Adaptive multi-terrain lower limb exoskeleton assistance control method and device

Abstract

This application provides an adaptive multi-terrain lower limb exoskeleton assistive control method and device. The device includes a backplate, a drive structure, a lower limb structure, sensors, a hip and knee joint transmission system, and a control system. The backplate provides stable support and reduces the device's weight; the lower limb structure enables a natural gait; sensors provide data for precise assistance through real-time monitoring; the hip and knee joint transmission system achieves dynamic torque distribution through pulleys and cams, improving assistance efficiency; the control system includes an electrically connected structure-adaptive allocation layer and a predictive adjustment layer. The predictive adjustment layer incorporates an offline-trained temporal convolutional network (TCN) torque prediction model, and the structure-adaptive allocation layer is electrically connected to the hip and knee joint transmission system. This application enables the lower limb exoskeleton assistive device to better adapt to the human body's natural gait, achieve dynamic torque distribution at the hip and knee joints, improve assistance efficiency and comfort, and enhance adaptability and practicality in complex sports scenarios.

Description

Technical Field

[0001] The invention relates to the field of exoskeleton assistance technology, and in particular to an adaptive multi-terrain lower limb exoskeleton assistance control method and device. Background Technology

[0002] In the current technological landscape, the deep integration of bionics and intelligent control technology has provided new opportunities for the research and development of lower limb assistive exoskeletons. With the increasing aging population, growing military personnel's load-bearing needs, and the urgent demand for physical assistance in industrial handling scenarios, lower limb exoskeleton technology has gradually become a research hotspot in rehabilitation medicine, military equipment, and industrial automation. Traditional lower limb exoskeleton systems mostly employ rigid linkages and single-joint drive modes. Their torque distribution mechanisms are static and difficult to adapt to the dynamic coordination needs of the hip and knee joints during the human gait cycle, resulting in low assist efficiency and poor wearing comfort. This is especially problematic in complex terrains (such as uphill or stair climbing), where movement lag and energy waste are common. Furthermore, existing systems largely rely on pre-programmed control strategies, lacking real-time perception and adaptive adjustment capabilities to human movement intentions, limiting their practicality and generalization capabilities in dynamic environments.

[0003] In recent years, advancements in biomimetic drive technology and artificial intelligence algorithms have provided new approaches to solving these problems. Biomimetic tendon-driven structures, by simulating the cross-joint coordination mechanism of human muscles, can significantly improve the natural motion adaptability of exoskeletons; the introduction of deep learning models (such as temporal convolutional networks TCN) provides data-driven solutions for real-time motion prediction and dynamic torque allocation. However, existing research still faces several technical challenges: First, existing bionic dual-joint drive structures struggle to balance mechanical efficiency and lightweight design, and cannot achieve continuous adaptive distribution of hip and knee joint torques through structural design, requiring frequent algorithmic adjustments and resulting in high response latency. Second, existing torque prediction schemes based on temporal convolutional networks (TCNs) do not customize network topologies for the characteristics of gait temporal data, have insufficient parameter disclosure, and their generalization ability and real-time performance are insufficient to meet the closed-loop control requirements of exoskeletons. Third, torque distribution often uses piecewise function mapping, which can easily lead to torque jumps at critical angle points, resulting in unstable assistance and even human-machine motion resistance. Fourth, existing schemes often rely on external terrain recognition sensors and off-track gait mode switching, which can easily lead to mode misjudgment in complex terrains, resulting in poor adaptability. Furthermore, they lack complete human-machine coupled dynamics simulation verification and quantitative experimental data to support the assistance effect, making the schemes insufficiently feasible and reliable.

[0004] Therefore, we propose an adaptive multi-terrain lower limb exoskeleton assistive control method and device. Summary of the Invention

[0005] Therefore, it is necessary to address the shortcomings of the existing production technologies by providing an adaptive multi-terrain lower limb exoskeleton assistive control method and device. This would enable the lower limb exoskeleton assistive device to better adapt to the natural gait of the human body, achieve dynamic distribution of hip and knee joint torque, improve assistive efficiency and comfort, and enhance adaptability and practicality in complex sports scenarios.

[0006] The first aspect of this application provides an adaptive multi-terrain lower limb exoskeleton assistive device, comprising: Backplate, which is fixed to the wearer's back; A drive structure is disposed on the back plate; The lower limb structure includes a hip joint strut link, a hip joint thigh link, a thigh link, and a lower leg link that are rotatably connected from top to bottom in the height direction; A sensor, mounted on the lower limb structure, is used to monitor spatial positional changes and kinematic time-series data of the lower limb structure. The hip and knee joint transmission system includes a traction line extending from a self-driving structure, a fixed pulley component mounted on a thigh link, and a movable pulley component. The movable end of the traction line passes around the movable pulley component and is connected to the thigh link. The hip and knee joint transmission system also includes a cam mechanism mounted at the connection between the thigh link and the lower leg link. The movable pulley component is connected to the cam mechanism via a transmission line. The control system includes an electrically connected structure adaptive allocation layer and a prediction adjustment layer. The prediction adjustment layer incorporates a temporal convolutional network (TCN) torque prediction model trained offline. The structure adaptive allocation layer is electrically connected to the hip and knee joint transmission system.

[0007] In other embodiments, the driving structure is a servo motor, used to pull the connecting line and output the total driving torque. The total output torque of the motor satisfies the formula:

[0008] in, Provide torque to the hip joint. Provides assist torque to the knee joint.

