A rehabilitation training device cluster scheduling method, system, device and medium

Abstract

This application relates to the field of rehabilitation training equipment scheduling technology, specifically to a method, system, device, and medium for cluster scheduling of rehabilitation training equipment. The method includes: acquiring rehabilitation task information and spatial location information of patients in the rehabilitation training area, and the status and spatial location information of each rehabilitation training device; based on the above information, generating a personalized scheduling scheme for each patient by performing multi-objective optimization calculations aimed at minimizing total rehabilitation training time, maximizing equipment utilization, minimizing patient waiting time, and minimizing energy consumption; generating and executing a sequence of control commands according to the personalized scheduling scheme, predicting and resolving potential conflicts in real time during execution, and dynamically adjusting task command parameters based on the patient's real-time motion and physiological data. This application can achieve collaborative scheduling and adaptive control of multiple patients and multiple devices, significantly improving the overall efficiency, safety, and personalization level of rehabilitation training.

Description

Technical Field

[0001] This application relates to the field of rehabilitation training equipment scheduling technology, specifically to a method, system, equipment, and medium for cluster scheduling of rehabilitation training equipment. Background Technology

[0002] In modern rehabilitation medicine, patients' demand for personalized and precise rehabilitation training is increasing, driving rehabilitation treatment towards intelligence and automation. Utilizing automated equipment such as robots to assist or perform rehabilitation training tasks has become an important way to alleviate manpower pressure and improve treatment standards.

[0003] To improve the efficiency and management of rehabilitation training, existing technologies mainly employ two methods for configuring and scheduling rehabilitation training equipment. One is to develop intelligent rehabilitation training equipment, such as active training robots with force feedback and trajectory guidance functions, which provide standardized training for individual patients through preset programs. The other is to introduce equipment management systems into rehabilitation departments, which manage the usage order of multiple devices through simple appointment queuing mechanisms or scheduling software based on fixed schedules, aiming to reduce patient waiting time and improve equipment turnover rate.

[0004] However, the configuration and scheduling methods of the aforementioned rehabilitation training equipment have inherent limitations when dealing with group-based and dynamic rehabilitation training scenarios: existing active training robots are mostly functionally isolated "standalone" versions, focusing on a single training task, lacking physical and informational linkage between devices, and unable to organize multi-device, whole-body collaborative training for the same patient; existing rehabilitation training equipment and its management system are usually independent of each other, with data not shared, forming "information silos," making it difficult for therapists to easily obtain complete training data of patients on different devices to form a unified rehabilitation assessment; existing configuration and scheduling modes mean that a device can usually only serve one patient at a time, while other patients must wait in line, resulting in both idle equipment and patient waiting, and low utilization of expensive equipment resources; due to the lack of real-time perception and collaborative control of the overall patient and equipment status, existing technologies cannot support multiple patients to receive safe, efficient, and parallel treatment in the same physical space, and cannot improve overall treatment efficiency through scheduling optimization, nor can they build a mutually motivating group rehabilitation environment. Summary of the Invention

[0005] To address the shortcomings of existing rehabilitation training equipment configuration and scheduling methods, which lack global awareness and coordination mechanisms, resulting in low equipment utilization, long patient waiting times, and an inability to safely and efficiently support parallel rehabilitation treatment for multiple patients, this application provides a rehabilitation training equipment cluster scheduling method, system, equipment, and medium. Through multi-objective optimization and real-time collaborative scheduling, heterogeneous rehabilitation equipment is integrated into an intelligent cluster, achieving efficient parallel scheduling and personalized adaptive treatment for multiple patients and multiple devices, thereby improving overall treatment efficiency, resource utilization, and treatment safety.

[0006] Firstly, this application provides a method for scheduling a cluster of rehabilitation training equipment, comprising the following steps: S1. Obtain rehabilitation task information for each patient in the rehabilitation training area, including patient identity information and digital rehabilitation prescription. The rehabilitation training area is equipped with rehabilitation training equipment, including active training robots, autonomous mobile robots and fixed rehabilitation training equipment. S2. Real-time acquisition of spatial location information of each patient in the rehabilitation training area and rehabilitation training equipment information of each rehabilitation training device; Information on rehabilitation training equipment includes the status and spatial location information of the equipment. S3. Based on the information of rehabilitation training equipment, the digital rehabilitation prescription and spatial location information of each patient, a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area is generated through multi-objective optimization calculation. The optimization objectives of the multi-objective optimization calculation include the shortest total rehabilitation training time, the highest equipment utilization rate, the least patient waiting time, and the lowest energy consumption. S4. Based on the personalized scheduling scheme, generate the control instruction sequence of the corresponding rehabilitation training equipment. The control instructions in the control instruction sequence include the movement guidance instruction for controlling the autonomous mobile robot to move and guiding the patient to the target location through human-computer interaction, and the rehabilitation task instruction for controlling the active training robot or fixed rehabilitation training to perform specific rehabilitation training actions. S5. Execute the control instruction sequence, and perform the following operations synchronously during the execution: Predict and resolve potential conflicts based on the spatial location information of each patient and each rehabilitation training device; The system acquires the patient's motion and physiological data corresponding to the rehabilitation task instructions in real time, and dynamically adjusts the execution parameters of the current rehabilitation task instructions based on the motion and physiological data.

[0007] Active training robots refer to robotic devices that can directly contact a patient's limbs and provide active force assistance, resistance, or trajectory guidance to complete specific rehabilitation training movements. They typically include multi-degree-of-freedom robotic arms, force / torque sensors, drive units, and control systems. Examples include multi-joint exoskeleton robots for upper limb rehabilitation, powered leg supports or ankle-foot rehabilitation robots for lower limb gait training, etc. The core function of these devices is to assist or guide patients to complete standardized active or active-assisted training movements under parameters preset by the therapist or adjusted in real time by the system.

[0008] Autonomous mobile robots refer to robotic devices capable of autonomous navigation and movement within rehabilitation training areas. Their core function is to perform spatial transfer tasks, such as guiding patients to designated training areas, transferring patients between different fixed devices, and delivering training aids or medications. These robots are typically equipped with autonomous localization and navigation modules (such as laser SLAM and visual navigation), obstacle avoidance sensors, human-machine interfaces (such as touchscreens and voice modules), and sensing units for safely following or guiding patients. For example, companion robots or intelligent transport vehicles with guidance capabilities belong to this category.

[0009] Fixed rehabilitation training equipment refers to non-mobile equipment that is permanently located and typically dedicated to a specific type or function of rehabilitation training. This type of equipment features adjustable parameters, data acquisition, and communication capabilities, and can receive and execute instructions from a central control system. Typical examples include: intelligent power bicycles with adjustable resistance and modes, balance trainers that quantify and provide feedback on training parameters, and isokinetic muscle strength trainers with data output capabilities. They are connected to a scheduling system through a unified interface, becoming schedulable resources within a cluster of equipment.

[0010] It should be further explained that, in step S1, obtaining the rehabilitation task information for each patient specifically involves: completing the patient's identity registration by swiping a card, facial recognition, or batch importing from the nurse's workstation, and loading the digital rehabilitation prescription corresponding to the patient's identity.

[0011] It should be further explained that in step S1, the digital rehabilitation prescription is a digital rehabilitation training instruction generated for the corresponding patient and received from the hospital information system. The content includes: at least one specified rehabilitation training item, the specific parameters of each training item, and the type of rehabilitation training equipment to be used.

