Rehabilitation robot adaptive assistance method based on intention response timing coupling

Abstract

The application discloses a rehabilitation robot adaptive assistance method based on intention response time sequence coupling and belongs to the field of robot intelligent control. In view of the fact that the movement intention of a patient with nerve injury and the limb movement exist individual response delay, the robot is prone to early compensation, the self-motion ability evaluation is distorted and other pains, the method synchronously collects multi-modal signals of brain electrical intention, electromyogram, limb movement and robot interaction; a personalized action response window is established at the starting moment of the movement intention; the contribution of external forces such as the robot and gravity is stripped through a dynamics model, and the completion degree of the self-action of the patient is accurately calculated. The system divides multiple assistance control stages, dynamically adjusts the intervention time, the assistance force size and the exit rate according to the time sequence characteristics, the self-completion degree and the safety state, iteratively updates the individual time sequence baseline after training and optimizes the subsequent assistance parameters. The method has multi-level safety control, is suitable for the upper and lower limb rehabilitation training of patients with stroke, spinal cord injury and the like, and has strong real-time performance and practicability.

Description

Technical Field

[0001] This invention belongs to the field of robot intelligent control technology, specifically relating to an adaptive assistance method for rehabilitation robots based on intent-response temporal coupling. Background Technology

[0002] Stroke, spinal cord injury, traumatic brain injury, and other neurological disorders can lead to motor dysfunction in the upper or lower limbs. High-intensity, repetitive, and task-oriented exercise training can help promote neural remodeling. Rehabilitation robots can output controllable motion trajectories, assistive forces, or assistive torques to help patients complete rehabilitation training for joints such as the shoulder, elbow, wrist, fingers, hip, knee, or ankle.

[0003] Existing rehabilitation robots typically employ passive control based on preset trajectories, compliant control based on positional or interaction force errors, intention-triggered control based on electromyography (EMG), or brain-computer interface control based on EEG-based motor imagery. Some on-demand assistive control methods adjust the amount of robot assistance based on the error between the patient's actual trajectory and the target trajectory.

[0004] However, there is often a response delay between a patient's intention to move and the formation of muscle activation and observable limb movement. This delay is related to the patient's degree of nerve damage, fatigue, muscle tone, attentional state, and rehabilitation stage. If the robot intervenes immediately after recognizing the intention to move, it may lead the limb before the patient can initiate voluntary movement, reducing the patient's chances of trying independently. If the robot always uses a fixed waiting time, problems such as late intervention, task failure, or patient fatigue may occur.

[0005] On the other hand, the joint displacement, velocity, and target trajectory completion measured after robot intervention simultaneously include contributions from the patient's voluntary movement, robot assistance, gravity, and passive joint tissues. Directly using the final trajectory completion as an indicator of the patient's ability could lead to the robot-driven movements being mistakenly interpreted as voluntary actions by the patient, resulting in insufficient assistance in the next round, distorted rehabilitation assessments, or safety risks.

[0006] Therefore, a closed-loop control method is needed that can establish an individual action response window based on the start time of the motion intention, identify the patient's autonomous action response and separate the robot-assisted contribution, and determine the timing and stage of assistance intervention based on the temporal coupling state and the patient's autonomous action completion. Summary of the Invention

[0007] The purpose of this invention is to provide an adaptive assistance method for rehabilitation robots based on intention-response temporal coupling, in order to solve the following problems: how to preserve the voluntary action response time of patients in accordance with their neuromotor abilities; how to identify different states such as intention without action, delayed action, and insufficient action; how to distinguish between patients' voluntary actions and actions caused by robot assistance; and how to adjust the timing of robot intervention, the amount of assistance, and the exit conditions according to the above temporal states and the voluntary completion status.

[0008] This invention adopts the following technical solution: an adaptive assist method for rehabilitation robots based on intent-response temporal coupling, comprising the following steps:

[0009] S101. Acquire the patient's motion intention signal, muscle activation signal, limb movement signal, and interaction information of the rehabilitation robot when performing the target rehabilitation movement, and synchronize the signals and interaction information in time.