[0009] In other embodiments, the fixed pulley component includes a fixed frame and a fixed pulley structure. The fixed pulley structure is mounted on the fixed frame, which is fixedly connected to the upper end of the thigh linkage near the hip joint. The connecting line is a Bowden line, one end of which is fixedly connected to the output end of the drive structure, and the other end passes sequentially around the fixed pulley structure and the movable pulley component before being fixedly connected to the fixed frame. The movable pulley component includes a slide rail mounted on the thigh linkage, a slider slidably mounted on the slide rail, and a movable pulley structure mounted on the slider. One end of the transmission line is rigidly connected to the slider, and the other end is connected to the cam mechanism via a guide pulley on the thigh linkage.

[0010] In other embodiments, the profile curve of the cam mechanism is directly fitted by a nonlinear mapping function from the knee joint angle to the target torque ratio of the hip and knee joint, the mapping function being implemented using a 10th-order high-order polynomial.

[0011] In other embodiments, the TCN torque prediction model of the prediction adjustment layer takes standardized lower limb kinematic time-series data as input and outputs the total assist torque command required for the current gait; the structural adaptive allocation layer is used to continuously allocate the total assist torque to the hip and knee joints according to the real-time knee joint angle and the geometric mapping relationship of the transmission system.

[0012] The second aspect of this application provides an adaptive multi-terrain lower limb exoskeleton assistive control method for controlling the aforementioned adaptive multi-terrain lower limb exoskeleton assistive device, comprising the following steps: Real-time acquisition of multimodal kinematic state sequences is achieved by using inertial measurement units (IMUs) deployed in the thigh and calf segments of the lower limb structure to collect real-time kinematic temporal data of the wearer's lower limb joints and then standardizing the acquired knee joint angle sequences. Based on the online generation of total driving torque of the TCN prediction adjustment layer, the standardized kinematic time series data is used as input and passed to the offline trained temporal convolutional network TCN to predict and output the total assist torque command required for the current gait in real time. Based on the continuous torque distribution of high-order polynomial mapping, after obtaining the total assist torque, the total torque is passively distributed to the hip and knee joints according to the real-time knee joint angle and the inherent geometric mapping relationship of the exoskeleton transmission system. The torque distribution ratio is calculated by the 10th-order high-order polynomial nonlinear mapping function of the knee joint angle to the target hip and knee joint torque ratio. Dynamic closed-loop control: The control system drives the motor to output adaptive torque according to the allocated hip and knee joint torque command, and completes the coordinated assistance of the hip and knee joints through the transmission system to achieve multi-terrain gait adaptive response.

[0013] In other embodiments, the TCN network comprises three one-dimensional convolutional layers, wherein the dilation factor of convolutional layer 1 is [1,1], the dilation factor of convolutional layer 2 is [2,1], and the dilation factor of convolutional layer 3 is [4,1]. The kernel size of each convolutional layer is uniformly set to [5×3]. ReLU is used as the activation function between network layers, and a Dropout layer with a ratio of 10% is configured. The end is connected to a global max pooling layer and a fully connected layer for regression output.

[0014] In other embodiments, the TCN network is trained using the Adam optimizer with a learning rate of 0.001 and a batch size of 64, and an early stopping strategy is introduced to prevent model overfitting; the training dataset uses a common gait kinematics and dynamics dataset that includes flat ground, slopes and stairs.

[0015] In other embodiments, the expression for the 10th-order higher-order polynomial nonlinear mapping function is:

[0016] in, These are the polynomial coefficients obtained by fitting experimental data using the least squares method; The knee joint angle is standardized with a mean of 35.78 and a standard deviation of 19.87. The target hip-knee joint torque ratio.

[0017] In other embodiments, the hip and knee joint torque distribution satisfies the following mechanical formula: Hip joint assist torque:

[0018] Knee joint assist torque:

[0019] Hip-knee joint torque ratio:

[0020] in, The tension of the connecting wire at the motor output end; The effective arm of the hip joint is constant. Let be the equivalent radius of the knee joint cam at the current angle. This is the real-time hip joint angle.

[0021] The beneficial effects of this application are as follows: This application proposes a two-layer control architecture that coordinates a structural adaptive allocation layer and a predictive adjustment layer, enabling adaptive assistance for multi-terrain gait without the need for external terrain recognition sensors or discrete mode switching. The structural layer is responsible for the continuous coordination of cross-joint moments, while the predictive layer is responsible for adjusting the intensity of the total moment. The two are fused through a unified control loop, fundamentally solving the pain points of existing technologies that rely on terrain recognition and mode switching, resulting in response delays and discontinuous assistance.

[0022] In addition, this application also has the following advantages: (1) A customized TCN network topology and training strategy adapted to gait time series data were designed. Through the gradient structure design of three-layer dilated convolution, the model’s ability to capture gait length-dependent features and real-time inference speed were taken into account. The network parameters, training hyperparameters and datasets were fully disclosed. The solution can be implemented directly, effectively improving the accuracy of torque prediction and multi-terrain generalization ability.

[0023] (2) The torque distribution mechanism of the 10th order high-order polynomial in the full angle range is adopted to replace the traditional piecewise function mapping, which eliminates the risk of torque jump at the angle critical point, realizes the continuous and smooth adjustment of the torque ratio of the hip and knee joints with the knee joint angle, and completes the torque distribution through the inherent geometric mapping of the mechanical structure, which greatly simplifies the algorithm calculation, improves the assist response speed and stability, and avoids human-machine motion confrontation.

[0024] (3) The bionic transmission system simulates the cross-joint synergistic mechanical characteristics of human muscles through the coordinated design of connecting line-fixed pulley-moving pulley-cam mechanism. The moving pulley realizes the force amplification effect, and the torque distribution ratio is dynamically adjusted through the cam profile. A single motor can realize the coordinated assistance of the hip and knee joints with dual degrees of freedom, which greatly simplifies the structure of the device, reduces the weight and energy consumption of the device, and improves the force transmission efficiency.

