Wearable exoskeleton motion control system based on AI man-machine cooperation

Abstract

The invention belongs to the technical field of intelligent control, and provides a wearable exoskeleton motion control system based on AI man-machine cooperation, and the system comprises a sensing unit which is used for collecting the multi-dimensional actual measurement data of joints, muscles, limbs and wearable exoskeletons when a human body moves; the AI processing unit is used for constructing a general model meeting general exercise requirements and a difference model meeting personalized exercise requirements, fusing the general model and the difference model to obtain a human body exercise model, and comparing and analyzing the multi-dimensional measured data and multi-dimensional reference data in the human body exercise model to obtain a human body exercise model; generating an adjustment instruction about the wearable exoskeleton; and the motion control unit is used for establishing a man-machine cooperative motion model and controlling the wearable exoskeleton to match human body motion according to the adjustment instruction through the man-machine cooperative motion model. According to the scheme, the wearable exoskeleton can be controlled more accurately and flexibly, and the reliability of a control link is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a wearable exoskeleton motion control system based on AI human-machine collaboration. Background Technology

[0002] With the rapid development of modern technology, innovative achievements are constantly emerging in the field of wearable devices, among which the development of wearable exoskeleton technology has attracted much attention. Wearable exoskeletons, as wearable robotic devices worn on the outside of the human body, are designed to work closely with the body to enhance motor function. In the field of medical rehabilitation, lower limb wearable exoskeletons can assist elderly people with reduced mobility and patients with limb disabilities in restoring their behavioral abilities or conducting rehabilitation training. They can also improve the strength and endurance levels of able-bodied individuals, thereby increasing work efficiency.

[0003] However, current motion control strategies for wearable exoskeletons typically employ a single sensor device, which can only capture limited motion or environmental information, making it difficult to meet the real-time control requirements in complex dynamic environments and resulting in insufficient control precision. At the same time, traditional control strategies lack the ability to adapt to individual differences, making it difficult to meet users' personalized needs.

[0004] Therefore, traditional wearable exoskeleton motion control strategies suffer from technical problems such as low control precision and insufficient control flexibility. Summary of the Invention

[0005] This invention provides a wearable exoskeleton motion control system based on AI human-machine collaboration, which solves the shortcomings of traditional wearable exoskeleton motion control strategies, such as low control accuracy and insufficient control flexibility.

[0006] On one hand, the present invention provides a wearable exoskeleton motion control system based on AI human-machine collaboration, comprising: The sensing unit is used to collect multi-dimensional measured data of joints, muscles, limbs, and wearable exoskeletons during human movement. The AI ​​processing unit is used to construct a general model that meets general movement needs and a differential model that meets personalized movement needs, and to fuse the general model and the differential model to obtain a human movement model. The unit compares and analyzes the multidimensional measured data with the multidimensional benchmark data in the human movement model to generate adjustment instructions for the wearable exoskeleton. The motion control unit is used to establish a human-machine collaborative motion model, and through the human-machine collaborative motion model, to control the wearable exoskeleton to match human movement according to the adjustment instructions.

[0007] According to the wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, the sensing unit includes: Joint position sensor, used to collect position data of joint movement; Force sensors are used to collect force data applied to wearable exoskeletons; Electromyography (EMG) sensors are used to collect electromyographic signals generated by muscle movement. An accelerometer is used to collect acceleration data of limb movements.

[0008] According to the wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, the AI ​​processing unit constructs a general model that meets general motion requirements and a differentiated model that meets personalized motion requirements, including: Obtain multidimensional sample data; The multidimensional sample data is subjected to denoising and normalization to obtain preprocessed sample data; Extract key features from the preprocessed sample data, perform motion pattern recognition based on the key features, and construct a sample dataset. A general model is obtained by training the basic model using the sample dataset. Individual difference data are obtained during the training process using a wearable exoskeleton. Based on the individual difference data, the general model is adjusted for differences to obtain a difference model.

[0009] According to the wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, the AI ​​processing unit fuses the general model and the differential model to obtain a human motion model, including: Extract the common parameters of human motion contained in the general model, and construct the motion infrastructure based on the common parameters; Extract the individual-specific parameters contained in the difference model, and embed the individual-specific parameters into the motion infrastructure to obtain the human motion model.

