Remote rehabilitation robot control method, system, equipment, medium and product

Abstract

The invention discloses a control method, system and equipment of a remote rehabilitation robot, a medium and a product, and belongs to the technical field of robots. The method comprises the steps of obtaining multi-modal data of a patient end, obtaining a time delay prediction value through prediction of a long short-term memory network model, dynamically adjusting admittance parameters of an admittance model according to the multi-modal data and the time delay prediction value, and finally controlling the remote rehabilitation robot according to the dynamically adjusted admittance parameters. According to the method, the admittance parameters of the admittance model are dynamically adjusted in real time based on the time delay predicted value predicted by the long-short-term memory network model and the electromyographic signals of the patient in the multi-modal number, dual adaptation of the network state and the physiological features of the patient is achieved, the admittance parameters are dynamically adjusted through the electromyographic signals, and the change requirement of the complete rehabilitation cycle of the patient is met; admittance parameters are dynamically adjusted through the time delay prediction value reflecting the network time delay state, and the time delay disturbance requirement in the remote communication environment is met.

Description

Technical Field

[0001] The present application relates to the field of robotics, and in particular to a control method, system, device, medium, and product of a remote rehabilitation robot. Background Art

[0002] In recent years, the number of stroke patients has increased year by year. The golden recovery period for patients is within three months after the onset of the disease. During this period, continuous training must be maintained, and rehabilitation equipment such as robots are often needed to assist in training.

[0003] In scenarios where robots are needed to assist in training, especially those requiring remote robots, related technologies generally use an admittance control solution. First, the force / torque signal applied by the patient is collected in real time through a force sensor. Then, a motion trajectory is generated based on preset fixed admittance parameters. Finally, the joint motors of the robotic arm are adjusted according to the motion trajectory to achieve trajectory tracking. However, the admittance parameters of this solution are fixed and cannot adapt to the changes in patient needs during different rehabilitation cycles or the time delay disturbances in remote communication environments. This will result in a single rehabilitation training model that cannot be personalized; the system will become unstable when directly transplanted to a remote scenario.

[0004] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0005] In order to solve at least one of the above technical problems, the main purpose of the embodiments of the present application is to propose a control method, system, device, medium and product of a remote rehabilitation robot, which can dynamically adjust the admittance parameters.

[0006] To achieve the above objectives, one aspect of an embodiment of the present application provides a control method for a remote rehabilitation robot, the method comprising the following steps:

[0007] Acquiring multimodal data from a patient, wherein the multimodal data includes a force signal applied by the patient and an electromyographic signal of the patient;

[0008] Obtain the historical delay sequence and then obtain the delay prediction value through the long short-term memory network model;

[0009] Dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value;

[0010] The remote rehabilitation robot is controlled according to the dynamically adjusted admittance parameters.

[0011] In some embodiments, the force signal comprises a force, and the method further comprises the steps of:

[0012] According to the time delay prediction value, a feedforward-feedback composite control architecture is used to dynamically adjust the trajectory compensation gain of the remote rehabilitation robot; in the feedforward-feedback composite control architecture, the feedforward term is used to offset the force differential deviation caused by the time delay, and the feedback term is used to suppress the trajectory tracking error. The feedforward term is determined according to the rate of change of the force and the time delay prediction value, and the feedback term is determined according to the trajectory tracking error of the remote rehabilitation robot and the rate of change of the trajectory tracking error.

[0013] In some embodiments, the admittance parameters include a damping coefficient and a stiffness coefficient. When the delay prediction value is greater than a preset delay threshold, dynamically adjusting the admittance parameters of the admittance model based on the multimodal data and the delay prediction value includes:

[0014] Adjust the stiffness coefficient to 0;

[0015] The damping coefficient is dynamically adjusted according to the electromyographic signal.

[0016] In some embodiments, the admittance parameters include a damping coefficient and a stiffness coefficient, and dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the delay prediction value includes:

[0017] Dynamically adjusting the stiffness coefficient according to the time delay prediction value;

[0018] The damping coefficient is dynamically adjusted according to the electromyographic signal.

[0019] In some embodiments, the formula used when dynamically adjusting the stiffness coefficient is:

[0020] K(t)=200×e ∧ (-0.023×ΔT_pred),

[0021] Where K(t) is the stiffness coefficient, e is the base of the natural logarithm, and ΔT_pred is the predicted delay value;

[0022] The formula used when dynamically adjusting the damping coefficient is:

[0023] B(t)=15+30 / [1+e ∧ (-0.12×(EMG_RMS-0.25))],

[0024] Where B(t) is the damping coefficient, e is the base of the natural logarithm, and EMG_RMS is the root mean square value of the electromyographic signal.

[0025] In some embodiments, the multimodal data further includes angles of mechanical arm joints in the remote rehabilitation robot, and the method further includes the following steps:

[0026] A 5G dual-channel transmission architecture is used for data transmission, wherein the 5G dual-channel transmission architecture includes a control channel and a detection channel. The control channel uses uRLLC slices to transmit control data, and the detection channel uses eMBB slices to transmit video streams. The video stream uses H.265 encoding combined with FEC forward error correction strategy, and the control data is encapsulated into a data packet using the UDP protocol. The structure of the data packet is: timestamp + value of the force signal + value of the electromyography signal + angle of the robotic arm joint.

[0027] In some embodiments, the method further comprises at least one of the following steps:

[0028] When the torque exceeds a preset torque threshold, mechanical limiting is performed by automatically tripping the elastic drive;

[0029] When the 5G signal strength is lower than a preset strength threshold for a preset time period, switching to local cache control mode; or when the temperature of the mechanical arm joint in the remote rehabilitation robot is higher than a preset temperature threshold, reducing the frequency of the motor power of the mechanical arm joint;

[0030] Heartbeat packets are sent through an independent NB-IoT channel, triggering the local emergency protocol when the network is disconnected.

[0031] In some embodiments, the long short-term memory network model is deployed on an edge computing node, and obtaining a historical delay sequence and then obtaining a delay prediction value through the long short-term memory network model includes:

[0032] Acquire, by the edge computing node, a delay sequence of the first duration before the current moment as the historical delay sequence;

[0033] Inputting the historical delay sequence into the long short-term memory network model through the edge computing node, and outputting a predicted value within a second time period in the future as the delay prediction value;

[0034] The dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value includes: dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value through the edge computing node.

[0035] To achieve the above objectives, another aspect of the present application provides a control system for a remote rehabilitation robot, the system comprising:

[0036] A data acquisition module, configured to acquire multimodal data from a patient, wherein the multimodal data includes a force signal applied by the patient and an electromyographic signal of the patient;

[0037] The delay prediction module is used to obtain the historical delay sequence and then obtain the delay prediction value through the long short-term memory network model;

[0038] A dynamic adjustment module, configured to dynamically adjust the admittance parameters of the admittance model according to the multimodal data and the delay prediction value;

[0039] The control module is used to control the remote rehabilitation robot according to the dynamically adjusted admittance parameters.