[0025] (4) The assistive effect was quantitatively verified through high-fidelity human-machine coupled dynamics simulation and rigorous electromyography control experiment: under flat terrain, the activation of the rectus femoris muscle was significantly reduced by 13.04%; under sloping terrain, the muscle activation was reduced by 9.02%; under high dynamic terrain of stairs, the muscle activation was reduced by 7.29%, which can completely cover the additional load brought by the mechanical structure of the exoskeleton itself, effectively reduce the metabolic load of human muscles, and has significant technological progress and engineering practical value. Attached Figure Description

[0026] Figure 1 A structural diagram of the lower limb exoskeleton assistive device of this application is shown.

[0027] Figure 2 The diagram shows the working state of the lower limb exoskeleton assistive device of this application.

[0028] Figure 3 A structural distribution diagram of the lower limb exoskeleton assistive device of this application is shown.

[0029] Figure 4 A side view of the lower limb exoskeleton assistive device of this application is shown.

[0030] Figure 5 A structural distribution diagram of the hip and knee joint transmission system of this application is shown.

[0031] Figure 6A detailed structural diagram of the hip and knee joint transmission system of this application is shown.

[0032] Figure 7 A simplified structural diagram of the hip and knee joint transmission system of this application is shown.

[0033] Figure 8 A flowchart illustrating the overall process of the control system of this application is shown.

[0034] in: 100. Backplate; 200. Drive structure; 300. Lower limb structure; 301. Hip joint cross link; 302. Hip joint thigh link; 303. Thigh link; 304. Lower leg link; 500. Hip and knee joint transmission system; 501. Connecting line; 502. Movable pulley assembly; 503. Cam mechanism; 504. Transmission line; 505. Fixed pulley assembly; 506. Fixed frame; 507. Fixed pulley structure; 508. Slide rail; 509. Slider; 510. Movable pulley structure; 511. Thigh pulley. Detailed Implementation

[0035] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0036] In the description of this application, it should be understood that if terms such as "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential" appear, these terms indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0037] Furthermore, where the terms "first" and "second" appear, these terms are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, where the term "multiple" appears, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0038] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0039] In this application, unless otherwise expressly specified and limited, the use of descriptions such as "above" or "below" the second feature indicates that the first and second features are in direct contact or indirect contact via an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. Similarly, "below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0040] It should be noted that if an element is referred to as being "fixed to" or "set on" another element, it can be directly on the other element or there may be an intervening element. If an element is considered to be "connected to" another element, it can be directly connected to the other element or there may be an intervening element. If so, the terms "vertical," "horizontal," "upper," "lower," "left," "right," and similar expressions used in this application are for illustrative purposes only and do not represent the only possible implementation.

[0041] Example 1 Reference Figures 1-7 As shown, this embodiment provides an adaptive multi-terrain lower limb exoskeleton assistive device, including a backplate 100, a drive structure 200, a lower limb structure 300, sensors, a hip and knee joint transmission system 500, and a control system.

[0042] The back panel 100 is fixed to the wearer's back.

[0043] The drive structure 200 is mounted on the back plate 100.

[0044] The lower limb structure 300 includes a hip joint strut link 301, a hip joint thigh link 302, a thigh link 303, and a lower leg link 304 that are rotatably connected from top to bottom in the height direction.

[0045] The sensor is mounted on the lower limb structure 300 to monitor the spatial position changes and kinematic time-series data of the lower limb structure 300.

[0046] The hip and knee joint transmission system 500 includes a connecting line 501 connecting the drive structure 200, a fixed pulley component 505 mounted on the thigh link 303, and a movable pulley component 502. The movable end of the connecting line 501 passes around the movable pulley component 502 and is connected to the thigh link 303. The hip and knee joint transmission system 500 also includes a cam mechanism 503 mounted at the connection between the thigh link 303 and the lower leg link 304. The movable pulley component 502 and the cam mechanism 503 are connected through the transmission line 504.

[0047] The control system includes an electrically connected structure adaptive allocation layer and a prediction adjustment layer. The prediction adjustment layer has a built-in temporal convolutional network (TCN) torque prediction model trained offline. The structure adaptive allocation layer is electrically connected to the hip and knee joint transmission system 500.

[0048] Specifically, the device is designed using biomimetic principles, mimicking the movement patterns of the human lower limbs to provide effective support for the wearer. The backplate 100, as the basic support structure of the entire device, not only needs sufficient strength to withstand the weight of the device itself and the various forces generated during movement, but also needs to be as lightweight as possible to reduce the burden on the wearer. The backplate 100 is typically made of high-strength and lightweight alloy materials, such as aluminum alloy or carbon fiber composite materials.

[0049] The drive structure 200 is the power source of the device. It is mounted on the back plate 100 and can stably output power, providing a guarantee for subsequent transmission and assistance.

[0050] The links of the lower limb structure 300 are connected by a rotating mechanism. This design allows the links to rotate relatively flexibly, thus better simulating the movement of human lower limb joints and achieving a natural and smooth gait. The sensor plays a crucial role; it acts as the device's "eyes," capable of sensing real-time changes in the spatial position and kinematic timing data of the lower limb structure 300. This provides accurate data to the control system, enabling adjustments to the assist strategy based on different movement conditions.

[0051] The hip and knee joint transmission system 500 is the core part of the entire device. The connecting line 501 transmits the power of the drive structure 200 to the movable pulley component 502. The movable pulley component 502 enhances the assist effect by amplifying the force. At the same time, it is connected to the cam mechanism 503 through the transmission line 504, realizing the dynamic distribution of torque between the hip and knee joints, enabling the device to adapt to different sports scenarios.

[0052] When the lower limb exoskeleton assistive device of this application is in operation, the wearer's lower limb movement causes the hip joint cross link 301, hip joint thigh link 302, thigh link 303 and lower leg link 304 in the lower limb structure 300 to rotate accordingly. Sensors installed on the lower limb structure 300 monitor the spatial position changes and kinematic timing data of each component in real time, providing real-time motion status feedback to the system.