[0010] The wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, after obtaining the human motion model, further includes: The system acquires real-time updated individual motion data and dynamically optimizes the model parameters in the human motion model based on this data.

[0011] According to the wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, the AI ​​processing unit compares and analyzes the multidimensional measured data with the multidimensional benchmark data in the human motion model to generate adjustment instructions for the wearable exoskeleton, including: Align the multidimensional measured data with the multidimensional benchmark data in terms of dimensions; For each dimension, the deviation between the measured value in the multidimensional measured data and the benchmark value in the multidimensional benchmark data is calculated to obtain the data deviation value for each dimension; The data deviation value under each dimension is compared with the corresponding preset deviation threshold to obtain the comparison result; Based on the comparison results, adjustment instructions for the wearable exoskeleton are generated.

[0012] According to the AI-based human-machine collaborative wearable exoskeleton motion control system provided by the present invention, based on the comparison results, adjustment instructions for the wearable exoskeleton are generated, including: Determine the target dimension in the comparison results where the data deviation value exceeds the corresponding preset deviation threshold. Based on the target dimension and the corresponding data deviation value, retrieve the deviation root causes from the pre-built root cause database; The adjustment strategy corresponding to the root cause of the deviation is determined, and the adjustment strategy is encoded into instructions to generate adjustment instructions for the wearable exoskeleton.

[0013] According to the wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, the motion control unit establishes a human-machine collaborative motion model, including: Acquire human motion parameters and exoskeleton structural parameters; Based on the human motion parameters and the exoskeleton parameters, human dynamics modeling, exoskeleton dynamics modeling and human-computer interaction force modeling are performed respectively to obtain kinematic model, dynamic model and human-computer interaction model; Extract typical motion features from the human motion parameters and establish a baseline motion pattern; The human motion parameters of different individuals are classified, the baseline motion pattern is adjusted according to the classification results, a personalized baseline pattern is established, and a baseline pattern library is generated. By integrating the kinematic model, the dynamic model, the human-computer interaction model, and the baseline pattern library, a human-computer cooperative motion model is obtained.

[0014] The wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, after obtaining the human-machine collaborative motion model, further includes: The exoskeleton mechanical limits and human safety were verified on the human-machine collaborative motion model, and the verification results were obtained. When the verification results show that both the mechanical limit verification and human safety verification of the exoskeleton are passed, a human-machine collaborative motion model that meets the application conditions is obtained.

[0015] According to the wearable exoskeleton motion control system based on AI human-machine collaboration provided by the present invention, the system further includes: The human-computer interaction unit is used to generate control commands for the wearable exoskeleton based on the target training mode selected in the mode switching command after receiving a mode switching command initiated by the user. The motion control unit is also used to control the wearable exoskeleton to match human movement according to the control instructions.

[0016] This invention provides an AI-based human-machine collaborative wearable exoskeleton motion control system. It collects multi-dimensional measured data of joints, muscles, limbs, and the wearable exoskeleton during human movement through a sensing unit. An AI processing unit constructs a general model to meet general movement needs and a differential model to meet personalized movement needs. The general model and the differential model are then fused to obtain a human movement model. The multi-dimensional measured data is compared and analyzed with multi-dimensional benchmark data in the human movement model to generate adjustment commands for the wearable exoskeleton. A human-machine collaborative motion model is established through a motion control unit. Using this model, the wearable exoskeleton is controlled to match human movement according to the adjustment commands. This solution, based on multi-dimensional measured data, can more accurately capture user control needs. Combined with a human movement model incorporating personalized movement needs, it enables more precise and flexible control of the wearable exoskeleton, improving the reliability of the control process. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of the wearable exoskeleton motion control system based on AI human-machine collaboration provided in an embodiment of the present invention; Figure 2 This is a partial schematic diagram of the control process in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] The following is combined Figure 1 and Figure 2This invention describes the detailed solution of a wearable exoskeleton motion control system based on AI human-machine collaboration provided in an embodiment of the present invention.

[0021] like Figure 1 As shown, the wearable exoskeleton motion control system based on AI human-machine collaboration provided in this embodiment of the invention mainly includes: The sensing unit 110 is used to collect multi-dimensional measured data of joints, muscles, limbs and wearable exoskeletons during human movement.

