Power-assisted exoskeleton robot based on brain-computer interface and control method thereof

Abstract

The invention relates to the technical field of brain-computer interfaces, and discloses a power-assisted exoskeleton robot based on a brain-computer interface and a control method thereof.The method comprises the steps that S1, electroencephalogram signals of a user are collected, and a mu rhythm power value, a beta rhythm power value and pre-starting signal duration in a motion pre-starting time window are extracted; s2, calculating a joint angle target value of the exoskeleton robot according to the variation amplitude of the mu rhythm power value and the beta rhythm power value and the duration of the pre-starting signal; s3, according to the joint angle target value, controlling each joint to output assistance torque and calculating the mismatch degree of each joint movement, when the mismatch degree exceeds a preset mismatch degree threshold value, adjusting the assistance strength and triggering an alarm, and when it is detected that alpha wave power in the electroencephalogram signal is suddenly reduced, cutting off assistance driving and locking the exoskeleton posture. The motion intention can be recognized 100-300 ms before the user actually initiates the motion, and the synchronism problem that in the prior art, the power-assisted motion lags behind the motion of the user is solved.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface technology, and in particular to a brain-computer interface-based assistive exoskeleton robot and its control method. Background Technology

[0002] Assistive exoskeleton robots can effectively reduce muscle load and prevent occupational injuries by providing mechanical assistance to users.

[0003] Existing assistive exoskeleton robots rely on electromyographic signals or manual control to achieve human-machine interaction. Control signals are only collected during movement, resulting in a significant response lag between the assistive action and the user's movement, affecting the continuity of work. The assist intensity is usually a fixed value or needs to be manually adjusted, and cannot be dynamically adapted to the user's actual load. It is insufficient in high-intensity work, while it has the problem of over-intervention in light-load scenarios. Summary of the Invention

[0004] This invention provides a brain-computer interface-based assistive exoskeleton robot and its control method. By collecting the power changes of μ rhythm and β rhythm within the motion pre-start time window, this invention can identify the motion intention 100-300ms before the user actually initiates the action, enabling the exoskeleton drive system to complete the assistive preparation in advance, thus solving the synchronization problem of the assistive action lagging behind the user's movement in the prior art.

[0005] In a first aspect, the present invention provides a brain-computer interface-based control method for a power-assisted exoskeleton robot, the brain-computer interface-based control method for a power-assisted exoskeleton robot comprising: S1: Collect the user's EEG signal and extract the μ rhythm power value and β rhythm power value within the motor pre-start time window, as well as the duration of the pre-start signal; S2: Calculate the target joint angle value of the exoskeleton robot based on the change amplitude of the μ rhythm power value and the β rhythm power value and the duration of the pre-start signal; S3: Control the output torque of each joint according to the target value of the joint angle and calculate the mismatch degree of each joint movement. When the mismatch degree exceeds the preset mismatch degree threshold, adjust the assist intensity and trigger an alarm. When a sudden drop in alpha wave power in the EEG signal is detected, cut off the assist drive and lock the exoskeleton posture.

[0006] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, step S1 includes: S11: Collect EEG signals at electrode sites in the user's motor cortex and parietal cortex; S12: Calculate the μ rhythm power value of the μ rhythm frequency band and the β rhythm power value of the β rhythm frequency band within the preset time window of the EEG signal, and record the time from the decrease of μ rhythm power to the increase of β rhythm power as the duration of the pre-start signal.

[0007] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, step S2 includes: S21: Calculate the decrease in the μ-rhythm power value and the increase in the β-rhythm power value, and determine the user's movement type based on the decrease and increase. S22: Assign corresponding motion control parameters according to the motion type; S23: Calculate the cognitive load index based on the μ rhythm power value, the β rhythm power value, and the duration of the pre-start signal, and calculate the assist intensity parameter based on the cognitive load index; S24: Define the position vectors of each joint in the exoskeleton robot, and calculate the target value of the joint angle based on the motion control parameters, the position vectors, and the assist strength parameters.

[0008] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, step S23 includes: S231: Multiply the ratio of the β-rhythm power value to the μ-rhythm power value by the duration of the pre-start signal to obtain the cognitive load index; S232: When the cognitive load index exceeds the first numerical threshold, the cognitive load level is determined to be high load level; when the cognitive load index is between the first numerical threshold and the second numerical threshold, the cognitive load level is determined to be medium load level; when the cognitive load index is below the second numerical threshold, the cognitive load level is determined to be low load level. S233: Based on the cognitive load level, calculate the assist intensity parameter corresponding to the cognitive load index and the rated assist intensity according to different linear functions.

[0009] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, step S24 includes: S241: Collect the actual angle and torque values ​​of the lumbar joint, bilateral hip joints and bilateral knee joints in the exoskeleton robot, establish a three-dimensional coordinate system with the user's waist center point as the origin, and define the position vectors of the lumbar joint, left hip joint, right hip joint, left knee joint and right knee joint; S242: Select the corresponding preset motion trajectory template according to the motion control parameters, and calculate the target angle value of each joint; S243: The μ rhythm power value and the β rhythm power value are used as motion trend prediction inputs, the assist intensity parameter is used as control inputs, and the actual angle value and the torque value are used as observation inputs to the Kalman filter to obtain the joint angle target value.

[0010] In conjunction with the first aspect, in a fifth implementation of the first aspect of the present invention, step S242 includes: S2421: Match a preset motion trajectory template according to the motion control parameters, and extract the angle offset of each joint corresponding to the motion type from the preset motion trajectory template; S2423: When performing bending or lifting movements, subtract the corresponding angle offset from the initial angle of the lumbar joint. When performing leg lifting or lifting movements, add the corresponding angle offset to the initial angle of the hip joint to obtain the target angle value of each joint.

