Gait control method of exoskeleton robot and related device

Abstract

This application provides a gait control method and related apparatus for an exoskeleton robot. The method includes: preprocessing, extracting features, and fusing exoskeleton robot motion data, user motion data, human-machine interaction force data, and environmental data collected by a multimodal sensor array to obtain fused features; when conflicting features exist among the multimodal features, reliability detection is performed, and fusion processing is conducted based on confidence levels; the user's motion intention type, motion state, and environmental state are determined based on the fused features; step length, step height, gait cycle, joint impedance parameters, and joint assist torque are adjusted accordingly, and the adjusted gait control parameters are subjected to smooth transition processing; the gait of the exoskeleton robot is controlled based on the processed gait control parameters. This application improves the reliability, continuity, and environmental adaptability of gait control in complex scenarios.

Description

Technical Field

[0001] This application relates to the field of control technology, and in particular to a gait control method and related device for an exoskeleton robot. Background Technology

[0002] Exoskeleton robots are wearable human-machine integrated devices with important application value in fields such as rehabilitation medicine and industrial assistance. For example, they can provide walking assistance for people with mobility impairments and help healthy people improve their load-bearing capacity.

[0003] Gait control technology is a crucial foundation for the stable assistance and environmental adaptation of exoskeleton robots. Its performance directly affects the wearer's stability, comfort, and adaptability to complex environments. Current gait control technologies control the exoskeleton robot's gait by pre-setting gait patterns (e.g., fixed stride length, stride height, and cycle). However, the gait parameters in this type of control are usually relatively fixed, making it difficult to dynamically adjust according to changes in the user's movement state and the external environment. This can easily lead to unnatural movements, insufficient human-robot coordination, and decreased balance. On the other hand, while some existing improvements have introduced multi-source sensor information for state perception and parameter adjustment, they still suffer from insufficient handling of multimodal information conflicts in complex scenarios, weak targeting of control parameter adjustments, and limited individual adaptability.

[0004] Therefore, there is an urgent need to provide a gait control scheme for exoskeleton robots that can adapt to complex motion states and environmental changes. Summary of the Invention

[0005] Based on the above problems, this application provides a gait control method and related device for exoskeleton robots, so as to at least solve the technical problems in the prior art, such as the difficulty of adapting to complex motion states and environmental changes, insufficient reliability of fusion results in multimodal information conflict scenarios, weak targeting of control parameter adjustment, and limited individual adaptability.

[0006] The embodiments of this application disclose the following technical solutions: In a first aspect, embodiments of this application provide a gait control method for an exoskeleton robot. This method involves preprocessing, extracting features, and fusing multi-dimensional data collected by a multimodal sensor array to obtain fused features. The multi-dimensional data includes motion data of the exoskeleton robot, motion data of the user wearing the exoskeleton robot, human-computer interaction forces, and environmental data of the user. Based on the fused features, the method determines the user's motion intention type, motion state, and environmental state. Based on the user's motion intention type, motion state, and environmental state, the method adjusts the stride length, stride height, and gait circumference of the exoskeleton robot. At least one of the following: adjusting the joint impedance parameters of the exoskeleton robot based on human-computer interaction force characteristics; adjusting the joint assist torque of the exoskeleton robot based on muscle activity intensity characterized by electromyographic signal characteristics; determining at least one of the adjusted step length, step height, and gait cycle, the adjusted joint impedance parameters, and the adjusted joint assist torque as adjusted gait control parameters; performing smooth transition processing on the adjusted gait control parameters based on the unadjusted gait control parameters and the adjusted gait control parameters to obtain processed gait control parameters; controlling the gait of the exoskeleton robot based on the processed gait control parameters.

[0007] The method provided in this application, in the process of adjusting the gait control parameters of the exoskeleton robot, comprehensively considers the user's movement intention type, movement state and the environmental state of the user. Based on this, the gait of the exoskeleton robot is controlled according to the adjusted gait control parameters, which can be adapted to complex changes in movement state and changes in the external environment, thereby improving the user's experience.

[0008] In one possible implementation, the preprocessing, feature extraction, and fusion processing of the multi-dimensional data collected by the multimodal sensor array to obtain fused features includes: performing time synchronization processing on multi-dimensional data with different sampling rates to obtain synchronized multi-dimensional data; performing noise reduction processing on the multi-dimensional data to obtain noise-reduced multi-dimensional data; performing standardization processing on the synchronized multi-dimensional data and the noise-reduced multi-dimensional data to obtain preprocessed multi-dimensional data; and performing feature extraction and fusion processing on the preprocessed multi-dimensional data to obtain fused features. In this embodiment, the deep fusion of multi-dimensional data collected by the multimodal sensor array provides accurate and reliable decision-making basis for the adaptive gait control of the exoskeleton robot, breaking the limitations of single sensor information, improving the exoskeleton robot's adaptability to complex environments and changes in user states, and effectively enhancing the user experience.

[0009] In one possible implementation, the step of performing feature extraction and fusion processing on the preprocessed multi-dimensional data to obtain fused features includes: performing feature extraction on the preprocessed multi-dimensional data to obtain multimodal features; when there are at least two conflicting features among the multimodal features, performing reliability detection on the conflicting features to determine the confidence level of the conflicting features; and performing fusion processing on the multimodal features based on the confidence level to obtain fused features.

[0010] In one possible implementation, when at least two conflicting features exist among the multimodal features, reliability detection is performed on the conflicting features to determine their confidence levels. This includes: when a first conflicting feature and a second conflicting feature exist among the multimodal features, reliability detection is performed on the first and second conflicting features based on the current motion state and / or environmental state to determine their confidence levels. The step of fusing the multimodal features based on the confidence levels to obtain fused features includes: when the confidence level of the first conflicting feature is greater than the confidence level of the second conflicting feature, filtering the second conflicting feature and fusing the remaining multimodal features to obtain the fused feature; or, when the confidence level of the second conflicting feature is greater than the confidence level of the first conflicting feature, filtering the first conflicting feature and fusing the remaining multimodal features to obtain the fused feature; or, based on the confidence levels of the first and second conflicting features, dynamically weighting and fusing the first and second conflicting features, and then fusing them with the remaining multimodal features to obtain the fused feature.

[0011] In this embodiment, the conflict arbitration mechanism effectively resolves the contradictions between signals from different sensors, ensuring the reliability and accuracy of the fusion results and enhancing the robustness of the exoskeleton robot under complex conditions.