[0040] To achieve the above objectives, another aspect of the present application provides an electronic device, including:

[0041] at least one processor;

[0042] at least one memory for storing at least one program;

[0043] When the at least one program is executed by the at least one processor, the at least one processor implements the aforementioned method.

[0044] To achieve the above objectives, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described above is implemented.

[0045] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides a computer program product, including a computer program, which implements the above-mentioned method when executed by a processor.

[0046] The embodiments of the present application include at least the following beneficial effects: The present application provides a control method, system, device, medium, and product for a remote rehabilitation robot. The solution obtains multimodal data from the patient end, obtains a delay prediction value through a long-short-term memory network model, and then dynamically adjusts the admittance parameters of the admittance model based on the multimodal data and the delay prediction value. Finally, the remote rehabilitation robot is controlled based on the dynamically adjusted admittance parameters. The solution dynamically adjusts the admittance parameters of the admittance model in real time based on the delay prediction value predicted by the long-short-term memory network model and the patient's electromyographic signal in the multimodal data, achieving dual adaptation of the network state and the patient's physiological characteristics. The dynamic adjustment of the admittance parameters through electromyographic signals adapts to the changing needs of the patient's entire rehabilitation cycle; the dynamic adjustment of the admittance parameters through the delay prediction value reflecting the network delay state adapts to the delay disturbance requirements in the remote communication environment and solves the remote instability problem caused by the fixed traditional admittance parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Schematic diagram of the system architecture and network element equipment of the 5G remote rehabilitation medical system provided in the embodiment of the present application;

[0048] Figure 2 This is a flow chart of a control method for a remote rehabilitation robot provided in an embodiment of the present application;

[0049] Figure 3 is a working principle diagram of the dynamic admittance controller provided in an embodiment of the present application;

[0050] Figure 4 This is a schematic diagram of the actual motion trajectory of a remote rehabilitation robot during rehabilitation training provided by an embodiment of the present application;

[0051] Figure 5 This is an interaction sequence diagram provided by a specific embodiment of the present application;

[0052] Figure 6 This is a fitting curve diagram of time delay-stiffness coefficient and myoelectricity-damping coefficient provided in a specific embodiment of the present application;

[0053] Figure 7 Schematic diagram of the control system of the remote rehabilitation robot provided in the embodiment of the present application;

[0054] Figure 8 A schematic diagram of a hardware structure of an electronic device provided in an embodiment of the present application;

[0055] Figure 9 This is another hardware structure diagram of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are merely examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0057] It will be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0058] The terms "at least one", "plurality", "each", "any", etc. used in this application include "at least one", "two" or more, "plurality" or "each", "any" or "any one", "each" or "any one" as used herein.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0060] Before explaining the embodiments of the present application in detail, some of the nouns and terms involved in the embodiments of the present application are first explained. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0061] 1) The fifth generation mobile communication technology (5th Generation Mobile Networks), referred to as 5G networks, has three major features: enhanced mobile broadband (eMBB), ultra-reliable low-latency communication (uRLLC), and massive machine-type communication (mMTC).

[0062] 2) Admittance model, a control model that describes the motion response of a mechanical system under the action of an external force. Its full English name is Force-to-Motion Admittance Control Model. Its core principle is to achieve precise force control by establishing a dynamic relationship between force input (F) and motion output (x). The mathematical expression is shown in the following formula (1):

[0063] F=M×a×(d 2 x / dt 2 )+B×v×(dx / dt)+K(x) (1)

[0064] Where F is the external force, a is the acceleration, v is the velocity, and the inertia coefficient M (Mass Coefficient) is a physical quantity that characterizes the system's resistance to acceleration changes (unit: kg·m 2 ); Damping Coefficient B: reflects the energy dissipation characteristics of the system during movement (unit: N·s / m); Stiffness Coefficient K: indicates the system's ability to resist deformation (unit: N / m).

[0065] In scenarios where robots are needed to assist stroke patients with rehabilitation training, the following two solutions are generally used:

[0066] (1) Traditional admittance control scheme: The robotic arm collects the force / torque signals applied by the patient in real time through a force sensor (such as ATIMini40), generates a motion trajectory based on preset fixed admittance parameters (i.e., M, B, K), and adjusts the motor of the robotic arm joint through a PID controller to achieve trajectory tracking. However, the admittance parameters of this scheme are fixed and cannot be dynamically adjusted according to the patient's recovery stage.

[0067] (2) Remote rehabilitation system solution: The TCP / IP protocol is used to transmit the robot arm's position data; the doctor receives the patient's motion status through the VR interface. For delay compensation, a fixed-gain feedforward control method is used, superimposing the predicted displacement in the control command, with a gain coefficient of γ = 0.8 (empirical value).

[0068] The admittance parameters (i.e., M, B, and K) of the above two solutions are fixed and cannot adapt to changes in patient needs during different rehabilitation cycles or stages (such as changes in stiffness requirements during muscle strength recovery) or to delay disturbances caused by network conditions in remote communication environments. This results in: a single rehabilitation training model that cannot be personalized; and system instability (e.g., reduced phase margin) when directly ported to remote scenarios.

[0069] In view of this, the embodiments of the present application provide a control method, system, device, medium and product for a remote rehabilitation robot. The solution obtains multimodal data from the patient end, obtains a delay prediction value through a long short-term memory network model prediction, and then dynamically adjusts the admittance parameters of the admittance model based on the multimodal data and the delay prediction value. Finally, the remote rehabilitation robot is controlled based on the dynamically adjusted admittance parameters. The solution dynamically adjusts the admittance parameters of the admittance model in real time based on the delay prediction value predicted by the long short-term memory network model and the patient's electromyographic signal in the multimodal data, achieving dual adaptation of the network state and the patient's physiological characteristics. The admittance parameters are dynamically adjusted through electromyographic signals to adapt to the needs of the patient's full rehabilitation cycle; the admittance parameters are dynamically adjusted through the delay prediction value reflecting the network delay state to adapt to the delay disturbance requirements in the remote communication environment, solving the remote instability problem caused by the fixed traditional admittance parameters.

[0070] The control method of the remote rehabilitation robot provided in the embodiment of the present application can be applied to Figure 1 In the 5G remote rehabilitation medical system shown, the system mainly includes a doctor control terminal (also called a doctor terminal), an edge computing node (also called an edge node), a patient execution terminal (also called a patient terminal) and a medical cloud platform.