[0012] It should be further noted that the specific parameters of the training program include at least one of the following: resistance magnitude, range of motion, duration, and number of repetitions.

[0013] It should be further explained that in step S2, the spatial location information of each patient and each rehabilitation training device is obtained in real time through a global perception network deployed in the rehabilitation training area. The global perception network includes a lidar for collecting environmental point cloud data and a camera for collecting image data.

[0014] It should be further explained that, in step S2, the specific process of obtaining the spatial location information of each patient and each rehabilitation training device through the global perception network includes: Use LiDAR to collect environmental point cloud data and use cameras to collect image data; The environmental point cloud data is denoised, filtered, and dynamically segmented to obtain point cloud clusters representing patients or rehabilitation training equipment and their locations. Target detection and recognition based on deep learning models are performed on image data to obtain the category, identity, and bounding box information of patients or rehabilitation training equipment; The location information of point cloud clusters is associated and matched with the bounding boxes of image recognition results, and each dynamic target is assigned a semantic label and a unique identity, thereby obtaining the spatial location information of each patient and each rehabilitation training device.

[0015] It should be further explained that the global perception network adopts a multimodal data cross-validation mechanism. When the lidar detects a potential near-field risk, it retrieves camera footage for visual confirmation and makes a final decision based on the confidence level of the fused information.

[0016] It should be further explained that the dynamic target segmentation uses the background subtraction method, which separates the dynamic point cloud clusters belonging to the patient and rehabilitation training equipment by comparing the current frame point cloud with the static background point cloud model. It should be further noted that step S2 also includes: A multi-target tracking algorithm is used to continuously track each dynamic target with a unique identifier across time frames, so as to output its continuous spatial position, velocity and motion trajectory information.

[0017] It should be further noted that, in step S2, the status information of the rehabilitation training equipment includes at least one of the following: working status, power information, and fault information.

[0018] It should be further explained that in step S3, based on the rehabilitation training equipment information, each patient's digital rehabilitation prescription, and spatial location information, a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area is generated through multi-objective optimization calculations. This specifically includes: S301. Decompose each patient's digital rehabilitation prescription into an atomic task sequence. The atomic task sequence contains a series of atomic tasks. Each atomic task is associated with a rehabilitation training item, the type of rehabilitation training equipment corresponding to the rehabilitation training item, and training parameters. S302. Construct a constrained optimization problem with the following optimization objectives: Minimize the total execution time of all atomic task sequences; Maximize the overall utilization rate of all rehabilitation training equipment; Minimize the overall wait time for all patients; The physiological rationality constraint must be met, meaning that there must be at least one low-intensity atomic task or rest period between two high-intensity atomic tasks for any patient; S303. Using semantic map information containing multiple semantic regions with different access attributes, real-time status and spatial location information of all rehabilitation training equipment, real-time spatial location information of all patients, and all atomic task sequences as input, an optimization algorithm based on multi-agent reinforcement learning is used to solve the constrained optimization problem and generate a personalized scheduling scheme for each patient.

[0019] It should be further noted that the rules for dividing high-load atomic tasks and low-load atomic tasks in step S302 include: An atomic task is classified as a high-load atomic task when the resistance value in the training parameters associated with the atomic task is greater than or equal to a preset high-load resistance threshold, or when the target duration of the atomic task is greater than or equal to a preset high-load duration threshold; otherwise, the atomic task is classified as a low-load atomic task.

[0020] It should be further noted that the high-load resistance threshold ranges from 60% to 80% of the patient's maximum voluntary contractile muscle strength. The high load duration threshold ranges from 90 to 180 seconds.

[0021] It should be further noted that in step S303, the semantic map information includes the location coordinates of each semantic region in the rehabilitation training area, and the semantic regions include the high-precision training area, the transition channel area, and the rest area. The high-precision training area is associated with the maximum permissible speed value of the first device, and the transition channel area is associated with the maximum permissible speed value of the second device, wherein the maximum permissible speed value of the first device is less than the maximum permissible speed value of the second device.

[0022] It should be further noted that in step S303, the optimization algorithm based on multi-agent reinforcement learning runs iteratively, and in each iteration decision, the joint state space at the current moment is obtained. And according to the state space Calculate rewards To optimize decision-making.

[0023] It should be further noted that in step S303, the state space of the optimization algorithm based on multi-agent reinforcement learning... Defined as:

[0024] in, It is a set of equipment locations consisting of the location information of all rehabilitation training equipment in the rehabilitation training area; The set of equipment velocities is comprised of the velocities of all rehabilitation training equipment in the rehabilitation training area. The patient location set is composed of the location information of all patients in the rehabilitation training area; The patient velocity set is composed of velocity information from all patients in the rehabilitation training area. It is a set of target location types consisting of the semantic region types to which the target locations of all currently scheduled atomic tasks belong in the semantic map; The fatigue index is a set of muscle fatigue indices composed of all patients in the rehabilitation training area. It is a set of compliance scores consisting of the exercise compliance scores of all patients in the rehabilitation training area.

[0025] It should be further noted that the fatigue index set In the middle, the first Muscle fatigue index of patients The calculation formula is:

[0026] in, Indicates the first Rate of change of average power frequency of surface electromyography signals in patients; Indicates based on the first The motion amplitude decay rate of each patient was calculated based on inertial data collected by the inertial measurement unit in the corresponding rehabilitation training equipment. The preset weighting coefficient for the rate of change of electromyographic signals; This is a preset weighting coefficient for the motion amplitude attenuation rate.

[0027] It should be further explained that the compliance rating set In the middle, the first Patient compliance scores The calculation method is as follows: Calculate the first The actual movement trajectory of the patients The preset ideal motion trajectory of the corresponding atomic task in its digital rehabilitation prescription Normalized cross-correlation coefficients between ,Will As .

[0028] It should be further explained that the reward The calculation formula is:

[0029] in, The efficiency bonus is calculated using the following formula:

[0030] in, The preset task completion weight coefficient, >0; The unit time interval; The number of atomic tasks completed per unit of time; The preset waiting time weighting coefficient, >0; Total number of patients; For the first A patient The cumulative waiting time at each moment; The preset equipment utilization rate weighting coefficient, 3>0; For safety bonus items, the calculation formula is as follows:

[0031] The preset safety weight coefficient, >0; This is an indicator function; its value is 1 when the condition inside the parentheses is true, and 0 otherwise. For the first A patient Adherence rating at specific times; The preset compliance score threshold ranges from 0.85 to 0.95. For quality awards, the calculation formula is as follows:

[0032] The preset quality weighting coefficients, >0; This represents the change in the muscle fatigue index; This represents the rate of change of the muscle fatigue index per unit time. The conflict penalty term is calculated using the following formula:

[0033] The default conflict penalty weighting coefficient is used. >0; Risk factor for potential conflict; The energy consumption penalty term is calculated using the following formula:

[0034] The preset energy consumption weighting coefficient, >0; The total number of rehabilitation training equipment; For the first Taiwan rehabilitation training equipment at all times Cumulative energy consumption; The penalty for abrupt movement is calculated using the following formula:

[0035] in, The preset abrupt motion weighting coefficient, >0; This represents the total number of autonomous mobile robots. The number of sampling points within the time window. The sampling interval is... For sampling point index; For the first Taiwan's autonomous mobile robots The acceleration of motion at any given moment.

