Upper limb exoskeleton robot with gravity compensation function, control method and system

Abstract

The invention provides an upper limb exoskeleton robot with a gravity compensation function and a control method and system. The upper limb exoskeleton robot comprises an upper limb exoskeleton mechanism and a gravity balancing device. The gravity balancing device is integrated in the upper limb exoskeleton mechanism and comprises a multi-connecting-rod mechanism and an elastic piece. The parameters of the rod length and the hinge point position of the multi-connecting rod mechanism of the gravity balancing device, the elastic force coefficient of the elastic piece and the initial length are optimized through an algorithm, so that self-adaptive gravity compensation is provided in the working space of the upper limb exoskeleton mechanism. The gravity balancing device optimized through an algorithm has the advantages of reducing motor energy consumption and improving gravity compensation precision and system flexibility, and the problems that the contact force sensing precision is insufficient, the human-computer interaction process is stiff and the requirement for flexibility of rehabilitation training cannot be met due to the fact that feedback adjustment is conducted only through current loop estimation torque are solved.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation medical robot technology, and in particular to an upper limb exoskeleton robot with gravity compensation, its control method, and system. Background Technology

[0002] In the field of medical rehabilitation, the number of patients with limb movement disorders continues to grow, among whom those with limited upper limb function due to factors such as stroke and spinal cord injury have an urgent need for rehabilitation training equipment. Existing upper limb exoskeleton robots face significant technical bottlenecks in neurorehabilitation, elderly assistance, and industrial assistance applications.

[0003] As a major weight-bearing joint, the shoulder joint generates gravitational torque during movement, which needs to be offset by the continuous output of redundant torque from the drive motor. This leads to a significant increase in motor energy consumption and weakens the dynamic response and motion compliance of the control system. Current gravity balancing technologies mostly use springs or linkage mechanisms, but the design of structural parameters generally relies on human experience and lacks systematic optimization methods. This results in large matching errors between gravitational torque and balancing torque at different joint angles, making it difficult to adapt to the needs of multi-posture movements.

[0004] Furthermore, to achieve safe and natural human-machine collaboration, existing admittance control strategies fail to effectively integrate the real-time compensation effect of mechanical balance torque, relying solely on torque estimation via a current loop for feedback adjustment. This results in insufficient accuracy in contact force sensing, a stiff human-machine interaction process, and an inability to meet the core requirement of flexibility in rehabilitation training. These issues collectively restrict improvements in training effectiveness, wearing comfort, and system energy efficiency of exoskeleton devices, necessitating the development of novel gravity compensation mechanisms and control architectures. Summary of the Invention

[0005] This invention provides an upper limb exoskeleton robot with gravity compensation, a control method, and a system. It has the advantages of reducing motor energy consumption, improving gravity compensation accuracy, and enhancing system compliance through an algorithm-optimized gravity balancing device. This solves the problem that relying solely on current loop to estimate torque for feedback adjustment leads to insufficient contact force sensing accuracy, a stiff human-machine interaction process, and an inability to meet the compliance requirements of rehabilitation training.

[0006] This invention provides an upper limb exoskeleton robot with gravity compensation, comprising an upper limb exoskeleton mechanism and a gravity balancing device. The exoskeleton mechanism includes a backplate and shoulder joint links. The shoulder joint links and the backplate are rotatably connected via a first shoulder joint module, which is configured to control the swing of the shoulder joint links. The gravity balancing device is configured to provide adaptive gravity compensation within the workspace of the upper limb exoskeleton mechanism. The gravity balancing device includes a first link, a second link, and a tension spring. The two ends of the first link are respectively hinged to the first shoulder joint module to form hinge point A and to one end of the second link to form hinge point B. The other end of the second link forms hinge point C with the shoulder joint link. The hinge center of the first shoulder joint is D, forming a four-bar linkage ABCD. The two ends of the tension spring form hinge point E with the first link and hinge point F with the shoulder joint link.

[0007] In one embodiment of the present invention, the upper limb exoskeleton mechanism further includes a second shoulder joint module, an upper arm, an elbow joint module, and a forearm connected in sequence, and the first shoulder joint module, the second shoulder joint module, and the elbow joint module provide pivoting for the shoulder joint link, the upper arm, and the forearm, respectively; a gravity balancing device is integrated in the first shoulder joint module to provide adaptive gravity compensation within the workspace of the first shoulder joint module.

[0008] In one embodiment of the present invention, the upper limb exoskeleton mechanism includes three degrees of freedom, namely, shoulder abduction and adduction pulled by the first shoulder joint module, upper arm swing and forearm swing, and a gravity balancing device loads the gravity pulled by the first shoulder joint module for shoulder abduction and adduction.

[0009] In one embodiment of the present invention, the distance between the hinge point EF at both ends of the tension spring and the hinge point A on the shoulder joint module is greater than the distance between the hinge point BC on the first link and the second link and the hinge point A on the shoulder joint module.

[0010] In one embodiment of the present invention, the link lengths, hinge point positions, elastic coefficients, and initial lengths of the elastic elements in the four-bar linkage ABCD are configured to be obtained through a parameter optimization algorithm, wherein the parameter optimization algorithm includes:

[0011] Establish the first shoulder joint gravitational moment with respect to the joint rotation angle gravitational torque function The elastic compensating torque function generated by the spring force transmitted through the four-bar linkage ),in The rotation angle of the target structure in the gravity balancing device. The parameters to be optimized include rod length, hinge point location, elastic coefficient of the elastic element, and initial length.

[0012] The objective function is constructed by minimizing the ratio of the maximum residual gravitational torque to the maximum gravitational torque; Based on the objective function, the particle swarm optimization algorithm is used to optimize the parameters. Global optimization is performed by iteratively updating the velocity and position of particles, tracking individual and global extrema, and finally converging to the optimal parameter combination P that minimizes the objective function T(P).

[0013] In one embodiment of the present invention, the gravitational torque function is: τ_g(θ) = m g r cos(θ), where m is the mass of the entire arm and load of the exoskeleton, and r is the distance from the center of mass to the hinge point A on the first shoulder joint module; The spring compensation torque is calculated by using geometric relationships to determine the spring length L_s(θ,P) and lever arm r_s(θ,P), resulting in τ_s(θ,P) = k. [L_s(θ,P) L0] r_s(θ,P), where L0 is the free length of the spring; The objective function is: , where the angle θ ranges from [-π / 2, π / 2].

[0014] This invention also provides a compliant control method for an upper limb exoskeleton robot with gravity compensation, applied to the aforementioned upper limb exoskeleton robot. The control method includes: The current, speed, and position signals of the motor encoder of the shoulder joint module are acquired in real time with a fixed sampling period. The current and speed values ​​are filtered to eliminate high-frequency noise and obtain the filtered current signal I_filtered. Based on the Newton-Euler recursive dynamics algorithm, the gravitational torque G(q) and Coriolis torque are calculated in real time by inputting the joint position θ and velocity V. In conjunction with the optimized parameters of the passive gravity balancing device, the compensation torque τ_s provided by it is calculated based on the current joint angle θ; Calculate the motor output torque τ_m=K_t using the motor torque constant K_t. I_filtered; then through the dynamic model Solve for the human-computer interaction torque τ_h, where It is the moment of inertia; The interaction torque τ_h is input into the admittance controller, and the dynamic equation of the second-order mass-damped-spring system is simulated as follows: The trajectory adjustment θ_e and its differential term are solved by numerical integration. The original desired trajectory θ_d is superimposed with the trajectory adjustment amount θe output by the admittance controller to generate the final joint position command θ_cmd=θ_d+θe. The motor is driven by the PID position closed-loop controller to accurately track the command, while the gravity balance device provides basic gravity compensation.

[0015] In one embodiment of the present invention, the calculation steps of the admittance controller include: The interactive torque τ_h based on the current loop estimation is used as the input of the admittance controller, where the signal-to-noise ratio of τ_h is improved due to passive gravity compensation, which can accurately reflect the user's intention. The admittance controller follows the second-order system equations Real-time solution is performed using Euler discretization, and the trajectory adjustment is output. Where M_d, B_d, and K_d are adjustable admittance parameters; The output of the admittance controller Used to correct the original expected trajectory θ_d and generate θ_cmd, enabling the exoskeleton to adapt to external interactive forces and achieve natural human-machine collaboration; The θ_cmd is input into the PID controller, which generates a motor drive signal through proportional, integral, and derivative operations to achieve high-precision position tracking. The passive gravity balancing device continuously counteracts gravity interference, reducing the motor load.

[0016] In one embodiment of the present invention, in the admittance control calculation step, the admittance parameters M_d, B_d, and K_d can be adjusted online to change the compliance of the exoskeleton with external interaction forces.

[0017] The present invention also provides a gravity-compensated compliant control system for an upper limb exoskeleton robot, applied to the above-mentioned upper limb exoskeleton robot, and the above-mentioned compliant control method for the upper limb exoskeleton robot. The control system includes: a signal filtering module, a calculation module, an interactive torque module, an admittance control module, and an instruction synthesis module. The signal filtering module is used to collect and filter encoder signals of current, speed, and position of the robot's joint motors in real time; the calculation module calculates the real-time dynamic torque, including gravitational torque, Coriolis torque, and inertial torque, based on the Newton-Euler algorithm, and integrates the passive compensation torque of the gravity balancing device; the interaction torque module calculates the motor output torque through the motor torque constant and the filtered current, and outputs a pure interaction torque by combining the real-time dynamic torque and the passive compensation torque; the admittance control module converts the pure interaction torque into a trajectory adjustment amount by simulating a second-order mass-damped-spring system; the instruction synthesis module superimposes the desired trajectory with the trajectory adjustment amount, and drives the joint motors to track the position through a PID controller.