[0071] Among them, the doctor control end and the patient execution end can both be terminals or software running on terminals. The terminal can be a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, and car terminal, etc., but is not limited to this. The doctor control end can run doctor control end software, which can be used for doctors to log in, select and load the patient electronic medical records to be loaded from the medical cloud platform, set the initial admittance parameters, establish a 5G dual-channel communication link, receive patient data, provide a VR operation interface, etc. In some embodiments, the doctor control end uses the control channel in the 5G dual-channel communication link to transmit instructions (including initial admittance parameters, etc.) to the edge computing node through the 5G uRLLC network slice.

[0072] The edge computing node is deployed with a long short-term memory network model (i.e., an LSTM model) and is used to establish a delay prediction model based on the LSTM and perform predictive calculations on network delays, dynamically adjust the admittance parameters, encrypt data, and issue configuration instructions such as acquisition configuration or compensation instructions to the patient execution end. In some embodiments, the edge computing node can use general-purpose edge computing devices (such as smart sensors, programmable logic controllers (PLCs), edge routers, ICT converged gateways, etc.), or it can use embedded NPUs / coprocessors, AI acceleration modules and FPGAs, cloud-based AI inference chips, etc. In some embodiments, the edge computing node can also transmit video streams to the patient execution end via 5G eMBB slices.

[0073] Medical cloud platforms are used to store various data (such as patient training data, AI model training data, and medical records), train AI models, and facilitate multi-center collaborative management. Medical cloud platforms store patient medical records through electronic medical record databases. A medical cloud platform can be a server or software running on a server. The server can be configured as a standalone physical server, a server cluster consisting of multiple physical servers, or a distributed system. It can also be configured to provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network.

[0074] The patient execution end is used to perform rehabilitation training movements through the robotic arm, collect biomechanical data of patient training (including but not limited to the force signal applied by the patient, the patient's electromyographic signal and the angle of the robotic arm joint, etc.), and local safety control. Figure 1As shown, the patient's execution end can collect the force signal (including force or torque) applied by the patient through the six-dimensional force sensor, the electromyographic signal of the patient (forearm flexor and extensor muscles) through the electromyographic acquisition module, and the dynamic admittance parameters through the dynamic admittance controller to calculate the motion trajectory of the remote rehabilitation robot arm, etc.

[0075] Figure 1 The key equipment description of the 5G remote rehabilitation medical system is shown in Table 1 below.

[0076] Table 1

[0077]

[0078] Figure 1 The specific application scenarios of the 5G remote rehabilitation medical system shown include:

[0079] (1) Cross-regional rehabilitation treatment: remote collaboration between tertiary hospitals and primary hospitals / community medical centers;

[0080] (2) Home rehabilitation monitoring: Patients use lightweight terminal devices at home and receive remote training guidance from the central hospital;

[0081] (3) Emergency contactless medical care: Contactless rehabilitation treatment is achieved during special periods such as epidemics and disasters;

[0082] (4) Community services for the elderly: Community elderly care centers deploy this system to provide convenient rehabilitation training and fall warning services for the elderly.

[0083] Figure 2 This is an optional flow chart of the control method of the remote rehabilitation robot provided in the embodiment of the present application, which can be applied to Figure 1 In the 5G remote rehabilitation medical system shown in the figure, Figure 1 The edge computing nodes shown are executed. Figure 2 The method may include but is not limited to steps S201 to S204.

[0084] Step S201, acquiring multimodal data from a patient, wherein the multimodal data includes a force signal applied by the patient and an electromyographic signal of the patient;

[0085] Step S202: Obtain a historical delay sequence, and then obtain a delay prediction value through a long short-term memory network model;

[0086] Step S203: dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the delay prediction value;

[0087] Step S204: Control the remote rehabilitation robot according to the dynamically adjusted admittance parameters.

[0088] In step S201, the multimodal data of the patient side can be obtained by Figure 1 The patient execution end can collect data, and the collection cycle can be selected according to actual needs, such as collecting data every 10ms. The multimodal data on the patient end may include but is not limited to the force signal applied by the patient, the patient's electromyographic signal (reflecting the patient's muscle activation state) and the angle or movement speed of the robotic arm joint (reflecting the posture), etc. By collecting multimodal data, the embodiment of the present application can more accurately identify the patient's movement intention and reduce training interruptions caused by mismovement; it supports precise control of complex rehabilitation movements (such as grasping-wrist rotation linkage).

[0089] For example, the specific process of collecting multimodal data on the patient side includes:

[0090] a) A six-axis force sensor is used to collect patient-applied force signals (including force and torque). These signals include the X / Y / Z triaxial forces F_x, F_y, and F_z, and the torques τ_x, τ_y, and τ_z about these three axes. The sampling rate is 500 Hz, and the data format is a floating-point array: [F_x, F_y, F_z, τ_x, τ_y, τ_z]. The measurement error is ±0.1 N.

[0091] b) Surface electromyographic (SEM) sensors were used to collect EMG signals from the forearm flexor and extensor muscles. 16 channels were sampled synchronously at a sampling rate of 2 kHz, and motion artifacts were removed by 50 Hz high-pass filtering.

[0092] c) The angles of the seven joints of the robotic arm are recorded using joint encoders with an accuracy of 0.01° and a data update rate of 1kHz.

[0093] In step S202, in order to accurately predict the state of the network, a long short-term memory network model (i.e., an LSTM model) can be used to construct a delay prediction model. In some embodiments, a delay sequence of a first duration before the current moment can be input into the LSTM model, and a delay prediction value within a second duration in the future is output. The first duration and the second duration can be selected according to actual needs, and this application does not make specific restrictions on this. For example, the first duration is 50ms and the second duration is 30ms.

[0094] In order to solve the problem that the admittance parameters in the related technology are statically solidified, in step S203, a dual adaptation of the network state and the patient's physiological characteristics is achieved through a dynamic parameter adjustment mechanism: the admittance parameters are dynamically adjusted through electromyographic signals to adapt to the changing needs of the patient's full rehabilitation cycle (such as changes in stiffness requirements during the patient's muscle strength recovery process); the admittance parameters are dynamically adjusted through the delay prediction value that reflects the network delay state to adapt to the delay disturbance requirements in the remote communication environment and avoid oscillation caused by phase lag. In some embodiments, the damping coefficient in the admittance parameter can be dynamically adjusted through the electromyographic signal so that the damping coefficient increases when the electromyographic signal is enhanced (i.e., the resistance of rehabilitation training increases) and the damping coefficient decreases when the electromyographic signal is weakened, thereby meeting the changes in stiffness requirements during the patient's muscle strength recovery process. In some embodiments, the stiffness coefficient in the admittance parameter can be dynamically adjusted through the delay prediction value so that the stiffness coefficient can dynamically decay with the network delay, reducing the impact of network delay disturbance.

