Massage robot personalized physiotherapy method and system based on cloud platform

Abstract

The invention provides a massage robot personalized physiotherapy method and system based on a cloud platform, and relates to the technical field of massage robots, and the method comprises the steps: carrying out the static posture scanning and dynamic posture capturing of a user; carrying out posture structure modeling based on a multi-modal fusion algorithm; calculating a security boundary, and marking a risk area; a massage path is planned through an improved RRT algorithm, and a personalized massage track is output; executing user massage, and monitoring massage pressure and spine displacement data in real time; abnormal values of massage pressure and spine displacement data are judged in real time, and personalized massage tracks are adjusted according to abnormal conditions. According to the method, an exclusive biomechanical digital twinborn model is established by adopting static and dynamic posture capture starting from individualization and fusing multi-dimensional data, and a safety constraint posture model is constructed through precisely quantified safety boundary calculation and risk area labeling, so that a unique massage path is planned; the individual difference requirements from the structure to the physiological tolerance are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of massage robots, and particularly relates to a massage robot individualized physiotherapy method and system based on a cloud platform. BACKGROUND

[0002] With the intensification of population aging, the increase of chronic pain population and the upgrading of health consumption, the demand for individualized physiotherapy services is rapidly growing. Massage robots are a kind of medical auxiliary equipment that combines robotics, biomechanics, artificial intelligence and human-computer interaction technology. Its technical foundation can be traced back to the development of fine control technology of industrial robots in the late 20th century. With the maturity of high-precision servo motors and multi-axis linkage control systems, robots have been able to simulate the basic mechanical properties of human hands. In terms of sensing technology, the application of six-axis force sensing and tactile array enables robots to perceive contact force and skin deformation in real time, while three-dimensional pose reconstruction technology based on optics or ultrasound provides data support for human body surface adaptation. Motion planning algorithms draw on the mechanical parameters decomposed from traditional massage techniques, and encode pushing, pulling, pressing and rubbing actions into control instruction sets containing force, frequency and trajectory. In recent years, the introduction of deep learning technology enables robots to analyze a large amount of clinical massage data and establish individualized force application models. The electromechanical system usually adopts modular design, and the end effector can be replaced with bionic fingers, rollers or thermal therapy modules to meet different therapy needs.

[0003] For diseases such as scoliosis, massage as an auxiliary means aims to regulate the tension balance of soft tissues around the spine through specific mechanical stimulation, relieve muscle spasm, improve local blood circulation, and promote compensatory adjustment of the mechanical structure of the spine.

[0004] Existing massage robots mostly use fixed programs or simple parameter adjustment, and lack the ability to analyze and dynamically adapt to user physiological state and health data. Moreover, the massage adaptability for users with diseases such as scoliosis is low, and it cannot meet the needs of most people with pain.

[0005] In order to solve the defects in the prior art, the present technical scheme provides a massage robot individualized physiotherapy method and system based on a cloud platform. SUMMARY

[0006] The present application provides a massage robot individualized physiotherapy method and system based on a cloud platform to solve the defects in the prior art.

[0007] In one aspect, the present application provides a massage robot individualized physiotherapy method based on a cloud platform, comprising: S1: performing static posture scanning and dynamic posture capturing on a user, and outputting an initial posture matrix and a dynamic posture sequence; S2: Based on the initial pose matrix and the dynamic pose sequence, a body structure model is established based on a multi-modal fusion algorithm, and a high-precision body model is output; S3: Based on the high-precision body model, a safety boundary is calculated, and a risk area is marked, and a safety-constrained body model is output; S4: Based on the safety-constrained body model, a massage path is planned by improving the RRT algorithm, and a personalized massage trajectory is output; S5: Based on the personalized massage trajectory, the user is massaged, and the massage pressure and spinal displacement data are monitored in real time; S6: The abnormal values of the massage pressure and spinal displacement data are judged in real time, and the personalized massage trajectory is adjusted according to the abnormal situation.

[0008] According to the massage robot personalized physiotherapy method based on the cloud platform provided by the application, the step S2 of outputting the high-precision body model comprises: S21: According to the initial pose matrix and the dynamic pose sequence, a parameterized digital twin body based on muscle-skeleton dynamics is established; S22: Based on the parameterized digital twin body, the physical properties of the user's subcutaneous soft tissue layer are simulated by using finite element analysis, and a mechanical response model is constructed; S23: The feature vectors of the input initial pose matrix and dynamic pose sequence are processed by using a deep neural network to obtain a body prediction result; S24: Based on the body prediction result, cross-validation and error function minimization are performed, and the parameters of the parameterized digital twin body and the parameters of the mechanical response model are adjusted until the optimal solution is converged to obtain a high-precision body model.