[0018] The beneficial effects of the present invention include at least the following: 1. This invention uses the particle swarm optimization (PSO) algorithm to optimize the global parameters of the passive balancing mechanism, breaking through the limitations of traditional design methods that can only achieve balancing at a single point or in a small range. It realizes efficient gravity compensation in the entire motion space of the shoulder joint, fundamentally reducing the power demand and operating energy consumption of the drive system, and achieving globally optimized passive compensation.

[0019] 2. Passive compensation eliminates most of the gravitational interference, significantly improving the signal-to-noise ratio of the interactive torque signal estimated by the current loop. This enables the admittance controller to respond more quickly and accurately to the user's true intentions, greatly improving the naturalness, safety, and comfort of human-computer interaction, and achieving purified force perception in human-computer interaction.

[0020] 3. Gravity load is borne by a passive gravity balancing mechanism, while the active control system handles dynamic interaction and motion tracking. This collaborative model not only saves energy but also reduces the peak torque requirements of the actuator, allowing for the use of smaller, lighter motors, which is beneficial for system miniaturization and weight reduction, realizing the synergistic advantages of the hybrid architecture.

[0021] 4. By fully utilizing the current feedback from the CAN communication of the motor when it is off for torque estimation, high-precision compliant control is achieved while avoiding the installation of expensive and easily interfered force sensors at the joints. This reduces system cost and complexity, improves reliability, and realizes a low-cost solution without additional force sensors. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0023] In the attached diagram: Figure 1 This is a schematic diagram of the structure of an upper limb exoskeleton robot provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the gravity balancing device provided in one embodiment of the present invention; Figure 3 This is a flowchart illustrating the particle swarm optimization algorithm used for mechanism parameter design in one embodiment of the present invention. Figure 4 This is an architecture diagram of a compliant control system provided in one embodiment of the present invention; Figure 5 This is a schematic diagram of a compliant control method provided in one embodiment of the present invention.

[0024] The attached figures are labeled as follows: 100. Gravity balancing device; 110. First link; 120. Second link; 130. Tension spring; 10. Back plate; 20. First shoulder joint module; 30. Shoulder joint link; 40. Second shoulder joint module; 50. Upper arm; 60. Elbow joint module; 70. Forearm. Detailed Implementation

[0025] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0026] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. The drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0027] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0028] Traditional upper limb exoskeleton robots commonly suffer from problems during rehabilitation training, including severe motor overload, high power loss, bulky mechanisms, and difficulty in accurately assisting patients in performing movements. Specifically, the gravitational torque generated by shoulder joint movements requires the motors to output a large amount of redundant torque to offset it, leading to a surge in system energy consumption and a reduction in control response speed. Furthermore, existing gravity balance schemes largely rely on human experience in design and lack systematic optimization methods, resulting in low gravity compensation accuracy and an inability to adapt to the needs of multi-posture movements.

[0029] For this, please see Figures 1 to 2This application proposes an upper limb exoskeleton robot with gravity compensation, including an upper limb exoskeleton mechanism and a gravity balancing device 100. The gravity balancing device 100 is integrated into the upper limb exoskeleton mechanism and includes a multi-link mechanism and an elastic element. The parameters of the gravity balancing device 100 are optimized by an algorithm to provide adaptive gravity compensation within the workspace of the upper limb exoskeleton mechanism. The parameters of the gravity balancing device 100 include the link length of the multi-link mechanism, the hinge point position, the elastic coefficient of the elastic element, and the initial length.

[0030] Specifically, the exoskeleton mechanism includes a backplate 10 and a shoulder joint link 30. The shoulder joint link 30 and the backplate 10 are rotatably connected through a first shoulder joint module 20, which is configured to control the swing of the shoulder joint link 30. A gravity balancing device 100 is configured to provide adaptive gravity compensation within the workspace of the upper limb exoskeleton mechanism. The gravity balancing device 100 includes a first link 110, a second link 120, and a tension spring 130. The two ends of the first link 110 are respectively hinged to the first shoulder joint module 20 to form hinge point A and to one end of the second link 120 to form hinge point B. The other end of the second link 120 is hinged to the shoulder joint link 30 to form hinge point C. The hinge center of the first shoulder joint module 20 is D, forming a four-bar linkage ABCD. The two ends of the tension spring 130 are respectively hinged to the first link 110 to form hinge point E and to the shoulder joint link 30 to form hinge point F.

[0031] The first link 110, a key component of the gravity balancing device 100, is typically made of lightweight, high-strength materials, such as aerospace aluminum alloy or carbon fiber composites. This link transmits force and motion within the mechanism, and its length and geometry are important parameters affecting the compensation effect. The second link 120, similar to the first link 110, is also made of lightweight, high-strength materials. Its main function is to form a linkage mechanism with the first link 110 and the shoulder joint link 30, achieving specific motion trajectories and torque transmission. The tension spring 130, as a concrete implementation of the elastic element, provides gravity compensation through its elastic force. Its elastic coefficient and initial length are key parameters optimized by algorithms to match the gravitational torque. The selection of the spring should consider its fatigue life and linearity within its operating range.

[0032] Hinge point A is the connection point between the first link 110 and the first shoulder joint module 20, typically a pivot point fixed to the first shoulder joint module 20, serving as one of the bases of the entire four-bar linkage. Hinge point B is the connection point between the first link 110 and the second link 120, allowing relative rotation between the two links. Hinge point C is the connection point between the second link 120 and the shoulder joint link 30, transmitting the force from the gravity balancing device 100 to the shoulder joint link 30. Hinge point D is the connection point between the shoulder joint link 30 and the pivot end of the first shoulder joint module 20, representing the actual rotation axis of the shoulder joint of the upper limb exoskeleton mechanism. Hinge point E is the connection point between the tension spring 130 and the first link 110, through which the spring's tension acts on the first link 110. Hinge point F is the connection point between the tension spring 130 and the shoulder joint link 30, through which the spring's tension acts on the shoulder joint link 30. The four-bar linkage ABCD refers to a planar four-bar linkage consisting of a first shoulder joint module 20 (as a frame), a first link 110, a second link 120, and a shoulder joint link 30 connected by hinge points A, B, C, and D. This mechanism has well-defined kinematic and dynamic characteristics. By precisely designing the length of each link and the position of the hinge points, an effective match between spring torque and gravitational torque can be achieved, thereby providing efficient gravity compensation throughout the entire range of motion of the shoulder joint.

[0033] By specifying the gravity balancing device 100 as a particular structure consisting of a first link 110, a second link 120, and a tension spring 130, and defining the connection methods of its hinge points A, B, C, D, E, and F, and particularly by designing the multi-link mechanism as a four-bar linkage ABCD, this application provides a gravity compensation scheme with a clear structure and controllable mechanical properties. This specific four-bar linkage design allows the elastic force of the tension spring 130 to generate a compensation torque highly matched to the gravitational torque of the upper limb exoskeleton mechanism during the pivoting process of the shoulder joint link 30 through precise lever arm conversion. Compared to a generalized multi-link mechanism, this specific structure simplifies the parameter optimization process and ensures the transmission efficiency and accuracy of the compensation torque, thereby achieving smoother and more effective adaptive gravity compensation throughout the workspace. This significantly reduces the gravitational load on the upper limb exoskeleton mechanism, lowers the power consumption of the drive motor, improves the comfort and smoothness of user operation, and extends battery life.

[0034] For ease of understanding, the following explains some key terms in this embodiment: An upper limb exoskeleton is a wearable robotic system whose structure corresponds to the human upper limb skeletal structure, enabling or enhancing the user's upper limb movement capabilities. This mechanism typically consists of multiple links and joints, with actuators driving the rotation of the joints, thereby moving the user's upper limbs.

[0035] Gravity balancing device 100 refers to a mechanical device used to counteract or reduce the effects of gravity on the upper limb exoskeleton mechanism and its load. This device aims to generate a torque opposite to the direction of gravity through a mechanical structure or elastic element, thereby reducing the output torque required by the actuator to maintain posture or movement.

[0036] A multi-link mechanism is a mechanical system consisting of multiple rigid links connected by hinges. This mechanism achieves complex motion trajectories and force transmission through the relative movement between the links, and is often used to convert the force of elastic elements into specific compensating torques.

[0037] An elastic element is a mechanical component with the ability to deform elastically, such as a spring. When subjected to an external force, the elastic element deforms and returns to its original shape after the force is removed. There is a specific relationship between the amount of deformation and the applied force. It can be used to store and release energy and provide compensating force.

[0038] Algorithm optimization refers to the calculation and adjustment of the structural parameters of the gravity balancing device 100 using mathematical algorithms to achieve its optimal state under specific performance indicators. This optimization process aims to systematically find the best combination of parameters, rather than relying on experience or trial and error.

[0039] Adaptive gravity compensation refers to the ability of the gravity balancing device 100 to automatically adjust the compensation torque it provides based on the different postures of the upper limb exoskeleton mechanism within the workspace, in order to continuously and effectively counteract the effects of gravity. This compensation method ensures high-precision gravity compensation throughout the entire range of motion.

[0040] This embodiment provides an upper limb exoskeleton robot with gravity compensation. The robot primarily includes an upper limb exoskeleton mechanism. This mechanism can be composed of multiple rigid links connected by rotary joints. For example, it may include a back support structure fixed to the user's torso, and multiple links connected to the support structure and extending to the user's arm. These links move via motor-driven joints. Alternatively, the upper limb exoskeleton mechanism can be designed as a modular structure, with each module corresponding to a major segment of the human upper limb, assembled via quick-connect fittings to accommodate users of different body types.

[0041] The upper limb exoskeleton integrates a gravity balancing device 100. This gravity balancing device 100 can be installed near any one or more joints of the upper limb exoskeleton, for example, it can be installed in the shoulder joint region to primarily compensate for the gravity borne by the shoulder joint. Alternatively, the gravity balancing device 100 can be distributed across multiple joints, for example, simultaneously installed in the shoulder and elbow joint regions to provide more comprehensive gravity compensation.