[0095] Furthermore, relevant variable information such as real-time network status and patient biometrics (delay prediction value, jitter standard deviation, packet loss rate, and movement speed and angle of the robotic arm joint) can be introduced as input into the dynamic adjustment process of the admittance parameters to further improve the adaptability of the admittance parameters.

[0096] After the stiffness coefficient, damping coefficient, etc. in the admittance parameters are obtained by real-time calculation in step S203, these parameters can be transmitted to the dynamic admittance controller in step S204. The dynamic admittance controller uses the dynamic relationship expression between the force input (F) and the motion output (x) in the admittance model shown in the above formula (1) according to these dynamically adjusted parameters, and then calculates the dynamic relationship between the force input (F) and the motion output (x) according to the following formula: Figure 3 The working principle shown is to calculate and generate the motion trajectory of the remote rehabilitation robot (i.e., rehabilitation training trajectory), and finally adjust the joint motors of the robotic arm according to the motion trajectory to achieve trajectory tracking.

[0097] In some embodiments, a rehabilitation training trajectory can be adaptively generated. Based on multimodal data collected in real time from the patient, such as electromyographic signal data, the angle or movement speed of the robotic arm joint, etc., the patient's real-time movement ability (such as maximum muscle strength, joint mobility, etc.) is evaluated. Then, combined with pre-trained AI models, a progressive rehabilitation training trajectory (such as elliptical, sinusoidal, custom paths, etc.) is automatically generated, and the movement amplitude and speed are dynamically adjusted. This eliminates the need to manually adjust the rehabilitation training trajectory parameters, improves training efficiency, shortens the rehabilitation cycle, avoids joint injuries caused by excessive movement amplitude, and improves safety.

[0098] Figure 4A schematic diagram of the actual motion trajectory of a remote rehabilitation robot during rehabilitation training is shown. Waypoint1 and waypoint2 are the two inflection points of the motion path, and base, joint1, joint2, joint3, joint4, joint5, and joint6 are the seven joints of the remote rehabilitation robot's mechanical arm.

[0099] In steps S201 to S204 shown in the embodiment of the present application, the admittance parameters of the admittance model are dynamically adjusted in real time based on the delay prediction value predicted by the long short-term memory network model and the electromyographic signal of the patient in the multimodal data, thereby achieving dual adaptation of the network status and the patient's physiological characteristics. The admittance parameters are dynamically adjusted by the electromyographic signal to adapt to the needs of the patient's full rehabilitation cycle; the admittance parameters are dynamically adjusted by the delay prediction value reflecting the network delay status to adapt to the delay disturbance requirements in the remote communication environment, thereby solving the remote instability problem caused by the fixed traditional admittance parameters.

[0100] In some embodiments, the force signal includes force, and the control method further includes the following steps S205:

[0101] S205 , dynamically adjusting the trajectory compensation gain of the telerehabilitation robot using a feedforward-feedback composite control architecture according to the time delay prediction value.

[0102] The admittance parameters of the related technology are fixed. Therefore, when tracking the trajectory of the remote rehabilitation robot, the compensation gain of the trajectory is also fixed. At this time, due to the limitations of the network status (such as delay jitter), there are errors in trajectory tracking. To this end, the embodiment of the present application adopts a feedforward-feedback composite control architecture, combining the delay prediction value to dynamically adjust the compensation gain (i.e. Figure 3 As shown, the posture deviation is dynamically adjusted according to the admittance control rate H(s) to reduce the trajectory tracking error.

[0103] In this feedforward-feedback composite control architecture, the feedforward term is used to offset force differential deviations caused by time delay, while the feedback term is used to suppress trajectory tracking errors. In some embodiments, the feedforward term can be determined based on the rate of change of force and the predicted time delay, while the feedback term can be determined based on the trajectory tracking error of the telerehabilitation robot and the rate of change of the trajectory tracking error.

[0104] For example, the calculation formula of the feedforward term F_ff is shown in formula (2), and the calculation formula of the feedback term F_fb is shown in formula (3):

[0105] F_ff = 0.65×(dF_ext / dt)× ΔT_pred (2)

[0106] F_fb = 120×e(t) + 18×(de / dt) (3)

[0107] Where: dF_ext / dt: rate of change of patient-applied force (N / ms); 0.65: feedforward gain factor, determined through empirical optimization; e(t) = x_desired - x_actual: tracking error (m), where x_desired is the planned target trajectory and x_actual is the actual trajectory; de / dt: error rate of change (m / ms); 120 and 18: proportional and differential coefficients, determined through Lyapunov stability analysis.

[0108] The embodiment of the present application adds a dynamic adjustment method for the trajectory compensation gain, establishes a closed-loop linkage mechanism of network status (delay) → admittance parameter → compensation gain, realizes adaptive bandwidth expansion of the control loop, and solves the problem of decoupling communication and control.

[0109] In some embodiments, the admittance parameters include a damping coefficient and a stiffness coefficient. When the delay prediction value is greater than a preset delay threshold, step S203 of dynamically adjusting the admittance parameters of the admittance model based on the multimodal data and the delay prediction value may specifically include steps S2031 and S2032:

[0110] Step S2031, adjusting the stiffness coefficient to 0;

[0111] Step S2032: Dynamically adjust the damping coefficient according to the electromyographic signal.

[0112] When the delay prediction value is greater than the preset delay threshold, it means that the network delay has a greater impact on the stability of the remote rehabilitation system, and it is easy to cause oscillation due to phase lag. In order to improve the stability of the remote rehabilitation system, the embodiment of the present application provides a safety constraint mechanism to perform delay compensation control. When the predicted delay exceeds the preset delay threshold, it automatically switches to the reduced-order admittance model to adjust the stiffness coefficient to 0. At this time, only the damping coefficient is retained, and the damping coefficient is dynamically adjusted according to the electromyographic signal. It can be understood that the preset delay threshold can be selected or adjusted according to actual needs, and this application does not limit this. For example, the preset delay threshold can be 80ms.

[0113] In some embodiments, step S203 of dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the delay prediction value may specifically include step S2033 and step S2034:

[0114] Step S2033: dynamically adjust the stiffness coefficient according to the delay prediction value;

[0115] Step S2034: dynamically adjust the damping coefficient according to the electromyographic signal.

[0116] It is understandable that when the delay prediction value is less than or equal to the preset delay threshold, it means that the network delay has little effect on the stability of the remote rehabilitation system at this time, and the stiffness coefficient and the damping coefficient can be dynamically adjusted at the same time. Optionally, the stiffness coefficient in the admittance parameter is dynamically adjusted by the delay prediction value, so that the stiffness coefficient can dynamically decay with the network delay, thereby reducing the delay sensitivity of the remote rehabilitation system. Optionally, the damping coefficient in the admittance parameter is dynamically adjusted by the electromyographic signal, so that the damping is increased when the electromyographic signal is enhanced (that is, the resistance of rehabilitation training is increased), to meet the changes in the stiffness requirements of the patient during muscle strength recovery, to achieve a stepless transition from passive training to active resistance training, and to adapt to the full cycle of patient rehabilitation needs.