[0053] When the wearer walks, runs, or performs other lower limb movements, the various links in the lower limb structure 300 rotate in sync with the movement of the body's joints. For example, during walking, the hip joint flexes, extends, and rotates, while the knee joint flexes and extends, causing the corresponding links to rotate. Sensors accurately capture information such as the positional changes, rotation angles, and movement speed of each link, converting this information into electrical signals that are transmitted to the control system. Based on the received signals, the control system can understand the wearer's movement status in real time, such as walking speed, stride length, and whether the user is on different terrains like uphill or downhill, thus providing a basis for subsequent adjustments to the assist system. This real-time motion status feedback mechanism allows the device to more accurately adapt to the wearer's movement needs, avoiding insufficient or excessive assistance, and improving the device's comfort and effectiveness.

[0054] In this embodiment, the drive structure 200 is fixed to the back plate 100 and serves as a power source, transmitting power to the hip and knee joint transmission system 500 via the connecting line 501. The end of the connecting line 501 furthest from the drive structure 200 passes around the movable pulley component 502 mounted on the thigh link 303 and connects to it. The movable pulley component 502 amplifies the driving force, enhancing the assist effect of the device, and simultaneously, in conjunction with the movement of the thigh link 303, achieves effective power transmission. After the motor starts, its rotational motion drives the connecting line 501. During power transmission, the connecting line 501 passes around the movable pulley component 502. According to the principle of the movable pulley, it can change the direction of force and save effort, thus amplifying the driving force. For example, when the motor outputs a certain pulling force, the force transmitted to the thigh link 303 after passing through the movable pulley component 502 will increase, thereby providing stronger assistance to the wearer's lower limbs. At the same time, the movable pulley component 502 works in conjunction with the movement of the thigh linkage 303. As the thigh linkage 303 rotates, the movable pulley component 502 will also move accordingly, ensuring that the power can be continuously and effectively transmitted to other parts of the lower limb structure 300, thus realizing the assist function of the entire device.

[0055] The movable pulley component 502 is connected to the cam mechanism 503 installed at the connection between the thigh link 303 and the lower leg link 304 via the transmission line 504. When the lower leg link 304 rotates relative to the thigh link 303, the cam mechanism 503 rotates accordingly, which can dynamically change the force point and lever arm length of the transmission line 504, thereby adjusting the torque distribution ratio between the hip joint and the knee joint. This allows the device to adaptively adjust the assist mode according to different movement states without relying on external environmental recognition equipment. It can switch the assist mode in different scenarios simply by changing the human gait. The whole process forms a coherent power transmission and adjustment system.

[0056] During human movement, the force distribution on the hip and knee joints differs and changes with the movement state. For example, when walking uphill, the knee joint needs to bear greater torque to overcome gravity; while when walking downhill, the hip joint requires more assistance to maintain balance. When the lower leg link 304 rotates relative to the thigh link 303, the cam mechanism 503 rotates accordingly. The cam mechanism 503 has a special profile shape; as it rotates, the contact point between the transmission line 504 and the cam mechanism 503 changes, resulting in changes in the force distribution point and lever arm length of the transmission line 504. According to the torque calculation formula, torque equals force multiplied by the lever arm; when the lever arm length changes, the torque also changes accordingly. In this way, the device can dynamically adjust the torque distribution ratio between the hip and knee joints, ensuring that each joint receives appropriate assistance under different movement states. This adaptive adjustment method does not require additional external environmental recognition equipment; it relies solely on the body's gait changes to switch assistance modes, greatly improving the device's practicality and convenience while reducing its complexity and cost.

[0057] The beneficial effects of this device are as follows: the backplate 100 provides stable support for the overall structure while reducing the device's weight; the cooperation between the drive structure 200 and the hip and knee joint transmission system 500, through the mechanical amplification of the movable pulley and the dynamic adjustment of the cam mechanism 503, ensures sufficient assist output while achieving flexible distribution of hip and knee joint torque, improving the device's adaptability and assist efficiency in complex movement scenarios; the rotatably connected lower limb structure 300 better adapts to the natural gait of the human body, reducing movement obstacles, and real-time monitoring by sensors provides data support for precise assistance. The stable support of the backplate 100 is the foundation for the normal operation of the entire device. If the backplate 100 is not stable enough, it may wobble or deform during movement, affecting the device's assist effect and potentially causing injury to the wearer. Reducing the device's weight reduces the wearer's burden, improves wearing comfort, and allows the wearer to use the device for longer periods. The cooperation between the drive structure 200 and the hip and knee joint transmission system 500 is key to the device's efficient assist. The mechanical amplification effect of the movable pulley enables greater assistance to the lower limbs with limited motor output power, meeting the assistance needs in different sports scenarios. The dynamic adjustment of the cam mechanism 503 ensures a more reasonable torque distribution between the hip and knee joints, allowing the device to better adapt to complex human movements. The rotatably connected lower limb structure 300 mimics the movement of human lower limb joints, naturally following the wearer's gait, reducing resistance during movement, and making the wearer's movement smoother. Real-time monitoring by sensors ensures precise assistance. Through real-time acquisition and analysis of motion information from the lower limb structure 300, the control system can adjust the assistance strategy in a timely manner, ensuring that the device provides appropriate assistance in different sports states, improving the device's intelligence and assistance effect.