[0009] According to the massage robot personalized physiotherapy method based on the cloud platform provided by the application, the step S3 of outputting the safety-constrained body model comprises: S31: Based on the high-precision body model and the mapped past medical history annotation points in the model, a conservative repulsive buffer zone is generated in three-dimensional space; S32: Based on the repulsive buffer zone, the physiological load boundary is set in combination with the biomechanical model and the tissue properties; S33: Through non-contact stimulation response monitoring of the preset points on the body surface, in combination with user subjective feedback, a pain-sensitive hot spot map is formed; S34: The repulsive buffer zone, the area exceeding the physiological load boundary, the past medical history annotation points and the area of the pain-sensitive hot spot map are marked as a risk area, and the high-precision body model is updated to obtain a safety-constrained body model.

[0010] According to the massage robot personalized physiotherapy method based on the cloud platform provided by the application, the step S32 of setting the physiological load boundary comprises: S321: divide the massage area outside the exclusion buffer zone into multiple local massage areas, and set a hierarchical safety threshold of the maximum allowable pressure of the local massage area; S322: set a global maximum force limit based on the hierarchical safety threshold; S323: for the spine area, according to the current spine shape, intervertebral disc state, and combined with the local muscle activity, the maximum allowable displacement of each segment of the spine in three-dimensional direction is calculated as the physiological load boundary.

[0011] According to the massage robot individualized physiotherapy method based on the cloud platform provided by the application, the setting method of the hierarchical safety threshold in step S321 is as follows:

[0012] In the formula, i is the index of the local massage area, is the maximum allowable pressure threshold of the i-th area, α is the gender adjustment factor, Fage is the age influence function, β is the age sensitivity index, is the muscle thickness estimated by the i-th area model, H ref is the reference muscle thickness, is the reference tolerance pressure based on the tissue type.

[0013] According to the massage robot individualized physiotherapy method based on the cloud platform provided by the application, the step of outputting the individualized massage trajectory in step S4 includes: S41: based on the body model and the preset path template, planning the main massage area and the massage route trunk to obtain an initial trajectory; S42: around each massage route trunk point, sampling planning is performed through the improved RRT algorithm to obtain a structured local trajectory data package; S43: based on the safety constraint body model, according to the structured local trajectory data package, extracting detailed spatial configurations and constraint conditions in multiple local massage areas; S44: re-planning the initial trajectory according to the detailed spatial configuration and the constraint condition to output the individualized massage trajectory.

[0014] According to the massage robot individualized physiotherapy method based on the cloud platform provided by the application, the step of more fine sampling planning through the improved RRT algorithm in step S42 includes: S421: initialize the RRT tree, and take the massage route trunk point as the root node of the RRT tree; S422: in the sampling space of the root node, execute biased random sampling based on the safety probability distribution of the body model to obtain a three-dimensional sampling point sequence; S423: adaptively expand each sampling node in the three-dimensional sampling node sequence to obtain a plurality of expansion path segments; and perform multi-level safety detection on each expansion path segment to output a set of safe path segments; S424: terminate iteration when the coverage rate of the sampling node reaches a preset range to obtain a structured local trajectory data package.

[0015] According to the massage robot individualized physiotherapy method based on the cloud platform provided by the application, the step S6 includes the following steps: The single-dimensional signal is used to preset a basic safety threshold value online to perform hard decision, and initial abnormality judgment data is output. According to the initial abnormality judgment data, a bad event mode library is defined, and persistent slight tremor of the massage area in an unexpected direction is identified. The control chart method is used to monitor the mean drift or variance surge of the persistent slight tremor, and an abnormality monitoring result is output.

[0016] According to the massage robot individualized physiotherapy method based on the cloud platform provided by the application, the step S6 includes the following steps:

[0017] The application also provides a massage robot individualized physiotherapy system based on the cloud platform, which includes: A body state information acquisition module is configured to perform static body state scanning and dynamic body state capturing on a user, and output an initial posture matrix and a dynamic posture sequence. An initial body state modeling module is configured to perform body state structure modeling based on a multi-modal fusion algorithm according to the initial posture matrix and the dynamic posture sequence, and output a high-precision body state model. A safety constraint marking module is configured to calculate a safety boundary based on the high-precision body state model, mark a risk area, and output a safety constraint body state model. A massage trajectory planning module is configured to perform massage path planning by improving an RRT algorithm based on the safety constraint body state model, and output an individualized massage trajectory. A massage data monitoring module is configured to perform user massage based on the individualized massage trajectory, and monitor massage pressure and spinal displacement data in real time. An abnormality detection module is configured to judge abnormal values of the massage pressure and the spinal displacement data in real time, and adjust the individualized massage trajectory according to abnormal conditions.