[0036] It should be further noted that the optimization algorithm based on multi-agent reinforcement learning adopts a hierarchical decision architecture, including a policy manager and an action executor. Among them, the policy manager uses a joint state space The input is processed through a policy network, and the output is a discrete high-level policy instruction, which includes continuing to execute the current atomic task, pausing the current task, or switching to a low-load training mode. Action executors utilize high-level policy instructions and a unified state space. The input is processed through a motion network to output specific, continuous values ​​of motion control parameters for rehabilitation training equipment, including target velocity, joint torque, or target position coordinates.

[0037] It should be further noted that in S303, the optimization algorithm maintains a dynamic priority parameter for each patient, and calculates the reward in each iteration of the optimization algorithm. At that time, the dynamic priority parameters of each patient are normalized and used as weights to weight the waiting time penalty terms of atomic tasks related to that patient. When a patient's exercise compliance score is detected in a continuous The threshold for determining the plateau period was lower than the threshold during each control cycle. When this occurs, it is determined to be in a plateau phase, and its dynamic priority parameter is temporarily increased; Among them, the threshold for determining the plateau period The value range is 0.75-0.85.

[0038] It should be further noted that in step S303, the optimization algorithm based on multi-agent reinforcement learning performs prospective conflict avoidance when solving constrained optimization problems, specifically including: Based on the spatial location and velocity information of all rehabilitation training equipment and patients, the movement trajectories of all rehabilitation training equipment and patients in the future are predicted under the candidate scheduling scheme. Generate a spatiotemporal cube for each predicted trajectory and detect whether there is an intersection between any two spatiotemporal cubes; If an intersection is detected, the risk coefficient of the potential conflict is calculated. The calculation formula is:

[0039] in, For the estimated collision time; The basic risk value is preset based on the type of conflict, which includes human-robot conflict and robot-robot conflict. The regional risk coefficient is set based on the region type in the semantic map where the conflict occurs; The preset collision time weighting coefficient; These are preset collision type weighting coefficients; These are the preset collision region weighting coefficients; When calculating the reward at each iteration of the optimization algorithm, the sum of the risk coefficients of all detected potential conflicts is used. As a penalty, the reward value is deducted, thereby guiding the optimization algorithm to generate a personalized scheduling scheme with lower conflict risk.

[0040] It should be further noted that in step S4, generating a control instruction sequence based on the personalized scheduling scheme specifically includes: S401. Compile the device allocation, start time, spatial path, and device parameters of each atomic task in the personalized scheduling scheme into control instructions that can be executed by the corresponding rehabilitation training device; S402. Based on the logical dependencies and spatiotemporal relationships between control instructions, arrange the compiled control instructions according to the execution sequence to form a control instruction sequence; The movement guidance instructions include the target location coordinates and the expected arrival time, while the rehabilitation task instructions include the specific training movement pattern, resistance parameters, movement trajectory, and duration.

[0041] It should be further noted that step S4 also includes: sending a sequence of control commands to the corresponding rehabilitation training equipment via an edge controller deployed in the rehabilitation training area and via 5G, Wi-Fi 6 or a time-sensitive network.

[0042] It should be further explained that in step S5, predicting and resolving potential conflicts based on spatial location information specifically includes: S501: Based on the real-time acquisition of the position and speed of each patient and each rehabilitation training device, predict their movement trajectory within a specified time window in the future. S502: Calculate the estimated collision time between any rehabilitation training device and any patient or other rehabilitation training device (excluding itself). If the estimated collision time is less than the real-time collision avoidance threshold... If so, then a potential conflict is determined to exist; S503: Based on the type of conflicting parties, select a resolution strategy from the strategy library, generate real-time adjustment instructions, and send them to the corresponding rehabilitation training equipment for execution; The strategy library should include at least speed adjustment, local path replanning, and task pause strategies.

[0043] It should be further explained that the real-time collision avoidance threshold The value range is 1-3s.

[0044] It should be further explained that in step S5, the execution parameters of the current rehabilitation task instruction are dynamically adjusted based on motion data and physiological data, specifically including: Real-time calculation of muscle fatigue index and exercise compliance score of patients currently undergoing training; If the rate of increase of the muscle fatigue index exceeds the preset fatigue increase threshold within a specified number of consecutive sampling periods, a first adjustment instruction is generated and sent to the rehabilitation training equipment currently used by the patient for execution. The first adjustment instruction specifically reduces the resistance parameter in the current rehabilitation task instruction by a preset adjustment amount. If the exercise compliance score is lower than the preset compliance maintenance threshold within a specified number of consecutive sampling periods, a second adjustment instruction is generated and sent to the rehabilitation training equipment currently used by the patient for execution. The second adjustment instruction specifically reduces the exercise speed or trajectory difficulty in the current rehabilitation task instruction by one level.

[0045] It should be further noted that the fatigue rise threshold ranges from 0.05 to 0.15 seconds. The compliance maintenance threshold ranges from 0.80 to 0.90.

[0046] It should be further explained that this also includes: generating a digital twin model of the rehabilitation training area, and establishing and maintaining a digital twin model of the rehabilitation training equipment that reflects the real-time status of each rehabilitation training device, and a digital twin model of the patient that reflects the real-time location and movement status of each patient within the digital twin model of the training area; The digital twin model of the training area is used to visualize and monitor the overall status of the rehabilitation training area.

[0047] It should be further explained that during the execution of step S5, the rehabilitation training equipment currently worn or used by the patient simultaneously collects the patient's physiological and motor data, including surface electromyography signals, inertial measurement unit data, joint angles and torques, and updates the patient's digital twin model based on the physiological and motor data. It should be further noted that step S5 also includes task termination and resource release steps: when all atomic tasks of a patient are completed, or when an emergency stop command is received, the system marks all rehabilitation training equipment occupied by the patient as idle and uploads its complete treatment data package to the cloud to generate an evaluation report.

[0048] Secondly, this application provides a rehabilitation training equipment cluster scheduling system for implementing the above-mentioned rehabilitation training equipment cluster scheduling method, including: The data acquisition module is used to acquire rehabilitation task information for each patient in the rehabilitation training area, acquire the spatial location information of each patient in the rehabilitation training area in real time, and acquire rehabilitation training equipment information for each rehabilitation training device. The scheduling scheme generation module is used to generate a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area based on rehabilitation training equipment information, each patient's digital rehabilitation prescription and spatial location information, through multi-objective optimization calculation. The control command sequence generation module is used to generate control command sequences for corresponding rehabilitation training equipment based on personalized scheduling schemes. The instruction execution module is used to execute control instruction sequences, and performs the following operations synchronously during execution: Predict and resolve potential conflicts based on the spatial location information of each patient and each rehabilitation training device; The system acquires the patient's motion and physiological data corresponding to the rehabilitation task instructions in real time, and dynamically adjusts the execution parameters of the current rehabilitation task instructions based on the motion and physiological data.

[0049] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described rehabilitation training equipment cluster scheduling method.

[0050] Fourthly, this application provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described rehabilitation training equipment cluster scheduling method.

[0051] As can be seen from the above technical solutions, this application has the following advantages: 1. This application solves the problems of inefficient resource utilization caused by fragmented equipment functions, information silos, and lack of scheduling in the prior art by acquiring the patient's digital rehabilitation prescription and location information, the status and location information of rehabilitation training equipment, and generating personalized scheduling schemes based on multi-objective optimization calculations to control the collaborative work of the equipment cluster. It realizes the integration of multiple heterogeneous rehabilitation equipment into a collaborative cluster system, thereby significantly improving the overall treatment capacity and operational efficiency of the rehabilitation training area.