[0011] In conjunction with the first aspect, in the sixth implementation of the first aspect of the present invention, step S243 includes: S2431: Calculate the motion trend coefficient based on the μ rhythm power value and the β rhythm power value, and multiply the motion trend coefficient with the assist intensity parameter and the target angle value of each joint to obtain the predicted angle value of each joint; S2432: Calculate the angle deviation between the predicted angle value and the actual angle value, and the torque deviation between the predicted torque and the torque value; calculate the Kalman gain coefficient based on the angle deviation value and the torque deviation value. S2433: Multiply the Kalman gain coefficient by the angle deviation value to obtain the angle correction amount, and add the predicted angle value to the angle correction amount to obtain the joint angle target value.

[0012] In conjunction with the first aspect, in the seventh implementation of the first aspect of the present invention, step S3 includes: S31: Multiply the assist strength parameter by the trigonometric function values ​​of the lever arm length and joint angle corresponding to each joint to obtain the assist torque of each joint; S32: Control each joint to output the assist torque according to the target value of the joint angle and calculate the mismatch of the joint movement; S33: When the mismatch degree exceeds the preset mismatch threshold, the assist strength parameter is multiplied by a preset reduction ratio, and the audible and visual alarm device is triggered to output an alarm signal. S34: Monitor the alpha wave power in the alpha wave band of the EEG signal. When the alpha wave power decreases by more than the sudden drop threshold within a preset time, control each joint to stop outputting assist torque and close the electromagnetic brake to lock the current angle of each joint of the exoskeleton.

[0013] In conjunction with the first aspect, the eighth implementation of the first aspect of the present invention further includes: S4: Collect the actual torque of each joint, calculate the target torque of each joint based on the assist strength parameters, lever arm length and joint angle, calculate the deviation between the actual torque and the target torque and convert it into a deviation percentage; S5: When the deviation percentage exceeds the preset deviation threshold, the actual torque is divided by the product of the assist strength parameter, the lever arm length and the trigonometric function value to obtain the adjustment coefficient. The assist strength parameter is then multiplied by the adjustment coefficient to obtain the adjusted assist strength parameter.

[0014] In a second aspect, the present invention provides a power-assisted exoskeleton robot, the power-assisted exoskeleton robot comprising: The acquisition module is used to acquire the user's EEG signals and extract the μ-rhythm power value and β-rhythm power value, as well as the duration of the pre-start signal, within the motor pre-start time window; The calculation module is used to calculate the target joint angle value of the exoskeleton robot based on the change amplitude of the μ rhythm power value and the β rhythm power value and the duration of the pre-start signal; The alarm module is used to control the output torque of each joint according to the target value of the joint angle and calculate the mismatch degree of each joint movement. When the mismatch degree exceeds the preset mismatch degree threshold, the assist intensity is adjusted and an alarm is triggered. When a sudden drop in alpha wave power in the EEG signal is detected, the assist drive is cut off and the exoskeleton posture is locked.

[0015] The technical solution provided by this invention, by collecting the power changes of μ-rhythm and β-rhythm within the pre-start time window, can identify the user's movement intention 100-300ms before the user actually initiates the movement, enabling the exoskeleton drive system to complete the assistance preparation in advance, thus solving the synchronization problem of the assistance action lagging behind the user's movement in the prior art. By calculating the cognitive load index by combining the β-rhythm to μ-rhythm power ratio with the duration of the pre-start signal, real-time assessment of the user's muscle load status is achieved, and the assistance intensity parameters are dynamically adjusted based on the cognitive load level, allowing the exoskeleton to adaptively output assistance according to the task intensity, avoiding insufficient assistance in high-load scenarios and excessive intervention in low-load scenarios. This invention fuses EEG pre-start signals, joint angles, and torque data through a Kalman filter, which improves the accuracy and anti-interference ability of movement trajectory planning compared to traditional solutions that rely on only a single signal source. By calculating the mismatch between brainwave commands and actual joint movements in real time, the assist intensity is automatically reduced and an alarm is triggered when the mismatch exceeds a preset threshold. At the same time, the system monitors for a sudden drop in alpha wave power to identify the user's stress state and executes emergency cut-off protection, ensuring that the exoskeleton's assist output always matches the user's actual needs. This improves the synchronization, adaptability, accuracy, and safety of the assist exoskeleton robot in complex working environments. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the steps of the brain-computer interface-based assistive exoskeleton robot control method in an embodiment of the present invention; Figure 2 This is a diagram showing the distribution of EEG signal acquisition electrode locations in an embodiment of the present invention; Figure 3 This is a time-frequency characteristic variation curve of the motion pre-start signal in an embodiment of the present invention; Figure 4 This is a piecewise linear mapping diagram of the cognitive load index and the assist intensity parameter in an embodiment of the present invention; Figure 5 This is a schematic diagram of the human joint positions in a standard standing posture according to an embodiment of the present invention. Figure 6 This is a schematic diagram of the joint angle changes during the bending motion in an embodiment of the present invention. Figure 7 This is a schematic diagram of the joint angle change during the leg-lifting action in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of the assistive exoskeleton robot in an embodiment of the present invention. Detailed Implementation

[0018] This invention provides a brain-computer interface-based assistive exoskeleton robot and its control method. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0019] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the brain-computer interface-based control method for a assisted exoskeleton robot in this invention includes: S1: Collect the user's EEG signal and extract the μ rhythm power value and β rhythm power value within the motor pre-start time window, as well as the duration of the pre-start signal; S2: Calculate the target joint angle value of the exoskeleton robot based on the change amplitude of the μ rhythm power value and the β rhythm power value and the duration of the pre-start signal; S3: Control the output torque of each joint according to the target value of the joint angle and calculate the mismatch degree of each joint movement. When the mismatch degree exceeds the preset mismatch degree threshold, adjust the assist intensity and trigger an alarm. When a sudden drop in alpha wave power in the EEG signal is detected, cut off the assist drive and lock the exoskeleton posture.