[0018] The cloud-based personalized therapy method and system for massage robots provided by the present invention adopts individualized static and dynamic posture capture, integrates multi-dimensional data to establish an exclusive biomechanical digital twin model, and plans a unique massage path to meet individual differences from structure to physiological tolerance. Through precise and quantitative safety boundary calculation and risk area labeling, a safety constraint posture model is constructed, and real-time monitoring, intelligent anomaly detection and multi-level rapid adjustment mechanisms are integrated to build a security protection system with active defense and dynamic response. By adopting a closed-loop system based on cloud computing, dynamic optimization can be performed based on the user's real-time physiological feedback and long-term therapy effects. The new data generated during each massage process becomes the fuel for optimization. The large-scale desensitized data accumulated in the cloud will drive the continuous evolution of the core algorithm, improving the accuracy, robustness, safety and effectiveness of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Fig. 1 This is a flow chart of a personalized physiotherapy method for a massage robot based on a cloud platform provided by an embodiment of the present invention; Fig. 2 It is a structural diagram of a cloud platform-based personalized massage robot therapy system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0022] Example 1: The following combination Figs. 1-2 The present invention describes the cloud platform-based personalized massage robot therapy method and system.

[0023] like Fig. 1 As shown, the cloud platform-based personalized massage robot therapy method provided by the embodiment of the present invention includes: S1: Perform static body scanning and dynamic body capture on the user, and output the initial posture matrix and dynamic posture sequence.

[0024] Static posture scanning usually adopts structured light or ToF (Time of Flight) depth camera system, combined with high-resolution RGB information, to conduct full-scan on the user in standard relaxed standing, sitting or lying posture. The scanning process generates millions of point cloud data, accurately depicting the user's body contour, key skeletal marker points (such as spinal process, scapular spine, anterior superior iliac spine, medial and lateral knee joint), subcutaneous fat thickness distribution and other three-dimensional information. Through point cloud registration and surface reconstruction algorithm, a high-fidelity user three-dimensional mesh model is generated. On this basis, through the inverse kinematics or machine learning model based on biomechanical constraints, the surface point cloud information is solved into the relative position and orientation of the internal skeletal structure (spine, ribs, pelvis, limbs), forming a mathematical representation describing the main joint angles, trunk lateral bending / rotation / inclination, pelvic tilt and other information, i.e. initial posture matrix. The matrix not only contains spatial coordinates, but also contains joint degree of freedom state information.

[0025] Dynamic posture capture usually uses a multi-view vision system composed of high-speed depth cameras and visible light cameras to capture the user's motion in simulated physiotherapy actions. By tracking the surface feature points with the vision system, the key point positions and joint poses that change continuously over time are obtained. After smoothing filter processing to remove noise and aligning according to the time stamp, a dynamic posture sequence accurately representing the specific physiological structure motion pattern of the user is generated. This sequence records the three-dimensional position, velocity and acceleration information of each segment of the body at each time point.

[0026] The initial posture matrix provides the baseline state of the user's physiological structure, including the physiological curvature of the spine, the tilt of the pelvis, and the deviation of the high and low shoulders formed by nature or after birth. The dynamic posture sequence reveals the kinetic characteristics of the structure in functional activities, including the range of motion of the spine, muscle compensation patterns, and joint stability.

[0027] S2: Based on the initial posture matrix and the dynamic posture sequence, a multi-modal fusion algorithm is used to model the posture structure, and a high-precision posture model is output, including the following steps: S21: On the cloud platform side, a parameterized digital twin body based on muscle-skeletal dynamics is established according to the initial posture matrix and the dynamic posture sequence. The initial posture matrix is used to set the initial skeletal morphology and joint initial pose parameters. The dynamic posture sequence is used to inversely deduce the activation state of the muscles, the synergistic contraction pattern, the ligament tension and the inter-joint force (achieved through inverse dynamics or forward dynamics solver). Group data prior is used to set a reasonable interval for unknown model parameters of individual users, reduce the solution space, and improve the accuracy of individual models.