[0042] The gravity balancing device 100 further includes a multi-link mechanism and an elastic element. The multi-link mechanism can be formed by connecting several rigid links via revolute joints; for example, it can be a three-bar linkage, where one link is fixed to the frame, and the other two links are connected by hinges and cooperate with the elastic element. The elastic element can be a compression spring, with one end fixed to a link of the multi-link mechanism and the other end fixed to the frame, generating a compensating force through compression or tension. Alternatively, the elastic element can be a torsion spring, directly mounted at a revolute joint of the multi-link mechanism, providing torque compensation through torsional deformation.

[0043] The parameters of the gravity balancing device 100 are optimized through an algorithm. This optimization process can employ numerical calculation methods, such as iteratively calculating the compensation effect under different parameter combinations and selecting the optimal parameters. Alternatively, a mathematical model of the gravity balancing device 100 can be established, and optimization algorithms such as gradient descent can be used to gradually adjust the parameters to approximate the optimal solution under preset performance indicators.

[0044] The algorithm optimization aims to enable the gravity balancing device 100 to provide adaptive gravity compensation within the workspace of the upper limb exoskeleton. For example, through optimization, the gravity balancing device 100 can continuously provide compensation matching the gravitational torque throughout the entire range of motion of the upper limb exoskeleton, from vertical descent to horizontal extension. As one implementation method, the compensation effect can be evaluated at multiple discrete joint angle points, and parameters can be adjusted through the optimization algorithm to minimize the compensation error at these discrete points.

[0045] The parameters of the gravity balancing device 100 include the link lengths, hinge point positions, elastic coefficients, and initial lengths of the multi-link mechanism. Specifically, the link lengths refer to the geometric lengths of each link constituting the mechanism; for example, they can be set to several preset values ​​and selected during optimization. The hinge point positions refer to the spatial coordinates of the connection points between the links in the multi-link mechanism and between the links and the frame; for example, they can be set as parameters that vary within a specific area. The elastic coefficient of the elastic element refers to the force or torque generated per unit deformation of the elastic element; for example, it can be selected from a series of standard springs. The initial length of the elastic element refers to its natural length when not under stress; for example, it can be adjusted according to the installation space.

[0046] Through the above technical solution, this embodiment effectively solves the problems of severe motor load, high energy consumption, and insufficient gravity compensation accuracy in traditional upper limb exoskeleton robots. This solution optimizes the structural parameters of the gravity balancing device 100 through algorithms, enabling the device to provide continuous and high-precision adaptive gravity compensation throughout the entire workspace of the upper limb exoskeleton mechanism. This significantly reduces the torque output required by the actuators to maintain posture and movement, thereby reducing system energy consumption and improving the overall motion performance and assistive effect of the exoskeleton mechanism, providing users with a more natural and efficient rehabilitation training or assistive experience.

[0047] In some embodiments described above, this application proposes an upper limb exoskeleton robot with gravity compensation. The parameters of its gravity balancing device 100 are optimized through algorithms to provide adaptive gravity compensation within the workspace of the upper limb exoskeleton mechanism. However, in practical applications, upper limb exoskeletons typically contain multiple joints and links, resulting in complex structures. The range of motion and the gravity load borne by each joint vary. If the integration location and compensation target of the gravity balancing device 100 are unclear, the compensation effect may be poor, failing to accurately offset the gravity load of specific joints. This can affect the overall motion performance of the exoskeleton, the energy consumption of the drive motor, and the user's comfort and natural human-computer interaction experience.

[0048] Please refer to the appendix for details. Figure 1 and Figure 2 This application further proposes an upper limb exoskeleton robot, whose upper limb exoskeleton mechanism also includes a second shoulder joint module 40, an upper arm 50, an elbow joint module 60 and a forearm 70 connected in sequence, and the first shoulder joint module 20, the second shoulder joint module 40 and the elbow joint module 60 provide pivoting for the shoulder joint link 30, the upper arm 50 and the forearm 70 respectively; a gravity balancing device 100 is integrated in the first shoulder joint module 20 to provide adaptive gravity compensation within the workspace of the first shoulder joint module 20.

[0049] Specifically, the backplate 10, serving as the base of the upper limb exoskeleton, is typically fixed to the user's torso. Its design should fully consider ergonomics, employing lightweight, high-strength materials such as carbon fiber composites or aerospace-grade aluminum alloys to ensure wearing comfort, stability, and overall structural lightness. The first shoulder joint module 20, connecting the backplate 10 to the shoulder joint linkage 30, is a key component simulating human shoulder movement, providing pivotal movements such as abduction or adduction. This module can integrate drive and sensing elements such as motors, reducers, and encoders to achieve active control and precise sensing of shoulder movements. The shoulder joint linkage 30 connects the first shoulder joint module 20 and the second shoulder joint module 40. Its length and geometry are optimized to simulate the connection between the scapula and humerus, ensuring the naturalness of the exoskeleton's movement trajectory. The second shoulder joint module 40, connecting the shoulder joint linkage 30 to the upper arm 50, provides another degree of pivotal freedom for the shoulder, such as internal or external rotation, and can also integrate drive and sensing elements. The upper arm 50 connects the second shoulder joint module 40 and the elbow joint module 60, serving as an intermediate link in the upper limb exoskeleton mechanism. Its rigid structure supports the weight of the elbow joint and forearm 70. The elbow joint module 60 connects the upper arm 50 and forearm 70, providing pivotal movements at the elbow, such as flexion or extension, enabling active control of the elbow. The forearm 70 connects to the elbow joint module 60, serving as the end link of the exoskeleton, which can be further connected to the hand or an end effector. These joint modules, by providing pivotal movements, collectively form the kinematic chain of the upper limb exoskeleton mechanism, enabling the exoskeleton to simulate the complex movements of the human upper limb.

[0050] Based on this, the gravity balancing device 100 is integrated into the first shoulder joint module 20. This integration means that the multi-link mechanism and elastic elements of the gravity balancing device 100 are designed to be integrated with the structure of the first shoulder joint module 20, for example, by directly hinged the links of the multi-link mechanism to the fixed or rotating parts of the first shoulder joint module 20. In this way, the gravity balancing device 100 can specifically compensate for the gravitational load borne by the first shoulder joint module 20 and caused by its movement. This compensation is adaptive, meaning that the compensation torque can be dynamically adjusted according to the real-time angular changes of the first shoulder joint module 20 throughout the workspace to always counteract or significantly reduce the gravitational torque borne by the joint.

[0051] By precisely integrating the gravity balancing device 100 into the first shoulder joint module 20 of the upper limb exoskeleton mechanism through the above-described technical solution, efficient and precise gravity compensation can be achieved for the shoulder joint, which bears the greatest weight in the upper limb exoskeleton mechanism. This targeted integration method allows the gravity balancing device 100 to focus on counteracting the gravitational torque generated by the first shoulder joint module 20 and its subsequent linkages (such as shoulder joint linkage 30, second shoulder joint module 40, upper arm 50, elbow joint module 60, and forearm 70) during movement, significantly reducing the burden on the motor driving the first shoulder joint module 20 and lowering energy consumption. Simultaneously, because the gravity balancing device 100 provides adaptive gravity compensation within the workspace of the first shoulder joint module 20, users can feel a significant reduction in arm weight when operating the exoskeleton, thereby improving wearing comfort and the smoothness of movement, making human-computer interaction more natural and efficient. This structured design also provides a better torque environment for subsequent compliant control, as passive gravity compensation has already offset most of the gravity interference, allowing the controller to focus more on recognizing and responding to user intentions.

[0052] In some embodiments described above in this application, an upper limb exoskeleton robot with gravity compensation is proposed, wherein the gravity balancing device 100 is integrated into the first shoulder joint module 20 to provide adaptive gravity compensation within the workspace of the first shoulder joint module 20. However, upper limb exoskeletons typically have multiple degrees of freedom of motion. If the main load direction targeted by the gravity balancing device 100 is not clearly defined, the compensation effect may not be sufficiently focused, especially in the direction of motion bearing the main gravity load, where the user may still experience significant fatigue.

[0053] In this regard, this application further proposes that the upper limb exoskeleton mechanism includes three degrees of freedom, namely the shoulder abduction and adduction pulled by the first shoulder joint module 20, the swing of the upper arm 50 and the swing of the forearm 70, and the gravity balancing device 100 mainly bears the gravity of the shoulder abduction and adduction pulled by the first shoulder joint module 20.

[0054] The three degrees of freedom contained in the upper limb exoskeleton are the core of realizing the basic motor functions of the upper limb. Shoulder abduction and adduction, pulled by the first shoulder joint module 20, refer to the movement of the arm in the coronal plane away from or towards the body midline. This direction of movement typically bears a large gravitational torque when the arm is raised. The swing of the upper arm 50 refers to the flexion and extension movement of the upper arm 50 in the sagittal plane, such as raising the arm forward or backward. The swing of the forearm 70 refers to the flexion and extension movement of the forearm 70 relative to the upper arm 50, such as bending and straightening the elbow joint. These three degrees of freedom together constitute the main motor capabilities of the upper limb in space and are the key motion axes that need to be considered in the design of the exoskeleton. The gravity balancing device 100 loads the gravity pulling the shoulder abduction and adduction of the first shoulder joint module 20, meaning that the design and optimization goal of the gravity balancing device 100 is specifically to compensate for the gravitational torque generated during shoulder abduction / adduction movements. Shoulder abduction and adduction movements are among the directions in which the upper limb bears the greatest gravitational force when raised above the horizontal plane. By precisely integrating and optimizing the structure and parameters (such as link length, hinge point position, elastic coefficient, and initial length of the elastic element) of the gravity balancing device 100 (e.g., multi-link mechanism and elastic element) into the mechanical structure of the first shoulder joint module 20, it can generate a compensating torque equal in magnitude and opposite in direction to the gravitational torque throughout the entire range of motion of shoulder abduction and adduction. For example, by adjusting the geometry of the multi-link mechanism and the installation position and preload of the elastic element, the tension or pressure of the elastic element during shoulder abduction can generate an upward lifting torque through the linkage mechanism to counteract the downward torque of the arm's own gravity.