[0010] S102. Determine the start time of the motion intention based on the motion intention signal, and establish a motion response time window corresponding to the patient with the start time of the motion intention as the starting point.

[0011] S103. Based on the muscle activation signal and limb movement signal, determine the muscle activation time and the voluntary movement start time within the movement response time window, and determine the movement intention-movement response temporal coupling characteristics based on the movement intention start time, muscle activation time and voluntary movement start time.

[0012] S104. Determine the contribution of robot-assisted movement based on the output parameters of the rehabilitation robot, the dynamic parameters of human limbs and the interaction information, and determine the patient's degree of autonomous movement completion based on the actual amount of movement and the contribution of robot-assisted movement.

[0013] S105. Based on the temporal coupling characteristics, the patient's degree of autonomous action completion, and safety status, determine the intervention time and assistance control stage of the rehabilitation robot.

[0014] S106. Generate target auxiliary parameters according to the assistance control stage, and control the rehabilitation robot to assist the patient in performing the target rehabilitation action according to the target auxiliary parameters;

[0015] S107. Update the action response time window and target auxiliary parameters based on the temporal coupling characteristics of subsequent action responses and the patient's degree of autonomous action completion.

[0016] S108. Establish an individual temporal baseline based on the neuro-motor response delay, patient's autonomous action completion rate, and average robot assistance obtained from multiple training sessions. Update the action response time window, initial assistance, maximum assistance, assistance growth rate, assistance exit rate, or task difficulty for the next training session based on the individual temporal baseline.

[0017] Furthermore, the motion intention signal includes at least one of electroencephalogram (EEG) signal, cerebral blood oxygenation signal, or cortical nerve electrical signal; the muscle activation signal includes surface electromyography (EMG) signal; the limb movement signal includes at least one of inertial measurement unit (IMU) signal, joint encoder signal, or visual-motor signal; and the interactive information includes at least one of interactive force, interactive torque, motor current, robot position, velocity, or acceleration.

[0018] The determination of the starting time of the motion intention includes: extracting sensorimotor rhythm, time-frequency energy, power spectral density, spatial filtering features or brain network features from the motion intention signal, obtaining the confidence level of the motion intention through the motion intention recognition model, and determining the moment when the confidence level of the motion intention is not lower than the intention threshold and the signal quality is not lower than the quality threshold for the first time in multiple consecutive analysis windows as the starting time of the motion intention.

[0019] Furthermore, the length of the action response time window is dynamically determined based on at least one of the patient's historical neuro-motor response delay, motor function level, fatigue level, muscle tone state, and recent training stability; the conditions for determining the start time of the voluntary action include at least the limb movement speed reaching a speed threshold, the movement displacement reaching a displacement threshold, the angle between the movement direction and the target direction not exceeding a direction threshold, and the movement duration reaching a duration threshold, and the action is not generated by the rehabilitation robot's active assistance, external propulsion, or abnormal spasticity alone.

[0020] Furthermore, the motion intention-action response temporal coupling feature includes at least two of the following: neuromuscular response delay, neuro-action response delay, and muscle-action response delay; wherein the neuromuscular response delay is the difference between the muscle activation time and the motion intention initiation time, the neuro-action response delay is the difference between the voluntary action initiation time and the motion intention initiation time, and the muscle-action response delay is the difference between the voluntary action initiation time and the muscle activation time.

[0021] Furthermore, the robot-assisted motion contribution is determined by the robot's output torque, robot's output power, or the amount of motion generated after robot intervention; the patient's autonomous output torque is obtained by subtracting the robot's output torque, gravitational torque, and passive joint torque from the measured total joint torque; the patient's autonomous movement completion degree is determined based on at least three of the following: autonomous movement range, target trajectory matching degree, autonomous movement speed, autonomous output torque, muscle activation degree, and temporal coupling degree, and the amount of motion caused solely by the robot-assisted motion contribution is not included in the patient's autonomous movement completion degree.

[0022] Furthermore, the assist control phase includes at least three of the following: autonomous waiting phase, micro-assist wake-up phase, progressive assist phase, assist phase based on completion gap, trajectory correction phase, assist exit phase, and safe takeover phase.

