stiffness value of the current contact area:

[0106]

[0107] The stiffness estimate will be used to dynamically adjust the control model parameters. When the muscles in the detection area are stiff, the system automatically increases the damping coefficient and decreases the stiffness coefficient to avoid discomfort caused by a violent response from the control system; when the detected tissue is soft, the stiffness and control precision are appropriately increased to improve the massage depth and effectiveness.

[0108] To enhance the system's responsiveness in complex environments, this invention further introduces a Model Reference Adaptive Control (MRAC) method, such as... Figure 3 As shown. Let the target reference model be:

[0109]

[0110] The actual system state is X, and the error is... -X, constructs Lyapunov functions

[0111]

[0112] Based on stability analysis, the following adaptive law is obtained.

[0113] Stiffness update law:

[0114]

[0115] Damping update rate:

[0116]

[0117] in, To adjust the gain, the velocity error response function f() is used, and all updated parameters are limited to the physical safety range to avoid excessive stiffness or response overshoot that could cause control instability.

[0118] Ultimately, the system dynamically adjusts the control gain based on the estimated stiffness, enabling the massage robot to adopt adaptive compliant adjustment strategies for different areas of the human body during execution. Users can then interact with the robot using a pre-trained voice module, such as... Figure 4 As shown, because the voice module has been embedded in the compliant control process, personalized interaction that adapts to the user's input can be achieved.

[0119] The entire control process is completely independent of external vision systems. Users only need to guide the robotic arm to the target massage area or use interactive methods such as dragging, voice commands, and teaching points to start control. Based on force feedback and stiffness judgment, the robotic arm can autonomously plan and execute massage strategies to complete a smooth, safe, and highly interactive intelligent massage process.

[0120] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A compliant control method for a massage robot with stiffness adaptive function, characterized in that, The specific plan is as follows: Step 1: Force sensor installation and force data acquisition: A six-dimensional force sensor is installed at the end of a six-degree-of-freedom robotic arm via a self-made flange and rigidly connected to the end effector through a self-made mechanism to ensure the accuracy of force signal transmission. During the robot's massage task, the sensor collects six-dimensional force data in real time when the robot's end is in contact with the human body, which serves as the basis for subsequent compliance control input. Step 2: Acquiring pose information: With a robotic arm having six degrees of freedom, the end effector consists of position and orientation. Using Euler angle modeling, a clear mapping relationship is established between orientation changes and angular velocity. To establish the mapping relationship, it is necessary to analyze the orientation rotation matrix of the end effector and introduce the Euler angular velocity transformation model. The orientation adopts the ZYX Euler angle modeling method, which is to first rotate around Z, then around Y, and finally around X. Step 3: Construct a compliant control model: By introducing a compliant control algorithm and combining impedance control with admittance control, the system collects the interactive force information generated when the robot comes into contact with the human body in real time through a six-dimensional force sensor installed at the end. Based on the error between the desired pose and the actual pose of the end, the system dynamically calculates the required response behavior, including displacement and attitude, and generates feedback control commands accordingly to achieve a compliant response effect that meets the needs of human-machine interaction and build a complete force-position closed-loop control system. Step 4. Actuator response and closed-loop control implementation; Based on the control target calculated in step 3, the controller generates the expected motion command at the end of the next cycle and controls the actuator to perform the motion through the joint actuator, thereby realizing complete force-pose closed-loop control. This closed loop does not rely on visual information and achieves compliant interaction entirely based on force and pose feedback. Step 5. Stiffness testing and dynamic response adjustment: Within each sampling period, the system calculates the stiffness of the current contact area based on the interaction force F and the pose change Δx, using the formula: k=F / Δx, where k is the stiffness of the human tissue. The stiffness result is used to adjust the controller stiffness and damping parameters in real time to adapt to the softness and hardness of the tissue and optimize the massage response characteristics. Step 6. Introduce a model reference adaptive control mechanism: By introducing a model reference adaptive control strategy, the parameters in compliant control are optimized and dynamically adjusted online. This method can adaptively adjust the control parameters according to the error between the target model and the actual system, ensuring that the robot can maintain the desired dynamic response characteristics when facing environments with different stiffness. Step 7. Integrate voice control and intelligent interaction; To improve the convenience and naturalness of human-computer interaction, enabling users to remotely control massage tasks or input high-level commands through natural language, this module is based on a trained voice recognition neural network model. It can semantically parse the commands issued by the user and convert them into a standard set of commands that can be executed by the robot control system. The system processes the voice signal through a voice big data model. Step 8: Control process integration and overall scheduling; All of the above control modules are integrated into the control host. Each hardware module communicates with the host via Ethernet. The entire control system can realize autonomous recognition of the massage area, adaptive adjustment of intensity, and natural language interaction, forming a complete closed-loop control process without the need for a vision system.