[0012] In one possible implementation, the feature extraction of the preprocessed multi-dimensional data includes: acquiring the user's historical wear data; the historical wear data includes at least one of the user's exercise habits, physiological characteristics, and environmental adaptation preferences; determining feature extraction weights based on the historical wear data; and performing feature extraction on the preprocessed multi-dimensional data based on the feature extraction weights to obtain multimodal features.

[0013] In this embodiment of the application, the user's historical wear data is considered in the process of obtaining multimodal features, so that the gait control strategy can be adapted to the different users' exercise habits and physiological characteristics (such as muscle strength and stride frequency preference), which is more in line with individual needs and expands the scope of application of the device.

[0014] In one possible implementation, the step of performing feature extraction and fusion processing on the preprocessed multi-dimensional data to obtain fused features includes: extracting features from the preprocessed multi-dimensional data to obtain multimodal features; enhancing the spatial and temporal features in the multimodal features based on an attention mechanism to obtain a first feature; the spatial features include pressure features of various regions of the user's foot and environmental data features of the user's location, and the temporal features include motion features of various parts of the user's body and electromyographic signal features; and fusing the first feature, human-computer interaction force features, and motion features of the exoskeleton robot to generate fused features.

[0015] In this embodiment, a deep learning model can be used to extract deep-level correlation features from multimodal features. At the same time, an attention mechanism is introduced to focus on key information. Combined with the closed-loop link of "sensor-processing-control-execution", a low-latency response from information perception to control output can be achieved, thereby improving the accuracy of gait control.

[0016] In one possible implementation, determining the user's motion intention type, motion state, and environmental state based on the fusion feature includes: determining the user's motion intention type based on the first feature; determining the user's motion state and environmental state based on the fusion feature; the user's motion state includes gait phase type and center of gravity state, and the environmental state includes terrain type and obstacle location information.

[0017] In one possible implementation, the smooth transition processing of the adjusted gait control parameters to obtain the processed gait control parameters includes: continuously switching between the gait control parameters before adjustment and the adjusted gait control parameters using a dynamic interpolation algorithm and / or an exponential smoothing algorithm to obtain the processed gait control parameters.

[0018] In this embodiment of the application, by performing a smooth transition process on the adjusted gait control parameters, abrupt gait changes are avoided, the risk of imbalance is reduced, and the user's wearing comfort and safety are effectively improved.

[0019] It should be noted that the embodiments of this application do not simply combine multimodal perception, state recognition and parameter adjustment in parallel, but establish a hierarchical adjustment mechanism for different control objects: adjusting gait trajectory parameters based on motion intention type, motion state and environmental state, adjusting joint impedance parameters based on human-computer interaction force characteristics, adjusting joint assist torque based on electromyographic signal characteristics, and ensuring the continuity of the above parameter switching through a smooth transition mechanism, thereby improving the control effect of the exoskeleton robot in complex scenarios.

[0020] Secondly, embodiments of this application provide a gait control device for an exoskeleton robot, comprising: The system comprises a first processing unit, a determining unit, an adjusting unit, a second processing unit, and a control unit. The first processing unit is used to preprocess, extract features, and fuse multi-dimensional data collected by the multi-modal sensor array to obtain fused features; the multi-dimensional data includes motion data of the exoskeleton robot, motion data of the user wearing the exoskeleton robot, human-computer interaction forces, and environmental data of the user. The determining unit is used to determine the user's motion intention type, motion state, and environmental state based on the fusion features. The adjustment unit is used to adjust at least one of the step length, step height, and gait cycle of the exoskeleton robot based on the user's movement intention type, movement state, and the environmental state in which the user is located; adjust the joint impedance parameters of the exoskeleton robot based on human-computer interaction force characteristics; adjust the joint assist torque of the exoskeleton robot based on the muscle activity intensity characterized by electromyographic signal characteristics; and determine the adjusted step length, step height, and gait cycle, the adjusted joint impedance parameters, and the adjusted joint assist torque as the adjusted gait control parameters. The second processing unit is used to perform smooth transition processing on the adjusted gait control parameters based on the gait control parameters before adjustment and the adjusted gait control parameters to obtain the processed gait control parameters. The control unit is used to control the gait of the exoskeleton robot based on the processed gait control parameters.

[0021] Thirdly, embodiments of this application provide an exoskeleton robot, the exoskeleton robot including: a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute, according to the instructions in the program code, the steps of the gait control method for an exoskeleton robot described in any of the embodiments of the first aspect above.

[0022] Compared with the prior art, the embodiments of this application have at least the following beneficial effects: This application embodiment improves the reliability and stability of the fusion results in complex scenarios by jointly processing exoskeleton robot motion data, user motion data, human-computer interaction forces, and environmental data, and by introducing a reliability detection and confidence arbitration mechanism when there are conflicts in multimodal features.

[0023] This application embodiment improves the targeting and coordination of gait control parameter adjustment by jointly identifying the type of movement intention, movement state and environmental state, and mapping different types of perception results to the adjustment process of gait trajectory parameters, joint impedance parameters and joint assist torque.

[0024] The embodiments of this application improve the adaptability of exoskeleton robots to individual differences among different users by introducing historical wear data to determine feature extraction weights.

[0025] The embodiments of this application can reduce the output abrupt changes caused by parameter switching by performing a smooth transition processing on the gait control parameters before and after adjustment, thereby improving control continuity, stability and safety.

[0026] In summary, the embodiments of this application can improve the reliability, continuity, and environmental adaptability of gait control of exoskeleton robots under complex motion states and environmental changes. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application 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 only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A schematic diagram illustrating an application scenario of a gait control method for an exoskeleton robot provided in an embodiment of this application; Figure 2 A flowchart illustrating a gait control method for an exoskeleton robot provided in an embodiment of this application; Figure 3 A schematic diagram of the hardware structure of an exoskeleton robot provided in an embodiment of this application; Figure 4 This is a schematic diagram of the gait control device for an exoskeleton robot provided in an embodiment of this application. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to be a limitation of this application.

[0030] As mentioned above, existing gait control technologies typically control the gait of exoskeleton robots by pre-setting gait patterns, such as fixed stride length, stride height, and gait cycle. However, the gait parameters in this type of control are usually relatively fixed, making it difficult to dynamically adjust according to changes in the user's movement state and the external environment. This can easily lead to unnatural movements, insufficient human-robot coordination, and decreased balance. On the other hand, while some existing improvements have introduced multi-source sensor information for state perception and parameter adjustment, they still suffer from insufficient handling of multimodal information conflicts, weak targeting of control parameter adjustments, and limited individual adaptability in complex scenarios.