[0058] In another embodiment, the drive structure 200 is a servo motor used to pull the connecting line 501 to output the total drive torque. The total output torque of the motor satisfies the formula:

[0059] in, Provide torque to the hip joint. Provides torque assistance to the knee joint. When the user wears the device for exercise, sensors installed on the lower limb structure 300 monitor in real time the spatial position changes and kinematic timing data of the hip joint strut link 301, hip joint thigh link 302, thigh link 303, and lower leg link 304. After the information is transmitted to the control system, the system controls the motor to start according to the motion state. The motor outputs power through the traction connection line 501. The end of the connection line 501 away from the motor passes around the movable pulley component 502 on the thigh link 303 and is connected to the thigh link 303. The movable pulley component 502 drives the thigh link 303 to move under the traction of the connection line 501. At the same time, the movable pulley component 502 pulls the cam mechanism 503 installed at the connection between the thigh link 303 and the lower leg link 304 through the transmission line 504. As the lower leg link 304 rotates relative to the thigh link 303, the cam mechanism 503 rotates synchronously and adjusts the force state of the transmission line 504 by changing its own contour. The motor's speed and torque can be precisely controlled by the control system, enabling it to output appropriate power according to different movement requirements. When the sensor detects the user starting to move, it transmits the information to the control system, which reacts quickly and starts the motor. After the motor starts, it pulls the connecting line 501 through rotation, which transmits power to the movable pulley component 502. Under the action of the connecting line 501, the movable pulley component 502 drives the thigh linkage 303, providing assistance to the thigh. Simultaneously, the movable pulley component 502 pulls the cam mechanism 503 through the transmission line 504. The cam mechanism 503 rotates synchronously with the rotation of the lower leg linkage 304 relative to the thigh linkage 303. The profile of the cam mechanism 503 is carefully designed; during rotation, it changes the force state of the transmission line 504, such as the position of the force point and the magnitude of the force. This change in force state further affects the torque distribution between the hip and knee joints, allowing the device to adjust the assistance mode in real time according to the movement of the lower leg, improving the device's assistance accuracy and adaptability.

[0060] In another embodiment, the sensor is an inertial measurement unit (IMU), mounted on the thigh link 303 and the lower leg link 304. The collected dataset includes spatial position changes of the hip and knee joints, real-time joint angles, angular velocities, and accelerations. An IMU sensor is a sensor capable of measuring the motion state of an object; it typically consists of an accelerometer, gyroscope, and magnetometer. An accelerometer measures an object's acceleration, a gyroscope measures its angular velocity, and a magnetometer measures its orientation. When the user wears the device, the thigh link 303 and the lower leg link 304 rotate. The IMU can sense these rotations in real time and accurately capture their spatial position changes, motion angles, and angular velocities. This information is crucial for the device's precise assistance. For example, by analyzing the motion angles and angular velocities, the control system can determine the user's current motion state, such as walking, running, or climbing stairs. Based on different motion states, the control system can adjust the motor's output power and assistance strategy, enabling the device to provide more appropriate assistance. Meanwhile, the high-precision measurement capability of the IMU sensor ensures the accuracy of the data, providing reliable data support for the intelligent control of the device.

[0061] In another embodiment, the fixed pulley component 505 includes a fixed frame 506 and a fixed pulley structure 507 mounted on the fixed frame 506. The fixed frame 506 is fixedly connected to the upper end of the thigh connecting rod 303 near the hip joint. The connecting line 501 is a Bowden line, one end of which is fixedly connected to the output end of the drive structure 200, and the other end passes through the fixed pulley structure 507 and the movable pulley component 502 in sequence before being fixedly connected to the fixed frame 506. The fixed pulley component 505 plays an important role in the device. The fixed frame 506 provides stable support for the fixed pulley structure 507, ensuring that the fixed pulley structure 507 will not shake or shift during operation. One end of the connecting line 501 is connected to the fixed frame 506, and the other end passes through the movable pulley component 502 and the fixed pulley structure 507 in sequence before being connected to the drive structure 200. The fixed pulley structure 507 can change the direction of force on the connecting line 501, which makes the installation position of the drive structure 200 more flexible. For example, the drive structure 200 can be installed in a more suitable position on the back plate 100 according to the overall layout and design requirements of the device, without being limited by the force direction of the connecting line 501. Simultaneously, the fixed pulley structure 507 can reduce friction and wear on the connecting line 501 during movement. During the movement of the connecting line 501, the fixed pulley structure 507 allows the connecting line 501 to slide more smoothly, avoiding direct friction between the connecting line 501 and other components, thus extending the service life of the connecting line 501. Furthermore, the cooperation between the fixed pulley and the movable pulley further improves the efficiency and stability of force transmission, making the assistive effect of the device more significant.

[0062] Bowden wire is used as the connecting wire 501. Its high-strength inner steel wire ensures efficient power transmission, while the outer sleeve effectively isolates external friction and interference, protects the inner steel wire, and extends its service life. The flexibility of Bowden wire allows it to adapt flexibly to the steering requirements of components such as fixed pulleys and movable pulleys. It can achieve smooth power transmission between different components without the need for a complex steering mechanism, simplifying the transmission system structure and reducing the overall weight of the device.

[0063] In another embodiment, the movable pulley component 502 includes a slide rail 508 mounted on the thigh connecting rod 303, a slider 509 slidably mounted on the slide rail 508, and a movable pulley structure 510 mounted on the slider 509. One end of the transmission line 504 is rigidly connected to the slider 509, and the other end is connected to the cam mechanism 503 via a guide pulley on the thigh connecting rod 303. This structural design of the movable pulley component 502 has many advantages. The slide rail 508, mounted on the thigh connecting rod 303, provides a stable track for the movement of the slider 509. The slider 509 can slide freely on the slide rail 508. When the connecting rope pulls the movable pulley structure 510, the slider 509 moves along the slide rail 508, thereby guiding the movement direction of the movable pulley structure 510. This guiding effect ensures that the movable pulley structure 510 does not deviate during movement, guaranteeing the accuracy of power transmission. A transmission line 504 is positioned between the slider 509 and the cam mechanism 503, transmitting the motion of the movable pulley structure 510 to the cam mechanism 503. When the movable pulley structure 510 moves under the tension of the connecting line 501, the slider 509 slides on the slide rail 508, simultaneously driving the cam mechanism 503 to rotate via the transmission line 504, thus achieving dynamic adjustment of the torque between the hip and knee joints. This structural design is reasonable, with tight fit between the components, improving the reliability and stability of the device.