[0052] 2. This application solves the problems of high equipment idle rate, long patient waiting time, and overall protracted treatment process caused by existing single-machine operation and simple queuing mode by constructing a multi-objective optimization problem covering total rehabilitation training time, equipment utilization rate, patient waiting time, and energy consumption, and generating personalized scheduling schemes. It can systematically shorten the total training time and improve the comprehensive utilization rate of expensive equipment.

[0053] 3. This application solves the problem that existing technologies cannot support safe parallel training of multiple patients and multiple devices due to the lack of global real-time perception by acquiring the spatial location information of each patient and each rehabilitation training device in real time, and predicting and resolving potential conflicts based on this information during the execution process. It provides a technical basis for multiple patients to receive efficient and non-interfering group rehabilitation treatment in the same space, and ensures the safety and order of the treatment process.

[0054] 4. This application solves the problem that existing equipment mechanically executes preset programs and cannot adapt to the real-time state of patients by acquiring the patient's motion and physiological data in real time when executing the control command sequence and dynamically adjusting the execution parameters of the rehabilitation task commands. It realizes personalized adaptive adjustment of the training process and improves the accuracy and safety of treatment.

[0055] 5. This application solves the efficiency bottleneck problem of patient transfer between different fixed devices relying on manual guidance or slow autonomous movement by generating movement guidance instructions that include control of an autonomous mobile robot. The mobile robot efficiently completes the patient's location transfer and guidance, optimizes the treatment process, reduces non-treatment time, and improves the patient experience.

[0056] 6. This application solves the problem of data silos in the treatment process by decomposing digital rehabilitation prescriptions into atomic task sequences for optimized scheduling. It establishes a correlation framework based on a unified task sequence for training data of the same patient on different devices, providing a structured data foundation for forming a coherent and complete digital profile of patient rehabilitation. Attached Figure Description

[0057] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of a rehabilitation training equipment cluster scheduling method in one embodiment of this application.

[0059] Figure 2 This is a flowchart of one embodiment of the present application for generating a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area.

[0060] Figure 3 This is a flowchart of generating a control instruction sequence based on a personalized scheduling scheme in one embodiment of this application.

[0061] Figure 4This is a flowchart of predicting and resolving potential conflicts based on spatial location information in one embodiment of this application.

[0062] Figure 5 This is a schematic block diagram of a rehabilitation training equipment cluster scheduling system in one embodiment of this application.

[0063] Figure 6 This is a schematic diagram of the hardware structure of an electronic device in one embodiment of this application. Detailed Implementation

[0064] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0065] The following describes in detail the cluster scheduling method for rehabilitation training equipment involved in this application. Specific details, such as particular system structures and technologies, are presented for illustrative purposes rather than limiting, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details.

[0066] In the rehabilitation training equipment cluster scheduling method involved in this application, the term "comprising" indicates the presence of the described feature, whole, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or sets thereof. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0067] To facilitate a clear description of the technical solutions of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" do not necessarily imply that they are different.

[0068] The terms "one embodiment" or "some embodiments" used in this application mean that one or more embodiments of this application include the specific features, structures, or characteristics described in that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this application do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0069] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0070] The rehabilitation training equipment cluster scheduling method provided in this application embodiment is executed by a computer device, and correspondingly, the rehabilitation training equipment cluster scheduling system runs in the computer device.

[0071] Figure 1 This is a flowchart of a rehabilitation training equipment cluster scheduling method according to an embodiment of this application. Figure 1 The implementing entity can be a cluster scheduling system for rehabilitation training equipment. Depending on different needs, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0072] like Figure 1 As shown, the method for scheduling the rehabilitation training equipment cluster includes: Step S1: Obtain rehabilitation task information for each patient in the rehabilitation training area, including patient identity information and digital rehabilitation prescription. The rehabilitation training area is equipped with rehabilitation training equipment, including active training robots, autonomous mobile robots and fixed rehabilitation training equipment.

[0073] In some specific embodiments, obtaining rehabilitation task information for each patient involves: completing patient identity registration by swiping a card, facial recognition, or batch importing from the nurse workstation, and loading the digital rehabilitation prescription corresponding to the patient's identity.

[0074] By using methods such as card swiping, facial recognition, or batch import at nurse workstations to complete patient registration and prescription loading, the convenience, accuracy, and efficiency of information entry are improved, manual input errors are reduced, and seamless integration with the hospital's existing information system is possible.

[0075] In some specific embodiments, a digital rehabilitation prescription is a digital rehabilitation training instruction received from a hospital information system and generated for the corresponding patient. The instruction includes: at least one specified rehabilitation training item, specific parameters for each training item, and the type of rehabilitation training equipment to be used.

[0076] By clearly defining digital rehabilitation prescriptions as digital instructions from the hospital information system and limiting their content to include training programs, specific parameters, and recommended equipment types, the authority, standardization, and resolvability of the scheduling basis are ensured, enabling automated scheduling to directly understand and execute clinical medical orders.

[0077] In some specific embodiments, the specific parameters of the training program include at least one of the following: resistance magnitude, range of motion, duration, and number of repetitions.

[0078] By specifically listing the resistance magnitude, range of motion, duration, and number of repetitions as key training parameters, the digital rehabilitation prescription provides a more quantitative and precise description of training intensity, offering directly usable and multi-dimensional parameter basis for subsequent task decomposition and optimization calculations.

[0079] Step S2: Real-time acquisition of spatial location information of each patient in the rehabilitation training area and rehabilitation training equipment information of each rehabilitation training device; Information on rehabilitation training equipment includes the status and spatial location information of the equipment.

[0080] In some specific embodiments, the spatial location information of each patient and each rehabilitation training device is acquired in real time through a global perception network deployed in the rehabilitation training area. The global perception network includes a lidar for collecting environmental point cloud data and a camera for collecting image data.

[0081] By utilizing a global perception network, including lidar and cameras, deployed in the rehabilitation training area to acquire spatial location information in real time, the system is provided with a stable, reliable, and comprehensive high-precision environmental data acquisition capability, which forms the perception foundation for achieving precise scheduling and safe obstacle avoidance.

[0082] In some specific embodiments, the process of acquiring the spatial location information of each patient and each rehabilitation training device through a global perception network includes: Use LiDAR to collect environmental point cloud data and use cameras to collect image data; The environmental point cloud data is denoised, filtered, and dynamically segmented to obtain point cloud clusters representing patients or rehabilitation training equipment and their locations. Target detection and recognition based on deep learning models are performed on image data to obtain the category, identity, and bounding box information of patients or rehabilitation training equipment; The location information of point cloud clusters is associated and matched with the bounding boxes of image recognition results, and each dynamic target is assigned a semantic label and a unique identity, thereby obtaining the spatial location information of each patient and each rehabilitation training device.

[0083] By specifically describing the technical process of integrating LiDAR point cloud data processing with camera image recognition and performing data association and matching, the system achieves semantic-level perception of patients and equipment that is not only accurately located but also clearly identified in terms of identity and category, greatly enhancing the system's understanding of complex dynamic environments.