[0117] In some embodiments, the formula used to dynamically adjust the stiffness coefficient is as shown in formula (4):

[0118] K(t)=200×e ∧ (-0.023×ΔT_pred) (4)

[0119] Where K(t) is the stiffness coefficient, e is the base of the natural logarithm (approximately 2.718), and ΔT_pred is the delay prediction value. The stiffness coefficient of the embodiment of the present application selects an exponential decay function with e as the base, which reduces the delay effect. 0.023: Attenuation coefficient (unit: ms -1 ), which determines the sensitivity of the stiffness coefficient to time delay changes.

[0120] In some embodiments, the formula used when dynamically adjusting the damping coefficient is as shown in formula (5):

[0121] B(t)=15+30 / [1+e∧(-0.12×(EMG_RMS-0.25))] (5)

[0122] Where B(t) is the damping coefficient, e is the base of the natural logarithm, and EMG_RMS is the root mean square value of the electromyographic signal. e^(-0.12(EMG_RMS-0.25)) is the Sigmoid function, which realizes the smooth mapping from the electromyographic signal to the damping coefficient; 0.12 is the slope adjustment (unit: mV). -1 ), the larger the value, the faster the damping coefficient changes with myoelectricity. 0.25: Myoelectric activation threshold (unit: mV), below which the damping coefficient remains at the minimum value.

[0123] In some embodiments, the multimodal data further includes angles of joints of a robotic arm in the remote rehabilitation robot, and the control method further includes the following step S206:

[0124] Step S206: Use 5G dual-channel transmission architecture for data transmission.

[0125] Remote rehabilitation systems in related technologies generally use 4G communication technology, which has a certain standard deviation in transmission delay, resulting in periodic oscillations in the force feedback loop; video streams and control streams compete for bandwidth, increasing the burst packet loss rate. To this end, the embodiment of the present application is based on 5G network slicing technology and adopts a 5G dual-channel transmission architecture (including a control channel and a detection channel) to dynamically allocate bandwidth resources according to real-time service needs. The control channel uses uRLLC slicing (priority QCI = 82) to transmit control data, and the detection channel uses eMBB slicing to transmit video streams. QoS parameters are dynamically adjusted through the protocol layer.

[0126] In specific implementation, after collecting multimodal data through the 5G communication module on the patient's execution end, the following operations are performed:

[0127] a) Control data encapsulation: UDP is used for transmission, with a 40% header compression. The data packet structure consists of: timestamp + force / torque + RMS value of the electromyographic signal + robot arm joint angle. After encapsulation, data can be sent at a fixed interval (e.g., 10ms) to reduce bandwidth usage.

[0128] b) Video stream encoding: Use H.265 encoding and dynamically adjust the bit rate as shown in formula (6), in Mbps.

[0129] R(t) =1+4×tanh(ΔT_pred / 25) (6)

[0130] Where ΔT_pred is the predicted delay (in milliseconds), and tanh is the hyperbolic tangent function, which is used to smooth bitrate changes. For example, when the predicted delay value ΔT_pred = 20ms, the bitrate is automatically reduced to 2.3Mbps to save bandwidth.

[0131] In some embodiments, video stream encoding may also be combined with a forward error correction (FEC) strategy to reduce the bit error rate of video stream transmission.

[0132] The embodiment of the present application is based on 5G network slicing technology and adopts a 5G dual-channel transmission architecture, which reduces the standard deviation of the transmission delay of control data (instructions) and ensures the stability of the force feedback loop; reduces the burst packet loss rate of the video stream and improves the operating experience of the doctor's control terminal.

[0133] In some embodiments, the control method further comprises at least one of the following steps:

[0134] When the torque exceeds the preset torque threshold, the elastic drive automatically trips to perform mechanical limiting;

[0135] When the 5G signal strength is lower than the preset strength threshold for a preset period of time, the system switches to the local cache control mode; or when the temperature of the robotic arm joint is higher than the preset temperature threshold, the motor power of the robotic arm joint is reduced;

[0136] Heartbeat packets are sent through an independent NB-IoT channel, triggering the local emergency protocol when the network is disconnected.

[0137] The remote rehabilitation system of related technologies has reliability bottlenecks and safety hazards: in emergency situations (such as patient spasms), it is impossible to quickly brake, causing safety problems; there is no emergency strategy, and when the 5G signal strength is extremely weak or the network is disconnected, rehabilitation training can no longer be continued. To this end, the embodiment of the application designs a three-level security protection mechanism, which is executed by the security monitoring module of the patient execution end (for monitoring the status of the network and equipment), specifically including:

[0138] a) Hardware layer protection: The series elastic actuator (SEA) has a built-in mechanical limiter, which automatically trips when the torque exceeds a preset torque threshold (such as ±50N·m), preventing the robotic arm from being damaged by overload.

[0139] b) Control layer protection: When the 5G signal strength (RSRP) is less than the preset strength threshold (such as -110dBm) and lasts for a preset duration (such as 200ms), it switches to local cache control mode to maintain the last valid parameters; when the temperature of the robotic arm joint is greater than the preset temperature threshold (such as 60°C), the motor power is reduced.

[0140] c) Protocol layer protection: Heartbeat packets are sent through an independent NB-IoT channel, triggering a local emergency protocol when the network is disconnected. Basic rehabilitation training can still be carried out within a certain period of time after the network is disconnected. An elliptical trajectory template (such as a long axis of 0.3m and a short axis of 0.2m) is used to track the trajectory of the remote robot.

[0141] In some embodiments, a low-power embedded controller (based on ARM Cortex-M7) is designed for the patient execution end, integrating a 5G communication module and a real-time operating system (RTOS), supporting local caching and rapid recovery during network disconnection, reducing the power consumption of terminal devices, extending battery life, and reducing hardware costs, making it suitable for popularization in primary hospitals and home scenarios.

[0142] The embodiment of the present application improves the security and reliability of the system through a three-level security protection mechanism.

[0143] In some embodiments, the long short-term memory network model can be deployed on an edge computing node, and the edge computing and AI reasoning capabilities of the edge computing node can be utilized to execute steps S202 and S203 to improve the computing efficiency and performance of the system.

[0144] The control method of the present application is described in detail below with reference to a specific embodiment.

[0145] As mentioned above, the related admittance control solutions or remote rehabilitation solutions have the following problems:

[0146] 1) Fixed admittance parameters cannot adapt to the following changes: changes in stiffness requirements during muscle recovery; and time delay disturbances in remote communication environments. This results in: a single rehabilitation training model that cannot be personalized; and system instability (phase margin reduction >40%) when directly transplanted to remote scenarios.

