Intelligent control system and method based on digital twin driven spine maintenance multi-modal method

Abstract

This invention proposes a digital twin-driven multimodal intelligent control system and method for spinal care, belonging to the field of smart home and health monitoring technology. The method includes: synchronously collecting physiological signal data and mechanical feedback data from the multimodal sensing unit of the smart mattress physical entity; performing noise reduction and standardization preprocessing on the collected raw data to generate a multimodal raw dataset; establishing a real-time bidirectional mapping between the smart mattress physical entity and a virtual model through digital twin technology, which can dynamically adjust the support strategy when the user's body shape changes or sleeping posture shifts, effectively avoiding the problem of pressure caused by preset modes for some users, and improving the personalized adaptability of spinal care.

Description

Technical Field

[0001] This invention proposes a digital twin-driven multimodal intelligent control system and method for spinal health maintenance, belonging to the field of smart home and health monitoring technology. Background Technology

[0002] In the field of smart mattresses or electric beds, the technology route based on pressure sensing and adaptive adjustment has shown significant advantages in improving sleep quality and spinal health management because it can actively adjust according to the real-time state of the human body. Traditional mattresses, due to their fixed firmness and inability to adapt to different body types, sleeping positions, and health needs, are prone to uneven static pressure distribution and even health problems such as blood circulation disorders. Currently, adaptive adjustment systems that combine pneumatic airbags and pressure sensor arrays have become the mainstream technology in this field. These systems typically include a pressure sensing module, a pressure distribution modeling module, a sleeping posture recognition module, and an adjustment actuator. They can detect pressure data from various parts of the body and map it into a three-dimensional pressure heat map, thereby identifying the user's sleeping posture and spinal curve, providing a basis for adjusting the airbag pressure in the lumbar region. The pressure distribution optimization method with multiple air pumps working together further introduces the analysis of neighboring air chamber pressure interference relationships and the comparison of spinal alignment curves, which can accurately optimize and control the mattress pressure distribution according to the user's individual characteristics and real-time posture, effectively improving the user's sleep comfort and mattress fit.

[0003] In the iterative development of smart mattresses, multimodal perception and intelligent linkage control have further expanded the perception dimensions of the system. For example, Chinese patent application number CN202510565558.2, entitled "Adaptive Adjustable Bed System Based on AI Multimodal Perception and Intelligent Linkage," proposes an adaptive adjustment system integrating a multimodal perception module, an AI processing module, and an intelligent linkage module. This system can collect user physiological data, sleep posture data, and environmental data, and generate adjustment control commands through the AI ​​processing module to drive the adjustment mechanism within the mattress to adjust parameters such as firmness, height, angle, and temperature. Furthermore, Chinese patent publication number CN120660983A, entitled "Optimization Method for Pressure Distribution of Smart Mattress with Multi-Pump Collaborative Operation," achieves precise optimization of pressure distribution based on individual user characteristics and real-time posture through neighboring air chamber pressure interference relationship analysis and spinal alignment curve comparison. This effectively solves the technical problem of existing technologies' difficulty in accurately optimizing pressure distribution based on individual user characteristics and real-time posture. However, in the smart mattress adjustment solutions proposed in the aforementioned patents, the system's control logic is still mainly driven by open-loop mode or preset rules: the AI-based multimodal perception adjustment system relies on preset physiological data models and sleep posture recognition models in actual operation, lacking a closed-loop feedback mechanism for millisecond-level linkages with high real-time requirements; although the multi-pump collaborative pressure distribution optimization method generates multi-chamber control parameters after posture recognition, a direct linkage loop is not established between physiological signals (such as heart rate, respiration, etc.) and physical deformation units such as air pumps and motors, making it impossible to achieve active intervention when the user enters deep sleep or other unconscious states. At the same time, existing pressure comfort adjustment solutions rely on the absolute magnitude of pressure in different human body zones for local adjustment during adaptive adjustment (e.g., reducing hip pressure when it is too high), ignoring the dynamic balance of human biomechanics, which can easily lead to compensatory problems such as insufficient lumbar support caused by local adjustment. Therefore, there is an urgent need for a method that, based on establishing a real-time mapping between physical entities and virtual models, uses multimodal data closed-loop adaptive control as the core to break through the limitations of traditional open-loop control and local adjustment, and achieve precise and dynamic spinal care support. In the method proposed in this invention, a digital twin framework with virtual and real bidirectional mapping is constructed to form a millisecond-level linkage feedback loop for pressure distribution optimization, physiological signal monitoring and airbag motor adjustment. This enables proactive and intelligent spinal protection regulation throughout the user's sleep process while ensuring precise adaptation of spinal support. Summary of the Invention

[0004] This invention provides a digital twin-driven multimodal intelligent control system and method for spinal cord care, to solve the problems mentioned in the background section above:

[0005] The present invention proposes a digital twin-driven multimodal intelligent control method for spinal cord care, the method comprising:

[0006] S1. Synchronously collect physiological signal data and mechanical feedback data from the multimodal sensing unit of the smart mattress physical entity, and perform noise reduction and standardization preprocessing on the collected raw data to generate a multimodal raw dataset.

[0007] S2. Based on the multimodal raw dataset, construct a digital twin virtual model of the physical entity of the smart mattress, establish a real-time bidirectional mapping relationship between the physical entity state and the virtual model state, and generate digital twin mapping data;

[0008] S3. Perform multimodal data fusion processing on the digital twin mapping data, combine physiological signal features and mechanical feedback features, use the intention recognition algorithm to analyze the spinal care regulation needs, and generate a regulation intention feature set;

[0009] S4. Generate adaptive control parameter instructions and control parameter instruction set based on the control intention feature set;

[0010] S5. Transmit the control parameter instruction set to the execution unit of the smart mattress physical entity to drive the air pump, massage motor, and angle adjustment mechanism to perform control actions. At the same time, collect real-time feedback data after execution, compare and analyze it with the expected state of the digital twin virtual model, dynamically adjust the control parameter instruction set, and generate closed-loop control feedback data.