[0031] Based on this, embodiments of this application provide a gait control method and related apparatus for an exoskeleton robot. The method preprocesses, extracts features, and fuses exoskeleton robot motion data, user motion data, human-machine interaction forces, and environmental data collected by a multimodal sensor array. When conflicts exist in the multimodal features, reliability detection is performed, and fusion processing is conducted based on confidence levels to obtain fused features. Based on the fused features, the user's motion intention type, motion state, and environmental state are determined. Based on the motion intention type, motion state, and environmental state, at least one of step length, step height, and gait cycle is adjusted. Joint impedance parameters are adjusted based on human-machine interaction force features, and joint assist torque is adjusted based on electromyographic signal features. Finally, the adjusted gait control parameters are smoothed to control the exoskeleton robot's gait.

[0032] The embodiments of this application improve the reliability, continuity, and environmental adaptability of gait control for exoskeleton robots under complex motion states and environmental changes by introducing multimodal information joint perception, conflict feature arbitration, hierarchical parameter adjustment, and smooth transition processing mechanisms during gait control.

[0033] like Figure 1 As shown in the figure, this is a schematic diagram of an application scenario for a gait control method for an exoskeleton robot provided in an embodiment of this application. This method can be implemented based on the interaction between the multimodal sensor array 101 and the processor 102.

[0034] The multimodal sensor array 101 collects multi-dimensional data and sends the multi-dimensional data to the processor 102.

[0035] The processor 102 preprocesses, extracts features, and fuses multi-dimensional data collected by the multimodal sensor array to obtain fused features. The multi-dimensional data includes motion data of the exoskeleton robot, motion data of the user wearing the exoskeleton robot, human-machine interaction forces, and environmental data of the user. Based on the fused features, the processor determines the user's motion intention type, motion state, and environmental state. Based on the user's motion intention type, motion state, and environmental state, the processor adjusts at least one of the exoskeleton robot's stride length, stride height, and gait cycle. Based on the human-machine interaction... The exoskeleton robot's joint impedance parameters are adjusted based on mutual force characteristics; the joint assist torque is adjusted based on muscle activity intensity characterized by electromyographic signal features; at least one of the adjusted stride length, stride height, and gait cycle, the adjusted joint impedance parameters, and the adjusted joint assist torque are determined as adjusted gait control parameters; the adjusted gait control parameters are then subjected to smooth transition processing based on the unadjusted gait control parameters and the adjusted gait control parameters to obtain processed gait control parameters; and the exoskeleton robot's gait is controlled based on the processed gait control parameters.

[0036] It is understood that the deployment relationship between the multimodal sensor array 101 and the processor 102 is not specifically limited in the embodiments of this application. For example, the multimodal sensor array and the processor can be deployed separately or as part of an exoskeleton robot in a joint deployment manner.

[0037] The following describes a gait control method for an exoskeleton robot provided by an embodiment of this application, with reference to the accompanying drawings. Figure 2 As shown in the figure, this figure is a flowchart of a gait control method for an exoskeleton robot provided in an embodiment of this application, including S201-S204.

[0038] S201. Preprocess, extract features, and fuse the multi-dimensional data collected by the multimodal sensor array to obtain fused features.

[0039] The multi-dimensional data may include, but is not limited to, the motion data of the exoskeleton robot, the motion data of the user wearing the exoskeleton robot, human-computer interaction forces, and the environmental data of the user.

[0040] For example, a multimodal sensor array can be installed at key joints of the exoskeleton robot body and at the user's physiological signal acquisition site, and it includes at least an inertial measurement unit (IMU), surface electromyography sensor, plantar pressure sensor, joint encoder, environmental perception sensor and force sensor.

[0041] Inertial measurement units can be used to collect some of the user's motion data, such as the user's posture and motion acceleration; surface electromyography (EMG) sensors are used to collect the user's electromyographic signal data; plantar pressure sensors are used to collect the user's plantar pressure distribution data; joint encoders are used to collect the angle information of the hip, knee and ankle joints of the exoskeleton robot; environmental perception sensors (such as depth cameras) are used to collect data on the user's environment; and force sensors are used to collect human-machine interaction forces.

[0042] Among them, the user's posture and motion acceleration, the user's electromyographic signal data, and the plantar pressure distribution data belong to the user's motion data; the angle information of the exoskeleton robot's hip joint, knee joint, and ankle joint belongs to the exoskeleton robot's motion data; environmental data can be depth image data.

[0043] In this embodiment, the collaborative work of multiple types of sensors enables the capture of relevant information about the exoskeleton robot and the user from different dimensions, achieving information complementarity and redundancy, and laying a solid data foundation for subsequent fusion processing and control decisions.

[0044] In this embodiment of the application, after obtaining multi-dimensional data collected by a multi-modal sensor array, multi-modal features can be obtained by extracting features from the multi-dimensional data; based on this, fused features can be obtained by fusing the multi-modal features.

[0045] In one possible implementation, the process of preprocessing, extracting features, and fusing multi-dimensional data acquired by a multi-modal sensor array to obtain fused features includes: performing time synchronization processing on multi-dimensional data with different sampling rates to obtain synchronized multi-dimensional data; performing noise reduction processing on the multi-dimensional data to obtain noise-reduced multi-dimensional data; performing standardization processing on the synchronized multi-dimensional data and the noise-reduced multi-dimensional data to obtain preprocessed multi-dimensional data; and performing feature extraction and fusion processing on the preprocessed multi-dimensional data to obtain fused features.

[0046] For example, time synchronization processing may include timestamp alignment of multi-dimensional data with different sampling rate signals. For instance, based on the timestamps of each sensor, signals with different sampling rates can be uniformly aligned to a 100Hz time axis to achieve time synchronization of multi-dimensional data.

[0047] The noise reduction process includes using Kalman filtering to reduce noise in the data collected by the inertial measurement unit; using wavelet transform filtering to reduce noise in the electromyographic signal data; and using mean filtering to reduce noise in the plantar pressure distribution data.

[0048] The Kalman filter is used to denoise the data collected by the inertial measurement unit. The process of removing noise is shown in Equation (1): (1) in, The state estimate at time k. Let K be the predicted value at time k based on time k-1. k For Kalman gain, z k Let H be the measurement value at time k, and H be the measurement matrix.