[0028] S22: Based on the parameterized digital twin, simulate the physical properties of the user's subcutaneous soft tissue layer using finite element analysis to construct a mechanical response model. In particular, by analyzing the deformation hysteresis of soft tissue in the dynamic sequence with skeletal movement, the viscoelastic characteristics of the tissue can be estimated.

[0029] S23: Use a deep neural network to process the input initial posture matrix and feature vectors of the dynamic posture sequence to obtain the body posture prediction result. This network architecture is trained on a large amount of anonymized body posture-physiological data in the cloud, which can learn the complex mapping relationship from structural information to deep biological mechanics parameters and physiological state, providing initial estimates or supplementary prediction models for the parameterized model. The fusion process is an iterative optimization process based on the parameterized model and neural network prediction results.

[0030] S24: Based on the body posture prediction result, cross-validation and error function minimization are performed to adjust the parameters of the parameterized digital twin and the parameters of the mechanical response model until the optimal solution is converged to obtain a high-precision body posture model. The high-precision body posture model is a comprehensive digital twin that includes: refined skeletal geometry and topological relationships, i.e., joint, ligament attachment points. Key muscle bundle origin and termination points, muscle force line directions, force-length-velocity characteristic estimates. Soft tissue layer distribution and regionalized mechanical properties, including Young's modulus, Poisson's ratio, and viscosity coefficient. Key biomechanical index quantification values such as physiological curvature, mobility, and stability. Feature vector encoding strongly related to user physiological state. This model not only provides a static description but also simulates the response of bones and soft tissues under external load, serving as the basis for subsequent safety boundary calculation and path planning.

[0031] S3: Based on the high-precision body posture model, calculate the safety boundary and mark the risk area, output the safety-constrained body posture model, including: S31: Based on the high-precision body posture model and the mapped past medical history annotation points in the model, generate a conservative repulsive buffer zone in three-dimensional space. The size of the buffer zone radius is dynamically calculated based on the tissue mechanical properties and the pre-set safety rule library. Use a fast collision detection algorithm to ensure that the planned trajectory points do not invade these buffer zones.

[0032] S32: Based on the repulsive buffer zone, set the physiological load boundary in combination with the biomechanical model and tissue properties, including: S321: Divide the massage area outside the repulsive buffer zone into multiple local massage areas and set the hierarchical safety threshold of the maximum allowed pressure in the local massage area, calculated as follows:

[0033] where, is the maximum allowable pressure threshold for the ith region, α is the gender adjustment factor, normal α > 1 for female, considering the lower muscle proportion of female. Fage is the age influence function, usually a piecewise function. β is the age sensitivity index, is the muscle thickness estimated by the ith region model, H ref is the reference muscle thickness, is the reference muscle thickness, is the reference muscle thickness,

[0034] The grading safety threshold can be dynamically adjusted according to the user's age, gender (obtained through the profile), current tissue hardness, and local tissue bearing capacity calculated in the high-precision body model.

[0035] S322: Set the global maximum force limit based on the grading safety threshold to prevent accidental jamming.

[0036] S323: For the spine region, according to the current spine shape, intervertebral disc state, combined with local muscle activity, calculate the maximum allowable displacement of each segment of the spine in three-dimensional direction as the physiological load boundary. The maximum displacement includes the relative displacement limit values of longitudinal traction, lateral movement and torsion.

[0037] Limit the speed of massage operation (moving speed, pressing depth change rate, kneading frequency) to prevent damage or trigger muscle defensive spasm caused by rapid impact.

[0038] S33: Form a pain-sensitive hotspot map by monitoring the non-contact stimulation response of the preset points on the body surface, combined with user subjective feedback. These areas are given higher obstacle avoidance level or lower pressure tolerance.

[0039] S34: Mark the exclusion buffer zone, areas exceeding the physiological load boundary, past medical history marked points, and areas of the pain-sensitive hotspot map as risk areas, and update the high-precision body model to obtain the safety-constrained body model. The safety-constrained body model is based on the original high-precision body model, and fully embeds the calculated physical space obstacle avoidance area, dynamic pressure / displacement / speed three-dimensional constraint envelope surface, pain-sensitive points and their level information. This safety-constrained body model is the only legal space for subsequent path planning, and all operation points planned must be located within the envelope surface and avoid all marked risk core areas.

[0040] S4: Based on the safety-constrained body model, the massage path is planned by improving the RRT algorithm, and the personalized massage trajectory is output, including the following steps: S41: Based on the body model and the preset path template, plan the main areas of massage and the main trunk of massage route, and obtain the initial trajectory.