[0011] The present invention proposes a digital twin-driven multimodal intelligent control system for spinal cord care, the system comprising:

[0012] One or more processors;

[0013] Memory, used to store one or more programs;

[0014] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0015] The beneficial effects of this invention are as follows: By establishing a real-time bidirectional mapping between the physical entity and virtual model of the smart mattress through digital twin technology, the support strategy can be dynamically adjusted when the user's body shape changes or sleeping posture shifts. This effectively avoids the problem of preset modes causing pressure on some users, improving the personalized adaptability of spinal care. Simultaneously, the fusion processing of multimodal data (physiological signals + mechanical feedback) enables millisecond-level linkage between physiological data such as heart rate and respiration and physical deformation parameters such as air pump inflation volume and massage unit force, reducing data silos and enhancing the real-time nature of the control. Furthermore, the closed-loop control mechanism can proactively intervene in unconscious states such as deep sleep, reducing the lag in the user's subjective adjustments and avoiding control failures caused by interaction delays. This achieves both precise spinal care support and an improved sleep experience for the user. Attached Figure Description

[0016] Figure 1 This is a diagram illustrating the steps of the method described in this invention;

[0017] Figure 2 This is a detailed flowchart of step S1 in the present invention;

[0018] Figure 3 This is a diagram of the internal structure of the mattress described in this invention;

[0019] Figure 4 These are schematic diagrams of various execution units described in this invention; Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] Example 1

[0022] One embodiment of the present invention, such as Figure 1 As shown, a digital twin-driven multimodal intelligent control method for spinal cord care is described, the method comprising:

[0023] S1. The physiological signal data (heart rate, respiration, body movement) and mechanical feedback data (spinal pressure distribution, mattress deformation degree) collected by the multimodal sensing unit of the smart mattress physical entity are collected synchronously. The collected raw data are denoised and standardized preprocessed to generate a multimodal raw dataset.

[0024] S2. Based on the multimodal raw dataset, construct a digital twin virtual model of the physical entity of the smart mattress, establish a real-time bidirectional mapping relationship between the physical entity state and the virtual model state, and generate digital twin mapping data;

[0025] S3. Perform multimodal data fusion processing on the digital twin mapping data, combine physiological signal features and mechanical feedback features, use intention recognition algorithm to analyze the spinal care regulation needs, and generate a regulation intention feature set containing regulation goals, priorities and constraints.

[0026] S4. Based on the feature set of control intentions, generate adaptive control parameter instructions, including specific parameters such as air pump inflation volume, massage unit force, and mattress angle adjustment amount, and generate a control parameter instruction set;

[0027] S5. Transmit the control parameter instruction set to the execution unit of the smart mattress physical entity to drive the air pump, massage motor, and angle adjustment mechanism to perform control actions. At the same time, collect real-time feedback data after execution, compare and analyze it with the expected state of the digital twin virtual model, dynamically adjust the control parameter instruction set, and generate closed-loop control feedback data.

[0028] The working principle and effects of the above technical solution are as follows: By synchronously acquiring and preprocessing multimodal physiological and mechanical data, the quality of raw data can be unified, reducing the impact of invalid interference data on subsequent analysis and ensuring the accuracy of subsequent state analysis. Through bidirectional state mapping using digital twins, real-time state synchronization between physical devices and virtual models can be achieved, improving the timeliness of perception of mattress operation and human body maintenance status. By combining multimodal data fusion with algorithmic analysis of maintenance needs, differentiated spinal maintenance requirements can be accurately captured, improving the degree of control matching. By adaptively generating multiple types of control parameters, it can adapt to different users' physical states and lying postures, enhancing the flexibility of control. Through a closed-loop control method of real-time feedback comparison and dynamic parameter correction, the control effect can be continuously optimized, reducing control deviation and avoiding the problem of inconsistent maintenance effects caused by fixed parameter control failing to adapt to the dynamic state of the human body. This satisfies personalized spinal maintenance needs while ensuring the stability of equipment control operation.

[0029] In one embodiment of the present invention, S1 includes:

[0030] S11, installed on a smart mattress, such as Figure 3 As shown, the multimodal sensing units at various locations continuously capture three types of physiological signals: human heart rate, respiration, and body movement. Simultaneously, they collect two types of mechanical feedback information: the distribution of spinal pressure on the mattress surface and the degree of overall deformation, thus completing the continuous collection of raw data in all dimensions.

[0031] S12. Perform full-domain noise filtering on all types of raw data that have been collected, remove invalid data caused by environmental interference signals and equipment jitter, and retain basic data content with effective analytical value.

[0032] S13. Perform a unified standard conversion operation on all types of data that have completed noise reduction processing, unify the numerical range and sampling frequency of the output data of different sensing units, and eliminate the analysis bias caused by data format differences.

[0033] S14. Integrate the denoised and standardized physiological signal data and mechanical feedback data to classify and integrate data from different dimensions, forming a complete and standardized multimodal raw dataset.

[0034] The working principle and effects of the above technical solution are as follows: By synchronously collecting physiological signals and mechanical feedback information through a multimodal sensing unit, the real-time state of the human body and mattress can be comprehensively captured, improving the completeness of data acquisition. By performing full-domain noise filtering on the raw data, various invalid signals can be eliminated, reducing data errors caused by environmental and equipment disturbances. By performing a unified standard conversion, differences in output data from different acquisition modules can be smoothed out, reducing analysis errors caused by inconsistent formats. By classifying and integrating the two types of data into a unified set, the data arrangement logic can be streamlined, enhancing the smoothness of subsequent data retrieval and calculation. This avoids the problem of analysis process bottlenecks and distorted results caused by directly using messy raw data, ensuring the comprehensiveness of the initial data acquisition and laying a solid data foundation for subsequent calculations and analyses.

[0035] In one embodiment of the present invention, S2 includes:

[0036] S21. Retrieve the stored multimodal raw dataset, extract the human body state information and mattress physical operation state information carried within the data, and sort out the current full-dimensional operation characteristics of the physical entity.

[0037] S22. Combine the extracted operational features to build a digital twin virtual model of the corresponding structure, replicate the hardware structure, sensor layout and force deformation characteristics of the smart mattress, and restore the complete form of the physical entity.

[0038] S23. Establish a data interaction and transmission link to open up the data channel between the physical entity of the smart mattress and the digital twin virtual model, ensuring that the data at both ends can flow and communicate without interruption.

[0039] S24. A state synchronization mechanism is built based on real-time data flow, so that the state changes generated by physical entities are synchronously mapped to the virtual model, and the state changes inferred by the virtual model are also synchronously transmitted back to the physical entities, forming a two-way linkage state system.