[0023] During the autonomous waiting phase, the rehabilitation robot maintains a state of zero active assistance, gravity compensation, or low impedance. When the patient does not produce an effective movement within the action response time window, it enters the micro-assistance awakening phase or the progressive assistance phase. When the patient produces an effective movement within the action response time window but the patient's autonomous movement completion rate is lower than the completion threshold, it enters the completion gap assistance phase.

[0024] During the trajectory correction phase, the output of the rehabilitation robot is decomposed into a propulsion component along the target trajectory direction and a correction component perpendicular to the target trajectory direction, and the correction component is preferentially used to limit the deviation of the patient's actual trajectory from the target trajectory.

[0025] When the patient's active contribution or the patient's completion of autonomous actions increases in multiple consecutive control cycles, the robot's assistance force is reduced according to the preset assistance withdrawal rate; when the patient's active contribution decreases but effective movement intention is still detected, the robot's assistance force is gradually increased according to the preset assistance growth rate.

[0026] Furthermore, when excessive interaction force, joint overstepping, abnormal reverse movement, spasm, loss of effective signal, communication abnormality, or emergency stop signal is detected, the system enters the safety takeover phase, performing force limiting, force unloading, braking, or stop control on the rehabilitation robot. An individual temporal baseline is established based on the neuro-motor response delay, patient's autonomous movement completion rate, and robot's average assistance amount obtained from multiple training sessions. The action response time window, initial assistance, maximum assistance, assistance growth rate, assistance exit rate, or task difficulty for the next training session are updated based on the individual temporal baseline.

[0027] This invention also provides an adaptive assistance system for rehabilitation robots based on intention-response temporal coupling, used to realize an adaptive assistance method for rehabilitation robots based on intention-response temporal coupling. The system includes: a multimodal signal acquisition unit, a time synchronization and preprocessing unit, a motion intention initiation time detection unit, an autonomous action response detection unit, a robot interaction information processing unit, a temporal coupling and autonomous contribution calculation unit, an assistance stage decision-making unit, a robot control unit, a safety monitoring unit, and an individual baseline update unit.

[0028] The present invention also provides a rehabilitation robot device for implementing an adaptive assistance method for rehabilitation robots based on intent-response temporal coupling. The rehabilitation robot includes a robot body, a drive device, a sensing device, a processor, and a memory. The memory stores a computer program, which is executed by the processor.

[0029] The present invention also provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor for a real-time intention-response-based adaptive assistance method for rehabilitation robots.

[0030] The beneficial effects achieved by the present invention using the above solution are as follows:

[0031] 1. By unifying motor intention, muscle activation, voluntary movement, and robotic intervention onto the same timeline, response delay can be used to characterize the patient's neuromotor state.

[0032] 2. Set an individual action response window after the movement intention appears to reduce the robot's premature intervention from replacing the patient's autonomous attempts.

[0033] 3. By separating the contribution of the robot output, the degree of completion of the patient's autonomous actions is calculated, avoiding the misjudgment of the robot-driven movements as the patient's ability.

[0034] 4. Switch between autonomous waiting, micro-assistance, progressive assistance, gap assistance, trajectory correction, and assistance withdrawal according to the patient's condition to increase active participation.

[0035] 5. Update the individual temporal baseline and assistance boundaries based on the results of multiple training sessions to adapt the control strategy to different stages of the patient's rehabilitation.

[0036] 6. Improve system reliability and safety through mechanisms such as signal quality evaluation, force limitation, joint limitation and spasm recognition. Attached Figure Description

[0037] Figure 1 This is a block diagram of the overall system structure of the present invention;

[0038] Figure 2 This is a flowchart of the adaptive assist control method of the present invention;

[0039] Figure 3 This is a timing diagram illustrating the motion intent, muscle activation, autonomous movement, and robot intervention of the present invention.

[0040] Figure 4 This is a phased assist control state transition diagram of the present invention;

[0041] Figure 5 A flowchart illustrating the patient voluntary movement contribution separation and voluntary movement completion calculation of the present invention;

[0042] Figure 6 This is a flowchart of the closed-loop update process for individual patient time-series baselines according to the present invention.