[0055] Through the aforementioned technical solution, the main degrees of freedom of motion of the upper limb exoskeleton mechanism are clarified, and the compensation effect of the gravity balancing device 100 is focused on the shoulder abduction and adduction movements tractioned by the first shoulder joint module 20. Since shoulder abduction and adduction are among the primary directions of motion for the upper limb against gravity, this targeted compensation significantly reduces muscle fatigue and energy consumption during actions such as raising the arm and maintaining posture. Furthermore, this focused compensation strategy avoids unnecessary complex compensation designs in non-primary gravity directions, thus simplifying the mechanism and improving compensation efficiency and accuracy. Combined with algorithmic optimization of the gravity balancing device 100 parameters, adaptive and precise gravity compensation is ensured throughout the entire workspace of shoulder abduction and adduction, thereby greatly improving user comfort and the practicality of the exoskeleton.

[0056] In some embodiments described above in this application, an upper limb exoskeleton robot with gravity compensation is proposed. Its gravity balancing device 100 integrates a multi-link mechanism and elastic elements, and uses algorithms to optimize parameters to provide adaptive gravity compensation. However, in practical applications, improper structural design of the multi-link mechanism may lead to low torque transmission efficiency or difficulty in achieving accurate and stable gravity compensation throughout the workspace, thereby affecting the overall performance of the exoskeleton and the user experience.

[0057] In some embodiments described above in this application, a gravity balancing device 100 composed of a multi-link mechanism and an elastic element is proposed to provide adaptive gravity compensation. However, in practice, if the geometric relationship between the connection point of the elastic element and the main pivot axis is not properly designed, the lever arm of the elastic element may be too small, resulting in low compensation torque efficiency. This necessitates the use of an elastic element with extremely high stiffness to achieve the desired compensation effect, or it may be difficult to maintain stable compensation performance throughout the entire workspace, thus affecting the overall compensation efficiency and compactness of the gravity balancing device 100.

[0058] Please refer to the appendix for details. Figure 2 This application further proposes that the distance between the hinge point EF at both ends of the tension spring 130 and the hinge point A on the shoulder joint module is greater than the distance between the hinge point BC on the first link 110 and the second link 120 and the hinge point A on the shoulder joint module.

[0059] Specifically, hinge point A is the pivotal connection point between the first link 110 and the first shoulder joint module 20, and is the core fulcrum around which the entire gravity balancing device 100 generates compensating torque. Hinge points B and C are the connection points between the first link 110 and the second link 120, and between the second link 120 and the shoulder joint link 30, respectively. These two points are key nodes for force transmission within the multi-link mechanism, and their positions relative to hinge point A determine the geometry and force transmission characteristics of the linkage mechanism. Hinge points E and F are the connection points between the tension spring 130 and the first link 110 and the shoulder joint link 30, respectively. These two points are the application points where the tension spring 130 applies compensating force. The aforementioned distance relationship refers to the fact that the radial distance between the two connection points E and F of the tension spring 130 relative to the main pivot point A is designed to be greater. This design allows the tension spring 130 to provide a larger lever arm while generating the same tension, thereby generating a larger compensating torque.

[0060] Through the aforementioned geometric configuration, the effective lever arm of the tension spring 130 is significantly increased. This means that when the tension spring 130 generates the same tension, it can provide a larger compensating torque to the exoskeleton mechanism. This not only improves the compensation efficiency of the gravity balancing device 100, allowing for the selection of a tension spring 130 with lower stiffness to achieve the same compensation effect, thereby reducing the overall stiffness of the system and improving compliance; at the same time, the larger lever arm also helps to maintain the compensation performance more stably throughout the workspace, reducing the pressure on the spring parameter optimization algorithm, and contributing to a more compact and lightweight design, further optimizing the adaptive gravity compensation performance of the upper limb exoskeleton robot.

[0061] In some embodiments described above in this application, an upper limb exoskeleton robot with gravity compensation is proposed, wherein the parameters of its gravity balancing device 100 are optimized by an algorithm to provide adaptive gravity compensation. However, in practical applications, ensuring that this parameter optimization can accurately control the residual gravitational torque within an acceptable range throughout the entire workspace, thereby achieving efficient and smooth gravity compensation, is a technical challenge that needs to be addressed.

[0062] Please refer to the appendix for details. Figure 3 This application further proposes that the link lengths, hinge point positions, elastic coefficients, and initial lengths of the elastic elements in the four-bar linkage ABCD are configured using a parameter optimization algorithm, which includes: Establish the first shoulder joint gravitational moment with respect to the joint rotation angle gravitational torque function The elastic compensating torque function generated by the spring force transmitted through the four-bar linkage ),in The rotation angle of the target structure in the gravity balancing device. The parameters to be optimized include rod length, hinge point location, elastic coefficient of the elastic element, and initial length.

[0063] The objective function is constructed by minimizing the ratio of the maximum residual gravitational torque to the maximum gravitational torque; Based on the objective function, the particle swarm optimization algorithm is used to optimize the parameters. Global optimization is performed by iteratively updating the velocity and position of particles, tracking individual and global extrema, and finally converging to the optimal parameter combination P that minimizes the objective function T(P).

[0064] Specifically, when optimizing the parameters of the gravity balancing device, it is first necessary to establish the gravity torque function τ_g(θ) and the spring compensation torque function τ_s(θ,P). The gravity torque function τ_g(θ) is used to accurately quantify the gravitational influence on the upper limb exoskeleton mechanism at different joint angles θ. Its establishment is typically based on the exoskeleton's geometry, the mass distribution of each component, and the kinematic model of the joints, derived through mechanical analysis. The spring compensation torque function τ_s(θ,P) describes the compensation torque provided by the elastic element in the gravity balancing device under different joint angles θ and the parameters to be optimized P (such as the elastic coefficient, initial length of the elastic element, and the link length and hinge point position of the multi-link mechanism). This function is constructed by analyzing the mechanical properties of the elastic element and its geometric relationship with the multi-link mechanism to reflect its ability to counteract the gravity torque. Here, θ, as the independent variable, represents the motion state of the exoskeleton joints, while P is a set of key design variables whose values ​​directly affect the effect of gravity compensation.

[0065] Based on this, an objective function needs to be constructed to minimize the ratio of the maximum residual gravitational torque to the maximum gravitational torque. The residual gravitational torque refers to the difference between the gravitational torque and the spring compensation torque, reflecting the accuracy of gravity compensation. By minimizing the ratio of the maximum residual gravitational torque to the maximum gravitational torque, it can be ensured that the relative error of gravity compensation is controlled to a minimum throughout the entire workspace, thereby achieving uniform and efficient compensation performance, rather than simply focusing on the absolute value of the residual torque.

[0066] To minimize the aforementioned objective function, this application employs a particle swarm optimization (PSO) algorithm to globally optimize the parameter P. PSO is a heuristic optimization algorithm whose advantage lies in its ability to effectively handle complex, nonlinear, and multivariate optimization problems. By simulating bird flock foraging behavior through iterative searching, it avoids getting trapped in local optima and finds the globally optimal or near-optimal parameter combination. This global optimization strategy ensures that the fluctuation of the residual gravitational torque is less than a set proportion of the maximum gravitational torque throughout the entire rotation range of the shoulder joint link, thus meeting the expected compensation accuracy requirements.

[0067] The above technical solution enables precise quantification of the gravitational influence of the exoskeleton under different postures and the compensation capability of the gravity balance device. By constructing an optimization function with the objective function of minimizing the ratio of the maximum residual gravitational torque to the maximum gravitational torque, the optimization process focuses on improving the relative accuracy of compensation, ensuring uniform and efficient compensation throughout the entire workspace. Particle swarm optimization is employed for global optimization, effectively avoiding local optima and ensuring that the found parameter combination controls the fluctuation of the residual gravitational torque within a preset, relatively small proportion across the entire rotation range of the shoulder joint linkage. This significantly improves the adaptability and accuracy of gravity compensation, reduces the perceived weight and muscle fatigue experienced by the user when operating the exoskeleton, thereby enhancing the comfort and smoothness of human-computer interaction.

[0068] In some embodiments described above in this application, an algorithm is proposed to optimize the parameters of the gravity balancing device to provide adaptive gravity compensation within the workspace. However, without an accurate mathematical model of the gravitational torque and the spring compensation torque, as well as a quantified objective function, the algorithm optimization process may struggle to converge efficiently and achieve the best compensation effect, thereby impacting the performance and user experience of the exoskeleton robot.

[0069] In this regard, this application further proposes specific expressions for the gravitational torque function, the spring compensation torque function, and the objective function. The gravitational torque function is defined as τ_g(θ) = m g r The function cos(θ) is used to accurately quantify the gravitational load on the upper limb exoskeleton mechanism under different postures. Specifically, m represents the total mass of the exoskeleton arm and its load, g is the acceleration due to gravity, r is the distance from the center of mass of the exoskeleton arm to the hinge point A on the first shoulder joint module, and θ represents the rotation angle of the shoulder joint link. This function can accurately calculate the magnitude of the torque generated by gravity on the exoskeleton system at a specific joint angle, providing precise input for subsequent compensation design. Its implementation is typically based on the exoskeleton's CAD model and mass distribution data, obtained through mechanical analysis.

