Intelligent service linkage scheduling method of pension community based on digital twinborn model

Abstract

The invention discloses a pension community intelligent service linkage scheduling method based on a digital twinborn model, and the method comprises the following steps: S1, carrying out the collection and preprocessing of multi-source perception data, and generating a standard data set; s2, constructing a pension community digital twinborn model, and performing digital modeling and virtual-real mapping; s3, performing data statistics and scene analysis, and generating a service request simulation data set; s4, performing global optimization scheduling based on an improved dung beetle optimization algorithm to generate candidate schemes; s5, performing virtual simulation on the digital twinborn model, screening an optimal scheme, and performing actual execution; and S6, real-time acquisition and feedback of operation data, closed-loop optimization and model updating are carried out. Intelligent, real-time and efficient cooperation of service scheduling of the pension community can be realized, and the resource utilization rate and the service response quality are comprehensively improved.

Description

Technical Field

[0001] The present invention relates to the field of smart elderly care technology, and in particular to a method for intelligent service linkage scheduling in elderly care communities based on a digital twin model. Background Art

[0002] With the accelerating aging of the population, retirement communities, as a crucial component of the social elderly care service system, meet the diverse needs of seniors, including daily care, health management, emergency response, and lifestyle services. In recent years, smart elderly care technologies have rapidly developed, with emerging technologies such as the Internet of Things, big data, and artificial intelligence gradually being applied to retirement community management. By deploying sensors and smart devices within the community, real-time awareness of seniors' health parameters, behavioral activities, environmental conditions, and service requests is achieved, providing the foundation for improving service efficiency and management. Traditional retirement community management systems rely heavily on manual scheduling and rule-based management, making it difficult to efficiently allocate and dynamically optimize complex service resources. Faced with challenges such as the diverse individual needs of seniors, dynamic service demands, and frequent emergencies, existing technologies struggle to fully meet the demands for multi-objective, personalized, and real-time coordinated services.

[0003] Although existing intelligent management technologies for retirement communities have initially achieved multi-source sensor data collection and intelligent scheduling of some services, they still have significant deficiencies in digital modeling, service process collaboration, and dynamic resource optimization. First, current digital modeling mostly remains at the basic static mapping of community physical space, personnel, and equipment, lacking comprehensive modeling and expression of multi-level service processes, dynamic attributes of entity objects, and behavioral evolution processes. Second, existing scheduling optimization methods mostly use single-objective or static weight mechanisms, which make it difficult to achieve adaptive trade-offs when faced with multiple objective requirements such as service response speed, resource utilization, task completion, and service balance. This results in scheduling results being easily affected by local constraints, resulting in insufficient overall efficiency and satisfaction. In addition, in special scenarios such as emergencies, resource conflicts, and high loads, traditional optimization algorithms often converge slowly, fall into local optimality, and lack global dynamic tuning capabilities.

[0004] Therefore, how to provide a smart service linkage scheduling method for elderly care communities based on digital twin models is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0005] One purpose of the present invention is to propose a method for intelligent service linkage scheduling in elderly care communities based on a digital twin model. This method fully utilizes advanced technologies such as multi-source perception, hierarchical nested digital twin modeling, intelligent behavior modeling, and swarm intelligence optimization. It describes in detail the entire process, from multidimensional perception data preprocessing, construction of a multi-level digital twin model for elderly care communities, service demand simulation analysis, to multi-objective service scheduling driven by an improved dung beetle optimization algorithm. This method achieves efficient global optimization of service dispatch, resource matching, personnel routing, and equipment scheduling. This method boasts advantages such as a high degree of intelligent service scheduling, high model expression sophistication, high resource utilization, good real-time response, and strong adaptive robustness in complex scenarios. It can significantly improve the overall quality and management efficiency of elderly care community service linkage.

[0006] According to an embodiment of the present invention, a method for intelligent service linkage scheduling in a retirement community based on a digital twin model includes the following steps:

[0007] S1. Collect multi-source perception data of the elderly care community through IoT sensing devices, pre-process the multi-source perception data of the elderly care community, and generate a standard data set;

[0008] S2. Based on the standard data set, a digital twin model of the elderly care community is constructed. The physical entities and service processes of the community are digitally modeled, and a real-time correspondence between the physical entities and the virtual twins is established.

[0009] S3. Based on the digital twin model of the elderly care community, conduct data statistics and scenario analysis on the elderly’s service needs, service processes, and emergencies, generate service request sequences and scheduling scenario simulation samples, and form a service request simulation dataset;

[0010] S4. Based on the service request simulation dataset, the dung beetle optimization algorithm is used to globally optimize service dispatch, resource matching, personnel path, and equipment scheduling, generating a candidate scheduling solution dataset. The candidate scheduling solution dataset is evaluated and evolved according to the fitness function.

[0011] S5. Use the digital twin model of the retirement community to conduct virtual simulation of the candidate scheduling plan data set to obtain scheduling effects and service execution results, select the optimal candidate scheduling plan, and issue it to the retirement community for actual implementation;

[0012] S6. Real-time collection of operational data during the actual scheduling execution process is performed and fed back to the digital twin model dataset to achieve continuous updating and optimization of the twin state, forming a closed-loop linkage between the physical and virtual worlds.