[0043] The components are as follows: 100—Multimodal signal acquisition unit; 200—Time synchronization and preprocessing unit; 310—Motion intention start time detection unit; 320—Autonomous action response detection unit; 330—Robot interaction information processing unit; 400—Temporal coupling and autonomous contribution calculation unit; 500—Assistance stage decision-making unit; 600—Robot control unit; 700—Safety monitoring unit; 800—Individual baseline update unit. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Equivalent substitutions made by those skilled in the art for signal types, thresholds, models, controllers, or robot structures without departing from the concept of the present invention can all be included in the technical solutions of the present invention. Example 1: System Overall Structure

[0045] like Figure 1 As shown, the system includes a multimodal signal acquisition unit 100, a time synchronization and preprocessing unit 200, a motion intention start time detection unit 310, an autonomous action response detection unit 320, a robot interaction information processing unit 330, a temporal coupling and autonomous contribution calculation unit 400, an assistance stage decision-making unit 500, a robot control unit 600, a safety monitoring unit 700, and an individual baseline update unit 800.

[0046] The multimodal signal acquisition unit 100 acquires EEG, EMG, inertia, joint position, and force or torque information. The time synchronization and preprocessing unit 200 maps data with different sampling rates to a unified clock. The motion intention initiation time detection unit 310 outputs the motion intention initiation time; the autonomous action response detection unit 320 outputs the muscle activation time and the autonomous action initiation time; the robot interaction information processing unit 330 calculates the robot output and the human-robot interaction state. The temporal coupling and autonomous contribution calculation unit 400 calculates the response delay, autonomous output torque, and autonomous action completion degree. The assistance stage decision-making unit 500 selects the control stage; the robot control unit 600 outputs control commands; the safety monitoring unit 700 has safety permissions that take precedence over general control commands; and the individual baseline update unit 800 updates the response window and assistance parameters after training. Example 2: Control Method Flow

[0047] like Figure 2 As shown, this method includes the following steps:

[0048] Step S101: Establish a target rehabilitation task

[0049] Training tasks are established based on the patient's disease type, affected side, range of motion of the joint, muscle strength grade, muscle tone, cognitive status, and rehabilitation goals. Target parameters include the target trajectory. Target speed Target joint angle Maximum permissible trajectory error Maximum Assist Force Maximum auxiliary torque and the range of safe joints.

[0050] Step S102: Acquire and synchronize multimodal signals

[0051] This embodiment uses a 32-channel EEG acquisition device to collect EEG data, surface electromyography electrodes to collect EMG data of the target and antagonistic muscle groups, an IMU and a robot encoder to collect limb movement status, and a six-dimensional force sensor or joint torque sensor to collect human-robot interaction information. Each device is timestamped via hardware triggering, synchronization pulses, or a unified software clock.

[0052]

[0053] Notch filtering, bandpass filtering, bad lead detection, and motion artifact processing are performed on EEG signals; bandpass filtering, rectification, envelope extraction, and individualized normalization are performed on EMG signals; and bias correction, coordinate transformation, and low-pass filtering are performed on motion and force signals. Quality indicators for each signal type are defined. When the quality is below the quality threshold, the modal weight is reduced or entry into the active assist phase is prohibited.

[0054] Step S103: Detect the start time of motion intention

[0055] Sensorimotor rhythms (ERD / ERS), time-frequency energy, power spectral density, CSP / FBCSP features, or deep spatiotemporal features are extracted from EEG signals. Motion intention recognition models can employ LDA, SVM, CNN, EEGNet, recurrent neural networks, Transformer, or graph convolutional networks. The model outputs a motion intention confidence score. .

[0056] Judgment rule: When within N consecutive analysis windows ( (Intention threshold), and EEG signal quality ( When the quality threshold is used, the sampling time at which all constraints are first met is defined as the motion intention start time. :

[0057]

[0058]

[0059] Continuous window constraints are used to filter out false triggers caused by transient noise; if the task is accompanied by external visual and auditory cues, the system performs additional checks. The reasonable time interval between the prompt trigger time.