[0070] Meanwhile, the spring compensation torque is defined as τ_s(θ,P) = k, obtained by calculating the spring length L_s(θ,P) and lever arm r_s(θ,P) through geometric relationships. [L_s(θ,P) L0] `r_s(θ,P)` is a spring compensation torque function used to describe the compensation torque provided by an elastic element (e.g., a tension spring) in a gravity balancing device. Here, `k` is the elastic coefficient of the elastic element, `L_s(θ,P)` is the actual length of the spring under the current joint angle `θ` and optimized parameter `P`, `L0` is the free length of the spring, and `r_s(θ,P)` is the lever arm from the line of action of the spring force to the pivot center. The calculation of the spring length `L_s(θ,P)` and the lever arm `r_s(θ,P)` depends on the geometry of the multi-link mechanism of the gravity balancing device and is obtained in real time by establishing an accurate geometric model (such as the vector method or coordinate transformation method). This function can accurately reflect the compensation capability of the elastic element under different elongations and lever arms.

[0071] Based on this, this application further proposes an objective function as follows: The objective function, θ, ranges from -π / 2 to π / 2. It guides the optimization algorithm to find the optimal parameter P for best gravity compensation. The core idea is to minimize the ratio of the maximum value of the residual gravitational torque (the difference between the gravitational torque τ_g and the spring compensation torque τ_s) to the maximum gravitational torque τ_g over the entire working angle range [θmin, θmax]. By minimizing this ratio, the gravity balancing device can provide highly adaptive compensation throughout the workspace, effectively suppressing residual gravitational torque fluctuations and significantly reducing the workload of the exoskeleton motor. The θ angle range [-π / 2, π / 2] typically represents the main range of motion of the shoulder joint linkage in a specific plane, ensuring compensation performance within this critical working area.

[0072] Through the above technical solution, this application provides a precise mathematical basis and quantitative evaluation standard for the parameter optimization of the gravity balancing device. By clearly defining the gravitational torque function, the spring compensation torque function, and the objective function, the optimization algorithm can perform optimization based on an accurate physical model, avoiding optimization deviations caused by model inaccuracies. Specifically, the introduction of the gravitational torque function τ_g(θ) enables the system to accurately sense and quantify the gravitational influence of the exoskeleton arm and load under different postures; the establishment of the spring compensation torque function τ_s(θ,P) accurately describes the compensation capability of the elastic element, ensuring effective matching between the compensation torque and the gravitational torque. On this basis, by constructing an objective function T(P) with the goal of minimizing the ratio of the maximum residual gravitational torque to the maximum gravitational torque, and limiting the optimization within the critical shoulder joint linkage angle range [-π / 2, π / 2], it can be ensured that the gravity balancing device provides highly adaptive and uniform gravity compensation throughout the entire workspace. This not only significantly reduces the load on the drive motor and improves the energy efficiency and battery life of the exoskeleton, but also reduces fluctuations in residual gravitational torque, allowing users to experience a more natural and effortless movement when operating the exoskeleton, thereby enhancing the smoothness and comfort of human-computer interaction.

[0073] Although the aforementioned upper limb exoskeleton robot achieves adaptive gravity compensation through a gravity balancing device, effectively reducing the gravity load on the exoskeleton itself, in actual human-computer interaction, due to the complexity and dynamism of human movement, passive gravity compensation alone cannot completely eliminate dynamic interferences such as residual gravitational torque, inertial torque, and Coriolis torque. This may lead to inaccurate estimation of human-computer interaction torque, affecting the compliance of the exoskeleton and the comfort and naturalness of user operation.

[0074] In response, this application further proposes a compliant control method for an upper limb exoskeleton robot with gravity compensation, applied to the aforementioned upper limb exoskeleton robot. This control method includes the following steps: First, the current, speed, and position signals from the motor encoder of the shoulder joint module are acquired in real time at a fixed sampling period. The current and speed values ​​are then filtered to eliminate high-frequency noise, resulting in a filtered current signal, I_filtered. Specifically, the fixed sampling period refers to data acquisition within a preset, constant time interval (e.g., every millisecond or every few milliseconds) to ensure the control system can acquire continuous and synchronous sensor data. The motor encoder measures the precise position and speed of the joint, while the current sensor measures the current flowing through the motor windings. The acquired raw current and speed signals typically contain high-frequency noise, which may originate from the sensor itself, electromagnetic interference, or motor commutation. To improve the accuracy and stability of subsequent calculations, these signals need to be filtered. Common filtering methods include low-pass filters (such as Butterworth filters), Kalman filters, or moving average filters. These filters effectively remove high-frequency noise, resulting in a smooth and reliable filtered current signal I_filtered and speed signal.

[0075] Secondly, based on the Newton-Euler recursive dynamics algorithm, the gravitational torque G(q) and Coriolis torque are calculated in real time by inputting the joint position θ and velocity V. The compensation torque τ_s provided is calculated based on the current joint angle θ, combined with the optimized parameters of the passive gravity balancing device. The Newton-Euler recursive dynamics algorithm is a widely used method for robot dynamics calculations, capable of accurately calculating the torques acting on each joint based on the robot's geometry, mass distribution, joint positions, and velocities. Here, G(q) represents the gravitational torque generated by the exoskeleton's own mass and its load in the gravitational field. This represents the Coriolis torque and centrifugal torque generated by joint movement. Simultaneously, based on the pre-optimized parameters of the passive gravity balancing device (including the link lengths of the multi-link mechanism, hinge point positions, elastic coefficients of the elastic elements, and initial lengths), combined with the current joint angle θ, the compensating torque τ_s provided by the passive device can be accurately calculated. This step aims to comprehensively quantify all known internal torques experienced by the exoskeleton system in its current state, laying the foundation for subsequent accurate estimation of human-machine interaction torques.

[0076] Next, using the motor torque constant K_t, calculate the motor output torque τ_m=K_t. I_filtered; then through the dynamic model Solve for the human-computer interaction torque τ_h, where The torque is the inertial torque. The motor torque constant K_t is an inherent parameter of the motor, characterizing the linear relationship between the motor current and the output torque. By multiplying the filtered current signal I_filtered by K_t, the actual output torque τ_m of the motor can be accurately obtained. Subsequently, this motor output torque τ_m is substituted into the complete dynamic model of the exoskeleton. This dynamic model considers the inertial torque. (Caused by joint acceleration), Coriolis torque The equations are: gravitational torque G(q) and compensating torque τ_s provided by the passive gravity balancing device. By rearranging these equations, the unknown human-machine interaction torque τ_h can be calculated in real time. This interaction torque represents the active force applied by the user to the exoskeleton and is a key input for achieving compliant control.

[0077] Next, the interaction torque τ_h is input into the admittance controller, and the dynamic equation of the second-order mass-damped-spring system is obtained by simulating it. The trajectory adjustment θ_e and its differential term are solved using numerical integration methods. An admittance controller is a control strategy that converts force input into motion output, enabling the robot to exhibit desired compliance in response to external forces. Here, the admittance controller simulates a virtual second-order mass-damped-spring system, where M_d represents the virtual mass, B_d represents the virtual damping, and K_d represents the virtual stiffness. When the human-robot interaction torque τ_h is applied to this virtual system, the system generates a corresponding displacement θ_e (i.e., trajectory adjustment) along with its velocity and acceleration. The second-order differential equation can be solved in real-time using numerical integration methods (e.g., Euler's method, Runge-Kutta method), thus obtaining the trajectory adjustment θ_e and its differential term at the current moment. These parameters M_d, B_d, and K_d are adjustable admittance parameters used to adjust the compliance response characteristics of the exoskeleton to external interaction forces online.

[0078] Finally, the original desired trajectory θ_d is superimposed with the trajectory adjustment θe output by the admittance controller to generate the final joint position command θ_cmd = θ_d + θ_e. This command is then used to drive the motors via a PID position closed-loop controller to precisely track the command, while a gravity balancing device provides basic gravity compensation. The original desired trajectory θ_d is typically the ideal motion path set by task planning or user intent. The trajectory adjustment θ_e calculated by the admittance controller reflects the deviation the exoskeleton needs to make to respond to user interaction forces. Superimposing both yields the final joint position command θ_cmd, which comprehensively considers task requirements and user intent. θ_cmd is then received by a PID (proportional-integral-derivative) position closed-loop controller, which generates corresponding motor drive signals based on the error between the actual joint position and the command, ensuring the motors can accurately and stably track the command. Throughout this process, the passive gravity balancing device continuously provides basic gravity compensation, thereby reducing the burden on the active control system and allowing it to focus more on achieving compliance and dynamic response.

[0079] Through the above technical solution, this application can accurately estimate the human-machine interaction torque. Since the passive gravity balancing device pre-counters most of the gravity, the interaction torque calculated in the dynamic model has a higher signal-to-noise ratio, thus more accurately reflecting the user's true intentions. By inputting this high-precision interaction torque into the admittance controller, the exoskeleton can simulate the desired compliance characteristics, dynamically adjusting its movement trajectory according to the user's pushing and pulling forces, achieving natural and intuitive human-machine collaboration. Furthermore, the continuous effect of passive gravity compensation significantly reduces the load on the motor in resisting gravity, allowing the motor to focus more on performing tasks and responding to user commands, improving the system's energy efficiency and response speed, while also reducing the user's operational burden and enhancing comfort and safety.

[0080] In some of the above embodiments, a compliant control method for an upper limb exoskeleton robot with gravity compensation is proposed. This method calculates the human-robot interaction torque using a dynamic model and inputs it into an admittance controller to achieve compliant control. However, in practical applications, due to factors such as sensor noise, model uncertainty, and external interference, the interaction torque directly calculated from the dynamic model may suffer from a low signal-to-noise ratio, which can affect the admittance controller's accurate perception and response speed to user intentions. Furthermore, the real-time calculation accuracy and efficiency of the admittance controller, as well as how to effectively utilize the compensation effect of the passive gravity balancing device, are crucial for achieving natural, smooth, and low-load human-robot collaboration.