[0022] In this embodiment, the sensing unit 110 integrates multiple types of sensors, which can collect data from multiple dimensions.

[0023] AI processing unit 120 is used to construct a general model that meets general movement needs and a differential model that meets personalized movement needs, and to fuse the general model and the differential model to obtain a human movement model. It compares and analyzes multidimensional measured data with multidimensional benchmark data in the human movement model to generate adjustment instructions for the wearable exoskeleton.

[0024] Understandably, the AI ​​processing unit 120 has model training and data analysis capabilities, enabling the automatic generation of modeling and adjustment instructions for human motion models. In practical applications, the AI ​​processing unit 120 can be implemented using a DSP (Digital Signal Processor).

[0025] The motion control unit 130 is used to establish a human-machine collaborative motion model, and through the human-machine collaborative motion model, to control the wearable exoskeleton to match human movement according to the adjustment instructions.

[0026] Understandably, the motion control unit 130 can control the wearable exoskeleton to execute adjustment commands to match human movement. In practical applications, the motion control unit 130 can be implemented using an FPGA (Field Programmable Gate Array).

[0027] Specifically, when a user walks using a wearable exoskeleton, the sensing unit can simultaneously collect multi-dimensional measured data. The AI ​​processing unit can input the multi-dimensional measured data into a human motion model containing multi-dimensional benchmark data for comparison and analysis of multiple parameters, thereby generating adjustment instructions for the wearable exoskeleton. The motion control unit can adjust the output torque curve of the knee joint drive motor on the wearable exoskeleton according to the adjustment instructions, so that the assist phase of the wearable exoskeleton is synchronized with the user's movement intention.

[0028] In one embodiment, such as Figure 2 As shown, the sensing unit specifically includes: The joint position sensor 210 is used to collect position data of joint movement.

[0029] Force sensor 220 is used to collect force data applied to the wearable exoskeleton.

[0030] Electromyography sensor 230 is used to collect electromyographic signals generated by muscle movement.

[0031] Accelerometer 240 is used to collect acceleration data of limb movements.

[0032] In other words, the multidimensional measured data specifically includes: position data, force data, electromyographic signals, and acceleration data.

[0033] In this embodiment, the joint position sensor 210 can accurately measure the angle changes of human joints (such as hip joints, knee joints, ankle joints, etc.), obtain the position information of joint movement, and provide basic data for analyzing human movement posture and exoskeleton joint coordination.

[0034] Force sensor 220 can monitor the forces applied to the exoskeleton, including the interaction forces between the human body and the exoskeleton, and the forces exerted by the exoskeleton on the external environment, in order to determine the human-machine collaboration status and the rationality of the forces applied to the exoskeleton.

[0035] The electromyography sensor 230 can collect electromyographic signals generated by muscle activity, reflecting the activation state and force exertion of muscles, and helping to determine the human body's movement intention, such as when the muscle contracts and how much force is exerted, thus providing a basis for the exoskeleton to respond to human movements.

[0036] The accelerometer 240 can detect the acceleration of various parts of the body (such as the trunk and limbs), and analyze the speed, direction, and movement pattern of human movement (such as the distinction between walking, running, and jumping) by combining the changes in acceleration, thus supplementing the information on the movement status.

[0037] In one embodiment, the AI ​​processing unit constructs a general model that meets general movement requirements and a differentiated model that meets individual movement requirements, specifically including: The first step is to obtain multidimensional sample data.

[0038] Understandably, the multidimensional sample data also includes real data from multiple dimensions, such as historically collected position data, force data, electromyographic signals, and acceleration data. This multidimensional sample data can also be collected by various sensors in the aforementioned sensing units, forming AI big data, which can be used to guide the training of differential and general models.

[0039] The second step is to perform noise reduction and normalization on the multidimensional sample data to obtain preprocessed sample data.

[0040] Denoising refers to eliminating high-frequency interference and baseline drift in sensor signals. This can be achieved using wavelet thresholding algorithms to improve data quality. Normalization maps sensor data with different dimensions to a unified numerical range. This can be achieved using maximum-minimum standardization methods to eliminate the interference of dimension differences on model training.