[0147] 2) Communication performance bottleneck: The standard deviation of 4G network latency, σ = 15ms (measured data), causes periodic oscillations in the force feedback loop; video streams and control streams compete for bandwidth, resulting in a burst packet loss rate > 3%; force perception transparency decreases (measured error > 20%); and in emergency situations (such as patient convulsions), rapid braking cannot be achieved, creating safety issues.

[0148] 3) Rigid coupling between the admittance model and the communication environment. Related technologies treat admittance control and network transmission as independent modules and do not establish a parameter linkage mechanism. The admittance model design does not incorporate network state variables (such as delay, jitter, and packet loss rate), and the control algorithm is not optimized for 5G features (such as network slicing and edge computing). This results in the stability of the remote force control system being strongly correlated with communication quality, making it difficult to meet the medical reliability requirement of an average trouble-free working time greater than 1000 hours.

[0149] like Figure 1 、 Figure 5 As shown in FIG, this embodiment realizes high-precision, low-latency remote rehabilitation training through dynamic admittance control and 5G network optimization technology. Figure 5 As shown, the system operation of this specific embodiment is divided into six steps: initialization, data acquisition, network transmission, dynamic calculation, real-time control and safety protection. Each step is closely coordinated to ensure that the robotic arm movements on the patient side are synchronized with the doctor's instructions in real time.

[0150] Step 1: Parameter initialization

[0151] Executing entity: software system at the doctor's control end;

[0152] Trigger condition: Triggered after the doctor logs into the system and selects the patient's electronic medical record.

[0153] Processing action:

[0154] a) Load patient historical data from the medical cloud platform, including Fugl-Meyer motor function score (e.g., 65 points) and maximum muscle strength (e.g., 28N for upper limb flexor muscles);

[0155] b) Set the initial admittance parameters: inertia coefficient M0 = 2.0 kg·m 2 , damping coefficient B0 = 15N·s / m, stiffness coefficient K0 = 200N / m;

[0156] c) Establish a 5G dual-channel communication link. The control channel allocates uRLLC network slices (5QI = 82) to ensure command transmission priority, with a fixed bandwidth of 2 Mbps. The monitoring channel (i.e., detection channel) uses eMBB slices to dynamically allocate bandwidth (1-5 Mbps) to transmit video streams and biosignals (such as force signals and electromyography).

[0157] In step 1, the initial parameters provide a baseline for subsequent dynamic adjustments. For example, for patients with poor muscle strength, the initial damping coefficient B0 can be set to 12 N·s / m to reduce training resistance. The dual-channel design prevents video streams from crowding out control bandwidth. Actual measurements show that the burst packet loss rate has been reduced from 3.2% with traditional solutions to 0.1%.

[0158] It is understandable that in step 1, the initial parameters can be recommended by the AI ​​model (based on the patient's age, medical history, etc.), but additional training of the neural network is required, which increases the complexity of the system.

[0159] Step 2: Multimodal data acquisition

[0160] Execution entity: embedded system at the patient execution end.

[0161] Trigger condition: Executed periodically every 10ms after the embedded system starts.

[0162] Processing action:

[0163] a) A six-axis force sensor (such as the ATI Mini45 sensor) collects force / torque signals applied by the patient, including X / Y / Z triaxial forces and torques around these three axes. The sampling rate is 500 Hz, and the data format is a floating-point array: [F_x, F_y, F_z, τ_x, τ_y, τ_z], with a measurement error of ±0.1 N.

[0164] b) Surface electromyography (EMG) sensors (e.g., the Delsys Trigno system) were used to collect EMG signals from the forearm flexor and extensor muscles. 16 channels were sampled synchronously at a 2 kHz sampling rate and then filtered at 50 Hz to remove motion artifacts.

[0165] c) Joint encoder (a type of position encoder) records the position of the 7 joints of the robot arm (such as Figure 4 The accuracy is 0.01° and the data update rate is 1kHz.

[0166] In step 2, the raw data is time-stamped and aligned (error < 0.5 ms) to provide input for dynamic admittance parameter calculation. For example, if a sudden spasm is detected (EMG RMS value > 0.5 mV), the system will trigger the protection mechanism within 5 ms.

[0167] Step 3: 5G Data Transmission

[0168] Executing entity: 5G communication module on the patient side.

[0169] Trigger condition: Triggered immediately after data collection is completed.

[0170] Processing action:

[0171] a) Control Data Encapsulation: UDP is used for transmission, with a 40% header compression. The data packet structure is: timestamp (4 bytes) + force / torque (24 bytes) + myoelectric RMS value (2 bytes) + joint angle (14 bytes). It is sent every 10ms, with a single packet size of 52 bytes, occupying approximately 41.6 Kbps of bandwidth.

[0172] b) Video stream encoding: Use H.265 encoding and dynamically adjust the bit rate as shown in the previous formula (6).

[0173] In step 3, the data bandwidth is controlled to be reduced to 40% of the traditional solution, and the measured network delay standard deviation is reduced from 15ms to 4.8ms; the dynamic bit rate strategy improves video smoothness and reduces the freeze rate from 8% to 0.5%.

[0174] Step 4: Dynamic admittance parameter calculation

[0175] Execution entity: edge computing node;

[0176] Trigger condition: Completed within 5ms after receiving data from the patient.

[0177] Processing action:

[0178] a) Delay prediction: Input the delay sequence of the past 50ms [ΔT -49 ,ΔT -48 ,...,ΔT0], the LSTM model is used to output the delay prediction value △T_pred for the next 30ms, with an error RMSE of 3.2ms.

[0179] b) Dynamic adjustment of stiffness coefficient, the formula is shown in the previous formula (4).

[0180] c) Dynamic adjustment of the damping coefficient, the formula is shown in the previous formula (5).

[0181] Combine Figure 6The fitting curves of delay-stiffness coefficient and electromyography-damping coefficient show that in step 4, the stiffness decreases by 18% for every 10 ms increase in delay, avoiding oscillation caused by phase lag. The damping increases when the electromyography signal strengthens. For example, after the patient's muscle strength recovers, the damping can gradually increase from 15 N·s / m to 45 N·s / m.

[0182] Step 5: Delay Compensation Control

[0183] Execution entity: real-time controller on the patient side.

[0184] Trigger condition: Execute immediately after receiving the edge computing node instruction (control frequency is 1kHz).

[0185] Processing action: Feedforward-feedback composite control. The feedforward term is used to offset the force differential deviation caused by time delay, and the feedback term suppresses trajectory tracking error. The formulas for the feedforward and feedback terms are shown in the previous equations (2)-(3).