[0020] Figure 2 This diagram shows the location and distribution of EEG signal acquisition electrodes. The outer ellipse represents the user's head as seen from above, and the frontal root, occipital region, left ear, and right ear are marked to indicate head orientation. Eight main electrode sites are marked: Fz is located in the central frontal region; FCz and Cz are located in the central region; C3 and C4 are located in the left and right motor cortex areas respectively (marked with dashed circles); and Pz, P3, and P4 are located in the parietal cortex area (marked with dashed circles). The motor cortex areas (C3, C4, Cz) primarily acquire motor pre-priming and execution signals, while the parietal cortex areas (P3, P4, Pz) primarily acquire signals related to cognitive load.

[0021] In one specific embodiment, step S1 includes: S11: Collect EEG signals at electrode sites in the user's motor cortex and parietal cortex; S12: Calculate the μ rhythm power value of the μ rhythm frequency band and the β rhythm power value of the β rhythm frequency band within the preset time window of the EEG signal, and record the time from the decrease of μ rhythm power to the increase of β rhythm power as the duration of the pre-start signal.

[0022] Specifically, a lightweight, multi-channel EEG helmet is used, with the internal electrode arrangement focusing on covering the motor cortex (e.g., C3, C4, Cz) and parietal cortex (e.g., P3, P4, Pz), areas that exhibit significant rhythmic changes during actual movement or motor imagery. The EEG helmet's sampling frequency is set above 500Hz to ensure the accuracy of time-domain and frequency-domain feature extraction, while the input impedance is ≥10MΩ to accommodate fluctuations in the contact impedance of human skin electrodes, and the anti-interference capability is ≥60dB to shield against the effects of power frequency, motion artifacts, or environmental electromagnetic interference. During EEG acquisition, the signal is bandpass filtered by the front-end acquisition module, with the filtering range set between 8Hz and 30Hz, effectively preserving μ and β rhythm components while filtering out low-frequency drift and high-frequency noise. At the processing end, short-time Fourier transform is used to perform time-frequency analysis on the filtered signal. The entire EEG time series is divided into several sliding time windows of 50ms each. Within each time window, the energy values ​​of the μ-rhythm band and the β-rhythm band are calculated, and the changes in the energy values ​​of the two bands over time are monitored in real time. When a decreasing trend is detected in the μ-rhythm band energy, i.e., a drop from a stable high value exceeding a preset threshold (e.g., a decrease exceeding 30% of the initial value), and simultaneously a significant increasing trend is detected in the β-rhythm band energy, i.e., an increase in energy value relative to the baseline level exceeding a certain proportion (e.g., 15%~25%), this indicates a pre-motor aura state in which the user is about to initiate movement. The time span between the start of the decrease in μ-rhythm energy and the start of the increase in β-rhythm energy is recorded as the "pre-start signal duration," which is between 100ms and 300ms. The specific value is related to the user's neural reaction speed, task complexity, and attention state, and exhibits individual differences.

[0023] Figure 3The graph shows the time-frequency characteristic changes of the pre-motor initiation signal. The horizontal axis represents time (in milliseconds), with the actual initiation time of movement as 0, ranging from -300ms to +500ms. The vertical axis represents the relative power (normalized value) of the EEG rhythm. The solid line represents the μ-rhythm power curve, which shows a desynchronization decreasing trend within the time window of 300ms to 100ms before movement, with the power decreasing from 1.0 to approximately 0.6. The dashed line represents the β-rhythm power curve, which shows an increasing trend in power within the time window of 200ms to 0ms before movement, rising from approximately 0.4 to close to 1.0. The intersection of the two curves is located at approximately -100ms, indicating that the pre-motor initiation signal characteristics are most significant at this moment. This invention achieves early decoding of movement intention by capturing the combined characteristics of the decreasing μ-rhythm and increasing β-rhythm within the time window of -300ms to -100ms. Compared to existing technologies that can only capture signals after 0ms, this invention can trigger exoskeleton assistance preparation 200-300ms in advance.

[0024] In one specific embodiment, step S2 includes: S21: Calculate the decrease in the μ-rhythm power value and the increase in the β-rhythm power value, and determine the user's movement type based on the decrease and increase. S22: Assign corresponding motion control parameters according to the motion type; S23: Calculate the cognitive load index based on the μ rhythm power value, the β rhythm power value, and the duration of the pre-start signal, and calculate the assist intensity parameter based on the cognitive load index; S24: Define the position vectors of each joint in the exoskeleton robot, and calculate the target value of the joint angle based on the motion control parameters, the position vectors, and the assist strength parameters.

