A Control Method and System for Lower Limb Exoskeleton Robots Based on Motion Intent Recognition

Abstract

This application provides a lower limb exoskeleton robot control method and system based on motion intention recognition, including: acquiring exoskeleton motion data and exoskeleton unloaded data; determining exoskeleton dynamic parameters through gait training based on the exoskeleton unloaded data; determining human lower limb dynamic parameters based on the exoskeleton motion data and exoskeleton dynamic parameters; calculating human joint torques through inverse dynamic equations based on the exoskeleton motion data, exoskeleton dynamic parameters, and human lower limb dynamic parameters; inputting human joint torques and predefined gait training joint trajectories into a control model to obtain the joint trajectory output by gait training control; the control model is constructed based on inertial terms, damping terms, nonlinear damping terms, and stiffness terms, and is used to dynamically adjust the predefined gait training joint trajectory under the drive of human motion intention, solving the technical problem of poor adaptability of human-computer interaction control due to insufficient motion intention perception accuracy in existing technologies.

Description

Technical Field

[0001] This application relates to the field of rehabilitation robots and human-computer interaction control technology, and in particular to a control method and system for a lower limb exoskeleton robot based on motion intention recognition. Background Technology

[0002] Lower limb exoskeleton robots, as core equipment for gait rehabilitation training, are of great significance in assisting individuals with lower limb motor dysfunction caused by stroke, spinal cord injury, and other diseases to regain their walking ability. Existing technologies typically employ position- or force-based control strategies, driving the exoskeleton to move the target's legs through predefined gait trajectories, or using bioelectrical signals such as electromyography (EMG) and electroencephalography (EEG) to identify the target's movement intentions for control. However, existing sensing methods largely rely on bioelectrical signals, making them susceptible to noise interference. Furthermore, they are mostly based on discrete motion pattern recognition, failing to fully consider the dynamic characteristics of human-machine coupling, making it difficult to achieve continuous and accurate motion intention capture. They also fail to adapt to individual biomechanical differences, resulting in control strategies that cannot accommodate the differences in dynamic characteristics between the support and swing phases, leading to an imbalance between training safety and naturalness. Therefore, a method is urgently needed to address the technical problem of insufficient motion intention perception accuracy in existing technologies, resulting in poor adaptability of human-machine interactive control. Summary of the Invention

[0003] This application provides a control method and system for a lower limb exoskeleton robot based on motion intention recognition, which solves the technical problem of poor adaptability of human-computer interaction control due to insufficient motion intention perception accuracy in the prior art.

[0004] To achieve the above objectives, this application adopts the following technical solution: Firstly, a method for controlling a lower limb exoskeleton robot based on motion intention recognition is provided, comprising: acquiring exoskeleton motion data and exoskeleton unloaded data; the exoskeleton motion data includes joint torque data, position information, and surface electromyography signals in the worn state; the exoskeleton unloaded data includes exoskeleton joint torque and position data in the unloaded state; determining exoskeleton dynamic parameters through gait training based on the exoskeleton unloaded data; determining human lower limb dynamic parameters based on the exoskeleton motion data and exoskeleton dynamic parameters; calculating human joint torques through inverse dynamic equations based on the exoskeleton motion data, exoskeleton dynamic parameters, and human lower limb dynamic parameters; inputting human joint torques and predefined gait training joint trajectories into a control model to obtain the joint trajectories output by gait training control; the control model is constructed based on inertial terms, damping terms, nonlinear damping terms, and stiffness terms, and is used to dynamically adjust the predefined gait training joint trajectories under the drive of human motion intention.

[0005] This application first identifies the dynamic parameters of the exoskeleton and the lower limbs of the human body, and then uses inverse dynamic equations to calculate the joint torques of the human body. This enables continuous and accurate perception of the human body's movement intentions. Based on a control model that includes inertia, damping, nonlinear damping, and stiffness terms, it dynamically adjusts the predefined trajectory, thereby achieving a precise response to the target object's movement intentions and integrating active and passive training, thus improving the adaptability of human-computer interaction and training effectiveness.

[0006] In conjunction with the first aspect mentioned above, in one possible implementation, after obtaining the gait training control output, the method further includes: calculating the deviation between the joint trajectory of the gait training control output and the predefined gait training joint trajectory; constructing virtual attraction and virtual repulsion based on the distance between the joint trajectory of the gait training control output and the preset boundary, and determining the virtual resultant force; adjusting the inertia coefficient and damping coefficient of the control model based on the virtual resultant force to obtain updated inertia coefficient and damping coefficient; acquiring plantar pressure sensor data of the exoskeleton, and adaptively adjusting the stiffness coefficient and nonlinear damping coefficient of the control model based on the plantar pressure data to obtain updated stiffness coefficient and nonlinear damping coefficient; and obtaining the updated joint trajectory of the gait training control output based on the updated control model with updated inertia coefficient, damping coefficient, stiffness coefficient, and nonlinear damping coefficient, driving the exoskeleton to perform active-passive gait training.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, based on the unloaded data of the exoskeleton, the exoskeleton dynamic parameters are determined through gait training, including: performing differential calculations on the position information in the unloaded state to obtain the joint angular velocity and joint angular acceleration; constructing an exoskeleton dynamic regression matrix based on the joint position information, joint angular velocity, and joint angular acceleration in the unloaded state; and identifying the exoskeleton dynamic parameters using the least squares method based on the regression matrix and the joint torque data in the unloaded state.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, the exoskeleton dynamics regression matrix... Satisfy the following formula:

[0009] in, The joint torque of the exoskeleton under no-load conditions. This is the location data under no-load conditions. Angular velocity, Angular acceleration; The exoskeleton dynamics parameter matrix is ​​given by T, which represents the transpose. The results are obtained through Lagrangian dynamics modeling and linear decomposition. and :

[0010]

[0011] By conducting parameter identification experiments, the regression matrix was sampled. and sampling torque vector Finally, the least squares method is used to obtain the optimal estimated parameter vector of the exoskeleton. , Individually identify exoskeleton dynamic parameters .

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, the human lower limb dynamic parameters are determined based on exoskeleton motion data and exoskeleton dynamic parameters, including: filtering data segments with low activation levels of surface electromyography signals under wearing conditions to obtain a first data segment; constructing a human lower limb dynamic regression model based on the first data segment and exoskeleton dynamic parameters; and identifying human lower limb dynamic parameters using the least squares method according to the human lower limb dynamic regression model.

[0013] In conjunction with the first aspect mentioned above, in one possible implementation, the human lower limb dynamics model satisfies the following formula:

[0014] in, The joint torque of the exoskeleton when worn. The joint torque of the exoskeleton under no-load conditions. For the resistance of human joints, This is the location data under no-load conditions. Angular velocity, Angular acceleration, To identify the regression matrix for lower limb dynamics, The lower limb dynamics regression matrix;

[0015]

[0016] By conducting parameter identification experiments, the regression matrix was sampled. and sampling torque vector Finally, the least squares method is used to obtain the optimal estimated parameter vector for the human lower limb. , Identifying the dynamic parameters of the human lower limbs .