[0040] S25. Record the correspondence information generated by each state synchronization between the physical entity and the virtual model, summarize all the related information, and generate complete digital twin mapping data.

[0041] The working principle and effects of the above technical solution are as follows: By extracting various state information from multimodal raw data, the operational characteristics of the smart mattress and the human body can be deeply explored, improving the detail of physical entity state perception. By replicating the mattress hardware structure and force characteristics to build a virtual model, the actual operating form of the device can be highly restored, narrowing the state gap between virtual simulation and physical device. By establishing a two-way data interaction transmission link, the continuous flow of data between the two ends can be maintained, improving the continuity and stability of data transmission. By establishing a two-way state synchronization system, real-time state linkage between the physical entity and the virtual model can be achieved, reducing state lag deviations between the two ends. By summarizing the state synchronization association information throughout the process, complete mapping data can be formed, avoiding the problem of inaccurate simulation analysis caused by the disconnect between virtual and real states. This ensures the simulation authenticity of the virtual model and provides reliable virtual-real linkage support for subsequent data fusion and intelligent control.

[0042] In one embodiment of the present invention, S24 includes:

[0043] It continuously receives real-time status data from the data interaction and transmission link, distinguishes between dynamic data from the physical entity of the smart mattress and inferred data from the digital twin virtual model, and completes the classification and aggregation of the two types of data streams.

[0044] Configure synchronous transmission rules for the collected physical entity dynamic data, set the data sampling frequency and forwarding logic, and push the real-time deformation, pressure, human body, physiological and other state changes of the mattress to the digital twin virtual model in real time to complete the synchronous mapping of the physical state to the virtual end.

[0045] Configure reverse feedback rules for the inference and calculation data output by the virtual model, match the running response rhythm of the physical entity, and continuously push the predicted state data calculated by the model to the physical entity of the smart mattress to complete the synchronous feedback of the virtual state to the physical end.

[0046] Continuously monitor the synchronization timeliness of data transmission and status updates at both ends, correct timing deviations that occur during data flow, and maintain the pace of status updates between the entity and the virtual model.

[0047] Integrate all operational aspects of bidirectional data flow and status synchronization, connect the complete operational process of status linkage between the two ends, and form a stable bidirectional linkage status system.

[0048] The working principle and effects of the above technical solution are as follows: By classifying and aggregating data streams from both the virtual and physical ends, the processing flow of data from different sources can be standardized, improving the regularity of data sorting. By configuring rules for the synchronous transmission of physical entity data, the real-time status of the equipment can be quickly synchronized to the virtual model, improving the real-time performance of entity state mapping. By setting rules for the reverse transmission of virtual data, the model's inferred state can be adapted to the equipment's operating rhythm, enhancing the adaptation effect of virtual-physical interaction. By continuously monitoring and correcting data timing deviations, the time difference between state updates at both ends can be reduced, minimizing inconsistencies caused by timing misalignments. By serializing the bidirectional linkage operation process, a stable virtual-physical collaborative operation system can be formed, avoiding deviations in simulation and control caused by asynchronous state updates. This maintains the dynamic balance of the virtual-physical system and continuously ensures the accuracy of subsequent intelligent control operations.

[0049] In one embodiment of the present invention, S3 includes:

[0050] S31. Retrieve the generated digital twin mapping data, separate the physiological signal feature components and mechanical feedback feature components contained in the data, and distinguish the data content of two different attributes.

[0051] S32. A multi-dimensional data fusion algorithm is used to deeply fuse the two types of feature components after splitting, break down the barriers between different data dimensions, and uncover the hidden state correlation patterns behind the data.

[0052] S33. Input the fused comprehensive feature data into the intent recognition algorithm module. The algorithm module deeply analyzes the actual needs of the human body for current spinal care and determines the corresponding adjustment direction.

[0053] S34. Based on the control direction obtained from the analysis, classify different types of control targets, determine the execution priority of each target according to the severity of the human body condition, and set corresponding constraints based on the operating limits of the equipment.

[0054] S35. Summarize all information related to the execution priority and constraints of all regulatory targets, integrate and summarize the feature information, and generate a complete feature set of regulatory intentions.

[0055] The working principle and effects of the above technical solution are as follows: By splitting the mapping data into different feature components, the classification of various information categories can be clarified, improving the orderliness of data classification and processing. Through algorithmic deep fusion of multi-dimensional features, the intrinsic relationship between signals and mechanical data can be uncovered, enhancing the comprehensiveness of state analysis. By analyzing actual maintenance needs through algorithmic modules, the control direction can be accurately identified, improving the relevance of demand interpretation. By dividing control targets, arranging execution sequences, and setting operating limits, the workload can be rationally allocated, reducing equipment over-limit operation. By integrating various information to form a feature set, standardized output content can be uniformly generated, avoiding the problem of scattered and chaotic demand analysis leading to deviations in control direction. This approach not only meets the maintenance needs of different users but also ensures that the overall control operation remains within a reasonable operating range.

[0056] In one embodiment of the present invention, S32 includes:

[0057] Receive the physiological signal feature components and mechanical feedback feature components that have been split and processed, unify the data sampling frequency and data precision of the two types of components, complete the pre-data normalization process, and generate normalized feature component data;

[0058] Run a multi-dimensional data fusion algorithm to perform cross-dimensional correlation matching on the normalized feature component data, connect the transmission links of different types of data, and generate preliminary fused feature data;

[0059] Dimensional correlation verification is performed on the preliminary fusion feature data to eliminate information barriers between different data dimensions, break down the interaction barriers formed by the independent existence of various types of data, and generate interconnected fusion data;

[0060] Full-domain feature mining operations are performed on interconnected and fused data to capture the correspondence between changes in physiological state and changes in mattress mechanical state, extract the intrinsic relationship between the interaction of various states, and generate correlation feature data.

[0061] By integrating all the information generated from fusion operations and pattern mining, a unified integration of multi-dimensional features is achieved, generating comprehensive feature data with deep fusion.