[0186] Through the delay compensation control in step 5, the trajectory tracking error is reduced from 12.7mm in the traditional solution to 3.2mm when the delay is 80ms; and the range of motion can be limited by an elliptical trajectory to prevent the out-of-control robotic arm from hitting the patient.

[0187] Step 6: Multi-layered security

[0188] Executing body: patient-side safety monitoring module.

[0189] Trigger conditions: Real-time monitoring of network and device status.

[0190] Processing action:

[0191] a) Hardware layer protection: The series elastic actuator (SEA) has a built-in mechanical limiter, which automatically trips when the torque exceeds ±50N·m.

[0192] b) Control layer protection: When the 5G signal strength (RSRP) is less than -110dBm for 200ms, the system switches to local cache control mode and maintains the last valid parameters. When the joint temperature is greater than 60°C, the motor power is reduced by 50%.

[0193] c) Protocol layer protection: Send heartbeat packets through an independent NB-IoT channel and trigger local emergency protocols when the network is disconnected.

[0194] The hardware (mechanical) limiter in step 6 prevents the robotic arm from overload damage. It can withstand a transient impact force of 100 N. Basic training can still be performed within 30 minutes after a network outage, and data transmission takes less than 3 seconds.

[0195] This specific embodiment has the following technical effects:

[0196] 1) Improved latency tolerance: In a 5G network jitter environment (latency standard deviation σ = 15ms), the system's maximum stable latency is extended from 35ms to 80ms, and the phase margin remains above 45° (traditional solutions <10°).

[0197] 2) Force transparency optimization: At a delay of 50ms, the force transparency error is reduced from 22.5% to 7.8%, close to the local control level (5%).

[0198] 3) Accelerated emergency response: When a patient experiences a sudden convulsion, the braking response time is shortened from 120ms to 38ms, avoiding secondary injuries.

[0199] This specific embodiment achieves three major breakthroughs through the coordinated design of dynamic admittance parameter adjustment and delay compensation, combined with 5G network slicing and edge computing technology:

[0200] 1) Precise control: Maintains sub-millimeter tracking accuracy with 80ms latency, meeting the medical requirements of stroke rehabilitation.

[0201] 2) Personalized adaptation: EMG-driven damping adjustment covers the patient's entire rehabilitation cycle, supporting a seamless transition from passive training to resistance training.

[0202] 3) High reliability: The three-level safety design ensures system reliability (MTBF) of 2,500 hours, far exceeding the Class III standard for medical equipment.

[0203] This specific embodiment was also experimentally verified: under extreme conditions such as simulated network jitter and sudden load, the system successfully completed a 72-hour continuous trouble-free operation test, and the average daily number of patients served increased to 4 times that of the traditional solution.

[0204] like Figure 7 As shown, the embodiment of the present application also provides a control system for a remote rehabilitation robot, including:

[0205] A data acquisition module 701 is used to acquire multimodal data from a patient, wherein the multimodal data includes a force signal applied by the patient and an electromyographic signal of the patient;

[0206] The delay prediction module 702 is used to obtain the historical delay sequence and then obtain the delay prediction value through the long short-term memory network model;

[0207] A dynamic adjustment module 703 is used to dynamically adjust the admittance parameters of the admittance model according to the multimodal data and the delay prediction value;

[0208] The control module 704 is used to control the remote rehabilitation robot according to the dynamically adjusted admittance parameters.

[0209] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0210] like Figure 8 As shown, an embodiment of the present application further provides an electronic device, including:

[0211] at least one processor 801;

[0212] At least one memory 802, configured to store at least one program;

[0213] When the at least one program is executed by the at least one processor 801 , the at least one processor 801 implements the aforementioned method.

[0214] See also Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0215] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0216] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the control method of the embodiments of this application.

[0217] Input / output interface 903, used to implement information input and output;

[0218] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0219] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0220] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0221] It can be understood that the contents of the above method embodiments are applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0222] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method described above is implemented.

[0223] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiment, the functions specifically implemented by the present storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0224] An embodiment of the present application further provides a computer program product, including a computer program, which implements the aforementioned method when executed by a processor.

[0225] It is understandable that the contents of the above method embodiments are all applicable to the present program product embodiments, the functions specifically implemented by the present program product embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0226] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0227] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0228] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0229] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0230] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0231] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0232] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0233] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0234] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0235] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0236] If the integrated unit is implemented in the form of 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 application, 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, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0237] In summary, this application discloses a control method, system, device, medium, and product for a telerehabilitation robot. Through multi-dimensional collaboration across "network, control, and biology," this application addresses the core pain points of traditional telerehabilitation systems, such as latency sensitivity, fixed parameters, and insufficient security. It also provides innovative solutions for 5G medical applications, specifically including:

[0238] 1) A dynamic admittance parameter adjustment method is provided. Based on network delay prediction (LSTM model) and patient electromyographic signals (EMG), the stiffness coefficient (K) and damping coefficient (B) of the admittance model are adjusted in real time, achieving dual adaptation of network status and patient physiological characteristics, solving the problem of remote instability caused by traditional fixed admittance parameters, and improving the system delay tolerance to 80ms (traditional solutions are only 35ms). By driving damping adjustment through electromyographic signals, a stepless transition from passive training to active resistance is achieved, adapting to the needs of patients throughout the entire rehabilitation cycle.

[0239] 2) A delay compensation control strategy is proposed, which adopts a feedforward-feedback composite control architecture, dynamically adjusts the compensation gain based on the delay prediction value, and ensures system stability under extreme delays through a safety reduction mechanism (stiffness coefficient zeroing + elliptical trajectory constraint). It breaks through the delay limitation of traditional fixed-gain compensation and reduces the trajectory tracking error by 74% under 80ms delay (compared to 12.7mm with the traditional solution); the emergency braking response time is shortened to 38ms (compared to 120ms with the traditional solution), significantly improving patient safety.

[0240] 3) The 5G dual-channel transmission architecture is adopted, with independent allocation of control channels (uRLLC slices) and detection channels (eMBB slices). Differentiated transmission is achieved through UDP header compression and dynamic bit rate technology, reducing the control data packet loss rate to 0.1% (3% for traditional solutions), ensuring the real-time and reliability of force control instructions, and increasing video stream bandwidth utilization by 40%, avoiding competition with control streams for network resources.

[0241] 4) It provides a multi-level security protection mechanism. Through the three-level coordinated protection of hardware limiting (SEA torque tripping), control degradation (local cache control mode), and protocol redundancy (NB-IoT heartbeat monitoring), it can continue training for 30 minutes after a sudden network outage, and the data transmission time is less than 3 seconds (traditional solutions require reinitialization).

[0242] 5) Integrating edge computing with biosignals, a lightweight LSTM model is deployed at edge nodes to fuse time-delay data with electromyographic signals to generate dynamic admittance parameters. The calculation latency is less than 5ms, meeting real-time control requirements. The system supports concurrent access from thousands of terminals, reducing the load on the central cloud. The overload protection response time of the robotic arm joint is less than 5ms, preventing equipment damage and patient injury.