[0041] S42: Around each massage route backbone point, a more refined sampling planning is performed by the improved RRT algorithm, including adjusting the posture of the end effector, the pressing force and depth, and the direction. The steps of the more refined sampling planning by the improved RRT algorithm include: S421: Initialize the RRT tree, and take the massage route backbone point as the root node of the RRT tree. Initialize the improved RRT algorithm parameters (including the maximum number of iterations, the step range, and the safety threshold). Output the initialized RRT root node (i.e. the backbone point coordinates) and the constraint-aware sampling space range.

[0042] S422: In the sampling space of the root node, perform biased random sampling based on the safety probability distribution of the body model. Dynamically adjust the sampling weight according to the danger level of the constraint region, generate 80% of the sampling points in the safe area (such as the muscle area), 15% of the sampling points in the dangerous buffer area (such as the skeleton area), and 5% of the sampling points in the absolute forbidden area (for boundary testing), to obtain a three-dimensional sampling point sequence. The calculation method of the sampling weight is as follows:

[0043] In the formula, is the sampling weight of the jth region in the sampling space, which is used to dynamically adjust the sampling probability so that the sampling density in the safe area is higher. is the Euclidean distance to the nearest risk area, k is the attenuation coefficient, usually k ∈ [0.1, 0.3]. N is the total number of partitions, and s is the sub-partition index.

[0044] S423: Perform adaptive tree expansion on each sampling node in the three-dimensional sampling node sequence to obtain a plurality of expansion path segments, including locating the nearest neighbor node in the RRT tree. Automatically scale the expansion step according to the body surface curvature at the current point. Generate a candidate node along the sampling direction. Perform multi-level safety detection on each expansion path segment to output a set of safe path segments. First, verify the geometric invasiveness, and then detect the physiological constraints (avoid the annotated nerve and blood vessel buffer area); if a collision occurs, start the surface redirection mechanism and deflect 10-15 degrees along the contact normal to recalculate the path.

[0045] S424: When the coverage rate of the sampling node reaches the preset range, terminate the iteration, extract the safe path from the RRT tree, the contact point cloud with a path point density > 5 points / cm², and the path smoothness evaluation matrix from the root node to all leaf nodes of the backbone point, to obtain a structured local trajectory data package containing three core data: coordinate set, normal vector, and safety score.

[0046] S43: Based on the safety constraint body model, extract the detailed spatial configuration and constraint conditions in multiple local massage areas, especially the accurate avoidance of the pressure envelope surface and the risk points.

[0047] S44: Re-plan the initial trajectory according to detailed spatial configuration and constraints, output personalized massage trajectory. The personalized massage trajectory is a highly customized spatiotemporal instruction set that defines the sequence of three-dimensional coordinates of the end effector center point in the user's body coordinate system (for describing point distribution), moving speed sequence, temporal variation of target pressure / depth on the user's body, end pose angle sequence (for determining contact surface angle), and timestamp. The trajectory fully reflects the user's unique physiological structure, tolerance level, risk aversion needs, and optimizes the smoothness, comfort, and preliminary expected physiotherapy stimulation effect of the operation. The planning algorithm is designed for online re-planning triggered by external signals, with interruption points and path parameter adjustment mechanisms.

[0048] S5: Based on the personalized massage trajectory, perform user massage and monitor massage pressure and spinal displacement data in real time.

[0049] The path derivative must change continuously to avoid sudden turns or accelerations, and is smoothed by path node interpolation and post-optimization.

[0050] In areas considering pressure, the planning node carries a pressure value attribute. The algorithm checks the pressure change gradient when evaluating path segment cost, penalizes severe pressure fluctuations, and ensures smooth pressure transitions, similar to a massage therapist's hand feel. This is achieved by expanding the state space and adding a pressure smoothing term to the objective function.

[0051] For deep techniques such as kneading, pressing, and pushing, path planning needs to consider the interaction between the end effector of the robotic arm and the deformation of soft tissue. The stress-strain relationship or simplified deformation relationship model of soft tissue in the safety constraint body model is integrated into collision detection and path feasibility judgment, rejecting paths that may cause local permanent deformation or extreme stretching of the tissue.

[0052] Based on spatial trajectory following, hybrid position / force control is formed by integrating real-time force feedback. Using robot torque sensors or high-precision six-axis force / torque sensors integrated at the end, the actual applied pressure, shear force, and torque are monitored and compared with the target pressure value set by the trajectory to generate control signals to dynamically adjust the position, especially in scenarios where soft tissue causes the position to change with pressure.