[0084] In some specific embodiments, the global perception network adopts a multimodal data cross-validation mechanism. When the lidar detects a potential near-range risk, it retrieves camera footage for visual confirmation and makes a final decision based on the confidence level of the fused information.

[0085] By introducing a multimodal data cross-validation mechanism into the global perception network, when the lidar detects a potential risk, the camera footage is retrieved for confirmation and decisions are made based on the confidence level of the fused information. This significantly improves the accuracy of the system in judging abnormal situations, effectively reduces the false alarm rate, and enhances robustness.

[0086] In some specific embodiments, dynamic target segmentation employs the background subtraction method, which separates dynamic point cloud clusters belonging to patients and rehabilitation training equipment by comparing the current frame point cloud with a static background point cloud model.

[0087] By specifying the use of background subtraction for dynamic target segmentation, an efficient and reliable point cloud data preprocessing method is provided, which can quickly separate dynamic point cloud clusters representing patients and devices from complex static backgrounds, providing clean input data for subsequent identification, tracking and localization.

[0088] In some specific embodiments, step S2 further includes: A multi-target tracking algorithm is used to continuously track each dynamic target with a unique identifier across time frames, so as to output its continuous spatial position, velocity and motion trajectory information.

[0089] In some specific embodiments, the status information of the rehabilitation training device includes at least one of the following: working status, power information, and fault information.

[0090] By explicitly extending the status information of rehabilitation training equipment to include working status, power information, and fault information, the scheduling system can more comprehensively and meticulously assess the real-time availability and health of the equipment, avoiding the assignment of tasks to equipment that is about to run out of power or has hidden faults, thus improving scheduling reliability.

[0091] Step S3: Based on the rehabilitation training equipment information, each patient's digital rehabilitation prescription and spatial location information, a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area is generated through multi-objective optimization calculation. The optimization objectives of the multi-objective optimization calculation include the shortest total rehabilitation training time, the highest equipment utilization rate, the least patient waiting time, and the lowest energy consumption.

[0092] Figure 2 A flowchart illustrating a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in a rehabilitation training area, as shown in one embodiment of this application, is provided. Figure 2 As shown, the process specifically includes: S301. Decompose each patient's digital rehabilitation prescription into an atomic task sequence. The atomic task sequence contains a series of atomic tasks. Each atomic task is associated with a rehabilitation training item, the type of rehabilitation training equipment corresponding to the rehabilitation training item, and training parameters. S302. Construct a constrained optimization problem with the following optimization objectives: Minimize the total execution time of all atomic task sequences; Maximize the overall utilization rate of all rehabilitation training equipment; Minimize the overall wait time for all patients; The physiological rationality constraint must be met, meaning that there must be at least one low-intensity atomic task or rest period between two high-intensity atomic tasks for any patient; S303. Using semantic map information containing multiple semantic regions with different access attributes, real-time status and spatial location information of all rehabilitation training equipment, real-time spatial location information of all patients, and all atomic task sequences as input, an optimization algorithm based on multi-agent reinforcement learning is used to solve the constrained optimization problem and generate a personalized scheduling scheme for each patient.

[0093] By specifying the process of generating scheduling schemes as a standardized process of decomposing prescriptions into atomic task sequences, constructing an optimization problem that includes total execution time, equipment utilization, patient waiting time, and physiological rationality constraints, and solving it using a multi-agent reinforcement learning-based algorithm, a systematic, computable, and automated decision-making method that conforms to the principles of rehabilitation medicine is provided.

[0094] In some specific embodiments, the rules for dividing high-load atomic tasks and low-load atomic tasks in step S302 include: An atomic task is classified as a high-load atomic task when the resistance value in the training parameters associated with the atomic task is greater than or equal to a preset high-load resistance threshold, or when the target duration of the atomic task is greater than or equal to a preset high-load duration threshold; otherwise, the atomic task is classified as a low-load atomic task.

[0095] By providing specific rules for classifying high-load and low-load atomic tasks based on resistance values ​​and target duration thresholds, a clear and quantifiable judgment standard is provided for the "physiological rationality constraint," enabling the optimization algorithm to accurately identify and arrange the load intervals of training tasks, thus ensuring the scientific nature of scheduling.

[0096] In some specific embodiments, the high-load resistance threshold is set at 60%-80% of the patient's maximum voluntary contractile muscle strength. The high load duration threshold ranges from 90 to 180 seconds.

[0097] In some specific embodiments, in step S303, the semantic map information includes the location coordinates of each semantic region in the rehabilitation training area, and the semantic regions include the high-precision training area, the transition channel area, and the rest area. The high-precision training area is associated with the maximum permissible speed value of the first device, and the transition channel area is associated with the maximum permissible speed value of the second device, wherein the maximum permissible speed value of the first device is less than the maximum permissible speed value of the second device.

[0098] By introducing semantic map information that includes different semantic regions such as high-precision training areas, transition channels, and rest areas, as well as the maximum allowable speeds of different devices, scheduling optimization can not only consider geometric paths but also understand the functional attributes of the environment, thereby generating device movement strategies that are more in line with regional safety standards and efficiency requirements.

[0099] In some specific embodiments, in step S303, the optimization algorithm based on multi-agent reinforcement learning runs iteratively, and in each iterative decision, the joint state space at the current moment is obtained. And according to the state space Calculate rewards To optimize decision-making.

[0100] By limiting the optimization algorithm based on multi-agent reinforcement learning to run iteratively and calculating the reward based on the current joint state space in each iteration to optimize the decision, an adaptive and self-learning algorithm framework is presented that can learn and improve the scheduling strategy step by step through continuous interaction with the environment, trial and error, and guidance by using reward signals.

[0101] In some specific embodiments, in step S303, the state space of the optimization algorithm based on multi-agent reinforcement learning... Defined as:

[0102] in, It is a set of equipment locations consisting of the location information of all rehabilitation training equipment in the rehabilitation training area; The set of equipment velocities is comprised of the velocities of all rehabilitation training equipment in the rehabilitation training area. The patient location set is composed of the location information of all patients in the rehabilitation training area; The patient velocity set is composed of velocity information from all patients in the rehabilitation training area. It is a set of target location types consisting of the semantic region types to which the target locations of all currently scheduled atomic tasks belong in the semantic map; The fatigue index is a set of muscle fatigue indices composed of all patients in the rehabilitation training area. It is a set of compliance scores consisting of the exercise compliance scores of all patients in the rehabilitation training area.

[0103] By explicitly defining the state space as consisting of a set of device positions and velocities, a set of patient positions and velocities, a set of target position semantic types, a set of patient muscle fatigue indices, and a set of exercise compliance scores, a high-dimensional information representation is constructed that can comprehensively depict the instantaneous dynamics of the rehabilitation training area, task objectives, and the physiological and psychological state of the patient. This provides rich and crucial input information for the intelligent agent to make informed decisions.

[0104] In some specific embodiments, the fatigue index set In the middle, the first Muscle fatigue index of patients The calculation formula is:

[0105] in, Indicates the first Rate of change of average power frequency of surface electromyography signals in patients; Indicates based on the first The motion amplitude decay rate of each patient was calculated based on inertial data collected by the inertial measurement unit in the corresponding rehabilitation training equipment. The preset weighting coefficient for the rate of change of electromyographic signals; This is a preset weighting coefficient for the motion amplitude attenuation rate.