[0243] 6) Dynamic scheduling and QoS optimization of network slicing are achieved: Based on 5G network slicing technology, bandwidth resources are dynamically allocated according to real-time business needs. The control channel adopts uRLLC slicing (priority QCI = 82), and the video channel adopts eMBB slicing (dynamic bit rate 1-5Mbps). QoS parameters are dynamically adjusted through the protocol layer, reducing the standard deviation of control command transmission delay from 15ms to 4.8ms, ensuring the stability of the power control loop; the burst packet loss rate of video streams is reduced to 0.5% (traditional solutions are 8%), improving the doctor's end operation experience.

[0244] 7) Provides a multimodal biosignal fusion algorithm that integrates three modal data, including electromyographic signals (EMG), six-dimensional force feedback, and joint motion angles. A weighted fusion algorithm is used to generate the patient's motion intention vector, which is input into the dynamic admittance controller. The accuracy of motion intention recognition is increased to 95% (80% for traditional solutions), reducing training interruptions caused by mismovements; and supports precise control of complex rehabilitation movements (such as grasping-wrist rotation linkage).

[0245] 8) Provide an adaptive rehabilitation trajectory generation method that automatically generates progressive training trajectories (such as elliptical, sinusoidal, and custom paths) based on the patient's real-time motor ability assessment (such as maximum muscle strength and joint range of motion), and dynamically adjusts the range of motion and speed, thereby improving training efficiency by 30% and shortening the rehabilitation cycle (traditional solutions require manual adjustment of trajectory parameters); avoiding joint injuries caused by excessive range of motion, and improving safety by 40%.

[0246] 9) The patient side supports lightweight terminal embedded systems, designed with a low-power embedded controller (based on ARMCortex-M7), integrated with a 5G communication module and a real-time operating system (RTOS), supports local caching and fast recovery during network disconnection, reduces terminal device power consumption by 60%, and extends battery life to 8 hours (3 hours for traditional solutions); hardware costs are reduced by 50%, making it suitable for popularization in primary hospitals and home scenarios.

[0247] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A control method for a remote rehabilitation robot, characterized in that: The method comprises the following steps: Acquiring multimodal data from a patient, wherein the multimodal data includes a force signal applied by the patient and an electromyographic signal of the patient; Obtain the historical delay sequence and then obtain the delay prediction value through the long short-term memory network model; Dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value; The remote rehabilitation robot is controlled according to the dynamically adjusted admittance parameters.

2. The method according to claim 1, characterized in that The force signal includes a force, and the method further includes the following steps: According to the time delay prediction value, a feedforward-feedback composite control architecture is used to dynamically adjust the trajectory compensation gain of the remote rehabilitation robot; in the feedforward-feedback composite control architecture, the feedforward term is used to offset the force differential deviation caused by the time delay, and the feedback term is used to suppress the trajectory tracking error. The feedforward term is determined according to the rate of change of the force and the time delay prediction value, and the feedback term is determined according to the trajectory tracking error of the remote rehabilitation robot and the rate of change of the trajectory tracking error.

3. The method according to claim 1, characterized in that The admittance parameters include a damping coefficient and a stiffness coefficient. When the delay prediction value is greater than a preset delay threshold, dynamically adjusting the admittance parameters of the admittance model based on the multimodal data and the delay prediction value includes: Adjust the stiffness coefficient to 0; The damping coefficient is dynamically adjusted according to the electromyographic signal.

4. The method according to claim 1, wherein The admittance parameters include a damping coefficient and a stiffness coefficient, and the dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value includes: Dynamically adjusting the stiffness coefficient according to the time delay prediction value; The damping coefficient is dynamically adjusted according to the electromyographic signal.

5. The method according to claim 4, characterized in that The formula used when dynamically adjusting the stiffness coefficient is: K(t)=200×e ∧ (-0.023×ΔT_prev), Where K(t) is the stiffness coefficient, e is the base of the natural logarithm, and ΔT_pred is the predicted delay value; The formula used when dynamically adjusting the damping coefficient is: B(t)=15+30 / [1+e ∧ (-0.12×(EMG_RMS-0.25))], Where B(t) is the damping coefficient, e is the base of the natural logarithm, and EMG_RMS is the root mean square value of the electromyographic signal.

6. The method according to claim 1, characterized in that The multimodal data also includes angles of mechanical arm joints in the remote rehabilitation robot, and the method further includes the following steps: A 5G dual-channel transmission architecture is used for data transmission, wherein the 5G dual-channel transmission architecture includes a control channel and a detection channel. The control channel uses uRLLC slices to transmit control data, and the detection channel uses eMBB slices to transmit video streams. The video stream uses H.265 encoding combined with FEC forward error correction strategy, and the control data is encapsulated into a data packet using the UDP protocol. The structure of the data packet is: timestamp + value of the force signal + value of the electromyography signal + angle of the robotic arm joint.

7. The method according to claim 1, characterized in that The force signal includes a torque, and the method further includes at least one of the following steps: When the torque exceeds a preset torque threshold, mechanical limiting is performed by automatically tripping the elastic drive; When the 5G signal strength is lower than a preset strength threshold for a preset time period, switching to local cache control mode; or when the temperature of the mechanical arm joint in the remote rehabilitation robot is higher than a preset temperature threshold, reducing the frequency of the motor power of the mechanical arm joint; Heartbeat packets are sent through an independent NB-IoT channel, triggering the local emergency protocol when the network is disconnected.

8. The method according to any one of claims 1 to 7, characterized in that The long short-term memory network model is deployed on the edge computing node. The acquisition of the historical delay sequence and the subsequent use of the long short-term memory network model to obtain the delay prediction value include: Acquire, by the edge computing node, a delay sequence of the first duration before the current moment as the historical delay sequence; Inputting the historical delay sequence into the long short-term memory network model through the edge computing node, and outputting a predicted value within a second time period in the future as the delay prediction value; The dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value includes: dynamically adjusting the admittance parameters of the admittance model according to the multimodal data and the time delay prediction value through the edge computing node.

9. A control system for a remote rehabilitation robot, characterized in that: The system comprises: A data acquisition module, configured to acquire multimodal data from a patient, wherein the multimodal data includes a force signal applied by the patient and an electromyographic signal of the patient; The delay prediction module is used to obtain the historical delay sequence and then obtain the delay prediction value through the long short-term memory network model; A dynamic adjustment module, configured to dynamically adjust the admittance parameters of the admittance model according to the multimodal data and the delay prediction value; The control module is used to control the remote rehabilitation robot according to the dynamically adjusted admittance parameters.

10. An electronic device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images