[0049] The process of using wavelet transform filtering to denoise electromyographic signal data and remove power frequency interference and motion artifacts is shown in Equation (2): (2) Where a is the scaling factor and b is the translation factor. The mother wavelet is denoted as x(t), and the electromyographic signal is denoted as x(t).

[0050] Mean filtering is used to reduce noise in the plantar pressure distribution data, and the process of smoothing noise is shown in Equation (3): (3) Where X(i) is the original pressure signal, y(i) is the filtered signal, and N is the size of the filtering window.

[0051] After performing synchronous processing and noise reduction on the multi-dimensional data, preliminary multi-dimensional data can be obtained. By standardizing the preliminary multi-dimensional data, it can be normalized to a unified range, resulting in pre-processed multi-dimensional data, as shown in equation (4): (4) Where x is the original data, , Let x' be the minimum and maximum values ​​of the data, respectively, and let x' be the preprocessed multi-dimensional data; the uniform range can be [-1, 1].

[0052] After obtaining the preprocessed multi-dimensional data, feature extraction can be performed on the preprocessed multi-dimensional data to obtain multimodal features, and the multimodal features can be fused to obtain fused features.

[0053] In one possible implementation, the multimodal features include at least spatial features, temporal features, human-computer interaction force features, and motion features of the exoskeleton robot.

[0054] The spatial features include pressure characteristics of different areas of the user's foot and environmental data characteristics of the user's location; the temporal features include motion characteristics and electromyographic signal characteristics of different parts of the user's body.

[0055] Specifically, the pressure characteristics of different areas of the user's foot were obtained by extracting features from the plantar pressure distribution data collected by the plantar pressure sensor; the environmental data characteristics of the user were obtained by extracting features from the environmental data collected by the environmental perception sensor; the motion characteristics of different parts of the user were obtained by extracting features from the user's posture and motion acceleration collected by the inertial measurement unit; the electromyographic signal characteristics were obtained by extracting features from the electromyographic signal data collected by the surface electromyography sensor; the human-computer interaction force characteristics were obtained by extracting features from the human-computer interaction force collected by the force sensor; and the motion characteristics of the exoskeleton robot were obtained by extracting features from the joint motion data of the exoskeleton robot collected by the joint encoder.

[0056] In this embodiment of the application, based on the attention mechanism, the spatial features and temporal features in the multimodal features can be enhanced to obtain the first feature; the first feature, the human-computer interaction force feature and the motion feature of the exoskeleton robot are fused to generate the fused feature.

[0057] In one example, in this embodiment of the application, a fusion model can be used to perform correlation analysis and feature extraction on multimodal features to generate fused features.

[0058] This fusion model can employ at least one of the deep learning models such as Convolutional Neural Network (CNN), Long Short-Term Memory Network (LSTM), or Transformer Network, and enhance the recognition ability of key features through an attention mechanism.

[0059] For example, CNN can be used to extract spatial features from plantar pressure distribution data and environmental data to obtain spatial features; LSTM can be used to extract temporal features from user posture and motion acceleration and electromyography signal data to obtain temporal features.

[0060] The cell state update formula for LSTM is shown in equation (5): (5) in, , , These are the input gate, forget gate, and output gate, respectively. In cellular state; σ represents the hidden state; σ is the sigmoid function; ⊙ represents element-wise multiplication; W is the weight matrix; b is the bias term.

[0061] The extracted spatial and temporal features are input into the attention mechanism module to enhance key features related to gait control (such as pressure features of different areas of the sole at the moment of heel strike and electromyographic signal features during muscle exertion), thus obtaining the first feature. The attention weight calculation formula is shown in equation (6): (6) Where Q, K, and V are the query matrix, key matrix, and value matrix, respectively; is the matrix dimension.

[0062] After obtaining the first feature, the first feature, the human-computer interaction force feature, and the motion feature of the exoskeleton robot are fused to generate a fused feature.

[0063] In this embodiment, a deep learning-based fusion model is used to perform feature-level fusion of multimodal features, which can automatically mine deep-level correlation features in heterogeneous data. At the same time, an attention mechanism is introduced to highlight key information on gait control, thereby improving the effectiveness and relevance of the fusion results.

[0064] In one possible implementation, when there are at least two conflicting features among the multimodal features, reliability detection can be performed on the at least two conflicting features to determine the confidence level of the at least two conflicting features; based on the confidence level of the at least two conflicting features, the multimodal features are fused to obtain fused features.

[0065] For example, the fusion model has a conflict arbitration mechanism to address contradictions in signals from different types of sensors. For instance, if the inertial measurement unit indicates that the user is showing signs of imbalance, but the plantar pressure distribution data shows that both feet are in stable contact with the ground, the reliability of the signal can be evaluated based on a combination of preset rules and a training model, prioritizing signals that conform to the current motion state or environmental context.

[0066] It is understood that the preset rules in this application embodiment are not specifically limited, and can be set according to actual needs. For example, the preset rules include: when the heel strikes the ground, the plantar pressure distribution data should be given priority.

[0067] Taking a feature with two conflicting characteristics (a first conflicting feature and a second conflicting feature) as an example, based on the current motion state and / or environmental state, the reliability of the first conflicting feature and the second conflicting feature is tested to determine the confidence level of the first conflicting feature and the confidence level of the second conflicting feature.

[0068] If the confidence level of the first conflicting feature is greater than the confidence level of the second conflicting feature, then the second conflicting feature is filtered out, and the other features are fused to obtain the fused feature; if the confidence level of the second conflicting feature is greater than the confidence level of the first conflicting feature, then the first conflicting feature is filtered out, and the other features are fused to obtain the fused feature.

[0069] In this embodiment, the data collected by the multimodal sensor array can form information complementarity and redundancy. Combined with the conflict arbitration mechanism, when there are abnormal or conflicting sensor signals, the reliability of the fusion result can be guaranteed by the combination of preset rules and training models, ensuring that the exoskeleton robot can still work stably in complex environments.

[0070] In another possible implementation, when there are at least two conflicting features among the multimodal features, the conflicting features can be dynamically weighted and fused to obtain fused features, as shown in equation (7): (7) in, The data collected by each sensor, For the corresponding weights, ; y represents the fusion feature.