[0060] Step S104: Establish individual action response window

[0061] Starting time of the intention to move Construct a timing decision window for the left boundary :

[0062] Window duration Instead of using a uniform fixed value for all patients, the historical response baseline is used. Motor function level Fatigue level Spasmodic muscle tone Stability of recent training data Dynamic solution of multivariable functions:

[0063] During the patient's first training, The parameters are set by pre-assessment motion data or manually by rehabilitation physicians; subsequent training is based on iterative updates of the recent average neuro-motor response delay. Within the complete response window, the robot maintains a state of zero active assistance, pure gravity compensation, or high compliance and low impedance by default.

[0064] Step S105: Detect muscle activation and voluntary movement response

[0065] Electromyographic activation indicators of target muscle groups The solution is obtained by fusing the root mean square of the signal, integral electromyography (EMG), envelope amplitude, and muscle synergy features; the adaptive EMG threshold is set based on the mean and standard deviation of resting EMG.

[0066] when The duration of continuous time exceeds the threshold The first moment when the condition is met is recorded as the muscle activation moment. .

[0067] Initiation time of autonomous action Simultaneously satisfying multi-dimensional kinematic constraints: motion velocity ≥ velocity threshold, displacement ≥ displacement threshold, angle between motion direction and target trajectory ≤ direction threshold, effective motion duration ≥ minimum duration; simultaneously adding an exclusion check: this motion is not generated solely by robot active output force, external thrust, or muscle spasm, ultimately resulting in:

[0068]

[0069] In the formula The angle between the real-time movement direction and the target rehabilitation trajectory.

[0070] Step S106: Calculate temporal coupling and autonomous completion degree

[0071] like Figure 3 As shown, the neuromuscular response delay was calculated separately. Neuro-motor response delay and muscle-motor response delay :

[0072] Neuromuscular response delay: Neuro-motor response delay: Muscle-motor response delay:

[0073] The aforementioned delay, motion intention confidence, signal quality of each modality, consistency of movement direction, and muscle coordination are combined to form a temporally coupled feature vector. The temporal coupling degree is obtained through weighting functions, fuzzy logic, probabilistic models, or machine learning models. .

[0074]

[0075] To separate the patient's voluntary movement contribution, a human-robot joint dynamics relationship is established, external force contributions are removed, and the estimated value of the patient's voluntary output torque is obtained. :

[0076]

[0077] In the formula: For the actual measured total joint torque; To output torque to the robot; For the gravitational torque of limbs and robotic devices; This refers to the passive stiffness, damping, and soft tissue resistance torque of the joint.

[0078] The patient's active contribution is obtained by weighting the torque autonomy, electromyographic initiative, and motor autonomy. The weighting coefficients satisfy :

[0079] Autonomous action completion rate The solution integrates five indicators: autonomous motion range, trajectory matching degree, autonomous velocity matching, autonomous output force, and temporal coupling degree, with a weighted sum of 1 and a value range of [missing information]. :

[0080] Define the gap for action completion The displacement and motion generated solely by the robot's assisted force do not participate. calculate.

[0081] Step S107: Determine the timing of power assistance intervention and the power assistance control phase.

[0082] Upon detecting a valid intention to move, the system first enters a voluntary waiting phase. If the patient produces sufficient voluntary movement within the response window, low impedance or gravity compensation is maintained; if electromyographic activation is detected but no valid movement is produced, micro-assisted awakening or progressive assistance is initiated; if partial movement is produced but the voluntary completion rate is below the threshold, assistance based on completion gap is initiated; if the trajectory deviates, trajectory correction is initiated; if the active contribution continues to increase, assistance is withdrawn; if a risk occurs, safe takeover is initiated.

[0083]

[0084] The robot-assisted intervention time t_R can be determined according to the following rules:

[0085] The response window did not perform any effective action throughout the entire process. ;

[0086] Actions are generated within the window, but the degree of autonomous completion is insufficient: ( (Observation time for autonomous movements).

[0087] Autonomous completion rate met, no safety risks: No proactive assistance provided, equivalent to

[0088] in For the observation time of autonomous movements, A threshold for autonomous completion is set. This rule ensures that robotic intervention depends on the patient's response process, rather than solely on a preset task timer.