[0064] In another embodiment, a thigh pulley 511 is provided on the thigh link 303, and the transmission line 504 contacts the thigh pulley 511. The thigh pulley 511 plays an important guiding role in the movement of the transmission line 504. During human movement, the lower limb structure 300 undergoes various complex movements, and the transmission line 504 is easily affected by various forces during these movements, causing it to deviate or slack. The thigh pulley 511 enables the transmission line 504 to move along a fixed trajectory, preventing deviation and ensuring accurate force transmission direction. Simultaneously, it prevents the transmission line 504 from slacking, ensuring it remains taut and thus improving force transmission efficiency. When the movement of the transmission line 504 is more stable and precise, the cam mechanism 503 becomes more sensitive in adjusting torque distribution. The cam mechanism 503 can adjust its rotation in a timely manner according to changes in the force on the transmission line 504, achieving dynamic torque distribution between the hip and knee joints. This precise torque distribution makes the coordinated assistance of the hip and knee joints more in line with the laws of human movement, reduces obstacles during movement, and improves the wearer's exercise comfort and the device's assistive effect.

[0065] In another embodiment, the profile curve of the cam mechanism 503 is directly fitted by a nonlinear mapping function from the knee joint angle to the target torque ratio of the hip and knee joint, and the mapping function is implemented using a 10th-order high-order polynomial.

[0066] In another embodiment, the TCN torque prediction model of the prediction adjustment layer takes standardized lower limb kinematic time series data as input and outputs the total assist torque command required for the current gait; the structural adaptive allocation layer is used to continuously allocate the total assist torque to the hip and knee joints according to the real-time knee joint angle and the geometric mapping relationship of the transmission system.

[0067] In another embodiment, the knee joint angle sequence acquired by the sensor needs to be standardized, with the standardized mean set to 35.78 and the standard deviation set to 19.87.

[0068] Example 2 This invention also proposes an adaptive multi-terrain lower limb exoskeleton assistive control method for the aforementioned adaptive multi-terrain lower limb exoskeleton assistive device, comprising the following steps: Real-time acquisition of multimodal kinematic state sequences is achieved by using inertial measurement units (IMUs) deployed in the thigh and calf segments of the lower limb structure 300 to collect real-time kinematic temporal data of the wearer's lower limb joints and then standardizing the acquired knee joint angle sequences. Based on the online generation of total driving torque of the TCN prediction adjustment layer, the standardized kinematic time series data is used as input and passed to the offline trained temporal convolutional network TCN to predict and output the total assist torque command required for the current gait in real time. Based on the continuous torque distribution of high-order polynomial mapping, after obtaining the total assist torque, the total torque is passively distributed to the hip and knee joints according to the real-time knee joint angle and the inherent geometric mapping relationship of the exoskeleton transmission system. The torque distribution ratio is calculated by the 10th-order high-order polynomial nonlinear mapping function of the knee joint angle to the target hip and knee joint torque ratio. Dynamic closed-loop control: The control system drives the motor to output adaptive torque according to the allocated hip and knee joint torque command, and completes the coordinated assistance of the hip and knee joints through the transmission system to achieve multi-terrain gait adaptive response.

[0069] In another embodiment, the TCN network comprises three one-dimensional convolutional layers, wherein the dilation factor of convolutional layer 1 is [1,1], the dilation factor of convolutional layer 2 is [2,1], and the dilation factor of convolutional layer 3 is [4,1]. The kernel size of each convolutional layer is uniformly set to [5×3]. ReLU is used as the activation function between network layers, and a Dropout layer with a ratio of 10% is configured. The final layer is connected to a global max pooling layer and a fully connected layer for regression output.

[0070] In another embodiment, the TCN network is trained using the Adam optimizer with a learning rate of 0.001 and a batch size of 64, and an early stopping strategy is introduced to prevent model overfitting; the training dataset uses a common gait kinematics and dynamics dataset that includes flat ground, slopes and stairs.

[0071] In another embodiment, the expression for the 10th-order higher-order polynomial nonlinear mapping function is:

[0072] in, These are the polynomial coefficients obtained by fitting experimental data using the least squares method; The knee joint angle is standardized with a mean of 35.78 and a standard deviation of 19.87. The target hip-knee joint torque ratio.

[0073] In another embodiment, the torque distribution at the hip and knee joints satisfies the following mechanical formula: Hip joint assist torque:

[0074] Knee joint assist torque:

[0075] Hip-knee joint torque ratio:

[0076] in, The tension of the motor output terminal connecting wire 501; The effective arm of the hip joint is constant. Let be the equivalent radius of the knee joint cam at the current angle. This is the real-time hip joint angle.

[0077] Polynomial coefficients obtained from fitting The following table shows the 95% confidence intervals for the following:

[0078] Through this polynomial mapping, the equivalent radius of the knee cam changes with the knee joint angle throughout the gait cycle. Synchronous dynamic adjustment enables continuous adaptive adjustment of the hip-knee joint torque ratio. The control system uses an STM32F4 series microcontroller as the main control core. The predictive adjustment layer has a built-in TCN torque prediction model trained offline. It receives standardized kinematic timing data collected by sensors and predicts and outputs total assist torque commands in real time. The structural adaptive allocation layer completes the continuous allocation of total torque to the hip-knee joint based on the real-time knee joint angle and the geometric mapping relationship of the transmission system, generates motor control commands, drives the motor to output adaptive torque, and realizes closed-loop control.