[0041] The third step is to extract key features from the preprocessed sample data, perform motion pattern recognition based on these key features, and construct a sample dataset.

[0042] In practical applications, key features such as gait cycle, stride length, joint motion angular velocity, and muscle activation sequence can be extracted. The movement patterns can be basic movement patterns, dynamic adjustment patterns, and physiological feature patterns.

[0043] The fourth step is to train the basic model using the sample dataset to obtain a general model.

[0044] Understandably, the general model covers typical parameters of normal human movement, joint-muscle coordination mechanisms, etc., and is used to initially match the user's movement state.

[0045] The fifth step is to obtain individual difference data during the training process using the wearable exoskeleton, and then adjust the general model based on the individual difference data to obtain the difference model.

[0046] In this embodiment, individual difference data refers to exoskeleton response data that reflects the user's gait characteristics and muscle force distribution. Specifically, it can be achieved by continuously collecting joint torque deviation and gait cycle differences during the interaction between the user and the exoskeleton, which is used to capture personalized movement characteristics.

[0047] The differential model then makes personalized adjustments based on individual characteristics, allowing wearable exoskeletons to adapt to the movement characteristics of different users.

[0048] In one embodiment, the AI ​​processing unit fuses a general model and a differential model to obtain a human motion model, specifically including: First, extract the common parameters of human motion contained in the general model, and then construct the motion infrastructure based on the common parameters.

[0049] In this embodiment, the trained general model is first invoked to extract common parameters of human movement, such as joint angle change curves of typical gait cycles, general laws of muscle force exertion timing, and physical constraints of limb movement, to form the basic framework of movement and ensure that the human movement model conforms to the basic physiological and physical logic of human movement.

[0050] Then, the individual-specific parameters contained in the difference model are extracted and embedded into the motion infrastructure to obtain the human motion model.

[0051] Understandably, the individual-specific parameters recorded in the differential model can include: height, weight, upper limit of joint range of motion, muscle strength threshold, habitual movement deviation, etc., which can serve as key variables for modifying the general model. For example, the normal range of motion of the knee joint in the general model is 0-120°, while the differential model may show that a certain individual can only reach 0-90° due to injury, and this can be used as a personalized constraint.

[0052] In practical applications, individual-specific parameters can be embedded into the framework of a general model using algorithms, while common parameters can be adjusted in a targeted manner, as follows: For movement timing, such as the ratio of heel strike to toe lift-off time in gait, the default values ​​in the general model can be corrected using individual habit data.

[0053] For the range of motion of a joint, the general range in the general model can be replaced with the individual's actual range of motion.

[0054] Regarding the intensity of force exertion, the force feedback coefficient in the general model can be adjusted by combining individual muscle strength data.

[0055] Ultimately, a human movement model can be formed that conforms to the general laws of human movement and is adapted to individual characteristics.

[0056] In some embodiments, after obtaining the human motion model, the following may also be included: It acquires real-time updated individual motion data and dynamically optimizes the model parameters in the human motion model based on this data.

[0057] In practical applications, such as wearable exoskeletons assisting in movement, the human motion model continuously receives individual motion data fed back from sensors, constantly compares the deviation between the current model output and the actual movement, and then dynamically fine-tunes the model parameters to ensure that the human motion model always maintains a high degree of matching with the individual's state.

[0058] In one embodiment, the AI ​​processing unit compares and analyzes the multidimensional measured data with the multidimensional benchmark data in the human motion model to generate adjustment instructions for the wearable exoskeleton, specifically including: The first step is to align the dimensions of the multidimensional measured data with the multidimensional benchmark data.

[0059] In this embodiment, multiple dimensions are involved, including joint motion dimension, biomechanical feature dimension, and temporal feature dimension. In the joint motion dimension, the maximum flexion angle of the knee joint and the peak rate of angle change can be calculated during the gait cycle. In the biomechanical feature dimension, the peak impact force at the moment of heel strike and the duration of quadriceps muscle exertion can be extracted. In the temporal feature dimension, the average stride length and stride frequency fluctuation coefficient of 10 consecutive steps can be statistically analyzed.

[0060] The second step is to calculate the deviation between the measured values ​​in the multidimensional measured data and the benchmark values ​​in the multidimensional benchmark data for each dimension, so as to obtain the data deviation value for each dimension.