[0053] For soft and flexible contact surface materials, a pressure balancing strategy is designed or a micro pressure sensor array is integrated to obtain a more realistic contact pressure distribution and prevent sharp point effects.

[0054] The steps of real-time monitoring of massage pressure and spinal displacement data include: Real-time monitoring of the positive pressure and its rate of change by the end six-dimensional force / torque sensor. For larger massage areas, multi-point pressure distribution information is also collected in real time.

[0055] Using a multi-sensor fusion strategy, the spinal displacement data is monitored, including the following steps: A high-frame-rate, high-resolution depth camera system is set up in a stable position near the massage bed or massage chair, continuously tracking the user's back surface pre-set or automatically identified stable marker points (such as spinal skin reflection marker points, stable and unchanging birthmarks / moles, etc.) highly related to deep skeletal movement. The three-dimensional movement of these marker points is calculated using optical tracking, indirectly estimating the macro or segment displacement / deformation of the spine. This method is suitable for observing larger displacements and overall posture changes. A thin, flexible, biocompatible film strain sensor network is laid on the user's skin. These sensors directly adhere to the skin and deform with the spinal movement, providing more accurate and high-frequency deformation / strain information at the segment level, and then inferring the relative movement of the vertebrae. The data processing focuses on eliminating the artifacts caused by skin sliding.

[0056] The visual tracking data and flexible sensor data are fused in the cloud or edge node to calculate the three-dimensional displacement vector of different spinal segments (thoracic vertebrae, lumbar vertebrae key segments) (including longitudinal relative displacement along the spinal axis, lateral displacement in the coronal plane, and rotation angle change in the horizontal plane). Set up an independent early warning threshold detector running on the data stream of each segment.

[0057] All core monitoring data and robot execution status are timestamped and packaged, uploaded to the cloud processing center through high-reliability low-latency communication.

[0058] S6: Real-time judgment of abnormal values of massage pressure and spinal displacement data, and adjustment of individualized massage trajectory according to abnormal conditions. The preset spinal segment displacement basic safety threshold is expressed as:

[0059] In the formula, represents the bth spinal segment displacement early warning threshold, is the basic threshold, 2mm for thoracic vertebrae and 3mm for lumbar vertebrae. is the current activity, is the average activity of healthy people.

[0060] The steps of real-time judgment of abnormal values of massage pressure and spinal displacement data include: S61: Hard decision by online single-dimension signal pre-set base safety threshold, output initial abnormality judgment data. Trigger abnormal event immediately when the monitored pressure value exceeds the maximum allowed pressure set for this point in the safety constraint posture model. Trigger abnormal event immediately when the monitored displacement value of any spinal segment exceeds its three-dimensional displacement threshold. Trigger response immediately based on user's pain feedback signal. This feedback signal has the highest priority.

[0061] S62: Define adverse event pattern library according to the initial abnormality judgment data, identify persistent micro-tremor in unintended direction in the massage area. Trigger warning as soon as these patterns are identified.

[0062] S63: Monitor the mean shift or variance surge of the persistent micro-tremor using control chart method, output abnormality monitoring result. These changes may indicate user tolerance decline or abnormal tissue response.

[0063] The trajectory dynamic adjustment strategy in adjusting the personalized massage trajectory according to abnormal conditions includes: When the hard safety threshold is exceeded, the user's emergency stop instruction is detected, or a high-risk pattern (such as severe spasm or excessive impact displacement) is identified, the system sends an immediate stop instruction with the highest priority to the robot through the pre-set emergency channel.

[0064] For abnormalities that do not immediately threaten safety but affect comfort or effectiveness, gradual adjustments are made, including: Pre-set rule library triggers rapid response: reduce local pressure, pause movement, or bypass the current micro-local area. These adjustments are directly executed on the edge computing node to ensure low latency and avoid lengthy processing time in the cloud.

[0065] For more significant abnormalities or when fine-tuning is ineffective, cloud analysis and calculation are triggered. The current abnormal condition and the current state of the robot are input into the improved RRT algorithm to quickly calculate a new local optimal or sub-optimal trajectory segment that bypasses the problem point and meets the new constraints.

[0066] If multiple adjustments occur during a single physiotherapy session, or online re-planning and calculation take too long to affect the experience, the system may trigger higher-level strategy adjustments, including skipping the current entire small area, switching to a more gentle pre-set soothing technique template, shortening the overall massage time, or reducing the pre-set intensity level.