[0106] By providing a specific formula for calculating the muscle fatigue index, namely, by combining the average power frequency change rate of surface electromyography signals and the motion amplitude attenuation rate based on inertial data and weighting them together, an objective, quantitative, and multi-dimensional integrated real-time physiological fatigue assessment method is provided, enabling the system to accurately "sensor" the patient's physical exertion status.

[0107] In some specific embodiments, compliance score sets In the middle, the first Patient compliance scores The calculation method is as follows: Calculate the first The actual movement trajectory of the patients The preset ideal motion trajectory of the corresponding atomic task in its digital rehabilitation prescription Normalized cross-correlation coefficients between ,Will As .

[0108] By defining the normalized cross-correlation coefficient between the patient's actual movement trajectory and the preset ideal trajectory as the exercise compliance score, a calculation standard is provided to directly quantify the accuracy and cooperation of the patient's movement execution, enabling the system to objectively measure the quality of rehabilitation training execution.

[0109] In some specific embodiments, rewards The calculation formula is:

[0110] in, The efficiency bonus is calculated using the following formula:

[0111] in, The preset task completion weight coefficient, >0; The unit time interval; The number of atomic tasks completed per unit of time; The preset waiting time weighting coefficient, >0; Total number of patients; For the first A patient The cumulative waiting time at each moment; The preset equipment utilization rate weighting coefficient, 3>0; For safety bonus items, the calculation formula is as follows:

[0112] The preset safety weight coefficient, >0; This is an indicator function; its value is 1 when the condition inside the parentheses is true, and 0 otherwise. For the first A patient Adherence rating at specific times; The preset compliance score threshold ranges from 0.85 to 0.95. For quality awards, the calculation formula is as follows:

[0113] The preset quality weighting coefficients, >0; This represents the change in the muscle fatigue index; This represents the rate of change of the muscle fatigue index per unit time. The conflict penalty term is calculated using the following formula:

[0114] The default conflict penalty weighting coefficient is used. >0; Risk factor for potential conflict; The energy consumption penalty term is calculated using the following formula:

[0115] The preset energy consumption weighting coefficient, >0; The total number of rehabilitation training equipment; For the first Taiwan rehabilitation training equipment at all times Cumulative energy consumption; The penalty for abrupt movement is calculated using the following formula:

[0116] in, The preset abrupt motion weighting coefficient, >0; This represents the total number of autonomous mobile robots. The number of sampling points within the time window. The sampling interval is... For sampling point index; For the first Taiwan's autonomous mobile robots The acceleration of motion at any given moment.

[0117] By designing specific reward function calculation formulas that include efficiency rewards, safety rewards, quality rewards, conflict penalties, energy consumption penalties, and abrupt movement penalties, a clear value orientation is provided for multi-agent reinforcement learning algorithms. This directly incentivizes agents to learn collaborative scheduling strategies that can simultaneously improve equipment utilization efficiency, training safety and quality, avoid conflicts, reduce energy consumption, and ensure movement comfort.

[0118] In some specific embodiments, the optimization algorithm based on multi-agent reinforcement learning adopts a hierarchical decision architecture, including a policy manager and an action executor; Among them, the policy manager uses a joint state space The input is processed through a policy network, and the output is a discrete high-level policy instruction, which includes continuing to execute the current atomic task, pausing the current task, or switching to a low-load training mode. Action executors utilize high-level policy instructions and a unified state space. The input is processed through a motion network to output specific, continuous values ​​of motion control parameters for rehabilitation training equipment, including target velocity, joint torque, or target position coordinates.

[0119] In some specific embodiments, in S303, the optimization algorithm maintains a dynamic priority parameter for each patient, and calculates the reward at each iteration step of the optimization algorithm. At that time, the dynamic priority parameters of each patient are normalized and used as weights to weight the waiting time penalty terms of atomic tasks related to that patient. When a patient's exercise compliance score is detected in a continuous The threshold for determining the plateau period was lower than the threshold during each control cycle. When this occurs, it is determined to be in a plateau phase, and its dynamic priority parameter is temporarily increased; Among them, the threshold for determining the plateau period The value range is 0.75-0.85.

[0120] By introducing dynamic priority parameters and setting a temporary increase in priority when a patient's exercise compliance score is detected to be consistently below the plateau threshold, the scheduling system can identify changes in the patient's training status and respond flexibly, prioritizing the allocation of resources to patients stuck in a bottleneck to help them break through, thus demonstrating humanized scheduling flexibility.

[0121] In some specific embodiments, in step S303, the optimization algorithm based on multi-agent reinforcement learning performs prospective conflict avoidance when solving constrained optimization problems, specifically including: Based on the spatial location and velocity information of all rehabilitation training equipment and patients, the movement trajectories of all rehabilitation training equipment and patients in the future are predicted under the candidate scheduling scheme. Generate a spatiotemporal cube for each predicted trajectory and detect whether there is an intersection between any two spatiotemporal cubes; If an intersection is detected, the risk coefficient of the potential conflict is calculated. The calculation formula is:

[0122] in, For the estimated collision time; The basic risk value is preset based on the type of conflict, which includes human-robot conflict and robot-robot conflict. The regional risk coefficient is set based on the region type in the semantic map where the conflict occurs; The preset collision time weighting coefficient; These are preset collision type weighting coefficients; These are the preset collision region weighting coefficients; When calculating the reward at each iteration of the optimization algorithm, the sum of the risk coefficients of all detected potential conflicts is used. As a penalty, the reward value is deducted, thereby guiding the optimization algorithm to generate a personalized scheduling scheme with lower conflict risk.

[0123] By implementing forward-looking conflict avoidance during the optimization algorithm solution process, namely predicting trajectories, detecting spatiotemporal cube intersections, calculating risk coefficients and incorporating them as penalty terms into reward calculations, the scheduling scheme can consider and avoid potential movement conflicts in advance during the generation stage, thereby reducing the collision risk in the execution stage from the source and improving the safety and reliability of the scheme.

[0124] Step S4: Based on the personalized scheduling scheme, generate a control instruction sequence for the corresponding rehabilitation training equipment. The control instructions in the control instruction sequence include a movement guidance instruction for controlling the autonomous mobile robot to move and guiding the patient to the target location through human-computer interaction, and a rehabilitation task instruction for controlling the active training robot or fixed rehabilitation training to perform specific rehabilitation training actions.

[0125] Figure 3 This is a flowchart illustrating the generation of a control instruction sequence based on a personalized scheduling scheme in one embodiment of this application, such as... Figure 3 As shown, the process specifically includes: S401. Compile the device allocation, start time, spatial path, and device parameters of each atomic task in the personalized scheduling scheme into control instructions that can be executed by the corresponding rehabilitation training device; S402. Based on the logical dependencies and spatiotemporal relationships between control instructions, arrange the compiled control instructions according to the execution sequence to form a control instruction sequence; The movement guidance instructions include the target location coordinates and the expected arrival time, while the rehabilitation task instructions include the specific training movement pattern, resistance parameters, movement trajectory, and duration.

[0126] By clearly defining the process of compiling the scheduling scheme into control instructions and arranging them logically and temporally to form an instruction sequence, and by clarifying the core content of the movement guidance instructions and rehabilitation task instructions, the instruction conversion process from decision-making to execution is ensured to be structured, unambiguous, and efficiently executed.

[0127] In some specific embodiments, step S4 further includes: sending a sequence of control commands to the corresponding rehabilitation training equipment via an edge controller deployed in the rehabilitation training area and via 5G, Wi-Fi 6 or a time-sensitive network.