[0089] Step S108: Generate auxiliary parameters and update individual baselines

[0090] The target assistance force for completing the gap-filling stage can be expressed as:

[0091] in This is a safety correction factor. The value is 1 under normal conditions, decreases when signal quality deteriorates, interaction force approaches its upper limit, or the patient experiences fatigue, and is set to 0 under severely abnormal conditions. To prevent helpful mutations, the help force is smoothed and its rate of change is limited.

[0092]

[0093] During the gradual assistance phase, the assistance starts with the minimum prompt and increases according to the growth rate. Increase; during the exit phase, according to the exit rate. The trajectory correction phase decomposes the output force into a component along the trajectory direction and a normal correction component, prioritizing the limitation of normal error and reducing the robot's substitution for the patient's active propulsion.

[0094] Robots can use impedance control:

[0095]

[0096] in The error between the target trajectory and the actual trajectory. For adaptive stiffness, Adaptive damping can be used. Admittance control, torque control, model predictive control, or position-force hybrid control can also be employed.

[0097] After each training session, the following parameters are calculated: neuromuscular response delay, neuromotor response delay, voluntary movement completion rate, active contribution, robot intervention time, and average assistance. The response baseline for the (k+1)th training session can be updated using the following formula:

[0098]

[0099] When patient response delays decrease and autonomy improves, the system postpones high-intensity assisted intervention, reduces maximum assistance, or increases task difficulty; when performance declines, the system appropriately extends the safety margin, reduces task speed, or increases assistance, but all updates are constrained by the safety boundaries set by the physician. If data is abnormal or unstable, the safety parameters from the previous training cycle are retained. Example 3: Elbow flexion training case

[0100] The patient performed training with the affected elbow joint flexed from an initial angle of 30° to a target angle of 90°. The system acquired EEG, sEMG of the biceps and triceps, forearm IMU, elbow joint encoder, and interactive torque data. Current individual response window. It takes 1.20 seconds.

[0101]

[0102] because Within the response window, the robot does not immediately provide active assistance. After the patient autonomously completes 35% of the target trajectory, their speed gradually decreases, and the system enters a gradual assistance phase after the autonomous action observation period ends. 0.86 40 N If the value is 1, the target assist is calculated based on the completion gap and output after smoothing and maximum rate of change limitation. With... , As the patient rises, the assistive force automatically decreases; once the patient regains stable voluntary movement, the assistive withdrawal phase begins. Example 4: Intentional but unobservable action

[0103] For patients with severe hemiplegia, the system detected an intention to maintain stable movement. However, no effective electromyography (EMG) or voluntary movement was detected within the action response window. The system first outputs visual or tactile cues; if the patient generates a valid movement intention again, it outputs a micro-assistance not exceeding the preset proportion of the maximum assist force to indicate the direction of movement. If target EMG activation occurs but the movement is still insufficient, the system enters progressive assistance; if there is still no valid intention, the robot does not perform the active task action to reduce false triggering. Example 5: Robot Contribution Separation Case

[0104] After robot intervention, the encoder detected a 20° increase in the elbow joint angle. Simultaneously, the robot's output torque was 5 N·m. Based on the dynamic model calculation, the combined torque from gravity and passive tissue forces was 2 N·m, and the total interactive torque was estimated at 7.3 N·m. Therefore, the patient's voluntary output torque was estimated at 0.3 N·m. No significant increase in target electromyography was observed.

[0105] Therefore, the system determines that this segment of motion was mainly generated by the robot, and does not include the entire 20° range in the patient's voluntary movement range, but only the portion corresponding to the patient's voluntary contribution. . Example 6: Security Control

[0106] The safety monitoring unit 700 performs parallel detection of joint angle, speed, interaction force, movement direction, signal quality, spasm, and communication status. When the interaction force reaches a warning threshold, it is reduced... When the stopping threshold is reached, let The value is 0, and unloading or braking is initiated. When the electromyographic co-contraction index suddenly increases, joint velocity decreases, and the reverse interaction force increases, it can be determined as suspected spasticity, and the safe takeover phase can be initiated. The safety control commands have higher priority than the general assistance phase commands. Example 7: Software and Device Implementation

[0107] The above method can be executed by a local industrial control computer, edge computing unit, embedded processor, or server. The motion intention recognition module and the assistive control module can be deployed on the same processor or distributed and communicate via wired or wireless networks. Program instructions are stored in a non-volatile computer-readable storage medium and are executed by the processor to complete the method. The rehabilitation robot can be an end-effector upper limb robot, exoskeleton robot, hand rehabilitation robot, lower limb exoskeleton, gait training robot, or a rehabilitation device combined with functional electrical stimulation.