[0062] The working principle and effects of the above technical solution are as follows: By unifying the sampling frequency and data precision of the two types of feature components, data specification differences can be eliminated, reducing fusion errors caused by inconsistencies in the initial data. By cross-dimensional correlation matching of various feature data, the correlation channels between different data dimensions can be opened, improving the adaptability of multi-type data collaborative processing. By conducting correlation verification on the preliminary fused data, information barriers between data can be eliminated, reducing the bias in judgment caused by analysis of a single data dimension. By using full-domain feature mining operations, the inherent correlation between human physiology and mattress biomechanics can be captured, improving the depth of state pattern mining. By integrating various computational information to generate comprehensive feature data, a complete basis for state analysis can be formed, avoiding the problem of judgment bias caused by the participation of scattered feature data in analysis. This not only enriches the dimensions of data analysis but also effectively improves the accuracy of subsequent maintenance needs analysis.

[0063] In one embodiment of the present invention, S34 includes:

[0064] The algorithm receives the control direction related data output by the algorithm module, classifies it according to different directions of spinal care, and separates the control directions into different categories such as posture adjustment, pressure relief, and posture correction, generating classified control direction data.

[0065] By matching the categorized control target data with the corresponding operational content, and matching specific operational content for each type of control target, we can divide the control targets into different levels and generate detailed control target data.

[0066] Collect information on the real-time physiological state of the human body and the stress state of the spine, compare the maintenance urgency corresponding to different states, and hierarchically arrange all subdivided control targets to generate target priority arrangement data.

[0067] Retrieve the operating threshold information of each functional unit of the smart mattress, combine the hardware operating boundaries and safety operating standards, match the corresponding operating restrictions for each control target, and generate operating constraint item data;

[0068] The data on the detailed control targets, the priority arrangement data of the targets, and the data on the operational constraints are combined and arranged to form a complete set of target priorities and constraint conditions.

[0069] The working principle and effects of the above technical solution are as follows: By categorizing and splitting the control direction data, different operational directions of spinal maintenance can be refined, improving the detail of control needs classification. By matching corresponding operational content to various control directions, clear hierarchical control targets can be refined, making the overall control operation more targeted. By combining the real-time status of the human body to prioritize target placement, urgent spinal maintenance problems that need improvement can be addressed first, enhancing the practical value of control operations. By setting constraints based on equipment operating thresholds, the operating range of each functional unit can be standardized, reducing safety hazards caused by equipment exceeding limits. By integrating various targets and constraints, a complete control standard system can be formed, avoiding problems such as poor maintenance effects or abnormal equipment operation caused by disordered control, which can both meet the dynamic maintenance needs of the human body and ensure the long-term stable and safe operation of the equipment.

[0070] One embodiment of the present invention, such as Figure 2 As shown, S4 includes:

[0071] S41. Read the information within the feature set of the control intention, analyze the corresponding control objectives and operational constraints under different priorities, and clarify the execution direction of the overall control work;

[0072] S42. Based on the analysis results, calculate the required air volume of the air pump under the corresponding working conditions, match the air inflation standards corresponding to different maintenance needs, and determine the relevant numerical content of the air pump operation.

[0073] S43. Calculate the required output force of the massage unit based on the stress state of the human spine, distinguish the massage intensity corresponding to different body areas, and complete the calculation of massage-related parameters.

[0074] S44. Calculate the overall angle adjustment range of the mattress according to the human lying posture and spinal care standards, define the reasonable range of angle adjustment, and determine the relevant operating parameters for angle adjustment;

[0075] S45 integrates various calculation parameters such as air pump inflation volume, massage unit force, mattress angle adjustment, etc., and arranges the parameters in an orderly manner according to the execution logic to generate a complete set of control parameter instructions.

[0076] The working principle and effects of the above technical solution are as follows: By reading and parsing the feature set of control intentions, it can accurately align the control targets and operational limitations at each level, ensuring that parameters are generated to meet actual maintenance needs. By specifically calculating the air pump inflation volume, the mattress support strength can be adapted to different spinal conditions, improving the fit of body support. By calculating the force of the massage unit in different zones, it can adapt to the differences in force on different parts of the body, improving the adaptability and comfort of the massage operation. By accurately calculating the mattress angle adjustment amount, it can correct poor lying postures and optimize the force distribution on the spine. By standardizing and integrating various control parameters into an instruction set, it can unify the equipment execution standards, reducing the problem of chaotic and disordered parameters leading to erratic control actions. This not only enables the coordinated operation of multi-dimensional maintenance functions but also improves the overall precision and operational stability of the smart mattress control.

[0077] In one embodiment of the present invention, S44 includes:

[0078] The system collects real-time data on human lying posture captured by the smart mattress, including the tilt angle of the human torso, the curvature of the spine, and the posture information of the area of ​​the body that is in contact with the mattress, and generates complete raw data on human lying posture.

[0079] Standardized spinal maintenance benchmark parameters are introduced. Real-time raw data of human lying posture is compared with maintenance benchmark parameters to calculate the spinal force imbalance deviation value under the current posture and generate posture deviation comparison data.

[0080] By combining posture deviation comparison data, mattress angle adaptation calculation is performed to calculate the mattress tilt adjustment value that can correct spinal force deviation and optimize human lying posture, and generate initial angle adjustment calculation data.

[0081] Based on the hardware operating threshold of the smart mattress angle adjustment mechanism, the initial angle adjustment calculation data is subjected to range constraint filtering to eliminate invalid values ​​that exceed the device's operating range, lock the legal fluctuation range of angle adjustment, and generate angle adjustment range data.

[0082] The angle adjustment values ​​within the legal range are precisely calibrated to match the optimal posture standard for human spinal relaxation and maintenance, generating mattress angle adjustment operation parameters that meet human maintenance needs.

[0083] The working principle and effects of the above technical solution are as follows: By comprehensively collecting data related to human lying posture, it can fully grasp the user's lying state and spinal shape, improving the comprehensiveness of posture recognition. By comparing real-time posture data with spinal maintenance benchmark parameters, it can accurately calculate the degree of spinal stress imbalance, improving the accuracy of posture deviation judgment. By calculating the mattress adjustment angle based on deviation data, it can specifically correct poor lying postures and improve abnormal spinal stress. By combining the device's hardware threshold to filter adjustment values, it can eliminate unreasonable adjustment parameters and reduce operational abnormalities caused by the mechanism exceeding its limits. By precisely calibrating the effective angle values, it can conform to the optimal state of spinal relaxation, improve the precision of angle control, and avoid the problems of poor maintenance effect and physical discomfort caused by coarse angle adjustment. It can adapt to different users' lying habits and continuously maintain a comfortable and healthy spinal support state.