[0079] System simulation and closed-loop verification: A high-fidelity human-machine coupled dynamics model was constructed in the MuJoCo physics simulation engine for closed-loop verification. The human musculoskeletal model was reconstructed based on the geometric and inertial parameters of the OpenSim open-source model library "gait2392," accurately reproducing the musculoskeletal dynamics characteristics of the lower limbs. A rigid body dynamics force transmission model between the exoskeleton and the human body was established to explicitly characterize the influence of the connection compliance of the binding interface on torque transmission. During simulation, the control algorithm accessed the mjData structure of MuJoCo to read the joint angle and angular velocity status of the virtual sensors in real time in a closed loop, and synchronously output the joint control torque constrained by the rated peak torque of the motor, realizing the simulation verification of multi-terrain gait adaptive response. Simulation results show that when walking on flat ground, the difference in hip joint angle between the wearing exoskeleton and the natural gait without the exoskeleton fluctuates within ±5°, and the difference in knee joint angle is about 10° during the gait transition phase, remaining within ±5° in other phases. On slopes and stairs, the maximum deviation of hip and knee joint angles does not exceed 10°, and the exoskeleton has minimal impact on the wearer's natural gait, exhibiting good motion consistency and terrain adaptability.

[0080] Experiment and Results Analysis: To verify the assistive effect of this application, four healthy adult subjects, aged 18–26 years, with a height of 173±3cm, a lower limb length of 90±5cm, and no history of lower limb musculoskeletal diseases, were recruited. Subjects completed three sets of controlled experiments at a natural walking pace on three typical terrains: flat ground, slopes, and stairs, without changing the control strategy and parameters: (1) natural walking without exoskeleton, serving as a physiological baseline; (2) wearing an exoskeleton without assistive function, quantifying the additional load of the mechanical structure; (3) wearing an exoskeleton with assistive function, verifying the net assistive effect. The experiment used surface electromyography (sEMG) sensors to collect electromyographic signals of the rectus femoris muscle. The root mean square (RMS) value was used to characterize muscle activation intensity, and all RMS values ​​were standardized using the individual's maximum spontaneous contraction value (MVC). Experimental results: Flat terrain: Wearing an unassisted exoskeleton increased rectus femoris muscle activation by 9.60%, while activating the assist control of this invention significantly reduced rectus femoris muscle activation by 13.04%. Sloping terrain: Wearing an unassisted exoskeleton increased muscle activation by 22.12%, while with assisted control enabled, muscle activation still decreased by 9.02%. Staircase high dynamic terrain: Wearing an unassisted exoskeleton increased muscle activation by 8.54%, while enabling assisted control reduced muscle activation by 7.29%. The experimental results fully demonstrate that the technical solution of this invention can effectively reduce the metabolic load on human muscles under different terrains, completely cover the additional load brought by the mechanical structure of the exoskeleton itself, and has excellent assist effect and adaptability to complex terrain.

[0081] This application proposes a two-layer control architecture that coordinates a structural adaptive allocation layer and a predictive adjustment layer, enabling adaptive assistance for multi-terrain gait without the need for external terrain recognition sensors or discrete mode switching. The structural layer is responsible for the continuous coordination of cross-joint moments, while the predictive layer is responsible for adjusting the intensity of the total moment. The two are fused through a unified control loop, fundamentally solving the pain points of existing technologies that rely on terrain recognition and mode switching, resulting in response delays and discontinuous assistance.

[0082] This application features a customized TCN network topology and training strategy adapted to gait time series data. Through a gradient structure design with three layers of dilated convolution, it balances the model's ability to capture gait length-dependent features with real-time inference speed. The network parameters, training hyperparameters, and dataset are fully disclosed, and the solution can be directly implemented, effectively improving the accuracy of torque prediction and multi-terrain generalization ability.

[0083] This application adopts a continuous torque distribution mechanism using a 10th-order high-order polynomial across the entire angle range, replacing the traditional piecewise function mapping. This eliminates the risk of torque jumps at critical angle points, enabling continuous and smooth adjustment of the hip-knee joint torque ratio with the knee joint angle. Simultaneously, torque distribution is completed through the inherent geometric mapping of the mechanical structure, significantly simplifying the algorithm's computational load, improving the assist response speed and stability, and avoiding human-machine motion resistance.

[0084] The bionic transmission system of this application simulates the cross-joint synergistic mechanical characteristics of human muscles through the coordinated design of connecting line-fixed pulley-moving pulley-cam mechanism. The moving pulley realizes the force amplification effect, and the torque distribution ratio is dynamically adjusted by the cam profile. A single motor can realize the coordinated assistance of the hip and knee joints with dual degrees of freedom, which greatly simplifies the structure of the device, reduces the weight and energy consumption of the device, and improves the force transmission efficiency.

[0085] This application quantifies and verifies the assistive effect through high-fidelity human-machine coupled dynamics simulation and rigorous electromyography comparative experiments: on flat terrain, the activation of the rectus femoris muscle is significantly reduced by 13.04%; on sloping terrain, the muscle activation is reduced by 9.02%; and on high-dynamic terrain such as stairs, the muscle activation is reduced by 7.29%. This can completely cover the additional load brought by the mechanical structure of the exoskeleton itself, effectively reduce the metabolic load on human muscles, and has significant technological progress and engineering practical value.

[0086] The above description is an explanation of the invention, not a limitation thereof. The scope of the invention is defined in the claims. Within the scope of protection of the invention, any form of modification may be made.