[0071] In this embodiment of the application, the conflict arbitration mechanism can effectively resolve the contradictions between different sensor signals, ensuring the reliability and accuracy of the fusion results and enhancing the robustness of the exoskeleton robot in complex situations.

[0072] In one possible implementation, to improve the personalized adaptation capability of exoskeleton robots, the method provided in this application embodiment can obtain the user's historical wearing data during the process of extracting features from multi-dimensional data collected by a multi-modal sensor array to obtain multi-modal features; determine feature extraction weights based on the historical wearing data; and extract features from the multi-dimensional data collected by the multi-modal sensor array based on the feature extraction weights to obtain multi-modal features.

[0073] The user's historical wear data may include, but is not limited to, exercise habits, physiological characteristics, and environmental adaptation preferences. Based on the historical wear data, the feature extraction weights and control parameters of the fusion model can be adjusted so that the gait control strategy can dynamically adapt to individual movement characteristics.

[0074] Because the embodiments of this application take into account the user's historical wearing data in the process of obtaining multimodal features, the exoskeleton robot can gradually adapt to the individual differences of different users, continuously optimize the control strategy, improve the personalized adaptation capability of the exoskeleton robot, and enhance the user experience and usage effect.

[0075] S202. Based on the fusion features, determine the user's motion intention type, motion state, and the environmental state of the user.

[0076] In one possible implementation, the user's motion intention type can be determined based on the first feature; and the user's motion state and the environmental state of the user can be determined based on the fusion feature.

[0077] The movement intention types include starting, stopping, turning, going up stairs, going down stairs, etc.; the user's movement state can include gait phase type (e.g., heel strike, toe lift) and center of gravity state (e.g., center of gravity position and movement trend); the environmental state can include terrain type (e.g., flat ground, stairs, slope) and obstacle location information.

[0078] In this embodiment, the process of determining the gait phase type considers data collected by the inertial measurement unit (such as the user's posture and motion acceleration), plantar pressure distribution data collected by the plantar pressure sensor, and angle information of the hip, knee, and ankle joints collected by the joint encoder. In determining the center of gravity state, the process considers lumbar acceleration information collected by the inertial measurement unit and angle information collected by the joint encoder. Combined with the kinematic model, the user's center of gravity position and motion trend can be determined. In determining the environmental state, the process considers environmental data collected by the environmental perception sensor. Combined with the semantic segmentation model, the terrain type can be determined and the spatial location information of obstacles can be obtained.

[0079] In the method provided in this application embodiment, based on fusion features, information covering multiple aspects closely related to gait control can be determined, which facilitates subsequent adaptive control parameter adjustment, realizes efficient connection from information perception to control decision, and ensures the pertinence and timeliness of control strategy.

[0080] S203. Based on the user's movement intention type, movement state, and the environmental state in which the user is located, adjust at least one of the following of the exoskeleton robot: stride length, stride height, and gait cycle.

[0081] For example, when the terrain is flat, the step length of the exoskeleton robot is adjusted to 0.6m, the step height to 0.02m, and the gait cycle to 1.2s; when the movement is climbing stairs, the step height of the exoskeleton robot is adjusted to 0.15m, the step length to 0.4m, and the gait cycle to 1.5s according to the height of the stairs; when the movement is turning, the inner step length can be shortened and the outer step length increased according to the turning angle.

[0082] It is understood that the specific values ​​for step length, step height, and gait cycle mentioned above are merely examples.

[0083] S204. Based on the human-computer interaction force characteristics, adjust the joint impedance parameters of the exoskeleton robot.

[0084] Based on the human-computer interaction force characteristics, the joint impedance parameters of the exoskeleton robot can be adjusted to obtain the adjusted joint impedance parameters.

[0085] The joint impedance parameters can include stiffness and damping. For example, when the human-machine interaction force characteristics indicate that the human-machine interaction force is less than or equal to the interaction force threshold, it means the human-machine interaction force is small. In this case, the joint stiffness and damping can be increased (e.g., the hip joint stiffness is set to 200N). m / rad, hip joint damping is 5N m (s / rad) to ensure the accuracy of gait trajectory; when the human-computer interaction force is greater than the interaction force threshold, it indicates that the interaction force is too large, and joint stiffness and damping can be reduced (e.g., hip joint stiffness is 100N). m / rad, hip joint damping is 3N m (s / rad) to improve the compliance of human-computer interaction.

[0086] S205. Adjust the joint assist torque of the exoskeleton robot based on the muscle activity intensity characterized by electromyographic signal features.

[0087] Based on the characteristics of electromyography signals, the joint assist torque of the exoskeleton robot can be adjusted to obtain the adjusted joint assist torque.

[0088] For example, the intensity of muscle activity can be determined based on the root mean square (RMS) value of the electromyographic signal, as shown in equation (8): (8) in, The signal is the electromyographic signal, and N is the number of sampling points.

[0089] When the RMS value is greater than the root mean square threshold, it indicates that the muscle activity intensity is high, and the joint assist torque can be increased (e.g., the maximum knee joint assist torque is 30N). When RMS is less than or equal to the root mean square threshold, it indicates that the muscle activity intensity is low, and the joint assist torque can be reduced to make the assist consistent with the human body's force exertion.

[0090] It should be noted that the execution order of S203, S204, and S205 is not specifically limited in this embodiment of the application. For example, they can be executed synchronously or sequentially.

[0091] S206. The adjusted stride length, stride height, and at least one of the gait cycle, the adjusted joint impedance parameter, and the adjusted joint assist torque are determined as the adjusted gait control parameters.

[0092] After obtaining at least one of the adjusted stride length, adjusted stride height, and adjusted gait cycle, the adjusted joint impedance parameter, and the adjusted joint assist torque, the adjusted stride length, adjusted stride height, and adjusted gait cycle, the adjusted joint impedance parameter, and the adjusted joint assist torque can be determined as the adjusted gait control parameters.

[0093] S207. Based on the gait control parameters before adjustment and the gait control parameters after adjustment, perform smooth transition processing on the adjusted gait control parameters to obtain the processed gait control parameters.

[0094] In this embodiment of the application, to ensure the continuity and stability of the gait control parameter adjustment process, the adjusted gait control parameters can be subjected to a smooth transition process based on the gait control parameters before adjustment and the adjusted gait control parameters to obtain the processed gait control parameters.

[0095] For example, the method provided in this application embodiment can use a dynamic interpolation algorithm and / or an exponential smoothing algorithm to continuously switch between the gait control parameters before adjustment and the gait control parameters after adjustment to obtain the processed gait control parameters.