[0081] In response, this application further proposes the solution steps for the aforementioned admittance controller, specifically including: using the interaction torque τ_h estimated based on the current loop as the input to the admittance controller, wherein the signal-to-noise ratio of the interaction torque τ_h is improved due to passive gravity compensation, and can accurately reflect the user's intent; the admittance controller follows the second-order system equations. The trajectory adjustment amount θe is calculated in real time using Euler discretization, where M_d, B_d, and K_d are adjustable admittance parameters. The trajectory adjustment amount θe output by the admittance controller is used to correct the original desired trajectory θ_d, generating the final joint position command θ_cmd, enabling the exoskeleton to adapt to external interaction forces and achieve natural human-machine collaboration. The final joint position command θ_cmd is input into the PID controller, which generates motor drive signals through proportional, integral, and derivative operations to achieve high-precision position tracking. The passive gravity balancing device continuously counteracts gravity interference, reducing the motor load.

[0082] Specifically, the estimation of the interaction torque τ_h is a key input for compliant control. This application uses a current loop to accurately measure the motor current and, combined with the motor torque constant K_t, can estimate the motor output torque τ_m in real time and directly. Based on this, the human-machine interaction torque τ_h is calculated in reverse using a dynamic model. Because the passive gravity balancing device continuously cancels out most of the gravity interference, the torque required by the motor to compensate for gravity is greatly reduced, thereby decreasing the dynamic range of the motor output torque τ_m. Furthermore, when calculating the interaction torque τ_h, the gravity term G(q) has been largely canceled out, significantly increasing the ratio of the residual interaction torque τ_h signal to noise, i.e., improving the signal-to-noise ratio. This ensures that the interaction torque τ_h can more accurately reflect the user's true intentions and avoids misjudgments caused by noise or gravity interference.

[0083] The admittance controller responds to the input interactive torque τ_h by simulating a second-order mass-damped spring system, and its dynamic equation is: To achieve real-time control, the equation is discretized using the Euler method. Specifically, within each sampling period, the second derivative of the trajectory adjustment θe is calculated based on the current interaction torque τ_h, trajectory adjustment θe, and its differential term. Then, the trajectory adjustment θe and its differential term for the next moment are obtained through Euler integration. Here, M_d, B_d, and K_d are admittance parameters, representing virtual mass, virtual damping, and virtual stiffness, respectively. These parameters can be adjusted online according to actual application requirements. For example, increasing M_d can increase the exoskeleton's inertia to external forces, increasing B_d can increase the damping response to external forces, and increasing K_d can increase the stiffness response to external forces. This adjustability allows the exoskeleton to adapt to different task scenarios and user preferences, flexibly adjusting its compliant response characteristics.

[0084] The trajectory adjustment θe output by the admittance controller represents the amount by which the exoskeleton deviates from the original desired trajectory θ_d in response to user interaction forces. By superimposing the trajectory adjustment θe with the original desired trajectory θ_d, the final joint position command θ_cmd is generated. This correction mechanism allows the exoskeleton to actively adapt to external interaction forces, rather than passively resisting them, thus achieving more natural and fluid human-machine collaboration. Users can guide the exoskeleton's movement by applying slight forces, and the exoskeleton will adjust its trajectory accordingly, creating a harmonious "force-motion" feedback between the user and the machine.

[0085] The final generated joint position command θ_cmd is input into a PID (Proportional-Integral-Derivative) controller. The PID controller calculates the error between the actual joint position and the commanded position, and combines the proportional, integral, and derivative terms of the error to generate the corresponding motor drive signal. This ensures that the motor can accurately track the command trajectory corrected by the admittance controller. Simultaneously, because the passive gravity balancing device continuously and effectively counteracts the gravitational interference from the exoskeleton itself and the load, the motor does not need to output a large amount of additional torque to counteract gravity when performing position tracking tasks, thus significantly reducing the motor load. This not only improves the efficiency and response speed of the control system but also helps extend the motor's lifespan and reduce energy consumption.

[0086] Through the above technical solutions, this application significantly improves the accuracy of human-computer interaction and the robustness of the system based on existing compliant control methods. First, by utilizing the continuous cancellation of gravity interference by a passive gravity balancing device, the signal-to-noise ratio of the interaction torque τ_h estimated through the current loop is significantly improved, enabling more accurate and sensitive capture of the user's true intentions and avoiding misjudgments caused by gravity or noise. Second, the admittance controller uses the Euler method for real-time discretization calculation, ensuring the control system can quickly respond to external interaction forces. Furthermore, through online adjustable admittance parameters M_d, B_d, and K_d, the compliant response characteristics of the exoskeleton can be flexibly adjusted according to actual needs, adapting to different task scenarios and user preferences. In addition, by superimposing the trajectory adjustment amount θe output by the admittance controller with the original desired trajectory θ_d, the final joint position command θ_cmd is generated, enabling the exoskeleton to adapt to external interaction forces and achieving more natural and fluid human-computer collaboration. Ultimately, the PID controller tracks the corrected command trajectory with high precision, while the passive gravity balancing device continuously shares the gravity load. This not only reduces the workload and energy consumption of the motor, but also improves the overall efficiency and stability of the control system, providing users with a more comfortable, safe, and efficient auxiliary experience.

[0087] In some embodiments described above, a compliant control method for an upper limb exoskeleton robot with gravity compensation is proposed, which uses an admittance controller to calculate the human-machine interaction torque and adjust the trajectory. However, in its implementation, parameters such as virtual mass M_d, virtual damping B_d, and virtual stiffness K_d in the admittance controller are usually preset to fixed values. This fixed parameter setting may cause the exoskeleton's compliant response characteristics to fail to achieve optimal matching when facing different users (e.g., rehabilitation patients with different physical fitness levels) or different task requirements (e.g., assisted training and active rehabilitation), thereby affecting the comfort and naturalness of the user experience and the efficiency of human-machine collaboration.

[0088] In this regard, this application further proposes that, in the admittance control calculation step, the admittance parameters M_d, B_d, and K_d can be adjusted online to change the compliance of the exoskeleton with external interaction forces.

[0089] Specifically, the admittance parameters M_d, B_d, and K_d represent the virtual mass, virtual damping, and virtual stiffness in the second-order mass-damped-spring system simulated by the admittance controller, respectively. The virtual mass M_d determines the exoskeleton's inertial response to changes in external torque; a larger value results in a less responsive exoskeleton to rapid torque changes. The virtual damping B_d determines the exoskeleton's damping effect to changes in external torque; a larger value results in greater resistance to motion velocity. The virtual stiffness K_d determines the exoskeleton's displacement resistance to changes in external torque; a larger value results in smaller displacement when subjected to the same torque. These parameters collectively determine the exoskeleton's compliance characteristics during human-robot interaction. "Online adjustment" refers to the ability to modify or update these admittance parameters in real time during robot operation. This can be achieved in several ways. For example, operators can manually input and adjust parameter values ​​via a graphical user interface (GUI) or physical knobs; the system can automatically load different parameter sets based on preset task modes or user profiles; or, more advancedly, adaptive algorithms can be used to dynamically optimize and adjust these parameters based on real-time collected user physiological signals (such as electromyography signals and heart rate) or interaction torque data to adapt to user intentions and environmental changes. By adjusting these parameters, the compliance of the exoskeleton can be directly controlled. For example, decreasing M_d, B_d, and K_d can make the exoskeleton "softer," more responsive to user operations, and easier for the user to pull; while increasing these parameters will make the exoskeleton "harder," providing greater support or resistance, suitable for scenarios requiring stability and precise control.

[0090] Through the above technical solution, this application can significantly improve the adaptability and personalization of upper limb exoskeleton robots in human-computer interaction. When user or task requirements change, the exoskeleton's compliant response characteristics can be adjusted in real time without stopping its operation. For example, in the early stages of rehabilitation training, lower virtual mass, damping, and stiffness can be set to make the exoskeleton more compliant, reducing obstacles to the patient's active movement, thereby reducing fatigue and encouraging active participation. In the later stages of rehabilitation or during strength training, these parameters can be appropriately increased to enhance the exoskeleton's resistance and provide more effective training results. This dynamic adjustment capability allows the exoskeleton to better adapt to different users' physical conditions, rehabilitation stages, and diverse task scenarios, thereby achieving a more natural, efficient, and comfortable human-computer collaboration experience and effectively improving the quality of rehabilitation training or assisted tasks.

[0091] Please see the appendix Figure 4 This application also proposes a gravity-compensated compliant control system for an upper limb exoskeleton robot, applied to the aforementioned upper limb exoskeleton robot, and the aforementioned compliant control method for the upper limb exoskeleton robot. The control system includes: a signal filtering module, a calculation module, an interactive torque module, an admittance control module, and an instruction synthesis module. The signal filtering module is used to collect and filter encoder signals of current, speed, and position of the robot's joint motors in real time; the calculation module calculates the real-time dynamic torque, including gravitational torque, Coriolis torque, and inertial torque, based on the Newton-Euler algorithm, and integrates the passive compensation torque of the gravity balancing device; the interaction torque module calculates the motor output torque through the motor torque constant and the filtered current, and outputs a pure interaction torque by combining the real-time dynamic torque and the passive compensation torque; the admittance control module converts the pure interaction torque into a trajectory adjustment amount by simulating a second-order mass-damped-spring system; the instruction synthesis module superimposes the desired trajectory with the trajectory adjustment amount, and drives the joint motors to track the position through a PID controller.