[0017] In conjunction with the first aspect mentioned above, one possible implementation involves solving for human joint torques using inverse dynamic equations based on exoskeleton motion data, exoskeleton dynamic parameters, and human lower limb dynamic parameters. This includes: constructing a joint dynamic model of the human-machine coupled system based on exoskeleton motion data; modeling the exoskeleton dynamics component of the joint dynamic model based on exoskeleton dynamic parameters; modeling the human lower limb dynamics component of the joint dynamic model based on human lower limb dynamic parameters; performing human-machine dynamic decoupling processing on the joint dynamic model to obtain a human dynamic sub-model; and constructing inverse dynamic equations based on the human dynamic sub-model and solving for the human joint torques.

[0018] In conjunction with the first aspect mentioned above, one possible implementation involves inputting human joint torques and predefined gait training joint trajectories into a control model to obtain the joint trajectory of the gait training control output. This includes: using human joint torques as the driving input and the predefined gait training joint trajectory as the reference input to construct a joint trajectory tracking error; constructing an inertial response term, a damping response term, a nonlinear damping response term, and a stiffness response term based on the joint trajectory tracking error and its derivative; weighting and fusing the inertial response term, damping response term, nonlinear damping response term, and stiffness response term to construct a joint control quantity; and correcting the predefined gait training joint trajectory based on the joint control quantity to obtain the joint trajectory of the gait training control output.

[0019] In conjunction with the first aspect mentioned above, one possible implementation is virtual synergy. Satisfy the following formula:

[0020] in, Represents virtual gravity. , The current trajectory is used for gait training; This refers to a predefined trajectory, i.e., location data under no-load conditions; Indicates distance, This is the gravitational gain coefficient; Represents virtual repulsion. , The repulsive force gain coefficient, For boundary thresholds; This refers to a predefined trajectory, i.e., location data under no-load conditions; The upper boundary, This is the lower boundary.

[0021] Secondly, a lower limb exoskeleton robot control system based on motion intention recognition is provided, comprising: a data acquisition module, an exoskeleton dynamics parameter determination module, a human lower limb dynamics parameter determination module, a human joint torque calculation module, and a control output generation module; wherein, the data acquisition module is used to acquire exoskeleton motion data and exoskeleton unloaded data; the exoskeleton motion data includes joint torque data, position information, and surface electromyography signals in the worn state; the exoskeleton unloaded data includes exoskeleton joint torque and position data in the unloaded state; the exoskeleton dynamics parameter determination module is used to determine the exoskeleton dynamics based on the exoskeleton unloaded data through gait training. The system includes: a lower limb dynamics parameter determination module, used to determine the lower limb dynamics parameters based on exoskeleton motion data and exoskeleton dynamics parameters; a joint torque calculation module, used to calculate the joint torque based on exoskeleton motion data, exoskeleton dynamics parameters, and lower limb dynamics parameters using inverse dynamic equations; and a control output generation module, used to input the joint torque and predefined gait training joint trajectories into the control model to obtain the joint trajectories of the gait training control output. The control model is constructed based on inertia, damping, nonlinear damping, and stiffness terms, and is used to dynamically adjust the predefined gait training joint trajectories under the drive of human movement intention.

[0022] This application provides a control method and system for a lower limb exoskeleton robot based on motion intention recognition. By identifying the dynamic parameters of the exoskeleton and the human lower limb, and using inverse dynamic equations to calculate the joint torques of the human body, it can continuously and accurately perceive the human body's motion intention, solving the problem of insufficient motion intention perception accuracy in existing technologies. At the same time, based on a control model that includes inertia, damping, nonlinear damping, and stiffness terms, and combined with virtual resultant force and plantar pressure to adaptively adjust the control parameters, it achieves adaptive matching of controller parameters to gait phase, taking into account both the impact resistance of the support phase and the compliance of the swing phase. This effectively improves the adaptability of human-computer interaction control and the safety and naturalness of the training process, solving the technical problem of poor adaptability of human-computer interaction control due to insufficient motion intention perception accuracy in existing technologies.

[0023] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0024] Figure 1 A system architecture diagram of a lower limb exoskeleton robot control system based on motion intention recognition is provided for embodiments of this application; Figure 2 A flowchart illustrating a lower limb exoskeleton robot control method based on motion intention recognition, provided in an embodiment of this application; Figure 3 A flowchart illustrating another lower limb exoskeleton robot control method based on motion intention recognition provided in this application embodiment; Figure 4 A flowchart illustrating another lower limb exoskeleton robot control method based on motion intention recognition provided in this application embodiment; Figure 5 A flowchart illustrating another lower limb exoskeleton robot control method based on motion intention recognition provided in this application embodiment; Figure 6 A flowchart illustrating another lower limb exoskeleton robot control method based on motion intention recognition provided in this application embodiment; Figure 7 This is a flowchart illustrating another lower limb exoskeleton robot control method based on motion intention recognition, provided in an embodiment of this application. Detailed Implementation

[0025] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0026] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0027] The lower limb exoskeleton robot control method based on motion intention recognition provided in this application embodiment can be applied to, for example... Figure 1 In the lower limb exoskeleton robot control system shown, based on motion intention recognition, such as Figure 1 As shown, the system includes: a data acquisition module 101, an exoskeleton dynamics parameter determination module 102, a human lower limb dynamics parameter determination module 103, a human joint torque calculation module 104, and a control output generation module 105. The data acquisition module 101 is used to acquire exoskeleton motion data and exoskeleton unloaded data. The exoskeleton motion data includes joint torque data, position information and surface electromyography signals in the wearing state. The exoskeleton unloaded data includes exoskeleton joint torque and position data in the unloaded state. The exoskeleton dynamics parameter determination module 102 is used to determine the exoskeleton dynamics parameters based on the exoskeleton's unloaded data and through gait training. The human lower limb dynamics parameter determination module 103 is used to determine the human lower limb dynamics parameters based on exoskeleton motion data and exoskeleton dynamics parameters; The human joint torque calculation module 104 is used to calculate the human joint torque based on exoskeleton motion data, exoskeleton dynamic parameters and human lower limb dynamic parameters by inverse dynamic equations. The control output generation module 105 is used to input human joint torque and predefined gait training joint trajectory into the control model to obtain the joint trajectory of gait training control output. The control model is constructed based on inertia term, damping term, nonlinear damping term and stiffness term, and is used to dynamically adjust the predefined gait training joint trajectory under the drive of human movement intention.

[0028] To address the technical problem of insufficient motion intent perception accuracy in existing technologies, leading to poor adaptability in human-computer interaction control, this application provides a lower limb exoskeleton robot control method based on motion intent recognition. Figure 2 This is a flowchart illustrating the lower limb exoskeleton robot control method based on motion intent recognition provided in an embodiment of this application, as shown below. Figure 2 As shown, the method includes: S201. Acquire exoskeleton motion data and exoskeleton unloaded data.