[0128] Step S5: Execute the control instruction sequence, and perform the following operations synchronously during the execution: Predict and resolve potential conflicts based on the spatial location information of each patient and each rehabilitation training device; The system acquires the patient's motion and physiological data corresponding to the rehabilitation task instructions in real time, and dynamically adjusts the execution parameters of the current rehabilitation task instructions based on the motion and physiological data.

[0129] Figure 4 This is a flowchart illustrating the prediction and resolution of potential conflicts based on spatial location information in one embodiment of this application, such as... Figure 4 As shown, the process specifically includes: S501: Based on the real-time acquisition of the position and speed of each patient and each rehabilitation training device, predict their movement trajectory within a specified time window in the future. S502: Calculate the estimated collision time between any rehabilitation training device and any patient or other rehabilitation training device (excluding itself). If the estimated collision time is less than the real-time collision avoidance threshold... If so, then a potential conflict is determined to exist; S503: Based on the type of conflicting parties, select a resolution strategy from the strategy library, generate real-time adjustment instructions, and send them to the corresponding rehabilitation training equipment for execution; The strategy library should include at least speed adjustment, local path replanning, and task pause strategies.

[0130] By clearly defining the steps involved in the execution process—predicting trajectories based on real-time position and velocity, calculating the estimated collision time and comparing it with a threshold to determine the conflict, and selecting a resolution strategy from a strategy library—a standardized and automated runtime real-time collision avoidance safety mechanism is provided, effectively addressing unplanned emergencies.

[0131] In some specific embodiments, the real-time collision avoidance threshold The value range is 1-3s.

[0132] In some specific embodiments, step S5, dynamically adjusting the execution parameters of the current rehabilitation task instruction based on motion data and physiological data, specifically includes: Real-time calculation of muscle fatigue index and exercise compliance score of patients currently undergoing training; If the rate of increase of the muscle fatigue index exceeds the preset fatigue increase threshold within a specified number of consecutive sampling periods, a first adjustment instruction is generated and sent to the rehabilitation training equipment currently used by the patient for execution. The first adjustment instruction specifically reduces the resistance parameter in the current rehabilitation task instruction by a preset adjustment amount. If the exercise compliance score is lower than the preset compliance maintenance threshold within a specified number of consecutive sampling periods, a second adjustment instruction is generated and sent to the rehabilitation training equipment currently used by the patient for execution. The second adjustment instruction specifically reduces the exercise speed or trajectory difficulty in the current rehabilitation task instruction by one level.

[0133] By specifying the adjustment logic of reducing resistance parameters when the rate of increase of the muscle fatigue index exceeds a threshold, and reducing exercise speed or trajectory difficulty when the exercise compliance score remains below a threshold, the adaptive adjustment principle based on physiological data is transformed into a clear and automatically executable closed-loop control rule.

[0134] In some specific embodiments, the fatigue rise threshold ranges from 0.05 to 0.15 / s; The compliance maintenance threshold ranges from 0.80 to 0.90.

[0135] In some specific embodiments, it also includes: generating a digital twin model of the rehabilitation training area, and establishing and maintaining a digital twin model of the rehabilitation training equipment that reflects the real-time status of each rehabilitation training device, and a digital twin model of the patient that reflects the real-time location and movement status of each patient within the digital twin model of the training area; The digital twin model of the training area is used to visualize and monitor the overall status of the rehabilitation training area.

[0136] By introducing and maintaining digital twin models of rehabilitation training areas, rehabilitation training equipment, and patients, a virtual monitoring environment synchronously mapped with the physical world has been constructed, providing managers with an intuitive, panoramic, real-time status visualization monitoring method, which greatly improves the efficiency of operation and maintenance and supervision.

[0137] In some specific embodiments, during the execution of step S5, the rehabilitation training device currently worn or used by the patient synchronously collects the patient's physiological and movement data, including surface electromyography signals, inertial measurement unit data, joint angles and torques, and updates the patient's digital twin model based on the physiological and movement data.

[0138] By using real-time physiological and motor data collected from rehabilitation training equipment to update the patient's digital twin model, it is ensured that the digital twin model can dynamically and realistically reflect the patient's real-time state, making it not only a static model, but also a continuously evolving digital profile of the patient.

[0139] In some specific embodiments, step S5 also includes a task termination and resource release step: when all atomic tasks of a patient are completed, or when an emergency stop command is received, the system marks all rehabilitation training equipment occupied by the patient as idle, and uploads its complete treatment data package to the cloud to generate an evaluation report.

[0140] By setting steps for task termination and resource release, including marking devices as idle and uploading complete data packets to generate reports, a standardized closed-loop treatment process and timely resource recovery are achieved. At the same time, the system automatically completes the archiving and preliminary analysis of treatment data, improving the level of management automation.

[0141] The following are embodiments of the rehabilitation training equipment cluster scheduling system provided in this application. This rehabilitation training equipment cluster scheduling system and the rehabilitation training equipment cluster scheduling method of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the rehabilitation training equipment cluster scheduling system, please refer to the embodiments of the rehabilitation training equipment cluster scheduling method described above.

[0142] like Figure 5 As shown, the rehabilitation training equipment cluster scheduling system includes: The data acquisition module is used to acquire rehabilitation task information for each patient in the rehabilitation training area, acquire the spatial location information of each patient in the rehabilitation training area in real time, and acquire rehabilitation training equipment information for each rehabilitation training device. The scheduling scheme generation module is used to generate a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area based on rehabilitation training equipment information, each patient's digital rehabilitation prescription and spatial location information, through multi-objective optimization calculation. The control command sequence generation module is used to generate control command sequences for corresponding rehabilitation training equipment based on personalized scheduling schemes. The instruction execution module is used to execute control instruction sequences, and performs the following operations synchronously during execution: Predict and resolve potential conflicts based on the spatial location information of each patient and each rehabilitation training device; The system acquires the patient's motion and physiological data corresponding to the rehabilitation task instructions in real time, and dynamically adjusts the execution parameters of the current rehabilitation task instructions based on the motion and physiological data.

[0143] The rehabilitation training equipment cluster scheduling system in this embodiment is used to implement the rehabilitation training equipment cluster scheduling method.

[0144] This application also provides an electronic device for implementing the various embodiments of this application. Figure 6 To illustrate the hardware structure of an electronic device according to various embodiments of this application, as shown in the following diagram... Figure 6 As shown, the electronic device includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor.

[0145] Those skilled in the art will understand that the electronic device structure involved in the embodiments of this application does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0146] In embodiments of this application, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0147] In this application embodiment, the processor can be implemented using at least one of an Application-Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a processor, a controller, a microcontroller, a microprocessor, or an electronic unit designed to perform the functions described herein. In some cases, such implementations can be implemented within a controller. For software implementations, implementations such as processes or functions can be implemented with separate software modules that allow the performance of at least one function or operation. The software code can be implemented by a software application (or program) written in any suitable programming language, and the software code can be stored in memory and executed by the controller.

[0148] In addition, the electronic device includes some functional modules not shown, which will not be described in detail here.

[0149] Those skilled in the art will understand that the various aspects of the electronic device provided in this application can be implemented as a system, method, or program product. Therefore, the various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "system."

[0150] This application also provides a storage medium storing a program product capable of implementing a method for cluster scheduling of rehabilitation training equipment. In some possible implementations, various aspects of this application can also be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this application.