[0084] In one embodiment of the present invention, step S5 includes:

[0085] S51. The established set of control parameter instructions is transmitted to various execution units of the smart mattress physical entity via a transmission link, such as... Figure 4 As shown, the command content has been distributed globally.

[0086] S52. After receiving the instruction, the execution unit drives the air pump massage motor angle adjustment mechanism to operate synchronously, and completes the corresponding air massage posture adjustment and other control actions according to the instruction.

[0087] S53. After the control action is completed, the multimodal sensing unit collects real-time information on the mattress operation status and human physiological status again to obtain real-time feedback data after the action is executed.

[0088] S54. Upload the real-time feedback data to the digital twin virtual model, compare the actual operating status with the model's preset expected status, and identify any state differences between the two.

[0089] S55. Combine the state difference to dynamically correct the original control parameter instruction set, repeatedly optimize various operating parameters, record the state changes and parameter adjustments throughout the process, and generate closed-loop control feedback data.

[0090] The working principle and effects of the above technical solution are as follows: By uniformly distributing control parameter instruction sets through the transmission link, it ensures that various execution units receive operating instructions synchronously, improving the overall efficiency of instruction transmission. Through the collaborative completion of control actions by multiple mechanisms, multiple maintenance operations can be implemented simultaneously, enhancing the synergy of overall functional execution. By re-collecting on-site status information, the actual situation after control can be grasped in a timely manner, improving the follow-up capability of status perception. By comparing the actual state with the expected state in the virtual model, operational differences can be quickly identified, improving the efficiency of deviation investigation. By continuously correcting parameters and retaining a complete record, the operating logic can be continuously optimized, reducing the decrease in adaptability caused by long-term use of fixed parameters, and avoiding the problem of gradually weakening maintenance effects due to a lack of follow-up adjustments after a single control operation. This ensures that the equipment operating status continuously conforms to maintenance standards and that the entire control system maintains long-term operational capability.

[0091] In one embodiment of the present invention, S55 includes:

[0092] Retrieve the deviation content corresponding to the state difference, and perform deviation adaptation and correction calculations for the air pump inflation parameters, massage unit force parameters, and mattress angle adjustment parameters to generate preliminary correction parameter data;

[0093] The preliminary correction parameter data is substituted into the operating conditions of the smart mattress for simulation and deduction, the control effect of the correction parameters is tested, the effective correction parameters that are suitable for the human spine maintenance state are selected, and the optimized parameter data is generated.

[0094] Replace the corresponding parameter content of the original control parameter instruction set with the optimized parameter data to complete the iterative update of the control parameter instruction set and generate the updated control instruction dataset.

[0095] It continuously records the entire process of parameter correction, command iteration, and state adaptation, collects corresponding data on changes in human body state and equipment parameter adjustments, and generates full-process control record data.

[0096] By integrating and iteratively updating control command data with full-process control record data, the system completes the collection of all-dimensional information for closed-loop control and generates complete closed-loop control feedback data.

[0097] The working principle and effects of the above technical solution are as follows: By combining state difference to correct various operating parameters item by item, it can specifically compensate for deviations in regulation and improve the accuracy of parameter correction. By simulating and deducing to verify the correction effect, unreasonable parameters can be screened out, reducing the fluctuations in user experience caused by improper parameter adjustments. By replacing the original content to complete the instruction set iteration, the operating rules can be continuously updated with state changes, enhancing the system's dynamic adaptability. By recording parameter changes and state changes throughout the process, a complete operating trajectory can be preserved, improving the traceability of overall operating data. By summarizing various information to form feedback data, the complete regulation chain can be connected, avoiding the problem of operational disorder caused by blindly changing parameters. This not only allows for continuous refinement of the regulation effect but also ensures that the entire closed-loop system maintains a stable and reliable operating state.

[0098] One embodiment of the present invention provides a digital twin-driven multimodal intelligent control system for spinal cord care, the system comprising:

[0099] One or more processors;

[0100] Memory, used to store one or more programs;

[0101] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0102] Example 2

[0103] In another embodiment of the present invention, during actual use, the support for the human body in a lying position is not determined independently by a single body area, but rather by multiple body areas such as the shoulders, back, waist, buttocks, and legs forming a continuous support chain extending along the spine. When the pressure in a certain local body area is too high, directly reducing the air volume of the corresponding area based solely on the pressure level may alleviate local pressure in that area, but it can easily disrupt the support transition relationship between adjacent body areas. For example, after the pressure in the buttocks area is reduced, the waist area may tend to be unsupported due to insufficient support connection; after the pressure in the shoulders and back area is released, the center of force on the torso may shift to the waist-hip junction area, causing compensatory pressure in adjacent areas.

[0104] Therefore, this embodiment further determines whether the local pressure release action will cause insufficient lumbar support or compensatory compression in adjacent areas, and optimizes the air pump inflation volume, massage unit force and mattress angle adjustment based on the judgment results, so that the final control parameter instruction set can simultaneously meet the requirements of local pressure release, lumbar support maintenance and overall force balance.

[0105] Specifically, the method includes the following steps S1-S8:

[0106] S1. The physiological signal data and mechanical feedback data collected by the multimodal sensing unit of the smart mattress physical entity are collected synchronously. The collected raw data is denoised and standardized preprocessed to generate a multimodal raw dataset. The physiological signal data includes human heart rate, respiration and body movement data. The mechanical feedback data includes the spinal pressure distribution on the mattress surface and the degree of overall deformation.

[0107] S2. Based on the multimodal raw dataset, construct a digital twin virtual model of the physical entity of the smart mattress, establish a real-time bidirectional mapping relationship between the state of the physical entity and the state of the virtual model, and generate digital twin mapping data; wherein, the digital twin virtual model is used to replicate the hardware structure, sensor layout, force deformation characteristics, and mattress support response corresponding to the human lying state of the smart mattress.

[0108] S3. Perform multimodal data fusion processing on the digital twin mapping data, combine physiological signal features and mechanical feedback features, and use an intent recognition algorithm to analyze the spinal cord care and regulation needs to generate a regulation intent feature set; wherein, the regulation intent feature set includes regulation targets, target priorities and operating constraints.

[0109] S4. Generate adaptive control parameter instructions based on the control intention feature set, and generate a control parameter instruction set; wherein, the control parameter instruction set includes air pump inflation volume, massage unit force, and mattress angle adjustment amount.