[0029] The exoskeleton motion data includes joint torque data, position information, and surface electromyography signals in the worn state, while the exoskeleton unloaded data includes joint torque and position data in the unloaded state.

[0030] In one possible implementation, data acquisition is fundamental to the entire control process. The system collects joint torque data in real time through torque sensors installed at the exoskeleton joints, acquires position information through encoders, and simultaneously collects surface electromyography (EMG) signals using EMG sensors attached to the surface of the lower limb muscles. Exoskeleton unloaded data refers to the torque and position data collected when the exoskeleton is not worn and is controlled to perform unloaded movements. It should be understood that the above data acquisition methods are merely examples; in other embodiments, other types of sensors or data acquisition devices can be used, as long as they can acquire relevant data reflecting the exoskeleton's state and the user's movement intentions.

[0031] It should be noted that the data in the wearable state reflects the overall state of the human-machine coupling system, while the data in the unloaded state is used to isolate the dynamic characteristics of the exoskeleton itself. The combination of the two provides data support for the subsequent accurate recognition of human intentions.

[0032] This step provides the necessary data foundation for subsequent parameter identification and intent calculation by acquiring comprehensive data including both wearing and idle states, ensuring the diversity and accuracy of data sources.

[0033] S202. Based on the unloaded data of the exoskeleton, determine the dynamic parameters of the exoskeleton through gait training.

[0034] Among them, exoskeleton dynamic parameters refer to parameters that describe the dynamic characteristics of the exoskeleton itself, such as mass, moment of inertia, and center of mass position.

[0035] In one possible implementation, the system utilizes exoskeleton unloaded data to drive exoskeleton movement through a pre-set gait training trajectory. Combined with dynamic modeling techniques, the system identifies the exoskeleton's own dynamic parameters. This process aims to establish a dynamic model of the exoskeleton itself, thereby enabling the separation of the exoskeleton's dynamic influence from the human-machine coupling system in subsequent processing. Specifically, the exoskeleton's dynamic parameters can be deduced by analyzing the relationship between joint torques and motion states (position, velocity, acceleration) under unloaded conditions.

[0036] Based on the above steps, the exoskeleton dynamic parameters were identified using unloaded data, laying the foundation for the subsequent construction of a human-machine coupling model and the separation of human dynamic parameters, thus improving the pertinence of parameter identification.

[0037] S203. Based on exoskeleton motion data and exoskeleton dynamic parameters, determine the dynamic parameters of the human lower limbs.

[0038] Among them, the human lower limb dynamic parameters refer to the parameters that describe the dynamic characteristics of the target object's lower limbs, such as the mass, moment of inertia, and length of the lower limbs.

[0039] In one possible implementation, based on known exoskeleton dynamic parameters, the system constructs a human-machine coupled dynamic model using exoskeleton motion data while worn. Since the total torque in the worn state includes both exoskeleton dynamic torque and human lower limb dynamic torque, the human lower limb dynamic parameters can be separated and identified by subtracting the identified exoskeleton dynamic influences and combining auxiliary information such as surface electromyography signals. This step fully considers the individual biomechanical differences of different target objects, enabling the control system to adapt to target objects of different body types.

[0040] It should be noted that the identification of human lower limb dynamic parameters is the key to realizing personalized rehabilitation training in this embodiment. It enables the control system to understand the limb characteristics of different target objects, thereby providing more precise assistance.

[0041] Based on the above steps, by combining exoskeleton motion data and identified exoskeleton parameters, the dynamic parameters of the human lower limbs are determined, achieving adaptation to individual differences of the target object and providing a personalized model basis for accurately perceiving human movement intentions.

[0042] S204. Based on exoskeleton motion data, exoskeleton dynamic parameters, and human lower limb dynamic parameters, the human joint torque is calculated by inverse dynamic equations.

[0043] Among them, the inverse dynamics equation refers to the dynamic equation for solving the required torque based on the motion state (position, velocity, acceleration) of an object. The human joint torque refers to the torque generated by the joints of the lower limbs, which reflects the human body's intention to move.

[0044] In one possible implementation, the system inputs real-time exoskeleton motion data (position, velocity, acceleration) into an established human-machine coupled dynamics model, which includes exoskeleton dynamics parameters and human lower limb dynamics parameters. Through inverse dynamics calculation, the system can calculate the total torque required for the current motion state, and then isolate the human joint torques. These human joint torques directly reflect the target object's intended action, such as the force and direction of lifting a leg or taking a step.

[0045] It should be noted that, compared with traditional position-based or simple force-based control methods, solving human joint torques through inverse dynamics equations can achieve continuous and quantitative intention perception, rather than just discrete action recognition, which greatly improves the accuracy and sensitivity of intention perception.

[0046] Based on the above steps, by solving the human joint torque through inverse dynamics equations, continuous and accurate perception of human movement intentions is achieved, solving the problem of insufficient intention perception accuracy in existing technologies and providing accurate input signals for subsequent active control.

[0047] S205. Input the human joint torque and the predefined gait training joint trajectory into the control model to obtain the joint trajectory of the gait training control output.

[0048] The control model is constructed based on inertia, damping, nonlinear damping, and stiffness terms, and is used to dynamically adjust the joint trajectories of predefined gait training under the drive of human movement intention.

[0049] In one possible implementation, the control model simulates the impedance characteristics of human joints, using the calculated joint torques as the driving input and a predefined standard gait trajectory as the reference input. When the target object generates a movement intention (i.e., joint torques), the control model dynamically adjusts the predefined trajectory based on these intention torques, combined with four response terms: inertia, damping, nonlinear damping, and stiffness, to generate a new training trajectory that matches the target object's intention. For example, when the target object wants to increase its walking speed, the intention torque increases, and the control model outputs a faster trajectory; when the target object wants to stop, the intention torque decreases, and the trajectory slows down accordingly.

[0050] It should be noted that this control model does not simply follow changes in torque, but simulates the dynamic response of real human joints through the weighted fusion of four response terms, making the exoskeleton's movement more fluid and natural, avoiding a mechanical and stiff response.

[0051] Based on the above steps, by constructing a control model that includes inertia, damping, nonlinear damping, and stiffness terms, dynamic adjustment of a predefined trajectory is achieved under the drive of human motion intention. This enables the exoskeleton to respond smoothly to the intention of the target object, improving the adaptability of human-computer interaction and the naturalness of training.

[0052] This application's embodiments acquire exoskeleton motion data and unloaded data, identify the dynamic parameters of the exoskeleton and human lower limbs respectively, and then use inverse dynamic equations to calculate human joint torques, achieving accurate perception of human movement intentions; at the same time, based on a control model including inertia, damping, nonlinear damping, and stiffness terms, it achieves compliant response and dynamic trajectory adjustment to the target object's intentions, solving the problem of insufficient intention perception accuracy leading to poor human-computer interaction adaptability in the prior art, and improving the safety and effectiveness of rehabilitation training.