[0025] Specifically, by calculating the percentage decrease of μ-rhythm energy relative to the baseline state and the percentage increase of β-rhythm energy, the magnitudes of decrease and increase are obtained. The decrease in μ-rhythm energy reflects the clarity of the user's movement intention, while the increase in β-rhythm energy reflects the intensity of neural activation leading to an impending active movement. These two magnitudes are input into a pre-trained motion recognition model. The model has multiple rhythmic change templates for typical movements. By comparing these templates with the current rhythmic change feature vector using Euclidean distance, the template with the highest matching degree is selected as the type of user's intended movement, and the corresponding action code is output to identify commands such as "bending over," "lifting a leg," "carrying," or "stopping." Based on the recognition results, a set of motion control parameters corresponding to the movement type is retrieved from a table. These parameters include preset posture curves, target joint combinations, drive priorities, and time synchronization windows. Based on the current μ-rhythm energy, β-rhythm energy, and the duration of the pre-start signal, a cognitive load index is calculated to reflect the degree of neural resource consumption faced by the user in the current task. Dividing the β-rhythm energy by the μ-rhythm energy and multiplying it by the duration of the pre-start signal yields a quantitative index indirectly corresponding to muscle load. Based on the cognitive load index's range, the current load state is divided into three levels: high, medium, and low. According to a preset mapping model, the cognitive load index is converted into specific assist intensity parameters. The assist intensity value determines the exoskeleton's maximum assist capability within the current cycle, limited to between 10kg and 16kg, and linearly adjusted according to the load level. Based on the joint movement requirements indicated by the identified motion type, the spatial position vectors of the joints involved in the exoskeleton robot are defined, including five degrees of freedom: lumbar joints, left and right hip joints, and left and right knee joints. A three-dimensional coordinate system is established with the user's lumbar spine center point as the reference origin, and the spatial vectors of each joint are defined. By combining the target posture requirements in the motion control parameters with the position vectors, and considering the influence of the assist intensity parameter on torque requirements, the target angle values ​​required for each joint are derived using geometric mapping and trigonometric function relationships, forming an array of joint angle target values.

[0026] Figure 4The table shows a piecewise linear mapping relationship between the cognitive load index and the assistance intensity parameter. The horizontal axis represents the cognitive load index (range 0-4.0), and the vertical axis represents the assistance intensity parameter (unit: kg, range 0-18). The mapping relationship is divided into three segments using 1.5 and 2.5 as dividing points: the solid line represents the low load segment (cognitive load index 0-1.5), with an assistance intensity range of 4.8-8 kg and a relatively small slope; the long dashed line represents the medium load segment (cognitive load index 1.5-2.5), with an assistance intensity range of 8-12.8 kg and a steepest slope; and the dotted dashed line represents the high load segment (cognitive load index 2.5-4.0), with an assistance intensity range of 12.8-16 kg. The three curves exhibit different growth rates, reflecting the core mechanism of this invention: providing moderate assistance to preserve the user's active movement ability at low loads, providing sufficient assistance to reduce muscle fatigue at high loads, and using the steepest transition curve to achieve smooth switching at medium loads. This solves the technical problem in existing technologies where fixed assistance intensity cannot adapt to different work scenarios.

[0027] In one specific embodiment, step S23 includes: S231: Multiply the ratio of the β-rhythm power value to the μ-rhythm power value by the duration of the pre-start signal to obtain the cognitive load index; S232: When the cognitive load index exceeds the first numerical threshold, the cognitive load level is determined to be high load level; when the cognitive load index is between the first numerical threshold and the second numerical threshold, the cognitive load level is determined to be medium load level; when the cognitive load index is below the second numerical threshold, the cognitive load level is determined to be low load level. S233: Based on the cognitive load level, calculate the assist intensity parameter corresponding to the cognitive load index and the rated assist intensity according to different linear functions.

[0028] Specifically, by dividing the energy value of the β rhythm band by the energy value of the μ rhythm band, the rhythm energy ratio is obtained, reflecting the proportional relationship between the user's current pre-movement cortical activation intensity and the degree of desynchronization in the resting state. Multiplying the rhythm energy ratio by the corresponding pre-priming signal duration yields a composite expression, namely the cognitive load index. The cognitive load index is compared with a preset first and second numerical threshold. When the cognitive load index is higher than the first threshold (e.g., 2.5), the user is determined to be in a high-load state, meaning that the intensity and duration of their brain activity are significantly enhanced, corresponding to situations such as heavy lifting or continuous work. When the cognitive load index is between the two thresholds (e.g., between 1.5 and 2.5), it is classified as a moderate-load level, representing that the user is performing a moderate-intensity task, such as leg raising, changing direction, or short-term lifting. If the cognitive load index is lower than the second threshold (e.g., less than 1.5), it indicates that the user is in a low-load state, corresponding to light activity or resting preparation. Based on different cognitive load levels, three independent linear mapping functions are introduced to correlate the cognitive load index with the rated assist intensity of the exoskeleton robot (e.g., a maximum of 16 kg), generating corresponding assist intensity parameters. For high load levels, a strong-response linear function is used, causing the assist intensity to increase rapidly with the index value, providing sufficient driving force under high load conditions, with the assist intensity reaching 80% to 100% of the rated value. For medium load levels, a medium-speed ascending function is used, corresponding to an assist intensity of 50% to 80% of the rated value, providing moderate support while preserving some autonomous movement space. For low load levels, a slow-ascending function is used, controlling the assist intensity between 30% and 50%, minimizing interference with the user's active movement capabilities.

[0029] In one specific embodiment, step S24 includes: S241: Collect the actual angle and torque values ​​of the lumbar joint, bilateral hip joints and bilateral knee joints in the exoskeleton robot, establish a three-dimensional coordinate system with the user's waist center point as the origin, and define the position vectors of the lumbar joint, left hip joint, right hip joint, left knee joint and right knee joint; S242: Select the corresponding preset motion trajectory template according to the motion control parameters, and calculate the target angle value of each joint; S243: The μ rhythm power value and the β rhythm power value are used as motion trend prediction inputs, the assist intensity parameter is used as control inputs, and the actual angle value and the torque value are used as observation inputs to the Kalman filter to obtain the joint angle target value.