[0096] Taking the continuous switching of gait control parameters before and after adjustment based on the exponential smoothing algorithm as an example, as shown in equation (9): (9) in, These are the smoothed values, i.e., the processed gait control parameters; These are the gait control parameters before adjustment; The value is the smoothed value from the previous time step; α is the smoothing coefficient used to avoid abrupt changes in gait.

[0097] It is understood that the value of α is not specifically limited in the embodiments of this application, and it can be set according to actual needs, such as α=0.3.

[0098] In this embodiment, by performing a smooth transition process on the adjusted gait control parameters, user discomfort or instability of the exoskeleton robot that may be caused by sudden gait changes is effectively avoided, thereby improving the user's wearing comfort and safety.

[0099] S208. Based on the processed gait control parameters, control the gait of the exoskeleton robot.

[0100] After obtaining the processed gait control parameters, the processed gait control parameters can be converted into execution instructions (PWM signals) to control the gait of the exoskeleton robot based on the execution instructions.

[0101] In summary, the method provided in this application improves the reliability and stability of the fusion results by jointly processing exoskeleton robot motion data, user motion data, human-computer interaction forces, and environmental data, and by introducing a reliability detection and confidence arbitration mechanism when conflicts exist in multimodal features. Furthermore, by jointly identifying motion intention types, motion states, and environmental states, and mapping different types of perception results to the adjustment processes of gait trajectory parameters, joint impedance parameters, and joint assist torques, the method enhances the targeting and coordination of gait control parameter adjustments. Finally, by performing smooth transition processing on the gait control parameters before and after adjustment, the method improves control continuity, stability, and environmental adaptability.

[0102] To facilitate understanding, the following description, in conjunction with the hardware structure of the exoskeleton robot, introduces a gait control method for an exoskeleton robot provided in this application. For example... Figure 3 As shown in the figure, this figure is a schematic diagram of the hardware structure of an exoskeleton robot provided in an embodiment of this application.

[0103] An exoskeleton robot 300 provided in this application includes an exoskeleton body 310, a multimodal sensor array 320, a processor 330, and a power supply module 340. The processor 330 includes a data processing unit 331 and a control unit 332. The data processing unit 331 is used to preprocess, extract features, and fuse multi-dimensional data collected by the multimodal sensor array 320. The control unit 332 is used to perform gait control parameter adjustment, smooth transition processing, and closed-loop control output based on the fusion results.

[0104] The exoskeleton body 310 adopts a lower limb exoskeleton structure, with a driveable joint structure and a corresponding actuator. The joint structure includes at least a hip joint, a knee joint, and an ankle joint, each of which can achieve multi-degree-of-freedom rotation; the actuator is a servo motor, whose rated torque meets the needs of human lower limb movement assistance and can accurately respond to the PWM control signal output by the control unit 332.

[0105] The multimodal sensor array 320 is used to collect multi-dimensional data, which includes at least the motion data of the exoskeleton robot 300, the motion data of the user wearing the exoskeleton robot 300, human-computer interaction forces, and the environmental data of the user.

[0106] The multimodal sensor array 320 includes at least an inertial measurement unit, a surface electromyography sensor, a plantar pressure sensor, a joint encoder, an environmental sensing sensor, and a force sensor, which are respectively installed at key joints of the exoskeleton body 310 and at the user's physiological signal acquisition sites.

[0107] The inertial measurement unit can be a 6-axis (3-axis accelerometer and 3-axis gyroscope) inertial measurement unit, which is installed on the user's waist, mid-thigh, mid-calf and foot respectively, with a sampling rate of 100Hz, to collect the user's limb posture and motion acceleration.

[0108] The surface electromyography (EMG) sensor can be a dry electrode surface EMG sensor, which is attached to the user's quadriceps, hamstrings, tibialis anterior, and gastrocnemius muscles, with a sampling rate of 1000 Hz, to collect the user's EMG signal data.

[0109] The plantar pressure sensor can be a thin-film pressure sensor array embedded in the user's shoe sole, with an array density of 16×16 and a sampling rate of 500Hz, used to collect the user's plantar pressure distribution data (e.g., the contact state between the foot and the ground).

[0110] The joint encoder can be a 16-bit resolution absolute encoder, integrated into the hip, knee and ankle joints of the exoskeleton, to collect joint motion data (such as angle information of the hip, knee and ankle joints) of the exoskeleton robot 300.

[0111] The environmental perception sensor can be a depth camera, mounted on the front shoulder of the exoskeleton robot 300, with a field of view of 90° and a frame rate of 30fps, used to collect environmental data of the user's surroundings.

[0112] The force sensor can be a strain gauge type force sensor, installed on the hip, thigh and calf support parts of the exoskeleton robot 300 that come into contact with the user, with a range of 0-500N, for collecting human-machine interaction forces.

[0113] The data processing unit 331 can consist of an FPGA module and a GPU module, and is connected to the multimodal sensor array 320 via an EtherCAT bus with a communication latency of less than 1ms. The FPGA module is used for preprocessing the raw data collected by each sensor, such as synchronization and noise reduction; the GPU module uses a high-performance graphics processor 330 to run the fusion model and perform the fusion processing of multi-dimensional data.

[0114] The control unit 332 can be a 32-bit DSP chip, which is connected to the data processing unit 331 and the actuator via a bus communication to perform adaptive control generation and closed-loop control output, forming a closed-loop link of "sensor-processing-control-execution".

[0115] For example, the control unit 332 has a control cycle of 5ms and can receive the fusion results (user's motion intention type, motion state, and environmental state of the user) output by the data processing unit 331 in real time, and generate corresponding control commands to send to the actuator.

[0116] In one example, the control unit 332 incorporates a reinforcement learning algorithm that optimizes the feature extraction weights and control parameter adjustment rules of the fusion model online based on the user's historical wear data (including but not limited to movement habits such as cadence and stride length preferences; physiological characteristics such as muscle strength; and environmental adaptation preferences such as movement speed in different terrains), enabling the gait control strategy to dynamically adapt to individual movement characteristics.

[0117] For example, for users with weaker muscle strength, the control unit 332 can appropriately increase the adjustment range of the joint assist torque; for users with a faster stride frequency, the control unit 332 can correspondingly shorten the adjustment interval of the gait cycle.