[0151] The storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0152] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for cluster scheduling of rehabilitation training equipment, characterized in that, include: S1. Obtain rehabilitation task information for each patient in the rehabilitation training area, including patient identity information and digital rehabilitation prescription. The rehabilitation training area is equipped with a cluster of rehabilitation training equipment, which includes several rehabilitation training devices, including active training robots, autonomous mobile robots and fixed rehabilitation training devices. S2. Real-time acquisition of spatial location information of each patient in the rehabilitation training area and rehabilitation training equipment information of each rehabilitation training device; Information on rehabilitation training equipment includes the status and spatial location information of the equipment. S3. Based on the information of rehabilitation training equipment, the digital rehabilitation prescription and spatial location information of each patient, a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area is generated through multi-objective optimization calculation. The optimization objectives of the multi-objective optimization calculation include the shortest total rehabilitation training time, the highest equipment utilization rate, the least patient waiting time, and the lowest energy consumption. S4. Based on the personalized scheduling scheme, generate the control instruction sequence of the corresponding rehabilitation training equipment. The control instructions in the control instruction sequence include the movement guidance instruction for controlling the autonomous mobile robot to move and guiding the patient to the target location through human-computer interaction, and the rehabilitation task instruction for controlling the active training robot or fixed rehabilitation training to perform specific rehabilitation training actions. S5. Execute the control instruction sequence, and perform the following operations synchronously during the execution: Predict and resolve potential conflicts based on the spatial location information of each patient and each rehabilitation training device; The system acquires the patient's motion and physiological data corresponding to the rehabilitation task instructions in real time, and dynamically adjusts the execution parameters of the current rehabilitation task instructions based on the motion and physiological data.

2. The rehabilitation training equipment cluster scheduling method as described in claim 1, characterized in that, In step S3, based on the rehabilitation training equipment information, each patient's digital rehabilitation prescription, and spatial location information, a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area is generated through multi-objective optimization calculations. This specifically includes: S301. Decompose each patient's digital rehabilitation prescription into an atomic task sequence. The atomic task sequence contains a series of atomic tasks. Each atomic task is associated with a rehabilitation training item, the type of rehabilitation training equipment corresponding to the rehabilitation training item, and training parameters. S302. Construct a constrained optimization problem with the following optimization objectives: Minimize the total execution time of all atomic task sequences; Maximize the overall utilization rate of all rehabilitation training equipment; Minimize the overall wait time for all patients; The physiological rationality constraint must be met, meaning that there must be at least one low-intensity atomic task or rest period between two high-intensity atomic tasks for any patient; S303. Using semantic map information containing multiple semantic regions with different access attributes, real-time status and spatial location information of all rehabilitation training equipment, real-time spatial location information of all patients, and all atomic task sequences as input, an optimization algorithm based on multi-agent reinforcement learning is used to solve the constrained optimization problem and generate a personalized scheduling scheme for each patient.

3. The rehabilitation training equipment cluster scheduling method as described in claim 2, characterized in that, In step S303, the semantic map information includes the location coordinates of each semantic region in the rehabilitation training area. The semantic regions include the high-precision training area, the transition channel area, and the rest area.

4. The rehabilitation training equipment cluster scheduling method as described in claim 1, characterized in that, In step S303, the optimization algorithm based on multi-agent reinforcement learning runs iteratively, and in each iteration decision, the joint state space at the current moment is obtained. And according to the state space Calculate rewards To optimize decision-making.

5. The rehabilitation training equipment cluster scheduling method as described in claim 1, characterized in that, In step S303, the optimization algorithm based on multi-agent reinforcement learning performs prospective conflict avoidance when solving constrained optimization problems, specifically including: Based on the spatial location and velocity information of all rehabilitation training equipment and patients, the movement trajectories of all rehabilitation training equipment and patients in the future are predicted under the candidate scheduling scheme. Generate a spatiotemporal cube for each predicted trajectory and detect whether there is an intersection between any two spatiotemporal cubes; If an intersection is detected, the risk coefficient of the potential conflict is calculated. The calculation formula is: in, For the estimated collision time; The basic risk value is preset based on the type of conflict, which includes human-robot conflict and robot-robot conflict. The regional risk coefficient is set based on the region type in the semantic map where the conflict occurs; The preset collision time weighting coefficient; These are preset collision type weighting coefficients; These are the preset collision region weighting coefficients; When calculating the reward at each iteration of the optimization algorithm, the sum of the risk coefficients of all detected potential conflicts is used. As a penalty, the reward value is deducted, thereby guiding the optimization algorithm to generate a personalized scheduling scheme with lower conflict risk.

6. The rehabilitation training equipment cluster scheduling method as described in claim 1, characterized in that, In step S4, a control instruction sequence is generated based on the personalized scheduling scheme, specifically including: S401. Compile the device allocation, start time, spatial path, and device parameters of each atomic task in the personalized scheduling scheme into control instructions that can be executed by the corresponding rehabilitation training device; S402. Based on the logical dependencies and spatiotemporal relationships between control instructions, arrange the compiled control instructions according to the execution sequence to form a control instruction sequence; The movement guidance instructions include the target location coordinates and the expected arrival time, while the rehabilitation task instructions include the specific training movement pattern, resistance parameters, movement trajectory, and duration.

7. The rehabilitation training equipment cluster scheduling method as described in claim 1, characterized in that, In step S5, potential conflicts are predicted and resolved based on spatial location information, specifically including: S501: Based on the real-time acquisition of the position and speed of each patient and each rehabilitation training device, predict their movement trajectory within a specified time window in the future. S502: Calculate the estimated collision time between any rehabilitation training device and any patient or other rehabilitation training device (excluding itself). If the estimated collision time is less than the real-time collision avoidance threshold... If so, then a potential conflict is determined to exist; S503: Based on the type of conflicting parties, select a resolution strategy from the strategy library, generate real-time adjustment instructions, and send them to the corresponding rehabilitation training equipment for execution; The strategy library should include at least speed adjustment, local path replanning, and task pause strategies.

8. A cluster scheduling system for rehabilitation training equipment, characterized in that, The method for implementing the cluster scheduling of rehabilitation training equipment as described in any one of claims 1-7 includes: The data acquisition module is used to acquire rehabilitation task information for each patient in the rehabilitation training area, acquire the spatial location information of each patient in the rehabilitation training area in real time, and acquire rehabilitation training equipment information for each rehabilitation training device. The scheduling scheme generation module is used to generate a personalized scheduling scheme for allocating and scheduling rehabilitation training equipment for each patient in the rehabilitation training area based on rehabilitation training equipment information, each patient's digital rehabilitation prescription and spatial location information, through multi-objective optimization calculation. The control command sequence generation module is used to generate control command sequences for corresponding rehabilitation training equipment based on personalized scheduling schemes. The instruction execution module is used to execute control instruction sequences, and performs the following operations synchronously during execution: Predict and resolve potential conflicts based on the spatial location information of each patient and each rehabilitation training device; The system acquires the patient's motion and physiological data corresponding to the rehabilitation task instructions in real time, and dynamically adjusts the execution parameters of the current rehabilitation task instructions based on the motion and physiological data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor is used to execute a computer program, it implements the steps of the rehabilitation training equipment cluster scheduling method as described in any one of claims 1-7.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the rehabilitation training equipment cluster scheduling method as described in any one of claims 1-7.

Images