[0030] Specifically, relying on high-precision angle and torque sensors installed in the various degrees of freedom structures of the exoskeleton, real-time angle and torque output values ​​of five joints—the waist, left hip, right hip, left knee, and right knee—are simultaneously collected at a sampling frequency of 200Hz. A right-handed three-dimensional coordinate system is established with the user's lumbar spine center as the origin, defining the front of the body as the positive X-axis, the left side as the positive Y-axis, and the top as the positive Z-axis. Within this coordinate system, spatial position vectors for the five key joints are constructed, creating a geometric representation consistent with the dynamic posture of the human body. Based on the decoded motion type identifier, a matching standard posture template is selected from the trajectory template library, such as the standard angle range corresponding to actions like "bending over," "lifting a leg," or "carrying." Combined with the posture baseline values ​​obtained during the initial calibration phase, the selected template is projected onto the current coordinate system. The target angle value that each joint should achieve is calculated based on the angle between spatial vectors, resulting in the posture control target under the ideal motion state. A multi-source data fusion module centered on a Kalman filter is constructed. The μ-rhythm energy and β-rhythm energy are input into the state prediction equation as feedforward inputs reflecting the user's intention generation trend. Simultaneously, the assist intensity parameter estimated from cognitive load is input as a control variable, driving the predictive model's response. Real-time acquired joint angle and torque values ​​are used as observation inputs to provide measurement feedback on the posture execution state. The target angle value is corrected and updated through the filter's state update equation. The Kalman state vector is set as a triplet containing the current joint angle, angular velocity, and output torque. Recursive calculations are performed using the system state transition matrix and observation matrix, outputting the fused and corrected optimal joint angle target value in each sampling period.

[0031] Figure 5 This diagram illustrates the positional distribution of human joints in a standard standing posture. The user's waist center is used as the reference point, and the coordinates are marked upwards, forwards, and to the left. The joint connections represent the human skeletal structure. This posture serves as the initial calibration baseline for the exoskeleton robot, requiring the waist to remain vertical (0° angle) and both hip and knee joints to be straight (0° angle). This invention uses the initial angle values ​​of each joint in this standard posture as baseline reference values ​​to ensure that the initial state of the exoskeleton is completely consistent with the user's body posture.

[0032] Figure 6This includes the changes in joint angles during a bending motion, with the positions of joints such as the head, waist, left hip, right hip, left knee, and right knee marked. In this movement, the waist joint tilts forward from an initial angle of 0° to a target angle of -60°, while both hip and knee joints remain essentially straight (within a ±5° fine-tuning range). After identifying the "bending" motion type, the target angle range of the waist joint is calculated based on a three-dimensional coordinate system as -60° to 0°. The exoskeleton's waist drive unit then outputs the corresponding assist torque to help the user complete the bending motion, thereby reducing the load on the waist muscles.

[0033] Figure 7 This includes the changes in joint angles during the leg-lifting motion, with the positions of joints such as the head, waist, left hip, right hip, left knee, and right knee marked. In this movement, the right hip joint rises from an initial angle of 0° to a target angle of +90°, while the right knee joint flexes from an initial angle of 0° to a target angle of +120°. The left lower limb, as the supporting leg, remains upright, and the waist maintains a vertical posture (within ±5°). After identifying the "leg-lifting" motion type, the system calculates the target angle range of the unilateral hip joint (+30° to +90°) and the corresponding knee joint target angle range (+20° to +120°) based on a three-dimensional coordinate system and joint position vectors. It then controls the corresponding hip and knee joint drive units of the exoskeleton to output assist torque, helping the user complete the leg-lifting motion. This is suitable for scenarios such as crossing obstacles and climbing stairs.

[0034] In one specific embodiment, step S242 includes: S2421: Match a preset motion trajectory template according to the motion control parameters, and extract the angle offset of each joint corresponding to the motion type from the preset motion trajectory template; S2423: When performing bending or lifting movements, subtract the corresponding angle offset from the initial angle of the lumbar joint. When performing leg lifting or lifting movements, add the corresponding angle offset to the initial angle of the hip joint to obtain the target angle value of each joint.

[0035] Specifically, a standardized trajectory template library is constructed. This library defines joint control strategies for typical movements according to a "motion type-joint type-angle offset" triple structure, including common movements such as bending, leg raising, lifting, and stopping. For each type of movement, the desired angle variation range is set in different degrees of freedom in the lumbar, hip, and knee joints. For example, for bending, the preset template sets the lumbar flexion angle offset to 60°, while for lifting, it includes a composite angle offset value of lumbar flexion, hip flexion, and slight knee flexion. For leg raising, it is a combination of unilateral hip elevation and ipsilateral knee flexion. Upon receiving motion control parameters, based on the included motion category field, the system automatically locates and extracts the corresponding angle offset data set from the template library, establishing joint angle adjustment rules matching the currently executed task. After successful template matching, the calculation phase begins, acquiring the initial angle values ​​of the five currently calibrated joints. These initial angle values ​​are recorded by angle sensors during the initial posture acquisition phase and stored in the system as individual baseline postures. Based on the action category, a corresponding angle correction strategy is implemented. When the action is identified as "bending over" or "lifting," since the action is mainly characterized by forward trunk leaning, a subtraction operation is performed on the lumbar joint, subtracting a preset forward flexion offset from the initial angle value to obtain the target angle of the lumbar joint. When the action is identified as "raising a leg" or "lifting," since the hip joint is the main force hub, an addition operation is performed on the hip joint, adding the corresponding flexion offset from the template to the initial angle value to obtain the target angle values ​​of the left and right hip joints. For the knee joint, a slight flexion angle offset is appropriately added to complex actions such as lifting, according to the coordination requirements of the specific action, to ensure that the overall action process conforms to ergonomic principles and has good balance. The target angle values ​​of each joint are compiled into a target angle array.