2. The compliant control method for a massage robot with stiffness adaptive function according to claim 1, characterized in that, The Euler angle modeling in step 2 is as follows: Euler angles ( , , corresponding angular velocity The relationship between it and its derivative can be written as: , ； , ； This matrix represents the Euler angle derivative. Mapped to angular velocity in rigid body coordinates Angular velocity input used to control attitude error feedback; To facilitate the control system in deriving the Euler angle derivative from the angular velocity acquired by the sensor, the inverse of this transformation matrix is ​​often used in numerical control: = 。 3. The compliant control method for a massage robot with stiffness adaptive function according to claim 1, characterized in that, Step 3 is described in detail below: 3.1 Impedance Model: Simulates the dynamic behavior of a mass-spring-damped system to describe the target response state input when the end effector is subjected to an external force. The input is the desired pose and the output is the control force. The impedance model takes the form of: ； in: ； ； ； The second derivative represents the error between the desired and actual acceleration at the end point of flight; the first derivative represents the velocity error; and the zeroth derivative term represents the static error between the end point position and attitude. The position representing the expected position; : These represent the desired mass, damping, and stiffness matrices, respectively; : The calculated force generated at the robot's end effector; By adjusting impedance parameters, including stiffness This allows for the control of response speed and compliance, enabling the robot to have sufficient adaptability in different areas or environments. 3.2 Admittance control: Unlike impedance control, admittance control takes an external force as input and outputs the desired displacement change. Its mathematical form is: ； The equivalent mass matrix defined by the admittance model; : Equivalent damping matrix, used to control the smoothing speed of the response; : Equivalent stiffness matrix, used to set the displacement sensitivity of the end to external forces; , , These represent terminal acceleration, velocity, and displacement, respectively. ： ； 3.3 Dynamic Model Compensation: The controller converts the end effector force into joint space torque using the Jacobian matrix, and compensates for it using the robot's dynamics model, which is as follows: ； Control torque in joint space Joint angle; : Joint angular velocity vector; Joint angular acceleration vector; The inertia matrix in joint space describes the effect of the robot's mass distribution on motion; Coriolis force / centrifugal force matrix, describing the dynamic effects of velocity changes; The gravity term represents the torque generated at each joint due to gravity. Torque-based control implementation: ; : The transpose of the Jacobian matrix from joint to Cartesian space, used to convert forces in Cartesian space into torques in joint space.

4. The compliant control method for a massage robot with stiffness adaptive function according to claim 1, characterized in that, Step 6 is as follows: 6.

1. Model reference structure setting; ； in , , These represent the reference mass, damping, and stiffness matrices, respectively. Given the desired output trajectory and the actual system state X, the state error between the two is defined as: -X； 6.

2. Construction and stability analysis of Lyapunov functions; Construct a Lyapunov function of the following form to analyze the stability of the system and derive the parameter update law: ； in : This represents the error in the estimation of impedance stiffness; >0: To adjust the gain, affecting the parameter convergence speed; : is the desired system quality matrix; By taking the derivative of V with respect to time and combining it with the system dynamics model, a stable adaptive rate can be obtained; 6.

3. Adaptive Rate Design; Based on the actual external force detected With pose error The adaptive parameter adjustment rate is designed as follows; Stiffness adaptive law: ； Damping adaptive rate: ； in , >0: Adjust the gain parameter; f() is the velocity error response function; All adjustment rates have physical constraint ranges. This prevents excessive stiffness or oscillation.

5. The compliant control method for a massage robot with stiffness adaptive function according to claim 1, characterized in that, Step 7 is described in detail below: 7.1 Speech Acquisition and Feature Extraction; Users input natural language commands through a microphone, the system collects voice data, and extracts Mel spectrograms, MFCCs, or other high-dimensional feature vectors as model input; 7.2 Command Recognition and Intent Parsing; The system uses a trained speech recognition model to map speech information into a standard semantic structure, and then uses the intent recognition module to convert it into predefined operation labels. 7.3 Control command generation and fusion; The voice commands mentioned above are converted into control commands and encapsulated into standard control functions, which are then embedded into the current compliant control flow. Specifically, the controller receives the control function command, updates the function parameters, and passes the command to the actuator. The actuator executes the command and receives feedback until the user says "okay" or "stop" to end the process. 7.4 Control process integration; After the voice-converted control commands are encapsulated, they are embedded into the compliant control process and work in conjunction with the end force feedback, stiffness detection, and adaptive adjustment modules to ensure that the system behavior remains dynamically stable and interactively compliant under the influence of language input.

Images