[0053] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 3 As shown, following S205 above, the method provided in this application further includes S301 to S305: S301. Calculate the deviation between the joint trajectory of the gait training control output and the predefined gait training joint trajectory.

[0054] Trajectory deviation refers to the positional difference between the actual generated trajectory and the preset standard trajectory, reflecting the degree to which the target object's movement intention deviates from the standard trajectory.

[0055] In one possible implementation, the system calculates the difference between the joint trajectory output by S205 and the predefined gait training joint trajectory in real time. This deviation value is not only used to evaluate the following performance of the target object, but also serves as the basic input data for subsequent construction of the virtual force field. Specifically, the larger the deviation, the stronger the target object's active intention or the greater the interference it is subjected to, requiring a more significant response from the control system.

[0056] It should be noted that this deviation calculation is bidirectional, including not only positional deviations but also deviations in higher-order derivatives such as velocity and acceleration, in order to comprehensively reflect the differences in motion states.

[0057] Based on the above steps, by calculating the trajectory deviation in real time, a precise quantitative indicator is provided for the subsequent construction of the virtual force field, so that parameter adjustment has a clear feedback basis.

[0058] S302. Based on the distance between the joint trajectory output by gait training control and the preset boundary, construct virtual attraction and virtual repulsion, and determine the virtual resultant force.

[0059] Among them, virtual gravity refers to the virtual force that guides the trajectory to converge toward a predefined trajectory, virtual repulsion refers to the virtual force that prevents the trajectory from exceeding the safety boundary, and virtual resultant force is the vector sum of the two.

[0060] In one possible implementation, the system constructs a virtual potential field environment. The magnitude of the virtual gravity is proportional to the trajectory deviation, aiming to pull the deviated trajectory back to the predefined trajectory, ensuring the standardization of training; the magnitude of the virtual repulsive force is inversely proportional to the distance from the trajectory to a preset boundary (such as the joint's limit position), aiming to push the trajectory away from the dangerous boundary, ensuring the safety of training. The system superimposes the two to obtain a virtual resultant force, and this resultant force vector comprehensively reflects the trajectory correction requirements.

[0061] Preferred, virtual synergy Satisfy the following formula:

[0062] in, Represents virtual gravity. , The current trajectory is used for gait training; This refers to a predefined trajectory, i.e., location data under no-load conditions; Indicates distance, This is the gravitational gain coefficient; Represents virtual repulsion. , The repulsive force gain coefficient, For boundary thresholds; This refers to a predefined trajectory, i.e., location data under no-load conditions; The upper boundary, This is the lower boundary.

[0063] Based on the above steps, by constructing virtual attraction and repulsion, a balance between training standardization and security is achieved, providing a physically meaningful input signal for the adaptive adjustment of control parameters.

[0064] S303. Adjust the inertia coefficient and damping coefficient of the control model based on the virtual resultant force to obtain the updated inertia coefficient and damping coefficient.

[0065] Among them, the inertia coefficient and damping coefficient are key parameters that determine the dynamic response characteristics of the system in the control model.

[0066] In one possible implementation, the system's virtual resultant force is mapped as an adjustment to the inertia and damping coefficients. Specifically, when the virtual resultant force is large, it indicates a large trajectory deviation or proximity to the boundary. In this case, the system automatically increases the damping coefficient to suppress oscillations or adjusts the inertia coefficient to change the response speed, thereby achieving rapid response and correction to trajectory deviations. This adjustment mechanism allows the exoskeleton to dynamically adjust its impedance characteristics based on real-time motion, simulating the human body's ability to adjust under different loads.

[0067] Preferably, the inertia coefficient after the control model update Satisfy the following formula:

[0068] in, The initial inertia coefficient, This is the inertia adjustment coefficient. This is a virtual combined force.

[0069] Preferably, the damping coefficient after the control model update Satisfy the following formula:

[0070] in, The initial damping coefficient, This is the damping adjustment coefficient. This is a virtual combined force.

[0071] Based on the above steps, by dynamically adjusting the inertia coefficient and damping coefficient through virtual resultant force, the control model achieves real-time response to motion intentions and boundary constraints, thereby improving the compliance of human-computer interaction.

[0072] S304. Acquire exoskeleton plantar pressure sensor data, and adaptively adjust the stiffness coefficient and nonlinear damping coefficient of the control model based on the plantar pressure data to obtain the updated stiffness coefficient and nonlinear damping coefficient.

[0073] Among them, the plantar pressure sensor data reflects the contact force between the foot and the ground, and is the direct basis for determining the gait phase (support phase or swing phase). The stiffness coefficient determines the magnitude of the restoring force caused by position error, and the nonlinear damping coefficient determines the velocity-related damping characteristics.

[0074] In one possible implementation, the system collects foot pressure data via pressure sensors mounted on the sole of the exoskeleton. When foot pressure is high, it is identified as the support phase, where the target's leg bears weight, requiring a high stiffness coefficient to provide support stability and a high nonlinear damping coefficient to absorb ground impact. When foot pressure is low, it is identified as the swing phase, where the leg is suspended, requiring a lower stiffness coefficient and nonlinear damping coefficient to ensure flexibility and compliance during leg swing. The system adjusts these two coefficients in real time based on the foot pressure data to achieve adaptive control of the gait phase.

[0075] Preferred, updated stiffness coefficient Satisfy the following formula:

[0076] in, This is the stiffness adjustment coefficient. This is data from the exoskeleton foot pressure sensor.

[0077] Preferred, updated nonlinear damping coefficient Satisfy the following formula:

[0078] in, Nonlinear adjustment coefficient, Data from exoskeleton foot pressure sensors Based on the above steps, by adaptively adjusting the stiffness coefficient and nonlinear damping coefficient using foot pressure data, a precise match between the control strategy and the gait phase is achieved, balancing the impact resistance of the support phase with the compliance of the oscillating phase.

[0079] S305. Based on the updated control model with updated inertia coefficient, damping coefficient, stiffness coefficient and nonlinear damping coefficient, the updated joint trajectory of the gait training control output is obtained, driving the exoskeleton to perform active and passive gait training.

[0080] Among them, active-passive gait training refers to a rehabilitation training model that combines the active movement intention of the target subject with passive exoskeleton assistance.

[0081] In one possible implementation, the system substitutes the four updated parameters (inertia coefficient, damping coefficient, stiffness coefficient, and nonlinear damping coefficient) from the above steps into the control model, recalculates, and outputs the updated joint trajectory. This trajectory incorporates the target object's subjective motion intention (represented by human joint torques), and is subject to safety boundary constraints (represented by virtual resultant forces) and gait phase adaptation (represented by foot pressure adjustment), ultimately driving the exoskeleton to move the target object during training. This control method allows the target object to fully utilize its subjective initiative within a safe range, while the exoskeleton provides appropriate assistance or resistance based on the actual situation.