[0118] The power supply module 340 can use a lithium battery pack with an output voltage of 12V and a capacity of 20Ah, providing power support for the above-mentioned units and actuators, and ensuring that the exoskeleton robot 300 can work continuously for more than 4 hours.

[0119] It is understood that the sampling rate of each sensor in the multimodal sensor array in this application embodiment (e.g., 100Hz for the inertial measurement unit and 1000Hz for the surface electromyography sensor), the parameter settings of the fusion model (e.g., weight allocation of the attention mechanism, number of network layers and neurons in the deep learning model), the adjustment threshold of adaptive control (e.g., the adjustment range of stride length and stride height for gait parameters, the range of stiffness and damping values ​​for joint impedance), and the coefficients of smooth transition processing (e.g., the α value of the exponential smoothing algorithm) are all adjustable parameterized settings. Those skilled in the art can adapt the application by modifying relevant parameters (e.g., the combination of sensor installation positions and the feature extraction weights of the fusion model) according to the actual application scenario.

[0120] This application provides a gait control device for an exoskeleton robot. See also... Figure 4 The figure is a schematic diagram of the gait control device for an exoskeleton robot provided in an embodiment of this application, and its specific implementation corresponds to the method embodiment described above.

[0121] This application provides a gait control device 4100 for an exoskeleton robot, comprising: The system comprises a first processing unit 4101, a determining unit 4102, an adjusting unit 4103, a second processing unit 4104, and a control unit 4105. The first processing unit 4101 is used to preprocess, extract features, and fuse multi-dimensional data collected by the multi-modal sensor array to obtain fused features; the multi-dimensional data includes motion data of the exoskeleton robot, motion data of the user wearing the exoskeleton robot, human-computer interaction forces, and environmental data of the user. The determining unit 4102 is used to determine the user's motion intention type, motion state, and environmental state based on the fusion features. The adjustment unit 4103 is used to adjust at least one of the step length, step height, and gait cycle of the exoskeleton robot based on the user's movement intention type, movement state, and the environmental state in which the user is located; adjust the joint impedance parameters of the exoskeleton robot based on human-computer interaction force characteristics; adjust the joint assist torque of the exoskeleton robot based on the muscle activity intensity characterized by electromyographic signal characteristics; and determine the adjusted step length, step height, and gait cycle, the adjusted joint impedance parameters, and the adjusted joint assist torque as the adjusted gait control parameters. The second processing unit 4104 is used to perform smooth transition processing on the adjusted gait control parameters based on the gait control parameters before adjustment and the adjusted gait control parameters to obtain the processed gait control parameters. The control unit 4105 is used to control the gait of the exoskeleton robot based on the processed gait control parameters.

[0122] In one possible implementation, the first processing unit is specifically used for: Time synchronization processing is performed on multi-dimensional data with different sampling rates to obtain synchronized multi-dimensional data. The multi-dimensional data is subjected to noise reduction processing to obtain noise-reduced multi-dimensional data. The multi-dimensional data after synchronous processing and the multi-dimensional data after noise reduction are standardized to obtain pre-processed multi-dimensional data. Feature extraction and fusion processing are performed on the preprocessed multi-dimensional data to obtain fused features.

[0123] In one possible implementation, the first processing unit is specifically used for: Feature extraction is performed on the preprocessed multi-dimensional data to obtain multimodal features; When there are at least two conflicting features among the multimodal features, the reliability of the conflicting features is tested to determine the confidence level of the conflicting features; The multimodal features are fused based on the confidence level to obtain fused features.

[0124] In one possible implementation, the first processing unit is specifically used for: When there is a first conflicting feature and a second conflicting feature among the multimodal features, based on the current motion state and / or environmental state, the reliability of the first conflicting feature and the second conflicting feature is tested to determine the confidence level of the first conflicting feature and the confidence level of the second conflicting feature. When the confidence level of the first conflicting feature is greater than the confidence level of the second conflicting feature, the second conflicting feature is filtered out, and the remaining multimodal features are fused to obtain the fused feature; or, when the confidence level of the second conflicting feature is greater than the confidence level of the first conflicting feature, the first conflicting feature is filtered out, and the remaining multimodal features are fused to obtain the fused feature. Alternatively, based on the confidence levels of the first conflicting feature and the second conflicting feature, the first conflicting feature and the second conflicting feature are dynamically weighted and fused, and then fused with the remaining multimodal features to obtain the fused feature.

[0125] In one possible implementation, the first processing unit is specifically used for: Acquire the user's historical wear data; the historical wear data includes at least one of the user's exercise habits, physiological characteristics, and environmental adaptation preferences. Based on the historical wear data, the feature extraction weights are determined; Based on the feature extraction weights, feature extraction is performed on the preprocessed multi-dimensional data to obtain multimodal features.

[0126] In one possible implementation, the first processing unit is specifically used for: Feature extraction is performed on the preprocessed multi-dimensional data to obtain multimodal features; Based on the attention mechanism, the spatial and temporal features in the multimodal features are enhanced to obtain the first feature; the spatial features include pressure features of various areas of the user's foot and environmental data features of the user's location, and the temporal features include motion features of various parts of the user's body and electromyographic signal features. The first feature, the human-computer interaction force feature, and the motion feature of the exoskeleton robot are fused to generate a fused feature.

[0127] In one possible implementation, the determining unit is specifically used for: Based on the first feature, the user's motion intention type is determined; Based on the fusion features, the user's motion state and the user's environmental state are determined; the user's motion state includes gait phase type and center of gravity state, and the environmental state includes terrain type and obstacle location information.

[0128] In one possible implementation, the second processing unit is specifically used for: The gait control parameters before and after adjustment are continuously switched using a dynamic interpolation algorithm and / or an exponential smoothing algorithm to obtain the processed gait control parameters.

[0129] In summary, the device provided in this application improves the reliability and stability of the fusion results by jointly processing exoskeleton robot motion data, user motion data, human-computer interaction forces, and environmental data, and by introducing a reliability detection and confidence arbitration mechanism when conflicts exist in multimodal features. Furthermore, by jointly identifying motion intention types, motion states, and environmental states, and mapping different types of perception results to the adjustment processes of gait trajectory parameters, joint impedance parameters, and joint assist torques, the device enhances the targeting and coordination of gait control parameter adjustments. Finally, by performing smooth transition processing on the gait control parameters before and after adjustment, the device improves control continuity, stability, and environmental adaptability.