[0061] In practical applications, the deviation between the measured value and the corresponding benchmark value can be calculated to achieve deviation calculation. For example, if the measured maximum angle of the knee joint is 100° and the corresponding upper limit benchmark value is 90°, then the data deviation value is +10°.

[0062] Alternatively, deviation can be calculated by analyzing the trend of the deviation between the measured value and the corresponding benchmark value. For example, if the measured step frequency is consistently lower than the benchmark range, such as a benchmark value of 80 steps / minute and a measured value of 60 steps / minute, and the deviation gradually increases.

[0063] Alternatively, deviations can be calculated by analyzing the timing discrepancies between measured values ​​and corresponding baseline values. For example, if the toe lift-off time in gait is 0.2 seconds earlier than the baseline timing, it can lead to abnormal gait coordination.

[0064] The third step is to compare the data deviation value under each dimension with the corresponding preset deviation threshold to obtain the comparison result.

[0065] The fourth step is to generate adjustment instructions for the wearable exoskeleton based on the comparison results.

[0066] In one specific implementation, based on the comparison results, adjustment instructions for the wearable exoskeleton are generated, specifically including: First, identify the target dimension in which the data deviation value in the comparison results exceeds the corresponding preset deviation threshold.

[0067] In other words, determining which dimension's measured value is deviating can provide a basis for subsequent root cause analysis of the deviation.

[0068] Then, based on the target dimension and the corresponding data deviation value, the deviation root causes are retrieved from the pre-built root cause database.

[0069] In practical applications, a root cause database can be pre-established, which stores the root causes of deviations under each dimension of deviation anomalies.

[0070] Finally, the adjustment strategies corresponding to the root causes of the deviations are determined, and the adjustment strategies are encoded into instructions to generate adjustment instructions for the wearable exoskeleton.

[0071] In practical applications, a preset control strategy library can be invoked based on the cause of the deviation to determine the specific adjustment strategy. For example, for the root cause of "excessive contact force", the adjustment strategy could be "reduce the auxiliary force of the exoskeleton hip joint and reduce the thrust on the human body"; if the root cause of the deviation is "too slow cadence", the adjustment strategy could be "increase the knee joint drive speed during the gait swing phase, shorten the swing time, and increase the cadence to 75 steps / minute".

[0072] The adjustment strategy obtains adjustment instructions through instruction encoding. The adjustment instructions can be converted into action signals of actuators such as motors and pneumatic cylinders by the controller (such as MCU or PLC) of the wearable exoskeleton, so as to correct the motion parameters of the wearable exoskeleton in real time.

[0073] In practical applications, new deviations between the adjusted multidimensional measured data and the multidimensional benchmark data can be compared and continuously fine-tuned, such as gradually increasing the assistance until the deviation enters a safe range, forming a closed loop of "collection-comparison-adjustment-recollection" to ensure that the wearable exoskeleton always conforms to the expected human motion model.

[0074] In some embodiments, when multiple root causes of deviations are identified through analysis, such as insufficient assistive force provided by the exoskeleton and user imbalance occurring simultaneously, a certain priority principle can be followed in generating adjustment instructions. That is, safety considerations take precedence over functionality, and issues that threaten user safety are addressed first. Therefore, the root cause of the imbalance problem is corrected first, and then the assistive force is adjusted, thereby ensuring the user's safety when using the wearable exoskeleton.

[0075] In one embodiment, the motion control unit establishes a human-machine cooperative motion model, specifically including: The first step is to obtain human motion parameters and exoskeleton structural parameters.

[0076] In practical applications, the acquisition of human motion parameters can be achieved through several methods. Firstly, 3D scanning and anthropometry can be used to collect anatomical parameters such as limb length (thigh, calf, upper arm), joint center point location (e.g., hip joint rotation center), and joint degrees of freedom (e.g., the knee joint primarily has flexion and extension degrees of freedom, while the hip joint has multiple degrees of freedom including flexion, extension, adduction, abduction, and rotation). Secondly, human kinematics databases can be consulted to obtain basic data such as joint angles, angular velocities, and angular accelerations under typical movement patterns. Simultaneously, motion capture systems (e.g., optical motion capture, inertial measurement units (IMUs)) and electromyography sensors can be used to collect real-time motion data of the human body in the target scenario, including joint movement trajectories, muscle activation sequences, and force characteristics, thus establishing an individual motion feature database.