[0036] In one specific embodiment, step S243 includes: S2431: Calculate the motion trend coefficient based on the μ rhythm power value and the β rhythm power value, and multiply the motion trend coefficient with the assist intensity parameter and the target angle value of each joint to obtain the predicted angle value of each joint; S2432: Calculate the angle deviation between the predicted angle value and the actual angle value, and the torque deviation between the predicted torque and the torque value; calculate the Kalman gain coefficient based on the angle deviation value and the torque deviation value. S2433: Multiply the Kalman gain coefficient by the angle deviation value to obtain the angle correction amount, and add the predicted angle value to the angle correction amount to obtain the joint angle target value.

[0037] Specifically, a motion trend coefficient is calculated based on the μ-rhythm energy value and the β-rhythm energy value. The β-rhythm energy value is divided by the μ-rhythm energy value and standardized to reflect the intensity of the movement intention and the clarity of the neural drive trend. The motion trend coefficient is used as a dynamic adjustment factor, multiplied by the assist intensity parameter obtained from the cognitive load assessment, and then multiplied by the corresponding joint target angle values ​​generated by the trajectory template to generate a set of predicted angle values ​​containing five degrees of freedom. Mathematically, the predicted angle values ​​are prior estimates of the Kalman filter, representing a trend prediction of changes in skeletal joint posture over several future sampling periods. The actual joint angle values ​​and actual torque values ​​acquired by the sensors at the current moment are collected and compared with the predicted angle values ​​and the predicted torque values ​​calculated from the assist intensity, lever arm length, and angle trigonometric functions, respectively, to obtain angle deviation values ​​and torque deviation values. The angle deviation value reflects the degree of posture deviation between the executed action and the predicted intention, while the torque deviation value reflects whether the output torque matches the user's actual muscle load. Based on the relationship between the estimated variance and the observation variance constructed using the classic Kalman filter algorithm, and considering the stability of the current observations and the system prediction error, the Kalman gain coefficient is dynamically calculated. The Kalman gain coefficient determines the weight ratio between the predicted and observed values ​​in the next fusion step. When system noise is high or the observation data is stable, the Kalman gain tends to be higher, enhancing the observation correction capability; conversely, when the observation data is unstable or the system model is highly accurate, the gain tends to be lower, maintaining the dominance of the predicted value. The Kalman gain coefficient is multiplied by the angle deviation value to obtain the angle correction amount, which is the dynamic compensation factor for the predicted angle result during actual execution. The predicted angle value and the angle correction amount are added together to obtain the final target angle value for each joint.

[0038] In one specific embodiment, step S3 includes: S31: Multiply the assist strength parameter by the trigonometric function values ​​of the lever arm length and joint angle corresponding to each joint to obtain the assist torque of each joint; S32: Control each joint to output the assist torque according to the target value of the joint angle and calculate the mismatch of the joint movement; S33: When the mismatch degree exceeds the preset mismatch threshold, the assist strength parameter is multiplied by a preset reduction ratio, and the audible and visual alarm device is triggered to output an alarm signal. S34: Monitor the alpha wave power in the alpha wave band of the EEG signal. When the alpha wave power decreases by more than the sudden drop threshold within a preset time, control each joint to stop outputting assist torque and close the electromagnetic brake to lock the current angle of each joint of the exoskeleton.

[0039] Specifically, based on the assist intensity parameter and the corresponding mechanical structural parameters of each joint, the lever arm lengths of the lumbar, hip, and knee joints are introduced respectively. The assist intensity parameter is then multiplied by the trigonometric function value corresponding to the current joint angle. For the lumbar and hip joints, a sine function is used to represent the vertical component of force around the center of rotation, while for the knee joint, a cosine function is used to express the force output trend in the support direction. This yields the theoretical assist torque value that each joint should output. The formula is directly derived from the torque calculation model in biomechanics, where torque equals the product of force and lever arm length, multiplied by the sine or cosine of the angle between the direction of action and the direction of bone segment rotation, to accurately express the effective assist output of the exoskeleton at different joint angles. The assist torque value is input into the drive controller corresponding to each joint. The drive module executes the output according to the set joint angle target value. During execution, real-time joint angle change data is simultaneously collected and compared with the angle target value within the current control cycle to calculate the actual angle deviation. Simultaneously, the theoretical torque value under the predicted state is derived from the torque output model and compared with the actual torque value collected by real-time sensors. The deviation between the two values, along with the joint angle deviation, is used to construct a comprehensive index, "Joint Motion Mismatch," to measure the consistency between the exoskeleton's motion trajectory and the user's neural commands under the current control state. When the mismatch exceeds a preset safety threshold (e.g., 30%), it is determined that there is a serious deviation between the current assist output and the user's actual movement intention. At this point, the mismatch protection mechanism is immediately activated. By multiplying the current assist intensity parameter by a preset reduction ratio (e.g., 0.7), the assist intensity in subsequent control cycles is reduced, thereby decreasing the risk of mechanical intervention due to mismatch. Simultaneously, the audible and visual alarm module is triggered, activating the buzzer to output an intermittent alarm sound and driving the LED to flash a red warning light to alert the user and surrounding monitoring personnel that the system is in an abnormal operating state. The energy level of the alpha wave band in the user's EEG signal is continuously monitored in real time. The alpha wave band, located around 8Hz, is an indicator reflecting the user's attention state and concentration level. When the alpha wave power is detected to drop more than the sudden drop threshold (e.g., a drop of more than 40%) within a preset time window (e.g., 100 milliseconds), it is determined that the user has experienced a sudden stress, mental distraction, or physiological abnormality. At this time, the system's emergency protection mechanism is activated, controlling all drive units to immediately stop the output command of the assist torque and simultaneously driving the electromagnetic brake to close, forcibly locking the current joint angles at the instantaneous posture at the time of the abnormality, in order to prevent collateral damage or malfunctions in the out-of-control state.