[0130] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A gait control method for an exoskeleton robot, characterized in that, include: The multi-dimensional data collected by the multimodal sensor array is preprocessed, feature extracted, and fused to obtain fused features; The multi-dimensional data includes the motion data of the exoskeleton robot, the motion data of the user wearing the exoskeleton robot, human-computer interaction forces, and the environmental data of the user. Based on the fusion features, the user's motion intention type, motion state, and environmental state are determined. Based on the user's movement intention type, movement state, and the environmental state of the user, adjust at least one of the following: stride length, stride height, and gait cycle of the exoskeleton robot; adjust the joint impedance parameters of the exoskeleton robot based on human-computer interaction force characteristics. The joint assist torque of the exoskeleton robot is adjusted based on the muscle activity intensity characterized by electromyographic signal features. The adjusted stride length, stride height, and at least one of the gait cycle, the adjusted joint impedance parameters, and the adjusted joint assist torque are determined as the adjusted gait control parameters. Based on the gait control parameters before adjustment and the gait control parameters after adjustment, the adjusted gait control parameters are subjected to a smooth transition process to obtain the processed gait control parameters. The gait of the exoskeleton robot is controlled based on the processed gait control parameters.

2. The method according to claim 1, characterized in that, The process of preprocessing, feature extraction, and fusion of multi-dimensional data acquired by the multimodal sensor array to obtain fused features includes: Time synchronization processing is performed on multi-dimensional data with different sampling rates to obtain synchronized multi-dimensional data. The multi-dimensional data is subjected to noise reduction processing to obtain noise-reduced multi-dimensional data; The multi-dimensional data after synchronous processing and the multi-dimensional data after noise reduction are standardized to obtain pre-processed multi-dimensional data. Feature extraction and fusion processing are performed on the preprocessed multi-dimensional data to obtain fused features.

3. The method according to claim 2, characterized in that, The step of extracting and fusing features from the preprocessed multi-dimensional data to obtain fused features includes: Feature extraction is performed on the preprocessed multi-dimensional data to obtain multimodal features; When at least two conflicting features exist among the multimodal features, the reliability of the conflicting features is tested to determine the confidence level of the conflicting features; The multimodal features are fused based on the confidence level to obtain fused features.

4. The method according to claim 3, characterized in that, When at least two conflicting features exist among the multimodal features, reliability testing is performed on the conflicting features to determine their confidence level, including: When there is a first conflicting feature and a second conflicting feature among the multimodal features, based on the current motion state and / or environmental state, the reliability of the first conflicting feature and the second conflicting feature is tested to determine the confidence level of the first conflicting feature and the confidence level of the second conflicting feature. The process of fusing the multimodal features based on the confidence level to obtain fused features includes: When the confidence level of the first conflicting feature is greater than the confidence level of the second conflicting feature, the second conflicting feature is filtered out, and the remaining multimodal features are fused to obtain the fused feature; or, when the confidence level of the second conflicting feature is greater than the confidence level of the first conflicting feature, the first conflicting feature is filtered out, and the remaining multimodal features are fused to obtain the fused feature. Alternatively, based on the confidence levels of the first conflicting feature and the second conflicting feature, the first conflicting feature and the second conflicting feature are dynamically weighted and fused, and then fused with the remaining multimodal features to obtain the fused feature.

5. The method according to claim 2, characterized in that, The feature extraction of the preprocessed multi-dimensional data includes: Acquire the user's historical wear data; the historical wear data includes at least one of the user's exercise habits, physiological characteristics, and environmental adaptation preferences. Based on the historical wear data, the feature extraction weights are determined; Based on the feature extraction weights, feature extraction is performed on the preprocessed multi-dimensional data to obtain multimodal features.

6. The method according to claim 2, characterized in that, The step of extracting and fusing features from the preprocessed multi-dimensional data to obtain fused features includes: Feature extraction is performed on the preprocessed multi-dimensional data to obtain multimodal features; Based on the attention mechanism, the spatial and temporal features in the multimodal features are enhanced to obtain the first feature; the spatial features include pressure features of various areas of the user's foot and environmental data features of the user's location, and the temporal features include motion features of various parts of the user's body and electromyographic signal features. The first feature, the human-computer interaction force feature, and the motion feature of the exoskeleton robot are fused to generate a fused feature.

7. The method according to claim 6, characterized in that, The process of determining the user's motion intention type, motion state, and environmental state based on the fusion features includes: Based on the first feature, the user's motion intention type is determined; Based on the fusion features, the user's motion state and the user's environmental state are determined; the user's motion state includes gait phase type and center of gravity state, and the environmental state includes terrain type and obstacle location information.

8. The method according to claim 1, characterized in that, The process of smoothing the adjusted gait control parameters to obtain processed gait control parameters includes: The gait control parameters before and after adjustment are continuously switched using a dynamic interpolation algorithm and / or an exponential smoothing algorithm to obtain the processed gait control parameters.

9. A gait control device for an exoskeleton robot, characterized in that, include: The system comprises a first processing unit, a determining unit, an adjusting unit, a second processing unit, and a control unit. The first processing unit is used to preprocess, extract features, and fuse multi-dimensional data collected by the multimodal sensor array to obtain fused features; The multi-dimensional data includes the motion data of the exoskeleton robot, the motion data of the user wearing the exoskeleton robot, human-computer interaction forces, and the environmental data of the user. The determining unit is used to determine the user's motion intention type, motion state, and environmental state based on the fusion features. The adjustment unit is used to adjust at least one of the step length, step height, and gait cycle of the exoskeleton robot based on the user's movement intention type, movement state, and the environmental state in which the user is located; and to adjust the joint impedance parameters of the exoskeleton robot based on human-machine interaction force characteristics. The joint assist torque of the exoskeleton robot is adjusted based on the muscle activity intensity characterized by electromyographic signal features. The adjusted stride length, stride height, and at least one of the gait cycle, the adjusted joint impedance parameters, and the adjusted joint assist torque are determined as the adjusted gait control parameters. The second processing unit is used to perform smooth transition processing on the adjusted gait control parameters based on the gait control parameters before adjustment and the adjusted gait control parameters to obtain the processed gait control parameters. The control unit is used to control the gait of the exoskeleton robot based on the processed gait control parameters.

10. An exoskeleton robot, characterized in that, The exoskeleton robot includes: a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the steps of the gait control method for an exoskeleton robot as described in any one of claims 1-8, according to the instructions in the program code.

Images