[0082] As an example, predefined gait training joint trajectories Define the coefficient of inertia To transfer the torque of human joints Angular acceleration is calculated as a controller input. According to angular acceleration The angular velocity is obtained by integration. and angle Then, define the damping coefficient. and nonlinear coefficients According to angular velocity Solving for the nonlinear damping moment Define the angle gain coefficient g, which is related to the angle. Perform multiplication to obtain the control angle output. Define stiffness coefficient According to the angle Solve for the elastic torque ; to nonlinear damping torque and elastic torque These feedbacks are respectively sent to the joint torque mechanism of the human body, forming a feedback closed loop, namely... Output the updated control model angle. The new gait training trajectory is obtained by summing the trajectory with the predefined gait training joint trajectory.

[0083] This application embodiment achieves dual adaptive adjustment of control model parameters for trajectory deviation and gait phase by constructing a virtual force field and introducing foot pressure feedback. It solves the problem that the control strategy in the prior art cannot adapt to gait phase changes, effectively balances the impact resistance of the support phase and the compliance of the swing phase, and greatly improves the safety and naturalness of human-computer interaction.

[0084] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 4 As shown, the above S202 can be implemented through the following S401, S402 and S403, which are explained in detail below: S401. Perform differential calculations on the position information under no-load conditions to obtain the joint angular velocity and joint angular acceleration.

[0085] Among them, position information refers to the angle data of the exoskeleton joints, which is usually collected by an encoder.

[0086] In one possible implementation, since the dynamic equations include velocity and acceleration terms, and the sensors typically collect position data directly, numerical differentiation processing of the position data is required. The system performs first-order differentiation calculations on the joint position data under no-load conditions to obtain the joint angular velocity, and then performs first-order differentiation calculations on the joint angular velocity to obtain the joint angular acceleration. It should be understood that the differentiation calculation can employ various numerical calculation methods such as the finite difference method and the filtered differentiation method; this embodiment does not specifically limit the method.

[0087] Based on the above steps, the joint angular velocity and angular acceleration were obtained through differential calculation, providing complete kinematic state variables for constructing the exoskeleton dynamics regression matrix.

[0088] S402. Construct an exoskeleton dynamics regression matrix based on joint position information, joint angular velocity, and joint angular acceleration under no-load conditions.

[0089] Among them, the exoskeleton dynamics regression matrix refers to the coefficient matrix extracted after linearizing the exoskeleton dynamics equation, which is related to the motion state and independent of the parameters to be identified.

[0090] In one possible implementation, the system is based on the Lagrange dynamics modeling method, establishing the dynamic equations of the exoskeleton and rewriting them in a linear regression form. Specifically, the exoskeleton joint moments are represented as the product of a regression matrix and a parameter vector. The elements in the regression matrix consist of nonlinear functions of joint position, angular velocity, and angular acceleration, reflecting the contribution structure of the exoskeleton's kinematic state to the joint moments.

[0091] Based on the above steps, by constructing an exoskeleton dynamics regression matrix, a linear mapping relationship between motion state and joint torque was established, providing a mathematical model basis for subsequent parameter identification.

[0092] S403. Based on the regression matrix and the joint torque data under no-load conditions, the least squares method is used to identify the exoskeleton dynamic parameters.

[0093] In one possible implementation, since the actual collected data often contains measurement noise and the system of equations is usually overdetermined (the number of equations exceeds the number of unknowns), direct solutions are often unsolvable or unstable. Therefore, the system uses the least squares method to find an optimal set of parameter vectors that minimizes the sum of squared errors between the torque values ​​predicted by the regression matrix and the actual measured torque values.

[0094] Preferred exoskeleton dynamics regression matrix Satisfy the following formula:

[0095] in, The joint torque of the exoskeleton under no-load conditions. This is the location data under no-load conditions. Angular velocity, Angular acceleration; The exoskeleton dynamics parameter matrix is ​​given by T, which represents the transpose. The results are obtained through Lagrangian dynamics modeling and linear decomposition. and :

[0096]

[0097] By conducting parameter identification experiments, the regression matrix was sampled. and sampling torque vector Finally, the least squares method is used to obtain the optimal estimated parameter vector of the exoskeleton. , Individually identify exoskeleton dynamic parameters .

[0098] It should be noted that the above formula directly solves for the optimal parameters through matrix operations, resulting in high computational efficiency and good noise resistance. The identified exoskeleton dynamic parameters accurately reflect the exoskeleton's own physical properties such as mass distribution and friction characteristics, providing precise model parameters for subsequent human-machine coupling dynamic decoupling.

[0099] Based on the above steps, the regression equation was solved by the least squares method, which enabled the accurate identification of the exoskeleton dynamic parameters and solved the problem that dynamic parameters are difficult to measure directly.

[0100] This application embodiment obtains the kinematic state by differential calculation of position information, constructs a regression matrix to establish a linear mapping relationship, and uses the least squares method to solve for the optimal parameters, thereby realizing the accurate identification of exoskeleton dynamic parameters. This provides an accurate model basis for subsequent human-machine coupling dynamic decoupling and ensures the accuracy of intent recognition.

[0101] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 5 As shown, the above S203 can be implemented through the following S501, S502 and S503, which are explained in detail below: S501. Filter the data segments with low activation of surface electromyography signals under the wearing state to obtain the first data segment.

[0102] Among them, the data segment with low activation of surface electromyography signals refers to the time period when the active contraction of the target object's lower limb muscles is weak or even completely relaxed. At this time, the human lower limb mainly exhibits passive dynamic characteristics.

[0103] In one possible implementation, the system monitors surface electromyography (EMG) signals collected in the wearable state in real time and calculates their amplitude or activation index. When this index is below a preset threshold, it is determined that the target object exerts relatively little active force during that time period, mainly relying on the exoskeleton to drive the lower limb movement. The system extracts the joint torque data and kinematic data corresponding to these time periods as the first data segment. It should be understood that the purpose of filtering the data segment is to eliminate the interference of active torque generated by the target object's subjective movement intention on parameter identification, thereby accurately identifying the inherent passive dynamic parameters of the human lower limb, such as limb mass and moment of inertia. If no filtering is performed and data containing active force is used directly, the identified parameters will contain active torque components and cannot reflect the true physical properties of the human limb.

[0104] It should be noted that the preset threshold can be adaptively adjusted according to the specific situation or recovery stage of the target object, and this embodiment does not impose specific limitations on it. The screening process can be a continuous time window or a discrete set of data points.

[0105] Based on the above steps, by filtering low electromyographic activation data segments, interference from active movement intentions is effectively eliminated, providing a clean data source for obtaining accurate human passive dynamic parameters.

[0106] S502. Based on the first data segment and exoskeleton dynamic parameters, construct a regression model of human lower limb dynamics.

[0107] Among them, the human lower limb dynamics regression model refers to a mathematical model that describes the relationship between the human lower limb motion state and joint torque. This model uses human lower limb dynamic parameters as variables to be identified.