[0092] Specifically, attached Figure 4 The data flow and module interactions of the control system are clearly demonstrated, with its core being a layered architecture based on passive compensation and utilizing active control. The implementation process can run with a fixed 1ms sampling period, and the specific steps are as follows: For the signal filtering module, see attached Figure 4 As shown on the left, the control system first acquires real-time current (I), velocity (V), and position (θ) signals from the joint motor via an encoder. The acquired signals undergo moving average filtering and second-order low-pass filtering to obtain the filtered current value (I_filtered) and joint state data. Simultaneously, the desired position (θ_d) is obtained from the trajectory generation module. This module ensures the purity of the input signal, providing a foundation for subsequent calculations and ensuring proper alignment. Figure 4The "encoder signal acquisition" stage.

[0093] For the calculation module, its corresponding appendix Figure 4 The "Real-time Dynamics Calculation" section describes the control system based on the Newton-Euler recursive dynamics algorithm. Inputs include joint position θ and velocity V (assuming acceleration). ≈0), real-time calculation of gravitational torque G(q) and Coriolis torque and inertial torque Simultaneously, combining the optimized parameters of the passive gravity compensation device (pre-optimized using the PSO algorithm), the compensation torque τ_s is calculated based on the current angle θ. This step integrates the physical compensation effect of the passive device, significantly reducing the burden on active control, such as... Figure 4 As shown in the "Compensation Torque Calculation" section.

[0094] For the interactive torque module, see attached... Figure 4 As shown in the middle section, this module calculates the motor output torque using the motor torque constant K_t and the filtered current I_filtered. Subsequently, based on the dynamic model The pure human-computer interaction torque τ_h is calculated. This process eliminates gravity interference, improves torque estimation accuracy, and closely matches the surrounding environment. Figure 4 The label "Human-computer interaction torque estimation".

[0095] For the admittance control module, this module corresponds to the attached... Figure 4 The "Admittance Compliance Control" section. The estimated interaction torque τ_h is input into the admittance controller, and its dynamic equation is: (M_d, B_d, and K_d are adjustable parameters). The system discretizes the equation using the Euler method and solves for the trajectory adjustment θ_e and its differential terms. By adjusting the parameters, the compliance of the exoskeleton can be customized online, such as... Figure 4 The data stream shown.

[0096] For the instruction synthesis module, see attached Figure 4 As shown on the right, this module modifies the desired trajectory θ_d with the trajectory adjustment amount of the admittance control output. Superimpose to generate the final position command The command is input to the PID position controller, based on the error ( -θ) generates a drive signal to accurately track the position. The entire process relies on a passive gravity compensation device to reduce the motor load and maintain close contact with the surrounding environment. Figure 4 The "command synthesis" and "position control" stages.

[0097] In this embodiment, the control system is via an attachment Figure 4The modular design enables highly efficient collaboration. The passive compensation of the gravity balance device bears most of the gravity load (keeping the residual torque <3%), while the active system focuses on dynamic response, reducing interaction force errors and energy consumption during rehabilitation training. The modular structure also improves system reliability and eliminates the need for additional force sensors.

[0098] The following example will provide a more detailed explanation of the above technical solution: In a rehabilitation training scenario, user A needs to use an upper limb exoskeleton robot for shoulder abduction and adduction rehabilitation training. Traditional upper limb exoskeleton robots require motors to continuously output a large amount of torque to offset the large gravitational torque borne by the shoulder joint, resulting in high motor energy consumption, slow response speed, and poor gravity compensation effect under different postures. This makes user A feel unnatural during training, and may even cause discomfort due to insufficient or excessive compensation.

[0099] This invention provides an upper limb exoskeleton robot with gravity compensation, which effectively solves the above-mentioned problems by integrating a gravity balancing device 100 and combining it with an optimized control method.

[0100] The upper limb exoskeleton robot includes an upper limb exoskeleton mechanism, which is sequentially connected to a backplate 10, a first shoulder joint module 20, a shoulder joint link 30, a second shoulder joint module 40, an upper arm 50, an elbow joint module 60, and a forearm 70. The first shoulder joint module 20, the second shoulder joint module 40, and the elbow joint module 60 all provide pivoting functionality, enabling the exoskeleton mechanism to achieve three degrees of freedom: shoulder abduction and adduction, upper arm 50 swing, and forearm 70 swing.

[0101] To address the issue of excessive gravitational torque at the shoulder joint, a gravity balancing device 100 is integrated into the first shoulder joint module 20. This gravity balancing device 100 specifically bears the weight of the shoulder joint module 20 during abduction and adduction. Specifically, the gravity balancing device 100 includes a multi-link mechanism and an elastic element. The multi-link mechanism is a four-bar linkage ABCD, consisting of a first link 110, a second link 120, and a shoulder joint link 30. The two ends of the first link 110 form hinge points A with the first shoulder joint module 20 and B with the second link 120, respectively. The other end of the second link 120 forms hinge point C with the shoulder joint link 30. The shoulder joint link 30 forms hinge point D with the pivot end of the first shoulder joint module 20. The elastic element is a tension spring 130, with its two ends forming hinge points E with the first link 110 and F with the shoulder joint link 30, respectively. The distance between the hinge points EF at both ends of the tension spring 130 and the hinge point A on the shoulder joint module is greater than the distance between the hinge points BC on the first link 110 and the second link 120 and the hinge point A on the shoulder joint module. This geometric configuration helps to effectively transmit the spring torque.

[0102] Unlike traditional gravity balancing devices 100 designed based on experience, this invention optimizes the parameters of the gravity balancing device 100 using algorithms to provide adaptive gravity compensation within the workspace of the first shoulder joint module 20. These parameters include the link lengths of the multi-link mechanism, the hinge point positions, the elastic coefficient of the elastic element, and its initial length. The optimization process is as follows: First, establish the gravitational torque function τ_g(θ) and the spring compensation torque function τ_s(θ,P). Here, θ is the rotation angle of shoulder joint link 4, and P is the parameter to be optimized. Gravitational torque function τ_g(θ) = m g r cos(θ), where m is the mass of the entire arm and load of the exoskeleton, and r is the distance from the center of mass to the hinge point A on the first shoulder joint module 3. The spring compensation torque τ_s(θ,P) = k [L_s(θ,P) L0] r_s(θ,P) calculates the spring length L_s(θ,P) and lever arm r_s(θ,P) through geometric relationships, where L0 is the free length of the spring.

[0103] Secondly, construct the objective function. To minimize the ratio of the maximum residual gravitational torque to the maximum gravitational torque, where the angle θ ranges from [-π / 2, π / 2] and the sampling interval is π / 1000.

[0104] Finally, a particle swarm optimization algorithm is used to globally optimize the parameter P. This algorithm iteratively updates the velocity and position of particles (parameter combinations), tracks individual and global extrema, and eventually converges to the optimal parameter combination P that minimizes the objective function T(P). For example, the variable dimension is set to 11 (the number of parameters P), the particle swarm size is 300, and the maximum number of iterations is 1000. Appropriate upper and lower bounds for parameter P are then set based on the actual situation of parameter P and the mechanism space, making it easier and faster to obtain the most suitable parameter P. Through further optimization, the algorithm outputs the optimal parameter combination P that minimizes T(P). After calculations using the mechanical model before and after compensation, with parameter P, the residual gravitational torque fluctuation range within the motion range is less than 3% of the maximum gravitational torque.

[0105] This global optimization method ensures that within the rotation range of the shoulder joint link, the residual gravitational torque fluctuation is less than a set proportion of the maximum gravitational torque, thereby achieving efficient and stable gravity compensation throughout the entire workspace. Compared to existing technologies that rely on manual experience or local linearization methods, this invention overcomes the limitations of traditional design methods that can only achieve balance at a single point or in a small range, fundamentally reducing the power requirements and operating energy consumption of the drive system.

[0106] In terms of control, based on the hybrid gravity-compensated compliant control method, this application constructs the following... Figure 4 The control system architecture shown has an outer loop of admittance control and an inner loop of position PID control.

[0107] The principle of the control method is as follows Figure 5 As shown. The entire system, with a real-time control system at its core, implements a hierarchical control strategy for the first shoulder joint (integrated with an optimized passive gravity balance device), employing "passive compensation as the basis and active regulation as the application." The control system operates with a fixed sampling period of 1ms, executing the following process in each cycle: 1. Encoder signal acquisition and processing: Read the joint encoder current value I, speed V, and position. The collected current and velocity values ​​are filtered to eliminate high-frequency noise. The filtering algorithm uses a combination of moving average filtering and second-order low-pass filtering. After two stages of filtering, I_filtered is obtained, and then the current desired position is obtained from the pre-designed trajectory generation module. .

[0108] 2. Real-time Dynamics and Compensation Torque Calculation: Due to the requirements of the overall mechanism control algorithm, a more precise dynamic analysis was performed on the entire mechanism. The gravity torque model designed with the gravity balance device was discarded. The Newton-Euler recursive dynamics algorithm was then invoked, and the input... and V (assuming) Therefore, the moment of inertia ), calculate the real-time gravitational torque G(q) and Coriolis torque Based on the optimized parameters of the designed passive gravity balancing device, combined with the force analysis mathematical model of the previous compensation device, and according to the current... Real-time calculation of the compensation torque it provides .

[0109] 3. Real-time torque estimation for human-machine interaction: Calculate the motor output torque: Based on the dynamic torque model Solve for the pure interaction torque: .

[0110] 4. Admittance compliance control solution: The estimated human-machine interaction torque is calculated... This data is fed into the admittance controller as input. The admittance controller simulates a second-order mass-damped spring system, whose dynamic equations are as follows: The system first directly calculates using this equation. Then, the equation is discretized using the real-time Euler method and solved. The desired trajectory adjustment caused by the interaction force is output through stepwise integration. and its differential terms By adjusting the admittance parameters M_d, B_d, and K_d, the "compliant" response characteristics of the exoskeleton joint to external forces can be altered online.