[0040] In one specific embodiment, it further includes: S4: Collect the actual torque of each joint, calculate the target torque of each joint based on the assist strength parameters, lever arm length and joint angle, calculate the deviation between the actual torque and the target torque and convert it into a deviation percentage; S5: When the deviation percentage exceeds the preset deviation threshold, the actual torque is divided by the product of the assist strength parameter, the lever arm length and the trigonometric function value to obtain the adjustment coefficient. The assist strength parameter is then multiplied by the adjustment coefficient to obtain the adjusted assist strength parameter.

[0041] Specifically, the actual output torque values ​​of the five joints of the exoskeleton are collected in real time, including the lumbar joint, left hip joint, right hip joint, left knee joint, and right knee joint, and these values ​​are used as the observation input for the current state of the system. Simultaneously, based on the currently determined assist intensity parameters, combined with the known lever arm length and angle values ​​of each joint, the target output torque is calculated for each of the five joints using the standard biomechanical torque calculation formula. The target torque equals the assist intensity parameter multiplied by the lever arm length, and then multiplied by the sine or cosine value corresponding to the current joint angle. The sine function is used to describe drive units rotating around a vertical axis, such as the lumbar and hip joints, while the cosine function is used to describe approximately linear flexion-extension structures such as the knee joint. After completing the target torque calculation, the difference between the actual torque value and the target torque value of each joint is calculated, and the difference is divided by the target torque and then multiplied by 100%, representing the deviation between the current torque output and the theoretical expectation as a percentage. The deviation percentage is compared with a preset deviation threshold. When the deviation percentage of any joint exceeds the threshold (e.g., 15%), it indicates a discrepancy between the current assist output state and the mechanical model prediction. In this case, the current assist intensity parameter is determined to deviate from the actual required support capacity, and adaptive adjustment is necessary. The actual torque value of the current joint is used as the divisor and divided by the theoretical torque calculated from the product of the current assist intensity parameter, the stress arm length, and the angle trigonometric function value. This yields a dimensionless adjustment coefficient, representing the proportion of the current actual output intensity relative to the theoretical output intensity. If the adjustment coefficient is greater than 1, it means the actual output exceeds the model's expectations, indicating that the current assist intensity setting is too high and needs to be corrected downwards; conversely, if it is less than 1, it indicates insufficient assist. The current assist intensity parameter is multiplied by the adjustment coefficient to obtain the adjusted assist intensity parameter.

[0042] The control method for the brain-computer interface-based assistive exoskeleton robot in the embodiments of the present invention has been described above. The assistive exoskeleton robot in the embodiments of the present invention will be described below. Please refer to [link / reference]. Figure 2 One embodiment of the assistive exoskeleton robot in this invention includes: The acquisition module 801 is used to acquire the user's EEG signals and extract the μ rhythm power value and β rhythm power value within the motor pre-start time window, as well as the duration of the pre-start signal. The calculation module 802 is used to calculate the target value of the joint angle of the exoskeleton robot based on the change amplitude of the μ rhythm power value and the β rhythm power value and the duration of the pre-start signal; The alarm module 803 is used to control the output torque of each joint according to the target value of the joint angle and calculate the mismatch degree of each joint movement. When the mismatch degree exceeds the preset mismatch degree threshold, the assist intensity is adjusted and an alarm is triggered. When a sudden drop in alpha wave power in the EEG signal is detected, the assist drive is cut off and the exoskeleton posture is locked.

[0043] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0044] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0045] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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. Such 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 control method for a brain-computer interface-based assistive exoskeleton robot, characterized in that, include: S1: Collect the user's EEG signal and extract the μ rhythm power value and β rhythm power value within the motor pre-start time window, as well as the duration of the pre-start signal; S2: Calculate the target joint angle value of the exoskeleton robot based on the change amplitude of the μ rhythm power value and the β rhythm power value and the duration of the pre-start signal; S3: Control the output torque of each joint according to the target value of the joint angle and calculate the mismatch degree of each joint movement. When the mismatch degree exceeds the preset mismatch degree threshold, adjust the assist intensity and trigger an alarm. When a sudden drop in alpha wave power in the EEG signal is detected, cut off the assist drive and lock the exoskeleton posture.

2. The brain-computer interface-based control method for an assistive exoskeleton robot according to claim 1, characterized in that, Step S1 includes: S11: Collect EEG signals at electrode sites in the user's motor cortex and parietal cortex; S12: Calculate the μ rhythm power value of the μ rhythm frequency band and the β rhythm power value of the β rhythm frequency band within the preset time window of the EEG signal, and record the time from the decrease of μ rhythm power to the increase of β rhythm power as the duration of the pre-start signal.