[0108] In one possible implementation, the system utilizes the kinematic data (position, velocity, acceleration) from the first data segment, along with the corresponding joint torque data under the wearing state, and combines this with the identified exoskeleton dynamic parameters to construct a dynamic equation. Since the total joint torque under the wearing state equals the sum of the exoskeleton dynamic torque and the human lower limb dynamic torque, and given the exoskeleton dynamic parameters and motion state, the dynamic torque generated by the exoskeleton can be calculated. Furthermore, by subtracting the exoskeleton dynamic torque from the total torque, the remaining portion contains the human lower limb dynamic parameter information. The system expresses this remaining torque as a function of the human lower limb motion state, constructing a regression model.

[0109] It should be noted that the process of constructing a regression model is to transform the complex nonlinear dynamic equations into a linear regression form, which makes it easier to solve the problem using the least squares method.

[0110] Based on the above steps, a mathematical relationship between the human lower limb movement state and the parameters to be identified was established by constructing a regression model, providing a model framework for parameter identification.

[0111] S503. Based on the human lower limb dynamics regression model, the least squares method is used to identify the human lower limb dynamic parameters.

[0112] In one possible implementation, the system collects multiple sets of kinematic and torque data within the first data segment, constructs an overdetermined set of equations, and uses the least squares method to solve for the optimal estimates of the dynamic parameters of the human lower limbs.

[0113] Preferably, the human lower limb dynamics model satisfies the following formula:

[0114] in, The joint torque of the exoskeleton when worn. The joint torque of the exoskeleton under no-load conditions. For the resistance of human joints, This is the location data under no-load conditions. Angular velocity, Angular acceleration, To identify the regression matrix for lower limb dynamics, The lower limb dynamics regression matrix;

[0115]

[0116] By conducting parameter identification experiments, the regression matrix was sampled. and sampling torque vector Finally, the least squares method is used to obtain the optimal estimated parameter vector for the human lower limb. , Identifying the dynamic parameters of the human lower limbs .

[0117] It should be noted that the regression matrix in the above formula contains a nonlinear combination of the kinematic states of the human lower limbs, while the parameter vector contains physical properties such as the mass, moment of inertia, and center of mass position of the human lower limbs. Using this formula, the system successfully separates the dynamic parameters of the human lower limbs from the human-machine coupling system.

[0118] Based on the above steps, the regression equation was solved by the least squares method, which enabled the accurate identification of the dynamic parameters of the human lower limbs. This fully considered the individual biomechanical differences of the target object and provided personalized model parameter support for the subsequent accurate calculation of human joint torques.

[0119] This application embodiment eliminates active interference by screening low electromyographic activation data segments, constructs a regression model, and solves it using the least squares method, thereby achieving accurate identification of human lower limb dynamic parameters. This solves the problem of existing technologies not adapting to individual biomechanical differences and provides an accurate model basis for personalized rehabilitation training.

[0120] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 6 As shown, the above S204 can be specifically implemented through the following S601 to S605, which are explained in detail below: S601. Based on exoskeleton motion data, construct a joint dynamics model of the human-machine coupling system.

[0121] Among them, the joint dynamics model of the human-machine coupling system refers to the mathematical model that treats the exoskeleton and the wearer's lower limbs as a whole system for dynamic description.

[0122] In one possible implementation, when the target subject wears the exoskeleton for training, the torque output by the exoskeleton joint motors needs to drive not only the movement of the exoskeleton itself but also the movement of the target subject's lower limbs. Therefore, the joint torque data collected by the sensors is a coupled superposition of the exoskeleton's dynamic torque and the human lower limb's dynamic torque. The system first establishes equations describing the overall system dynamics based on the exoskeleton motion data (including joint position, velocity, and acceleration).

[0123] Based on the above steps, by constructing a human-machine coupled joint dynamics model, a complete system description framework is provided for separating human intention from the total torque.

[0124] S602. Based on the exoskeleton dynamic parameters, model the exoskeleton dynamics part in the joint dynamics model.

[0125] The exoskeleton dynamics part refers to the sub-item in the joint dynamics model that describes the dynamic behavior of the exoskeleton itself.

[0126] In one possible implementation, the system substitutes the identified exoskeleton dynamic parameters (such as exoskeleton link mass and moment of inertia) into the corresponding exoskeleton terms in the joint dynamics model. This step essentially uses known physical parameters to calculate the torque component required for the exoskeleton's own motion under the current motion state. Specifically, this torque includes the exoskeleton's gravitational torque, inertial torque, and frictional torque. Through this modeling process, the system clarifies the dynamic burden of the exoskeleton itself within the total torque.

[0127] It should be noted that the accuracy of the exoskeleton dynamic parameters directly affects the accuracy of this part of the modeling, and thus affects the accuracy of the subsequent human body torque calculation.

[0128] Based on the above steps, by modeling the exoskeleton dynamics, a quantitative description of the exoskeleton component in the total torque is achieved, providing a computational basis for subsequent stripping.

[0129] S603. Based on the dynamic parameters of the human lower limbs, model the dynamic part of the human lower limbs in the joint dynamic model.

[0130] The lower limb dynamics component refers to the sub-item in the joint dynamics model that describes the dynamic behavior of the target object's lower limbs.

[0131] In one possible implementation, the system substitutes the identified lower limb dynamic parameters into the corresponding human component in the joint dynamics model. This step aims to calculate the torque component required to drive the target object's lower limb movement under the current motion state. This torque includes the gravitational torque and inertial torque of the lower limbs. Since the lower limb parameters are individualized, this modeling process fully considers the differences in the physical properties of different target objects' limbs.

[0132] Based on the above steps, by modeling the dynamics of the human lower limbs, a quantitative description of the passive dynamic components of the human lower limbs is achieved.

[0133] S604. Perform human-machine dynamics decoupling processing on the joint dynamics model to obtain the human dynamics sub-model.

[0134] Among them, human-machine dynamics decoupling processing refers to the mathematical calculation process of separating the human active torque term from the overall dynamic equation.

[0135] In one possible implementation, the system, based on an established joint model, subtracts the torque calculated from the exoskeleton dynamics and the passive torque calculated from the lower limb dynamics from the total joint torque collected by sensors. The remaining portion is the active joint torque of the human body. Specifically, the total torque consists of three parts: exoskeleton dynamic torque, lower limb passive dynamic torque, and active joint torque. Since the first two parts can be calculated through the model, the active joint torque can be separated through decoupling operations.

[0136] Based on the above steps, through human-machine dynamics decoupling processing, the main torque component reflecting the human body's movement intention was successfully separated from the total coupled torque, thus solving the technical problem of difficulty in intention perception in human-computer interaction.

[0137] S605. Based on the human body dynamics sub-model, construct the inverse dynamics equation and solve for the human body joint torque.

[0138] Among them, the inverse dynamics equation refers to the mathematical equation that maps motion state to joint torque based on the human body dynamics sub-model.

[0139] In one possible implementation, the system transforms the decoupled human dynamics sub-model into an inverse dynamics form and substitutes the real-time collected motion data (position, velocity, acceleration) into the equation to finally obtain the numerical value of the human joint torque. This human joint torque value directly reflects the intensity and direction of the target object's current motion intention; for example, a positive value may represent a flexion intention, and a negative value may represent an extension intention, with the magnitude of the value representing the strength of the intention.