[0108] In summary, this invention can be used in hospital rehabilitation departments, neurology departments, rehabilitation centers, community rehabilitation institutions, and home rehabilitation equipment. It is suitable for upper limb, hand, or lower limb training for patients with stroke, traumatic brain injury, spinal cord injury, and other neuromuscular dysfunctions. The EEG, EMG, inertial, encoder, and force sensing devices used in the system can all be implemented using existing engineering components, and the related calculation and control methods can be executed in real time on a processor, thus possessing industrial applicability.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0110] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An adaptive assist method for rehabilitation robots based on intent-response temporal coupling, characterized in that, Includes the following steps, S101. Acquire the patient's motion intention signal, muscle activation signal, limb movement signal, and interaction information of the rehabilitation robot when performing the target rehabilitation movement, and synchronize the signals and interaction information in time. S102. Determine the start time of the motion intention based on the motion intention signal, and establish a motion response time window corresponding to the patient with the start time of the motion intention as the starting point. S103. Based on the muscle activation signal and limb movement signal, determine the muscle activation time and the voluntary movement start time within the movement response time window, and determine the movement intention-movement response temporal coupling characteristics based on the movement intention start time, muscle activation time and voluntary movement start time. S104. Determine the contribution of robot-assisted movement based on the output parameters of the rehabilitation robot, the dynamic parameters of human limbs and the interaction information, and determine the patient's degree of autonomous movement completion based on the actual amount of movement and the contribution of robot-assisted movement. S105. Based on the temporal coupling characteristics, the patient's degree of autonomous action completion, and safety status, determine the intervention time and assistance control stage of the rehabilitation robot. S106. Generate target auxiliary parameters according to the assistance control stage, and control the rehabilitation robot to assist the patient in performing the target rehabilitation action according to the target auxiliary parameters; S107. Update the action response time window and target auxiliary parameters based on the temporal coupling characteristics of subsequent action responses and the patient's degree of autonomous action completion. S108. Establish an individual temporal baseline based on the neuro-motor response delay, patient's autonomous action completion rate, and average robot assistance obtained from multiple training sessions. Update the action response time window, initial assistance, maximum assistance, assistance growth rate, assistance exit rate, or task difficulty for the next training session based on the individual temporal baseline.

2. The adaptive assist method for rehabilitation robots based on intent-response temporal coupling according to claim 1, characterized in that, The motion intention signal includes at least one of electroencephalogram (EEG) signal, cerebral blood oxygenation signal, or cortical nerve electrical signal; the muscle activation signal includes surface electromyography (EMG) signal; the limb movement signal includes at least one of inertial measurement unit (IMU) signal, joint encoder signal, or visual-motor signal; and the interactive information includes at least one of interactive force, interactive torque, motor current, robot position, velocity, or acceleration. The determination of the starting time of the motion intention includes: extracting sensorimotor rhythm, time-frequency energy, power spectral density, spatial filtering features or brain network features from the motion intention signal, obtaining the confidence level of the motion intention through the motion intention recognition model, and determining the moment when the confidence level of the motion intention is not lower than the intention threshold and the signal quality is not lower than the quality threshold for the first time in multiple consecutive analysis windows as the starting time of the motion intention.

3. The adaptive assist method for rehabilitation robots based on intent-response temporal coupling according to claim 1, characterized in that, The length of the action response time window is dynamically determined based on at least one of the patient's historical neuro-motor response delay, motor function level, fatigue level, muscle tone state, and recent training stability. The conditions for determining the start time of the voluntary action include at least the limb movement speed reaching a speed threshold, the movement displacement reaching a displacement threshold, the angle between the movement direction and the target direction not exceeding a direction threshold, and the movement duration reaching a duration threshold. Furthermore, the action is not generated solely by the rehabilitation robot's active assistance, external propulsion, or abnormal spasticity.