Claims

1. An adaptive multi-terrain lower limb exoskeleton assistive device, characterized in that, include: Backplate, which is fixed to the wearer's back; A drive structure is disposed on the back plate; The lower limb structure includes a hip joint strut link, a hip joint thigh link, a thigh link, and a lower leg link that are rotatably connected from top to bottom in the height direction; A sensor, mounted on the lower limb structure, is used to monitor spatial positional changes and kinematic time-series data of the lower limb structure. The hip and knee joint transmission system includes a traction line extending from a self-driving structure, a fixed pulley component mounted on a thigh link, and a movable pulley component. The movable end of the traction line passes around the movable pulley component and is connected to the thigh link. The hip and knee joint transmission system also includes a cam mechanism mounted at the connection between the thigh link and the lower leg link. The movable pulley component is connected to the cam mechanism via a transmission line. The control system includes an electrically connected structure adaptive allocation layer and a prediction adjustment layer. The prediction adjustment layer incorporates a temporal convolutional network (TCN) torque prediction model trained offline. The structure adaptive allocation layer is electrically connected to the hip and knee joint transmission system.

2. The adaptive multi-terrain lower limb exoskeleton assistive device according to claim 1, characterized in that: The drive structure is a servo motor, used to pull the connecting line and output the total driving torque. The total output torque of the motor satisfies the formula: in, Provide torque to the hip joint. Provides assist torque to the knee joint.

3. The adaptive multi-terrain lower limb exoskeleton assistive device according to claim 1, characterized in that: The fixed pulley component includes a fixed frame and a fixed pulley structure. The fixed pulley structure is mounted on the fixed frame, which is fixedly connected to the upper end of the thigh linkage near the hip joint. The connecting line is a Bowden line, one end of which is fixedly connected to the output end of the drive structure, and the other end passes through the fixed pulley structure and the movable pulley component in sequence before being fixedly connected to the fixed frame. The movable pulley component includes a slide rail mounted on the thigh linkage, a slider slidably mounted on the slide rail, and a movable pulley structure mounted on the slider. One end of the transmission line is rigidly connected to the slider, and the other end is connected to the cam mechanism via a guide pulley on the thigh linkage.

4. The adaptive multi-terrain lower limb exoskeleton assistive device according to claim 1, characterized in that: The profile curve of the cam mechanism is directly fitted by a nonlinear mapping function from the knee joint angle to the target torque ratio of the hip and knee joint, and the mapping function is implemented using a 10th-order high-order polynomial.

5. The adaptive multi-terrain lower limb exoskeleton assistive device according to claim 1, characterized in that: The TCN torque prediction model of the prediction and adjustment layer takes standardized lower limb kinematic time-series data as input and outputs the total assist torque command required for the current gait. The structural adaptive allocation layer is used to continuously allocate the total assist torque to the hip and knee joints according to the real-time knee joint angle and the geometric mapping relationship of the transmission system.

6. An adaptive multi-terrain lower limb exoskeleton assistive control method, characterized in that: A method for controlling an adaptive multi-terrain lower limb exoskeleton assistive device according to any one of claims 1-5 includes the following steps: Real-time acquisition of multimodal kinematic state sequences is achieved by using inertial measurement units (IMUs) deployed in the thigh and calf segments of the lower limb structure to collect real-time kinematic temporal data of the wearer's lower limb joints and then standardizing the acquired knee joint angle sequences. Based on the online generation of total driving torque of the TCN prediction adjustment layer, the standardized kinematic time series data is used as input and passed to the offline trained temporal convolutional network TCN to predict and output the total assist torque command required for the current gait in real time. Based on the continuous torque distribution of high-order polynomial mapping, after obtaining the total assist torque, the total torque is passively distributed to the hip and knee joints according to the real-time knee joint angle and the inherent geometric mapping relationship of the exoskeleton transmission system. The torque distribution ratio is calculated by the 10th-order high-order polynomial nonlinear mapping function of the knee joint angle to the target hip and knee joint torque ratio. Dynamic closed-loop control: The control system drives the motor to output adaptive torque according to the allocated hip and knee joint torque command, and completes the coordinated assistance of the hip and knee joints through the transmission system to achieve multi-terrain gait adaptive response.

7. The adaptive multi-terrain lower limb exoskeleton assistive control method according to claim 6, characterized in that: The TCN network consists of three one-dimensional convolutional layers, where the dilation factor of convolutional layer 1 is [1,1], the dilation factor of convolutional layer 2 is [2,1], and the dilation factor of convolutional layer 3 is [4,1]. The kernel size of each convolutional layer is uniformly set to [5×3]. ReLU is used as the activation function between network layers, and a Dropout layer with a ratio of 10% is configured. The final layer is connected to a global max pooling layer and a fully connected layer for regression output.

8. The adaptive multi-terrain lower limb exoskeleton assistive control method according to claim 6, characterized in that: The TCN network was trained using the Adam optimizer with a learning rate of 0.001 and a batch size of 64. An early stopping strategy was introduced to prevent the model from overfitting. The training dataset used a common gait kinematics and dynamics dataset that included flat ground, slopes, and stairs.

9. The adaptive multi-terrain lower limb exoskeleton assistive control method according to claim 6, characterized in that: The expression for the 10th-order higher-order polynomial nonlinear mapping function is: in, These are the polynomial coefficients obtained by fitting experimental data using the least squares method; The knee joint angle is standardized with a mean of 35.78 and a standard deviation of 19.

87. The target hip-knee joint torque ratio.

10. The adaptive multi-terrain lower limb exoskeleton assistive control method according to claim 6, characterized in that: The torque distribution of the hip and knee joints satisfies the following mechanical formula: Hip joint assist torque: Knee joint assist torque: Hip-knee joint torque ratio: in, The tension of the connecting wire at the motor output end; The effective arm of the hip joint is constant. Let be the equivalent radius of the knee joint cam at the current angle. This is the real-time hip joint angle.

Images