[0111] 5. Command Synthesis and High-Precision Position Tracking: The system will synthesize the original desired motion trajectory. Trajectory adjustment amount output by admittance controller The commands are superimposed to generate the final joint position commands. The instruction is input into a high-performance PID position closed-loop controller, which calculates... With actual feedback location The error between the parameters is calculated and then processed using proportional, integral, and differential operations to generate a motor drive signal, thereby driving the articulated motor to accurately and quickly track the synthesized command. Throughout the process, the passive gravity balancing device consistently provides a basic compensating force opposite to the gravitational torque at the physical level, allowing the motor to primarily provide the dynamic torque needed to overcome inertia, damping, and respond to interactive forces, significantly reducing the load on the driver and system energy consumption.

[0112] This upper limb exoskeleton robot employs a hybrid gravity-compensated compliant control method. This method acquires the current, velocity, and position signals from the motor encoder of the shoulder joint module in real time at a fixed sampling period. The current and velocity values ​​are then filtered to eliminate high-frequency noise, resulting in a filtered current signal I_filtered.

[0113] Based on the Newton-Euler recursive dynamics algorithm, the gravitational torque G(q) and Coriolis torque are calculated in real time by inputting the joint position θ and velocity V. Simultaneously, based on the optimized parameters of the passive gravity balancing device 2, the compensation torque τ_s it provides is calculated according to the current joint angle θ.

[0114] Calculate the motor output torque τ_m=K_t using the motor torque constant K_t. I_filtered. Then, through the dynamic model... Solve for the human-computer interaction torque τ_h, where The torque is the inertial torque. Since the passive gravity balancing device 2 has already counteracted most of the gravitational interference, the signal-to-noise ratio of the interaction torque τ_h estimated based on the current loop is significantly improved, enabling it to more accurately reflect the true intention of user A. This contrasts with existing technologies that do not consider real-time fusion of mechanical balance torques and rely solely on current loop estimation for torque feedback adjustment, resulting in large contact force errors.

[0115] The interaction torque τ_h is input into the admittance controller. The admittance controller simulates a second-order mass-damped spring system, whose dynamic equation is: The trajectory adjustment θ_e and its differential term are solved using numerical integration. The admittance parameters M_d, B_d, and K_d can be adjusted online to change the compliance of the exoskeleton with external interaction forces.

[0116] The original desired trajectory θ_d is superimposed with the trajectory adjustment θ_e output by the admittance controller to generate the final joint position command θ_cmd = θ_d + θ_e. The motor is then driven by a PID position closed-loop controller to precisely track the command. During this process, the gravity balancing device 2 continuously provides basic gravity compensation, reducing the motor load so that the motor primarily needs to provide the dynamic torque required to overcome inertia, damping, and respond to interaction forces. The synergistic advantages of this hybrid architecture not only save energy but also reduce the peak torque requirements of the actuator, allowing for the use of smaller, lighter motors, which is beneficial for system miniaturization and weight reduction.

[0117] Through the above-described solution, User A experiences a more natural and fluid assistance when using the upper limb exoskeleton robot for rehabilitation training. The gravity balance device effectively counteracts the arm's weight throughout the entire range of motion, significantly reducing motor load and system energy consumption. Simultaneously, the optimized control method enables more precise human-machine interaction torque perception, and the admittance controller can quickly respond to User A's intentions, achieving high-precision, fast-response, compliant human-machine interaction control, thus improving training effectiveness and wearing comfort.

[0118] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A gravity-compensated upper limb exoskeleton robot, characterized in that, include: An upper limb exoskeleton mechanism, the exoskeleton mechanism including a back plate and a shoulder joint link, the shoulder joint link and the back plate being rotatably connected by a first shoulder joint module, the first shoulder joint module being configured to control the swing of the shoulder joint link; A gravity balancing device is configured to provide adaptive gravity compensation within the workspace of the upper limb exoskeleton mechanism, the gravity balancing device including a first link, a second link, and a tension spring; The two ends of the first connecting rod are respectively hinged to the first shoulder joint module to form hinge point A, and to one end of the second connecting rod to form hinge point B; The other end of the second link forms a hinge point C with the shoulder joint link; The hinge center of the first shoulder joint module is D, so as to form a four-bar linkage ABCD; The two ends of the tension spring form hinge points E with the first connecting rod and hinge points F with the shoulder joint connecting rod, respectively.

2. The upper limb exoskeleton robot according to claim 1, characterized in that, The upper limb exoskeleton mechanism also includes a second shoulder joint module, an upper arm, an elbow joint module, and a forearm connected in sequence, and the first shoulder joint module, the second shoulder joint module, and the elbow joint module provide pivoting for the shoulder joint link, the upper arm, and the forearm, respectively; the gravity balancing device is integrated in the first shoulder joint module to provide adaptive gravity compensation within the workspace of the first shoulder joint module.

3. The upper limb exoskeleton robot according to claim 2, characterized in that, The upper limb exoskeleton mechanism includes three degrees of freedom: shoulder abduction and adduction pulled by the first shoulder joint module, swinging of the upper arm, and swinging of the forearm. The gravity balancing device loads the gravity pulled by the first shoulder joint module for shoulder abduction and adduction.

4. The upper limb exoskeleton robot according to claim 1, characterized in that, The distance between the hinge point EF at both ends of the tension spring and the hinge point A on the shoulder joint module is greater than the distance between the hinge point BC on the first link and the second link and the hinge point A on the shoulder joint module.

5. The upper limb exoskeleton robot according to claim 1, characterized in that, The link lengths, hinge point positions, elastic coefficients, and initial lengths of the elastic elements in the four-bar linkage ABCD are configured using a parameter optimization algorithm, which includes: Establish the first shoulder joint gravitational moment with respect to the joint rotation angle gravitational torque function The elastic compensating torque function generated by the spring force transmitted through the four-bar linkage ),in The rotation angle of the target structure in the gravity balancing device. The parameters to be optimized include rod length, hinge point location, elastic coefficient of elastic element, and initial length; The objective function is constructed by minimizing the ratio of the maximum residual gravitational torque to the maximum gravitational torque; Based on the objective function, the particle swarm optimization algorithm is used to optimize the parameters. Global optimization is performed by iteratively updating the velocity and position of particles, tracking individual and global extrema, and finally converging to the optimal parameter combination P that minimizes the objective function T(P).

6. The upper limb exoskeleton robot according to claim 5, characterized in that, The gravitational torque function is: , where m is the mass of the entire arm and load of the exoskeleton, and r is the distance from the center of mass to the hinge point A on the first shoulder joint module; The spring compensation torque is calculated using geometric relationships to determine the spring length. lever arm ,get , where L0 is the free length of the spring; The objective function is: ,in, The angle range is [-π / 2, π / 2].

7. A compliant control method for an upper limb exoskeleton robot with gravity compensation, characterized in that, The control method, applied to the upper limb exoskeleton robot according to any one of claims 1-6, comprises: The current, speed, and position signals of the motor encoder of the shoulder joint module are acquired in real time with a fixed sampling period. The current and speed values ​​are filtered to eliminate high-frequency noise and obtain the filtered current signal I_filtered. Based on the Newton-Euler recursive dynamics algorithm, the gravitational torque G(q) and Coriolis torque are calculated in real time by inputting the joint position θ and velocity V. In conjunction with the optimized parameters of the passive gravity balancing device, the compensation torque provided by it is calculated based on the current joint angle θ. ; Utilizing the motor torque constant Calculate the motor output torque Then through the dynamic model Calculate the torque of human-computer interaction ,in It is the moment of inertia; Interaction torque The input admittance controller, through simulation of a second-order mass-damped-spring system, has the following dynamic equation: The trajectory adjustment θ_e and its differential term are solved by numerical integration. Original expected trajectory Trajectory adjustment amount output by admittance controller Superimpose to generate the final joint position command The motor is driven by a PID position closed-loop controller to accurately track commands, while a gravity balance device provides basic gravity compensation.

8. The compliant control method for an upper limb exoskeleton robot according to claim 7, characterized in that, The calculation steps for the admittance controller include: The interactive torque τ_h based on the current loop estimation is used as the input of the admittance controller, where the signal-to-noise ratio of τ_h is improved due to passive gravity compensation, which can accurately reflect the user's intention. The admittance controller follows the second-order system equations Real-time solution is performed using Euler discretization, and the trajectory adjustment is output. Where M_d, B_d, and K_d are adjustable admittance parameters; The output of the admittance controller Used to correct the original expected trajectory θ_d and generate θ_cmd, enabling the exoskeleton to adapt to external interactive forces and achieve natural human-machine collaboration; The θ_cmd is input into the PID controller, which generates a motor drive signal through proportional, integral, and derivative operations to achieve high-precision position tracking. The passive gravity balancing device continuously counteracts gravity interference, reducing the motor load.

9. The compliant control method for an upper limb exoskeleton robot according to claim 8, characterized in that, In the admittance control calculation step, the admittance parameters M_d, B_d, and K_d can be adjusted online to change the compliance of the exoskeleton with external interaction forces.

10. A compliant control system for an upper limb exoskeleton robot with gravity compensation, characterized in that, An upper limb exoskeleton robot according to any one of claims 1-6, and an upper limb exoskeleton robot compliant control method according to any one of claims 7-9, wherein the control system comprises: The signal filtering module is used to collect and filter encoder signals of current, speed and position of robot joint motors in real time. The calculation module calculates real-time dynamic torques, including gravitational torque, Coriolis torque, and inertial torque, based on the Newton-Euler algorithm, and integrates the passive compensation torque of the gravity balance device. The interactive torque module calculates the motor output torque using the motor torque constant and the filtered current, and outputs a pure interactive torque by combining the real-time dynamic torque and the passive compensation torque. The admittance control module converts the pure interactive torque into a trajectory adjustment amount by simulating a second-order mass-damped-spring system; and The instruction synthesis module superimposes the desired trajectory with the trajectory adjustment amount and drives the joint motor to track the position through a PID controller.

Images