[0077] In the exoskeleton structural parameter acquisition stage, physical parameters of the wearable exoskeleton can be obtained, including joint type (such as rotational joints and sliding joints), number and range of degrees of freedom (such as knee flexion and extension angle limitations), drive method (motor / hydraulic), transmission ratio, maximum drive torque, weight distribution, etc. Then, a mechanical structural model of the wearable exoskeleton is constructed using CA software or dynamic simulation tools to clarify its connection points with the human body (such as the binding positions of the hip and knee joints) and motion constraints.

[0078] The second step involves performing human dynamics modeling, exoskeleton dynamics modeling, and human-computer interaction force modeling based on human motion parameters and exoskeleton parameters, respectively, to obtain kinematic models, dynamic models, and human-computer interaction models.

[0079] In this embodiment, a dynamic model of the human limb can be established based on the Lagrange equation or the Newton-Euler method to calculate the driving torque required by the joint during the movement. Factors such as limb mass, rotational inertia, gravity, and inertial force are considered. The muscle force characteristics are analyzed by combining electromyographic signals, and the muscle activation degree is correlated with the joint torque. For example, muscle force can be estimated by using a muscle-skeleton dynamic model.

[0080] In practical applications, dynamic equations for the exoskeleton drive system can be established to describe the relationship between the output torque and speed of the motor / hydraulic actuator and joint movement. Considering factors such as friction and transmission losses, a dynamic model of the wearable exoskeleton can be obtained.

[0081] In addition, force sensors can be used to collect the interaction forces at the contact points between the exoskeleton and the human body, such as strap pressure and joint thrust, to establish a mapping relationship between the interaction forces and the human body's movement intentions. For example, changes in the interaction forces can reflect the human body's intention to accelerate or decelerate, thus obtaining a human-computer interaction model and providing a basis for subsequent collaborative control.

[0082] The third step is to extract typical motion characteristics from human motion parameters and establish a benchmark motion pattern.

[0083] Understandably, baseline movement patterns for the target scenario can be extracted from human motion parameters, such as the temporal curve of joint angles during walking, like the knee joint changing from 0°→60°→0° during the gait cycle; and the statistical patterns of key parameters such as cadence, stride length, and center of gravity transfer trajectory can be analyzed, such as the cadence of a normal adult walking on flat ground being about 50-80 steps / minute.

[0084] The fourth step is to classify the human motion parameters of different individuals, adjust the baseline motion pattern based on the classification results, establish a personalized baseline pattern, and generate a baseline pattern library.

[0085] In practical applications, machine learning algorithms such as cluster analysis can be used to classify and process the human motion parameters of different individuals. This allows for the creation of personalized baseline models based on individual differences such as height, weight, and exercise habits. This establishes unique personalized baseline models for different individual characteristics, improving the accuracy and adaptability of human motion models. For example, for tall individuals who habitually walk briskly, cluster analysis of their human motion parameters can create personalized motion baseline models suitable for this group, rather than using a uniform model applicable to everyone. This allows exoskeletons and other devices to better match individual exercise needs.

[0086] The fifth step is to integrate the kinematic model, dynamic model, human-computer interaction model, and benchmark pattern library to obtain the human-computer cooperative motion model.

[0087] By integrating the kinematic model, dynamic model, human-computer interaction model, and benchmark model library, a complete human-computer cooperative motion model can be constructed. This model can be digitally modeled using simulation software such as MATLAB, Simulink, and OpenSim, that is, the human-computer cooperative motion model can be presented and constructed in digital form through these software programs.

[0088] In some embodiments, after obtaining the human-machine cooperative motion model, the following may also be included: First, the exoskeleton mechanical limits and human safety of the human-machine collaborative motion model were verified, and the verification results were obtained.

[0089] Then, when the verification results show that both the mechanical limit verification of the exoskeleton and the human safety verification are passed, a human-machine collaborative motion model that meets the application conditions is obtained.