[0067] Each abnormality trigger and adjustment action is recorded along with the original posture model and constraint conditions at the time, and is stored in the cloud user profile for a long time, anonymized and merged into the global database. It is used to optimize the user's personalized model accuracy, adjust safety boundary parameters, and improve path planning rule library in the future.

[0068] In summary, the massage robot individualized physiotherapy method and system based on the cloud platform provided by the application, through the static and dynamic body posture capture starting from individualization, the establishment of a special biomechanics digital twin model by fusing multi-dimensional data, the planning of a unique massage path, and the meeting of individual difference requirements from structure to physiological tolerance. Through the accurate quantitative safety boundary calculation and risk area labeling, a safety constraint body posture model is constructed, and real-time monitoring, intelligent anomaly detection and multi-level rapid adjustment mechanism are integrated to build a safety protection system with active defense and dynamic response. Through the use of a closed-loop system based on cloud computing, the system can be dynamically optimized according to the user's real-time physiological feedback and long-term physiotherapy effect. The new data generated in each massage process becomes the fuel for optimization. And the large-scale desensitization data accumulated in the cloud will drive the continuous evolution of the core algorithm, improving the accuracy, robustness, safety and effectiveness of the entire system.

[0069] As shown in Fig. 2 The application also provides a massage robot individualized physiotherapy system based on a cloud platform, which comprises: A body posture information acquisition module is configured to perform static body posture scanning and dynamic body posture capture on a user, and output an initial posture matrix and a dynamic posture sequence.

[0070] An initial body posture modeling module is configured to perform body posture structure modeling based on a multi-modal fusion algorithm according to the initial posture matrix and the dynamic posture sequence, and output a high-precision body posture model.

[0071] A safety constraint labeling module is configured to calculate a safety boundary and label a risk area based on the high-precision body posture model, and output a safety constraint body posture model.

[0072] A massage trajectory planning module is configured to perform massage path planning by improving an RRT algorithm based on the safety constraint body posture model, and output an individualized massage trajectory.

[0073] A massage data monitoring module is configured to perform user massage based on the individualized massage trajectory, and monitor massage pressure and spinal displacement data in real time.

[0074] An anomaly detection module is configured to determine abnormal values of the massage pressure and the spinal displacement data in real time, and adjust the individualized massage trajectory according to the abnormal situation.

[0075] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0076] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cloud platform-based massage robot personalized physiotherapy method, characterized in that, Comprise: S1: static body scan and dynamic body capture for users, output initial posture matrix and dynamic posture sequence; S2: according to the initial posture matrix and dynamic posture sequence, based on multi-modal fusion algorithm for body structure modeling, output high-precision body model; S3: based on the high-precision body model, calculate the safety boundary, and mark the risk area, output the safety constraint body model; S4: based on the safety constraint body model, improve the RRT algorithm for massage path planning, output personalized massage trajectory; S5: based on the individualized massage trajectory, execute user massage, and monitor massage pressure and spinal displacement data in real time; S6: real-time judgment of the abnormal value of the massage pressure and the spinal displacement data, and adjustment of the individualized massage trajectory according to the abnormal situation.

2. The cloud platform-based massage robot individual physiotherapy method according to claim 1, characterized in that, In step S2, the step of outputting the high-precision body model comprises: S21: according to the initial posture matrix and dynamic posture sequence, establish a parameterized digital twin body based on muscle-skeleton dynamics; S22: based on the parameterized digital twin body, use finite element analysis to simulate the physical properties of the user's subcutaneous soft tissue layer and construct its mechanical response model; S23: use deep neural network to process the feature vectors of the input initial posture matrix and dynamic posture sequence to get body posture prediction results; S24: based on the body posture prediction results, cross-validation and minimum error function are carried out, the parameters of the parameterized digital twin body and the parameters of the mechanical response model are adjusted until the optimal solution is converged to get the high-precision body model. 3.The cloud platform-based massage robot individual physiotherapy method of claim 2, wherein, In step S3, the step of outputting the safety constraint body model comprises: S31: based on the high-precision body model and the mapping of the past medical history annotation points in the model, a conservative repulsive buffer zone is generated in three-dimensional space; S32: based on the repulsive buffer zone, combined with the biomechanical model and tissue properties, set the physiological load boundary; S33: through non-contact stimulation response monitoring of the preset points on the body surface, combined with user subjective feedback, form a pain sensitive hotspot map; S34: mark the areas of the repulsive buffer zone, the areas exceeding the physiological load boundary, the past medical history annotation points and the pain sensitive hotspot map as risk areas, and update the high-precision body model to get the safety constraint body model.