[0110] S5. Transmit the control parameter instruction set to the execution unit of the smart mattress physical entity to drive the air pump, massage motor, and angle adjustment mechanism to perform control actions. At the same time, collect real-time feedback data after execution, compare and analyze it with the expected state of the digital twin virtual model, dynamically adjust the control parameter instruction set, and generate closed-loop control feedback data.

[0111] S6. Based on closed-loop control feedback data and expected state data from the digital twin virtual model, the human body is divided into multiple body zones along the spinal extension direction. The pressure deviation degree, deformation response deviation degree, and support continuity mutation degree of each body zone are obtained. Based on the pressure deviation degree, deformation response deviation degree, and support continuity mutation degree, a spinal support continuity imbalance evaluation result is generated. The spinal support continuity imbalance evaluation result is used to characterize whether there is a risk of breakage, suspension, or mutation in the support chain along the spinal extension direction after local pressure adjustment, thereby identifying the risk of insufficient lumbar support before further adjustment of the air pump inflation volume.

[0112] S7. Based on the pressure deviation degree, deformation response deviation degree, and the degree of compensatory transfer in adjacent areas derived from the digital twin virtual model, a compensatory pressure risk assessment result is generated; wherein, the compensatory pressure risk assessment result is used to characterize whether there is a risk of increased force, concentrated deformation, or pressure transfer in adjacent body areas after the target body area is decompressed, thereby identifying the compensatory pressure risk in adjacent areas before the force of the massage unit and the amount of mattress angle adjustment are determined.

[0113] S8. The evaluation results of spinal support continuity imbalance and the evaluation results of compensatory compression risk are used as optimization constraints input to the digital twin virtual model. Within the legal operating range of air pump inflation volume, massage unit force, and mattress angle adjustment, a set of spinal support balance control parameters is obtained. The set of spinal support balance control parameters replaces or modifies the corresponding parameters in the original control parameter instruction set to generate the final control parameter instruction set. The final control parameter instruction set is used to ensure that local pressure release, lumbar support maintenance, and adjacent area compression inhibition simultaneously meet the control requirements.

[0114] Further, S6 includes:

[0115] S61. Call the spinal pressure distribution and overall deformation degree on the mattress surface from the closed-loop control feedback data, and call the target pressure distribution and target deformation state corresponding to the current human lying posture in the digital twin virtual model. Divide the human body along the spinal extension direction into shoulder and back areas, waist areas, hip areas and leg areas to form multiple body areas.

[0116] S62. For each body region, the actual pressure state collected by the multimodal sensing unit is compared with the target pressure state output by the digital twin virtual model to generate the pressure deviation degree of the body region; wherein, the pressure deviation degree is used to indicate the degree to which the actual pressure state of the body region deviates from the target pressure state. The higher the pressure deviation degree, the less the support pressure of the body region conforms to the current spinal maintenance and control target.

[0117] S63. For each body section, the actual deformation state collected by the multimodal sensing unit is compared with the target deformation state output by the digital twin virtual model to generate the deformation response deviation degree of the body section; wherein, the deformation response deviation degree is used to indicate the degree to which the actual deformation response of the mattress of the body section deviates from the target deformation response. The higher the deformation response deviation degree, the lower the degree of matching of the mattress deformation of the body section with the human spine support.

[0118] S64. Based on the pressure change trend and deformation change trend between adjacent body sections, generate the support continuity abrupt change degree; wherein, the support continuity abrupt change degree is used to indicate the abrupt change in the support transition between adjacent body sections. The higher the support continuity abrupt change degree, the easier it is for support breakpoints, waist suspension or local collapse to form between adjacent body sections.

[0119] S65. Based on the degree of pressure deviation, deformation response deviation, and abrupt change in support continuity of each body region, and combined with the importance of each body region in the evaluation of spinal support continuity, a comprehensive evaluation of the support status of each body region is conducted to generate a spinal support continuity imbalance evaluation result. Among them, in the comprehensive evaluation, the evaluation weight of the lumbar region and adjacent lumbar regions is higher than that of body regions far from the lumbar region, so that the evaluation result can reflect the risk of insufficient lumbar support first.

[0120] The function of S6 mentioned above is to move beyond solely relying on whether the pressure in a particular body region is too high as the basis for regulation. Instead, it incorporates the degree of pressure deviation, the degree of deformation response deviation, and the degree of abrupt changes in the continuity of support between adjacent body regions into the judgment scope. This allows the system to determine whether the adjustment action will disrupt the continuous support state along the spine before local decompression. In particular, when the pressure deviation in the lumbar region is not significant, but significant deformation changes occur in the adjacent hip or shoulder-back regions, the evaluation result of spinal support continuity imbalance can still reflect a high risk, thus prompting the system to avoid directly performing single-point decompression actions and reducing the probability of insufficient lumbar support.

[0121] Further, S7 includes:

[0122] S71. Using the control parameter instruction set generated in S4 as the candidate control parameter instruction set, the candidate control parameter instruction set is input into the digital twin virtual model to simulate and extrapolate the mattress support state after the changes in air pump inflation volume, massage unit force, and mattress angle adjustment, and to generate candidate extrapolation state data.

[0123] S72. Extract the predicted pressure changes and predicted deformation changes of each body region from the candidate simulation state data, and compare them with the pressure state and deformation state in the current closed-loop control feedback data to determine the force transfer trend of adjacent body regions after local body region decompression.

[0124] S73. Generate the degree of compensatory transfer in adjacent areas based on the force transfer trend; wherein, the degree of compensatory transfer in adjacent areas is used to indicate the degree of compensatory force transfer that a certain body part bears after the adjacent body part performs a pressure release action. The higher the degree of compensatory transfer in adjacent areas, the more likely the body part is to experience increased pressure, concentrated deformation, or overloaded support due to the decompression of other body parts.

[0125] S74. Call the pressure deviation degree and deformation response deviation degree generated in S6, and combine them with the degree of compensation transfer in adjacent areas to comprehensively evaluate the compensatory compression risk of each body part and generate the compensatory compression risk evaluation result. Among them, in the comprehensive evaluation, the pressure deviation degree and deformation response deviation degree are used to reflect the existing support anomalies, and the degree of compensation transfer in adjacent areas is used to reflect the possible new compression transfer risk after local adjustment.