4. The adaptive assist method for rehabilitation robots based on intent-response temporal coupling according to claim 1, characterized in that, The motor intention-action response temporal coupling feature includes at least two of the following: neuromuscular response delay, neuro-action response delay, and muscle-action response delay; wherein the neuromuscular response delay is the difference between the muscle activation time and the motor intention initiation time, the neuro-action response delay is the difference between the voluntary action initiation time and the motor intention initiation time, and the muscle-action response delay is the difference between the voluntary action initiation time and the muscle activation time.

5. The adaptive assist method for rehabilitation robots based on intent-response temporal coupling according to claim 1, characterized in that, The contribution of robot-assisted motion is determined by the robot's output torque, robot's output power, or the amount of motion generated after robot intervention; the patient's voluntary output torque is obtained by subtracting the robot's output torque, gravitational torque, and passive joint torque from the measured total joint torque; The patient's autonomous movement completion rate is determined based on at least three of the following: autonomous movement range, target trajectory matching degree, autonomous movement speed, autonomous output torque, muscle activation degree, and temporal coupling degree, and the amount of movement contributed solely by robot-assisted movement is not included in the patient's autonomous movement completion rate.

6. The adaptive assist method for rehabilitation robots based on intent-response temporal coupling according to claim 1, characterized in that, The assist control phase includes at least three of the following: autonomous waiting phase, micro-assist wake-up phase, gradual assist phase, assist phase based on completion gap, trajectory correction phase, assist exit phase, and safe takeover phase. During the autonomous waiting phase, the rehabilitation robot maintains a state of zero active assistance, gravity compensation, or low impedance. When the patient does not produce an effective movement within the action response time window, it enters the micro-assistance awakening phase or the progressive assistance phase. When the patient produces an effective movement within the action response time window but the patient's autonomous movement completion rate is lower than the completion threshold, it enters the completion gap assistance phase. During the trajectory correction phase, the output of the rehabilitation robot is decomposed into a propulsion component along the target trajectory direction and a correction component perpendicular to the target trajectory direction, and the correction component is preferentially used to limit the deviation of the patient's actual trajectory from the target trajectory. When the patient's active contribution or the patient's completion of autonomous actions increases in multiple consecutive control cycles, the robot's assistance force is reduced according to the preset assistance withdrawal rate; When the patient's active contribution decreases but effective movement intention is still detected, the robot's assistive force is gradually increased according to the preset assist growth rate.

7. The adaptive assist method for rehabilitation robots based on intent-response temporal coupling according to claim 1, characterized in that, When excessive interaction force, joint overstepping, abnormal reverse movement, spasm, missing effective signal, communication abnormality, or emergency stop signal is detected, the system enters the safety takeover phase, performing force limiting, force unloading, braking, or stop control on the rehabilitation robot. An individual temporal baseline is established based on the neuro-motor response delay, patient's autonomous movement completion rate, and robot's average assistance amount obtained from multiple training sessions. The action response time window, initial assistance, maximum assistance, assistance growth rate, assistance exit rate, or task difficulty for the next training session are updated based on the individual temporal baseline.

8. An adaptive assist system for a rehabilitation robot based on intent-response temporal coupling, used to execute the adaptive assist method for a rehabilitation robot based on intent-response temporal coupling as described in any one of claims 1 to 7, characterized in that, include: The system includes a multimodal signal acquisition unit, a time synchronization and preprocessing unit, a motion intention start time detection unit, an autonomous action response detection unit, a robot interaction information processing unit, a temporal coupling and autonomous contribution calculation unit, an assistance phase decision-making unit, a robot control unit, a safety monitoring unit, and an individual baseline update unit.

9. A rehabilitation robot device for implementing the adaptive assist method for rehabilitation robots based on intention-response temporal coupling as described in any one of claims 1 to 7, characterized in that, It includes a robot body, a drive unit, a sensor, a processor, and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the adaptive assistance method for rehabilitation robots based on intent-response temporal coupling as described in any one of claims 1 to 7.

Images