[0140] As an example, during gait training, the joint torques of the exoskeleton in the worn state are considered. Based on inverse dynamics, the exoskeleton joint dynamics term is calculated. This includes gravitational torque, frictional torque, and inertial torque, namely: Based on inverse dynamics, the dynamics of the human lower limb joints are calculated. This includes gravitational torque and inertial torque, namely: Calculate the difference between the joint torques of the worn exoskeleton and the exoskeleton dynamics and the human lower limb dynamics, i.e. Based on gait phase division, the joint torques of the lower limbs during the swing phase are obtained. .

[0141] Based on the above steps, by constructing and solving the inverse dynamic equation, the quantitative calculation of human movement intention is realized, providing accurate input signals for subsequent active-passive integrated control.

[0142] This application embodiment constructs a human-machine coupled joint dynamic model, substitutes exoskeleton and human body parameters for modeling, and removes the passive dynamic influence of exoskeleton and human body through decoupling processing. Finally, the human body joint torque is obtained, realizing the accurate extraction of human body active intention from the total torque. This solves the problem of inaccurate intention perception in the prior art and lays the foundation for highly responsive human-machine interaction control.

[0143] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 7 As shown, the above S205 can be specifically implemented through the following S701 to S704, which are explained in detail below: S701. Using human joint torque as the driving input and predefined gait training joint trajectory as the reference input, construct the joint trajectory tracking error.

[0144] Among them, the driving input refers to the main excitation source of the control system, the reference input refers to the target trajectory that the system expects to track, and the joint trajectory tracking error refers to the deviation between the actual trajectory and the target trajectory.

[0145] In one possible implementation, the control model employs an impedance-based control architecture. The system's joint torques serve as input to the drive controller, meaning that the greater the intended torque of the target object, the stronger the controller's response. Simultaneously, the system uses a predefined standard gait trajectory as a reference input, providing a basic framework and guidance for the movement. The system calculates the deviation between the current actual trajectory and the predefined trajectory in real time, constructing a tracking error signal. It should be understood that this error signal includes not only positional error but also velocity and acceleration errors to comprehensively reflect the deviations in the motion state.

[0146] Based on the above steps, by constructing an intent-driven input architecture, the basic logic of the control model responding to the subjective intentions of the target object is established.

[0147] S702. Based on the joint trajectory tracking error and its derivative, inertial response term, damping response term, nonlinear damping response term and stiffness response term are constructed respectively.

[0148] Among them, the inertial response term simulates the inertial characteristics of limb movement, the damping response term simulates the viscous resistance characteristics, the nonlinear damping response term simulates the velocity-related nonlinear resistance characteristics, and the stiffness response term simulates the elastic restoring force characteristics.

[0149] In one possible implementation, the system constructs four-dimensional response terms based on the dynamic characteristics of human joints. The inertial response term is related to angular acceleration, simulating the inertial resistance when a limb starts or stops, giving the exoskeleton's movement a sense of inertia and avoiding discomfort caused by sudden acceleration. The damping response term is related to angular velocity, simulating the viscous resistance of joint movement to smooth the movement process and eliminate oscillations. The nonlinear damping response term is also related to angular velocity, but its damping coefficient changes with velocity; for example, it provides greater damping at high speeds to prevent overshoot and less damping at low speeds to ensure flexibility. The stiffness response term is related to position error, simulating the restoring force of a spring to pull the trajectory back to a predefined path, ensuring the standardization of training.

[0150] Based on the above steps, by constructing four response terms, the impedance characteristics of human joints are fully simulated, making the control output of the exoskeleton more in line with the laws of human biomechanics and improving the compliance and naturalness of human-computer interaction.

[0151] S703. The inertial response term, damping response term, nonlinear damping response term, and stiffness response term are weighted and fused to construct the joint control quantity.

[0152] Weighted fusion refers to the process of superimposing the values ​​of four response items according to preset weighting coefficients or adjustment mechanisms to obtain the final control command.

[0153] In one possible implementation, the system superimposes the calculation results of the four response terms to form a total joint control quantity. Specifically, the coefficient (i.e., weight) before each term can be dynamically adjusted according to the target object's rehabilitation stage or movement state. For example, for a target object with weak muscle strength, the weight of the stiffness term can be increased to provide stronger guidance; for a target object with good muscle strength, the weight of the stiffness term can be decreased and the weight of the damping term increased to encourage active movement. It should be understood that the weighted fusion process essentially integrates multi-dimensional dynamic responses into a unified control command, which comprehensively reflects the influence of inertia, damping, nonlinear damping, and stiffness on motion.

[0154] It should be noted that this fusion method is not a simple linear superposition; there are complex dynamic coupling relationships between the response terms, which together determine the dynamic performance of the system.

[0155] Based on the above steps, by weighted fusion of the four response terms, comprehensive regulation of multi-dimensional dynamic characteristics is achieved, enabling the generated control quantity to accurately match the target object's movement intentions and rehabilitation needs.

[0156] S704. Based on joint control quantities, the predefined gait training joint trajectory is modified to obtain the joint trajectory of the gait training control output.

[0157] Among them, correction refers to superimposing or applying the calculated joint control values ​​onto a predefined trajectory to generate a new motion trajectory.

[0158] In one possible implementation, the system's joint control variables act on predefined gait training joint trajectories to generate the final gait training control output trajectory. Specifically, the joint control variables can be considered as trajectory correction variables, superimposed on the predefined trajectory. This output trajectory retains the basic gait framework of the predefined trajectory while incorporating the target object's movement intentions, achieving an organic combination of passive guidance and active participation. It should be understood that this trajectory is updated in real time, dynamically adjusting as the target object's intentions change.

[0159] It should be noted that this correction mechanism ensures the safety of training because the correction amount is constrained by various impedance parameters, and will not cause drastic changes in the trajectory or exceed the safe range.

[0160] Based on the above steps, the predefined trajectory is modified by joint control, which realizes real-time response to the target object's movement intention while ensuring the basic gait framework, thus achieving the effect of active and passive integrated training.

[0161] This application embodiment simulates the impedance characteristics of human joints by constructing a control model that includes four response terms: inertia, damping, nonlinear damping, and stiffness. It uses the torque of human joints as the driving input to achieve dynamic correction of predefined trajectories, enabling the exoskeleton to respond smoothly to the movement intentions of the target object like a real limb, effectively improving the naturalness of training and the target object's sense of participation.

[0162] As an example, the initial parameters and coefficient settings of the exoskeleton control model of this application are shown in Table 1 below.