4. The cloud platform-based massage robot individual physiotherapy method according to claim 3, characterized in that, In step S32, the step of setting the physiological load boundary comprises: S321: divide the massage area outside the repulsive buffer zone into multiple local massage areas, and set the hierarchical safety threshold of the maximum allowable pressure of the local massage area; S322: set the global maximum force limit based on the hierarchical safety threshold; S323: for the spine area, according to the current spine shape, intervertebral disc state, combined with the local muscle activity, calculate the maximum allowable offset of each segment of the spine in three-dimensional direction as the physiological load boundary. 5.The cloud platform-based massage robot individual physiotherapy method of claim 4, wherein, In step S321, the hierarchical safety threshold is set as follows: where i is the index of the local massage area, is the maximum allowable pressure threshold for the ith region, a is a gender adjustment factor, Fageis an age effect function, and β is an age sensitivity index, is the muscle thickness estimated by the ith region model, H ref is the reference muscle thickness, is the reference tolerance pressure based on tissue type. 6.The cloud platform-based massage robot individual physiotherapy method of claim 4, wherein, In step S4, the step of outputting the individualized massage trajectory comprises: S41: based on the body model and the preset path template, plan the main area of massage and the main trunk of massage route, get the initial trajectory; S42: Around each of the massage route backbone points, sample planning is performed through an improved RRT algorithm to obtain a structured local trajectory data package; S43: Based on the safety constraint body model, detailed spatial configurations and constraint conditions in a plurality of local massage areas are extracted from the structured local trajectory data package; S44: The initial trajectory is re-planned according to the detailed spatial configurations and constraint conditions, and the personalized massage trajectory is output.

7. The cloud platform-based massage robot individual physiotherapy method according to claim 6, characterized in that, In step S42, the step of more fine sampling planning through the improved RRT algorithm includes: S421: The RRT tree is initialized, and the massage route backbone point is taken as a root node of the RRT tree; S422: In the sampling space of the root node, biased random sampling is performed based on the safety probability distribution of the body model to obtain a three-dimensional sampling point sequence; S423: Each sampling node in the three-dimensional sampling node sequence is adaptively tree expanded to obtain a plurality of expansion path segments, and each expansion path segment is subjected to multi-level safety detection to output a safety path segment set; S424: Iteration is terminated when the coverage rate of the sampling node reaches a preset range, and a structured local trajectory data package is obtained. 8.The cloud platform-based massage robot individual physiotherapy method of claim 1, wherein, In step S6, the step of judging the abnormal value of the massage pressure and the spinal displacement data in real time includes: S61: Hard decision is performed through one-dimensional signal online preset basic safety threshold value to output initial abnormal judgment data; S62: According to the initial abnormal judgment data, a bad event mode library is defined, and it is identified that the massage area appears persistent slight tremor in an unexpected direction; S63: The mean shift of the persistent slight tremor is monitored using a control chart method, and an abnormal monitoring result is output. 9.The cloud platform-based massage robot individual physiotherapy method of claim 1, wherein, In step S6, the adjustment mode of the personalized massage trajectory according to the abnormal condition includes: when the hard safety threshold value is out of limit, a user emergency stop instruction is detected, or a high-risk mode is identified, an immediate stop instruction of the highest priority is sent to the robot through a preset emergency channel; for the abnormality that does not immediately threaten safety but affects comfort or effect, gradual adjustment is performed.

10. A cloud platform-based massage robot individualized physiotherapy system, which adopts the cloud platform-based massage robot individualized physiotherapy method according to any one of claims 1 to 9, characterized in that, It includes: The body information acquisition module is used for static body scanning and dynamic body capturing of the user, and outputs an initial posture matrix and a dynamic posture sequence; The initial body modeling module is used for body structure modeling based on a multi-modal fusion algorithm according to the initial posture matrix and the dynamic posture sequence, and outputs a high-precision body model; The safety constraint marking module is used for calculating a safety boundary and marking a risk area based on the high-precision body model, and outputs a safety constraint body model; The massage trajectory planning module is used for massage path planning based on the safety constraint body model through an improved RRT algorithm, and outputs a personalized massage trajectory; The massage data monitoring module is used for user massage based on the personalized massage trajectory, and real-time monitoring of massage pressure and spinal displacement data; The abnormality detection module is used for real-time judgment of the abnormal value of the massage pressure and the spinal displacement data, and adjustment of the personalized massage trajectory according to the abnormal condition.

Images