[0090] In this embodiment, after obtaining the human-machine collaborative motion model, it is necessary to further confirm whether the human-machine collaborative motion model can meet two requirements: on the one hand, it must meet the mechanical limit conditions of the wearable exoskeleton itself. For example, the maximum driving torque set by the model cannot exceed the rated value of the exoskeleton motor. If it exceeds this value, it may cause problems such as motor damage. This process is the mechanical limit verification step of the exoskeleton.

[0091] On the other hand, it is necessary to ensure human safety. For example, the interactive forces involved in the model should not exceed the threshold of human comfort, otherwise it will cause discomfort or even harm to the user. This process is the human safety verification stage.

[0092] Through the above two verifications, it can be ensured that the wearable exoskeleton will not malfunction due to the model setting exceeding mechanical capabilities during actual operation, and that the user's safety and comfort can be guaranteed.

[0093] In one embodiment, the above-mentioned wearable exoskeleton motion control system based on AI human-machine collaboration may further include: The human-computer interaction unit is used to generate control commands for the wearable exoskeleton based on the target training mode selected in the mode switching command after receiving the mode switching command initiated by the user.

[0094] Furthermore, the motion control unit is also used to control the wearable exoskeleton to match human movement according to control commands.

[0095] In practical applications, the human-computer interaction unit (HCI) can receive mode switching commands initiated by users through interactive methods such as touch screens, voice commands, or physical buttons. For example, users can select target training modes such as rehabilitation training mode or strength enhancement mode. The HCI has a built-in mode mapping table that stores basic control parameters corresponding to different training modes, such as joint range of motion, assist force coefficient, and gait cycle baseline values. Upon receiving a mode switching command, it can retrieve the basic control parameters corresponding to the target training mode from the mode mapping table to generate initial commands. These initial commands contain the motion framework that the wearable exoskeleton needs to execute; for example, in rehabilitation training mode, joint movement amplitude is smaller and assist force is stronger. Then, the basic control parameters in the initial commands can be integrated with the human motion model. For example, when the target training mode is climbing stairs, the step length baseline value in the initial commands will be corrected based on the user's leg length parameters in the human motion model. Simultaneously, the joint angle timing curves in the target training mode will be retrieved, such as larger knee flexion and extension angles and faster speeds, to obtain the final control commands. This ensures that the control commands are adapted to the user's individual characteristics, avoiding conflicts between general parameters and individual differences.

[0096] Subsequently, the control commands are transmitted by the motion control unit to the actuators of the wearable exoskeleton, such as motors and hydraulic devices. More preferably, the motion control unit can monitor the actual movement state of the wearable exoskeleton in real time through a closed-loop feedback mechanism, compare the actual movement state with the target parameters in the control commands, and if there is a deviation, such as the knee joint angle not meeting the command requirements when climbing stairs, the output of the actuator can be adjusted through algorithms such as PID to dynamically correct the movement trajectory. Ultimately, this achieves precise matching between the wearable exoskeleton and human movement in the target training mode, thus taking into account both mode characteristics and individual adaptability.

[0097] In summary, the wearable exoskeleton motion control system based on AI human-machine collaboration provided in this embodiment of the invention collects multi-dimensional measured data through a multi-sensor network. Relying on the collaborative processing capabilities of DSP and FPGA, it can construct and optimize general and differential models of human motion, achieving personalized adaptation. It can precisely control the wearable exoskeleton to match human motion according to adjustment commands, while integrating human-machine interaction and supporting mode switching. It can not only respond to the user's motion intentions in real time and dynamically optimize training difficulty and mode to ensure training effect and comfort, but also improve the system's real-time performance, intelligence, and personalization level. It effectively solves the problems of insufficient accuracy and flexibility of traditional exoskeleton control solutions, and can provide efficient, safe, and personalized motion assistance and training support.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wearable exoskeleton motion control system based on AI human-machine collaboration, characterized in that, include: The sensing unit is used to collect multi-dimensional measured data of joints, muscles, limbs, and wearable exoskeletons during human movement. The AI ​​processing unit is used to construct a general model that meets general movement needs and a differential model that meets personalized movement needs, and to fuse the general model and the differential model to obtain a human movement model. The unit compares and analyzes the multidimensional measured data with the multidimensional benchmark data in the human movement model to generate adjustment instructions for the wearable exoskeleton. The motion control unit is used to establish a human-machine collaborative motion model, and through the human-machine collaborative motion model, to control the wearable exoskeleton to match human movement according to the adjustment instructions.

2. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 1, characterized in that, The sensing unit includes: Joint position sensor, used to collect position data of joint movement; Force sensors are used to collect force data applied to wearable exoskeletons; Electromyography (EMG) sensors are used to collect electromyographic signals generated by muscle movement. An accelerometer is used to collect acceleration data of limb movements.

3. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 1, characterized in that, The AI ​​processing unit constructs a general model that meets general exercise needs and a differentiated model that meets personalized exercise needs, including: Obtain multidimensional sample data; The multidimensional sample data is subjected to denoising and normalization to obtain preprocessed sample data; Extract key features from the preprocessed sample data, perform motion pattern recognition based on the key features, and construct a sample dataset. A general model is obtained by training the basic model using the sample dataset. Individual difference data are obtained during the training process using a wearable exoskeleton. Based on the individual difference data, the general model is adjusted for differences to obtain a difference model.

4. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 1, characterized in that, The AI ​​processing unit fuses the general model and the differential model to obtain a human motion model, including: Extract the common parameters of human motion contained in the general model, and construct the motion infrastructure based on the common parameters; Extract the individual-specific parameters contained in the difference model, and embed the individual-specific parameters into the motion infrastructure to obtain the human motion model.

5. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 4, characterized in that, After obtaining the human motion model, the following is also included: The system acquires real-time updated individual motion data and dynamically optimizes the model parameters in the human motion model based on this data.

6. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 1, characterized in that, The AI ​​processing unit compares and analyzes the multidimensional measured data with the multidimensional benchmark data in the human motion model to generate adjustment instructions for the wearable exoskeleton, including: Align the multidimensional measured data with the multidimensional benchmark data in terms of dimensions; For each dimension, the deviation between the measured value in the multidimensional measured data and the benchmark value in the multidimensional benchmark data is calculated to obtain the data deviation value for each dimension; The data deviation value under each dimension is compared with the corresponding preset deviation threshold to obtain the comparison result; Based on the comparison results, adjustment instructions for the wearable exoskeleton are generated.

7. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 6, characterized in that, Based on the comparison results, adjustment instructions for the wearable exoskeleton are generated, including: Determine the target dimension in the comparison results where the data deviation value exceeds the corresponding preset deviation threshold. Based on the target dimension and the corresponding data deviation value, retrieve the deviation root causes from the pre-built root cause database; The adjustment strategy corresponding to the root cause of the deviation is determined, and the adjustment strategy is encoded into instructions to generate adjustment instructions for the wearable exoskeleton.

8. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 1, characterized in that, The motion control unit establishes a human-machine collaborative motion model, including: Acquire human motion parameters and exoskeleton structural parameters; Based on the human motion parameters and the exoskeleton parameters, human dynamics modeling, exoskeleton dynamics modeling and human-computer interaction force modeling are performed respectively to obtain kinematic model, dynamic model and human-computer interaction model; Extract typical motion features from the human motion parameters and establish a baseline motion pattern; The human motion parameters of different individuals are classified, the baseline motion pattern is adjusted according to the classification results, a personalized baseline pattern is established, and a baseline pattern library is generated. By integrating the kinematic model, the dynamic model, the human-computer interaction model, and the baseline pattern library, a human-computer cooperative motion model is obtained.

9. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 8, characterized in that, After obtaining the human-machine cooperative motion model, the following is also included: The exoskeleton mechanical limits and human safety were verified on the human-machine collaborative motion model, and the verification results were obtained. When the verification results show that both the mechanical limit verification and human safety verification of the exoskeleton are passed, a human-machine collaborative motion model that meets the application conditions is obtained.

10. The wearable exoskeleton motion control system based on AI human-machine collaboration according to claim 1, characterized in that, The system also includes: The human-computer interaction unit is used to generate control commands for the wearable exoskeleton based on the target training mode selected in the mode switching command after receiving a mode switching command initiated by the user. The motion control unit is also used to control the wearable exoskeleton to match human movement according to the control instructions.

Images