[0126] The function of S7 mentioned above is to pre-simulate the impact of the decompression action on the pressure transfer of adjacent body regions using a digital twin virtual model before the local decompression control parameters are officially issued. The degree of pressure deviation, deformation response deviation, and the degree of compensatory transfer in adjacent areas are then used together to assess the risk of compensatory compression. Therefore, the system can avoid introducing new pressure concentrations in adjacent regions while only addressing compression in the target region, thus improving the overall stability of spinal health control.

[0127] Further, S8 includes:

[0128] S81. Construct a set of candidate control parameters, which includes multiple candidate air pump inflation volumes, candidate massage unit forces, and candidate mattress angle adjustments; wherein each candidate control parameter satisfies the hardware operating thresholds of the air pump, massage motor, and angle adjustment mechanism, as well as the operating constraints in the control intention feature set.

[0129] S82. Input the candidate control parameter set into the digital twin virtual model one by one to perform state simulation, and obtain the evaluation results of spinal support continuity imbalance and compensatory compression risk assessment results corresponding to each candidate control parameter set.

[0130] S83. Based on the evaluation results of spinal support continuity imbalance and compensatory compression risk assessment results, the candidate control parameter set is jointly screened. Among them, candidate control parameters that will lead to a decrease in lumbar support continuity are first eliminated, and then candidate control parameters that will lead to an increase in compensatory compression in adjacent areas are eliminated. From the remaining candidate control parameters, the candidate control parameters that can take into account local pressure release, lumbar support maintenance and adjacent area compression inhibition are selected as the spinal support balance control parameter set.

[0131] S84. Replace the corresponding parameters in the original control parameter instruction set with the air pump inflation volume, massage unit force, and mattress angle adjustment amount in the spinal support balance control parameter set to generate the final control parameter instruction set.

[0132] S85. The final control parameter instruction set is transmitted to the execution unit of the smart mattress physical entity, so that the air pump, massage motor and angle adjustment mechanism execute control actions according to the final control parameter instruction set. The multimodal sensing unit re-collects the real-time feedback data after execution and transmits the real-time feedback data back to the digital twin virtual model to continue the next round of spinal support balance constraint calculation.

[0133] The function of S8 is to combine the spinal support continuity imbalance evaluation results obtained from S6 and the compensatory compression risk evaluation results obtained from S7 as the final parameter selection criteria. This ensures that the final control parameter instruction set does not simply pursue a reduction in pressure in the target body area, but rather, while satisfying local pressure release, it also considers the continuity of lumbar support and the risk of compression in adjacent areas. This approach avoids problems such as insufficient lumbar support, force transfer to adjacent areas, or support breakage between body areas caused by local pressure regulation, transforming the smart mattress's control action from a single-point pressure response to overall support balance control along the spinal direction.

[0134] The working principle and effect of this embodiment are as follows: In scenarios of abnormal local pressure, if the air pump inflation volume is directly adjusted based on the local pressure deviation, problems may arise such as a decrease in local pressure leading to a decline in the continuity of lumbar spine support and an increase in compensatory compression in adjacent areas. First, an evaluation result of spinal support continuity imbalance is generated using the degree of pressure deviation, the degree of deformation response deviation, and the degree of abrupt change in support continuity. Then, a compensatory compression risk evaluation result is generated using the same degree of pressure deviation, the degree of deformation response deviation, and the degree of compensatory transfer in adjacent areas. Finally, the spinal support continuity imbalance evaluation result and the compensatory compression risk evaluation result are jointly input into a digital twin virtual model for parameter optimization, resulting in a set of spinal support balance control parameters. Therefore, this approach retains the original technical links of multimodal perception, digital twin mapping, control parameter generation, and closed-loop feedback from Embodiment 1, while further addressing the problem of support imbalance in other areas caused by local pressure adjustment, thereby improving the overall integrity, stability, and safety of the intelligent mattress's spinal care and control.

[0135] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A digital twin-driven multimodal intelligent control method for spinal cord care, characterized in that, The method includes: S1. Synchronously collect physiological signal data and mechanical feedback data from the multimodal sensing unit of the smart mattress physical entity, and perform noise reduction and standardization preprocessing on the collected raw data to generate a multimodal raw dataset. S2. Based on the multimodal raw dataset, construct a digital twin virtual model of the physical entity of the smart mattress, establish a real-time bidirectional mapping relationship between the physical entity state and the virtual model state, and generate digital twin mapping data; S3. Perform multimodal data fusion processing on the digital twin mapping data, combine physiological signal features and mechanical feedback features, use the intention recognition algorithm to analyze the spinal care regulation needs, and generate a regulation intention feature set; S4. Generate adaptive control parameter instructions and control parameter instruction set based on the control intention feature set; S5. Transmit the control parameter instruction set to the execution unit of the smart mattress physical entity to drive the air pump, massage motor, and angle adjustment mechanism to perform control actions. At the same time, collect real-time feedback data after execution, compare and analyze it with the expected state of the digital twin virtual model, dynamically adjust the control parameter instruction set, and generate closed-loop control feedback data.

2. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 1, characterized in that, S1 includes: S11. Multimodal sensing units deployed throughout the smart mattress continuously capture three types of physiological signals: human heart rate, breathing, and body movement. Simultaneously, they collect two types of mechanical feedback information: the distribution of spinal pressure on the mattress surface and the degree of overall deformation, thus completing the continuous collection of raw data in all dimensions. S12. Perform full-domain noise filtering on all types of raw data that have been collected, remove invalid data caused by environmental interference signals and equipment jitter, and retain basic data content with effective analytical value. S13. Perform a unified standard conversion operation on all types of data that have completed noise reduction processing, unify the numerical range and sampling frequency of the output data of different sensing units, and eliminate the analysis bias caused by data format differences. S14. Integrate the denoised and standardized physiological signal data and mechanical feedback data to classify and integrate data from different dimensions, forming a complete and standardized multimodal raw dataset.

3. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 1, characterized in that, S2 includes: S21. Retrieve the stored multimodal raw dataset, extract the human body state information and mattress physical operation state information carried within the data, and sort out the current full-dimensional operation characteristics of the physical entity. S22. Combine the extracted operational features to build a digital twin virtual model of the corresponding structure, replicate the hardware structure, sensor layout and force deformation characteristics of the smart mattress, and restore the complete form of the physical entity. S23. Establish a data interaction and transmission link to open up the data channel between the physical entity of the smart mattress and the digital twin virtual model, ensuring that the data at both ends can flow and communicate without interruption. S24. A state synchronization mechanism is built based on real-time data flow, so that the state changes generated by physical entities are synchronously mapped to the virtual model, and the state changes inferred by the virtual model are also synchronously transmitted back to the physical entities, forming a two-way linkage state system. S25. Record the correspondence information generated by each state synchronization between the physical entity and the virtual model, summarize all the related information, and generate complete digital twin mapping data.

4. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 3, characterized in that, S24 includes: It continuously receives real-time status data from the data interaction and transmission link, distinguishes between dynamic data from the physical entity of the smart mattress and inferred data from the digital twin virtual model, and completes the classification and aggregation of the two types of data streams. Configure synchronous transmission rules for the collected physical entity dynamic data, set the data sampling frequency and forwarding logic, and push the real-time state changes of the mattress to the digital twin virtual model in real time to complete the synchronous mapping of the physical state to the virtual end. Configure reverse feedback rules for the inference and calculation data output by the virtual model, match the running response rhythm of the physical entity, and continuously push the predicted state data calculated by the model to the physical entity of the smart mattress to complete the synchronous feedback of the virtual state to the physical end. Continuously monitor the synchronization timeliness of data transmission and status updates at both ends, correct timing deviations that occur during data flow, and maintain the pace of status updates between the entity and the virtual model. Integrate all operational aspects of bidirectional data flow and status synchronization, connect the complete operational process of status linkage between the two ends, and form a stable bidirectional linkage status system.

5. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 1, characterized in that, The S3 includes: S31. Retrieve the generated digital twin mapping data, separate the physiological signal feature components and mechanical feedback feature components contained in the data, and distinguish the data content of two different attributes. S32. A multi-dimensional data fusion algorithm is used to deeply fuse the two types of feature components after splitting, break down the barriers between different data dimensions, and uncover the hidden state correlation patterns behind the data. S33. Input the fused comprehensive feature data into the intent recognition algorithm module. The algorithm module deeply analyzes the actual needs of the human body for current spinal care and determines the corresponding adjustment direction. S34. Based on the control direction obtained from the analysis, classify different types of control targets, determine the execution priority of each target according to the severity of the human body condition, and set corresponding constraints based on the operating limits of the equipment. S35. Summarize all information related to the execution priority and constraints of all regulatory targets, integrate and summarize the feature information, and generate a complete feature set of regulatory intentions.

6. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 5, characterized in that, S32 includes: Receive the physiological signal feature components and mechanical feedback feature components that have been split and processed, unify the data sampling frequency and data precision of the two types of components, complete the pre-data normalization process, and generate normalized feature component data; Run a multi-dimensional data fusion algorithm to perform cross-dimensional correlation matching on the normalized feature component data, connect the transmission links of different types of data, and generate preliminary fused feature data; Dimensional correlation verification is performed on the preliminary fusion feature data to eliminate information barriers between different data dimensions, break down the interaction barriers formed by the independent existence of various types of data, and generate interconnected fusion data; Full-domain feature mining operations are performed on interconnected and fused data to capture the correspondence between changes in physiological state and changes in mattress mechanical state, extract the intrinsic relationship between the interaction of various states, and generate correlation feature data. By integrating all the information generated from fusion operations and pattern mining, a unified integration of multi-dimensional features is achieved, generating comprehensive feature data with deep fusion.

7. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 5, characterized in that, S34 includes: The algorithm receives the control direction-related data output by the algorithm module, classifies it according to different directions of spinal care, separates the control directions of different categories, and generates classified control direction data. By matching the categorized control target data with the corresponding operational content, and matching specific operational content for each type of control target, we can divide the control targets into different levels and generate detailed control target data. Collect information on the real-time physiological state of the human body and the stress state of the spine, compare the maintenance urgency corresponding to different states, and hierarchically arrange all subdivided control targets to generate target priority arrangement data. Retrieve the operating threshold information of each functional unit of the smart mattress, combine the hardware operating boundaries and safety operating standards, match the corresponding operating restrictions for each control target, and generate operating constraint item data; The data on the detailed control targets, the priority arrangement data of the targets, and the data on the operational constraints are combined and arranged to form a complete set of target priorities and constraint conditions.

8. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 1, characterized in that, The S4 includes: S41. Read the information within the feature set of the control intention, analyze the corresponding control objectives and operational constraints under different priorities, and clarify the execution direction of the overall control work; S42. Based on the analysis results, calculate the required air volume of the air pump under the corresponding working conditions, match the air inflation standards corresponding to different maintenance needs, and determine the relevant numerical content of the air pump operation. S43. Calculate the required output force of the massage unit based on the stress state of the human spine, distinguish the massage intensity corresponding to different body areas, and complete the calculation of massage-related parameters. S44. Calculate the overall angle adjustment range of the mattress by referring to the human lying posture and spinal care standards, define the range of angle adjustment, and determine the relevant operating parameters for angle adjustment. S45 integrates various calculation parameters such as air pump inflation volume, massage unit force, mattress angle adjustment, etc., and arranges the parameters in an orderly manner according to the execution logic to generate a complete set of control parameter instructions.

9. The method for intelligent control of spinal cord maintenance based on digital twin-driven multimodal sensing according to claim 1, characterized in that, The S5 includes: S51. The established set of control parameter instructions is transmitted to various execution units of the smart mattress physical entity through the transmission link to complete the full-domain distribution of instruction content. S52. After receiving the instruction, the execution unit drives the air pump massage motor angle adjustment mechanism to operate synchronously and completes the corresponding control action according to the instruction. S53. After the control action is completed, the multimodal sensing unit collects real-time information on the mattress operation status and human physiological status again to obtain real-time feedback data after the action is executed. S54. Upload the real-time feedback data to the digital twin virtual model, compare the actual operating status with the model's preset expected status, and identify any state differences between the two. S55. Combine the state difference to dynamically correct the original control parameter instruction set, repeatedly optimize various operating parameters, record the state changes and parameter adjustments throughout the process, and generate closed-loop control feedback data.

10. A digital twin-driven multimodal intelligent control system for spinal cord maintenance, characterized in that: The system includes: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images