[0163] Table 1 Initial parameters and coefficient settings for the exoskeleton control model

[0164] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0165] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A control method for a lower limb exoskeleton robot based on motion intention recognition, characterized in that, include: Acquire exoskeleton motion data and exoskeleton unloaded data; The exoskeleton motion data includes joint torque data, position information, and surface electromyography signals in the worn state; The exoskeleton unloaded data includes the joint torques and position data of the exoskeleton in the unloaded state; Based on the exoskeleton's unloaded data, gait training is used to determine the exoskeleton's dynamic parameters. Based on the exoskeleton motion data and exoskeleton dynamic parameters, the dynamic parameters of the human lower limbs are determined; Based on the exoskeleton motion data, the exoskeleton dynamic parameters, and the human lower limb dynamic parameters, the human joint torque is calculated by inverse dynamic equations. The human joint torque and the predefined gait training joint trajectory are input into the control model to obtain the joint trajectory of the gait training control output. The control model is constructed based on inertia, damping, nonlinear damping and stiffness terms, and is used to dynamically adjust the predefined gait training joint trajectory under the drive of human movement intention.

2. The method according to claim 1, characterized in that, After obtaining the gait training control output, the method further includes: Calculate the deviation between the joint trajectory output of gait training control and the predefined gait training joint trajectory; Based on the distance between the joint trajectory output by the gait training control and the preset boundary, virtual attraction and virtual repulsion are constructed, and virtual resultant force is determined; The inertia coefficient and damping coefficient of the control model are adjusted based on the virtual resultant force to obtain the updated inertia coefficient and damping coefficient. Data from the plantar pressure sensor of the exoskeleton is acquired, and the stiffness coefficient and nonlinear damping coefficient of the control model are adaptively adjusted based on the plantar pressure data to obtain the updated stiffness coefficient and nonlinear damping coefficient. Based on the updated control model with updated inertia coefficient, damping coefficient, stiffness coefficient, and nonlinear damping coefficient, the updated joint trajectory of the gait training control output is obtained, driving the exoskeleton to perform active and passive gait training.

3. The method according to claim 1, characterized in that, The process of determining exoskeleton dynamic parameters through gait training based on the exoskeleton's unloaded data includes: The position information under no-load conditions is differentiated to obtain the joint angular velocity and joint angular acceleration. An exoskeleton dynamics regression matrix is ​​constructed based on the joint position information, joint angular velocity, and joint angular acceleration under the unloaded state. Based on the regression matrix and the joint torque data under no-load conditions, the least squares method is used to identify the exoskeleton dynamic parameters.

4. The method according to claim 3, characterized in that, The exoskeleton dynamics regression matrix Satisfy the following formula: in, The joint torque of the exoskeleton under no-load conditions. This is the location data under no-load conditions. Angular velocity, Angular acceleration; The exoskeleton dynamics parameter matrix is ​​given by T, which represents the transpose. The results are obtained through Lagrangian dynamics modeling and linear decomposition. and : By conducting parameter identification experiments, the regression matrix was sampled. and sampling torque vector Finally, the least squares method is used to obtain the optimal estimated parameter vector of the exoskeleton. , Individually identify exoskeleton dynamic parameters .

5. The method according to claim 1, characterized in that, The determination of human lower limb dynamic parameters based on the exoskeleton motion data and exoskeleton dynamic parameters includes: The first data segment was obtained by filtering out data segments with low activation levels of surface electromyography signals under wear conditions. Based on the first data segment and the exoskeleton dynamic parameters, a regression model of human lower limb dynamics is constructed. Based on the aforementioned human lower limb dynamics regression model, the least squares method is used to identify the human lower limb dynamic parameters.

6. The method according to claim 5, characterized in that, The human lower limb dynamics model satisfies the following formula: in, The joint torque of the exoskeleton when worn. The joint torque of the exoskeleton under no-load conditions. For the resistance of human joints, This is the location data under no-load conditions. Angular velocity, Angular acceleration, To identify the regression matrix for lower limb dynamics, The lower limb dynamics regression matrix; By conducting parameter identification experiments, the regression matrix was sampled. and sampling torque vector Finally, the least squares method is used to obtain the optimal estimated parameter vector for the human lower limb. , Identifying the dynamic parameters of the human lower limbs .

7. The method according to claim 1, characterized in that, The calculation of human joint torques based on the exoskeleton motion data, the exoskeleton dynamic parameters, and the human lower limb dynamic parameters, using inverse dynamic equations, includes: Based on exoskeleton motion data, a joint dynamics model of the human-machine coupling system is constructed. Based on the exoskeleton dynamic parameters, the exoskeleton dynamics component in the joint dynamics model is modeled. Based on the dynamic parameters of the human lower limbs, the dynamic component of the human lower limbs in the joint dynamic model is modeled. The joint dynamics model is subjected to human-machine dynamics decoupling processing to obtain a human dynamics sub-model; Based on the aforementioned human dynamics sub-model, inverse dynamic equations are constructed and solved to obtain the joint torques of the human body.

8. The method according to claim 1, characterized in that, The step of inputting the human joint torques and predefined gait training joint trajectories into the control model to obtain the joint trajectories output by the gait training control includes: Using human joint torque as the driving input and predefined gait training joint trajectories as the reference input, a joint trajectory tracking error is constructed. Based on the joint trajectory tracking error and its derivative, inertial response terms, damping response terms, nonlinear damping response terms, and stiffness response terms are constructed respectively. The inertial response term, damping response term, nonlinear damping response term, and stiffness response term are weighted and fused to construct the joint control quantity; Based on the joint control values, the predefined gait training joint trajectory is modified to obtain the joint trajectory of the gait training control output.

9. The method according to claim 2, characterized in that, The virtual combined force Satisfy the following formula: in, Represents virtual gravity. , The current trajectory is used for gait training; This refers to a predefined trajectory, i.e., location data under no-load conditions; Indicates distance, This is the gravitational gain coefficient; Represents virtual repulsion. , The repulsive force gain coefficient, For boundary thresholds; This refers to a predefined trajectory, i.e., location data under no-load conditions; The upper boundary, This is the lower boundary.

10. A control system for a lower limb exoskeleton robot based on motion intention recognition, characterized in that, include: The system includes a data acquisition module, an exoskeleton dynamics parameter determination module, a human lower limb dynamics parameter determination module, a human joint torque calculation module, and a control output generation module. The data acquisition module is used to acquire exoskeleton motion data and exoskeleton unloaded data; the exoskeleton motion data includes joint torque data, position information, and surface electromyography signals in the worn state; the exoskeleton unloaded data includes exoskeleton joint torque and position data in the unloaded state. The exoskeleton dynamics parameter determination module is used to determine the exoskeleton dynamics parameters based on the exoskeleton unloaded data and through gait training. The human lower limb dynamics parameter determination module is used to determine the human lower limb dynamics parameters based on the exoskeleton motion data and exoskeleton dynamics parameters. The human joint torque calculation module is used to calculate the human joint torque based on the exoskeleton motion data, the exoskeleton dynamic parameters, and the human lower limb dynamic parameters by using inverse dynamic equations. The control output generation module is used to input the human joint torque and the predefined gait training joint trajectory into the control model to obtain the joint trajectory of the gait training control output; the control model is constructed based on inertia, damping, nonlinear damping and stiffness terms, and is used to dynamically adjust the predefined gait training joint trajectory under the drive of human movement intention.

Images