3. The brain-computer interface-based assistive exoskeleton robot control method according to claim 1, characterized in that, Step S2 includes: S21: Calculate the decrease in the μ-rhythm power value and the increase in the β-rhythm power value, and determine the user's movement type based on the decrease and increase. S22: Assign corresponding motion control parameters according to the motion type; S23: Calculate the cognitive load index based on the μ rhythm power value, the β rhythm power value, and the duration of the pre-start signal, and calculate the assist intensity parameter based on the cognitive load index; S24: Define the position vectors of each joint in the exoskeleton robot, and calculate the target value of the joint angle based on the motion control parameters, the position vectors, and the assist strength parameters.

4. The brain-computer interface-based assistive exoskeleton robot control method according to claim 3, characterized in that, Step S23 includes: S231: Multiply the ratio of the β-rhythm power value to the μ-rhythm power value by the duration of the pre-start signal to obtain the cognitive load index; S232: When the cognitive load index exceeds the first numerical threshold, the cognitive load level is determined to be high load level; when the cognitive load index is between the first numerical threshold and the second numerical threshold, the cognitive load level is determined to be medium load level; when the cognitive load index is below the second numerical threshold, the cognitive load level is determined to be low load level. S233: Based on the cognitive load level, calculate the assist intensity parameter corresponding to the cognitive load index and the rated assist intensity according to different linear functions.

5. The brain-computer interface-based control method for an assistive exoskeleton robot according to claim 3, characterized in that, Step S24 includes: S241: Collect the actual angle and torque values ​​of the lumbar joint, bilateral hip joints and bilateral knee joints in the exoskeleton robot, establish a three-dimensional coordinate system with the user's waist center point as the origin, and define the position vectors of the lumbar joint, left hip joint, right hip joint, left knee joint and right knee joint; S242: Select the corresponding preset motion trajectory template according to the motion control parameters, and calculate the target angle value of each joint; S243: The μ rhythm power value and the β rhythm power value are used as motion trend prediction inputs, the assist intensity parameter is used as control inputs, and the actual angle value and the torque value are used as observation inputs to the Kalman filter to obtain the joint angle target value.

6. The brain-computer interface-based assistive exoskeleton robot control method according to claim 5, characterized in that, Step S242 includes: S2421: Match a preset motion trajectory template according to the motion control parameters, and extract the angle offset of each joint corresponding to the motion type from the preset motion trajectory template; S2423: When performing bending or lifting movements, subtract the corresponding angle offset from the initial angle of the lumbar joint. When performing leg lifting or lifting movements, add the corresponding angle offset to the initial angle of the hip joint to obtain the target angle value of each joint.

7. The brain-computer interface-based control method for an assistive exoskeleton robot according to claim 6, characterized in that, Step S243 includes: S2431: Calculate the motion trend coefficient based on the μ rhythm power value and the β rhythm power value, and multiply the motion trend coefficient with the assist intensity parameter and the target angle value of each joint to obtain the predicted angle value of each joint; S2432: Calculate the angle deviation between the predicted angle value and the actual angle value, and the torque deviation between the predicted torque and the torque value; calculate the Kalman gain coefficient based on the angle deviation value and the torque deviation value. S2433: Multiply the Kalman gain coefficient by the angle deviation value to obtain the angle correction amount, and add the predicted angle value to the angle correction amount to obtain the joint angle target value.

8. The brain-computer interface-based control method for an assistive exoskeleton robot according to claim 1, characterized in that, Step S3 includes: S31: Multiply the assist strength parameter by the trigonometric function values ​​of the lever arm length and joint angle corresponding to each joint to obtain the assist torque of each joint; S32: Control each joint to output the assist torque according to the target value of the joint angle and calculate the mismatch of the joint movement; S33: When the mismatch degree exceeds the preset mismatch threshold, the assist strength parameter is multiplied by a preset reduction ratio, and the audible and visual alarm device is triggered to output an alarm signal. S34: Monitor the alpha wave power in the alpha wave band of the EEG signal. When the alpha wave power decreases by more than the sudden drop threshold within a preset time, control each joint to stop outputting assist torque and close the electromagnetic brake to lock the current angle of each joint of the exoskeleton.

9. The brain-computer interface-based control method for an assistive exoskeleton robot according to claim 1, characterized in that, Also includes: S4: Collect the actual torque of each joint, calculate the target torque of each joint based on the assist strength parameters, lever arm length and joint angle, calculate the deviation between the actual torque and the target torque and convert it into a deviation percentage; S5: When the deviation percentage exceeds the preset deviation threshold, the actual torque is divided by the product of the assist strength parameter, the lever arm length and the trigonometric function value to obtain the adjustment coefficient. The assist strength parameter is then multiplied by the adjustment coefficient to obtain the adjusted assist strength parameter.

10. A power-assisted exoskeleton robot, characterized in that, A method for controlling a brain-computer interface-based assistive exoskeleton robot as described in any one of claims 1-9, comprising: The acquisition module is used to acquire the user's EEG signals and extract the μ-rhythm power value and β-rhythm power value, as well as the duration of the pre-start signal, within the motor pre-start time window; The calculation module is used to calculate the target joint angle value of the exoskeleton robot based on the change amplitude of the μ rhythm power value and the β rhythm power value and the duration of the pre-start signal; The alarm module is used to control the output torque of each joint according to the target value of the joint angle and calculate the mismatch degree of each joint movement. When the mismatch degree exceeds the preset mismatch degree threshold, the assist intensity is adjusted and an alarm is triggered. When a sudden drop in alpha wave power in the EEG signal is detected, the assist drive is cut off and the exoskeleton posture is locked.

Images