[0013] Optionally, the multi-source perception data of the retirement community specifically includes health parameters of the elderly, service requests, personnel location information, service equipment status and community environment data.

[0014] Optionally, preprocessing the multi-source perception data of the retirement community specifically includes denoising, standardization, missing value filling and format unification of the multi-source perception data of the retirement community.

[0015] Optionally, the S2 specifically includes:

[0016] S21. Classify and organize the physical space, service objects, service personnel, service equipment, and service process information of the retirement community in the standard dataset, and mark them as physical layer, logical layer, and behavioral layer input datasets respectively;

[0017] S22. At the physical layer, a 3D modeling method is used to digitally represent the buildings, rooms, and hardware facilities in the retirement community, and to construct a physical node set V phy , each node space coordinate is p i =(x i ,y i ,z i ) and append the entity type tag, where x i is the X-axis coordinate, y i is the Y-axis coordinate, z i is the Z-axis coordinate;

[0018] S23. At the logic layer, based on the service process information, define the process node set S and the directed connection relationship R. The service process is modeled with a directed graph P = (S, R). All standard processes are modeled with module objects F. j Form packaging, ready for pluggable operation;

[0019] S24. At the behavior layer, a behavior node set B is constructed for all service objects, including the elderly, caregivers, and equipment, and a multi-channel attribute tensor X is established for each behavior node. i , where each channel corresponds to health parameters, location information, status labels, and interaction history;

[0020] S25. Establish a virtual-real mapping function φ, link the physical layer nodes, logical layer process nodes, and behavioral layer subject nodes through mapping relationships. The mapping from physical entities to virtual twins is φ:V phy →V virt , where V phy is the physical layer node set of the retirement community, V virt A collection of virtual twin nodes for the retirement community;

[0021] S26, standard service process unit F j Register in the logic layer in the form of modular objects to form a pluggable process unit pool, and configure the logic layer process management engine to support dynamic loading, replacement and reconstruction of service process units. The process execution sequence is recorded as

[0022] S27. Embed a lifecycle state set L = {create, activate, run, sleep, deregister} for each twin entity and define a lifecycle state transition function in is the state of entity i at time t, For transfer conditions;

[0023] S28. Real-time monitoring of multi-channel attribute changes of behavior nodes, detection of lifecycle state transition events, and automatic triggering of dynamic adjustment of logic layer service processes and process unit switching based on state transition results;

[0024] S29. Integrate all nodes, attributes, processes, and mapping relationships of the physical layer, logical layer, and behavioral layer into a unified data structure;

[0025] S210. Finally, a digital twin model of the retirement community is formed that is hierarchically nested, multi-channel in attributes, pluggable in processes, and lifecycle-driven.

[0026] Optionally, the S3 specifically includes:

[0027] S31. Obtain the behavior layer node set B in the digital twin model of the retirement community, and calculate the multi-channel attribute tensor X of each behavior node. i Conduct time series collection to form a time series attribute data set;

[0028] S32, the behavior layer node set B and the physical layer node set V phy The multi-channel time series attribute data is processed by using an adaptive weighted dynamic aggregation method, and the weight coefficient α of each attribute channel is set. j , integrating health parameters, location information, status tags and interaction history to build multi-dimensional service demand statistics

[0029]

[0030] in, is the value of attribute channel j at time s, α j is the adaptive weight of attribute channel j, d is the number of attribute channels, and w is the sliding window length;

[0031] S33. Based on the process node set S and directed connection relationship R of the logical layer service, the occurrence frequency of various service processes is counted to form the service process statistical matrix F ij ;

[0032] S34. Mark the emergencies in the digital twin model dataset of the retirement community and establish the emergency event set E event , record the behavior node, physical location, state parameters and timestamp when each event occurs;

[0033] S35. Using a scenario analysis algorithm, perform cluster analysis and scenario division on the time series attribute data, service demand statistics, service process statistical matrix, and emergency event data, and map common service scenarios and emergency situations in the elderly care community into a scenario category set C;

[0034] S36, combine the multi-channel attributes of behavior nodes and service process nodes to j The typical service request, resource distribution, process status and event characteristics under the scenario are encoded to generate the scenario feature vector z j ;

[0035] S37, based on the scene category set C and the scene feature vector z j , construct service request sequence RQ={rq1,rq2,...,rq h}, where rq i Represents a service request, including the request body, request content, time, location and priority information, where h is the total number of service requests;

[0036] S38. For each service request sequence RQ and scenario category C, simulate different scheduling response paths, record the process node activation sequence and service execution status under each path, and form a scheduling scenario simulation sample data set S sim ;

[0037] S39, compare the service request sequence RQ with the scheduling scenario simulation sample data set S sim Integration, output is service request simulation dataset S RQsim ={(rq i ,sim j )}.

[0038] Optionally, the S4 specifically includes:

[0039] S41, the service request simulation data set S RQsim As optimization input, service assignment, resource matching, personnel path and equipment scheduling are set as decision variables, and the dung beetle population is initialized with an individual size of N;

[0040] S42, divide the dung beetle population into several clusters, and each cluster independently initializes the decision variable set (A g ,M g ,P g ,E g ), where g is the cluster number, A is the decision variable vector of the service dispatching solution, M is the decision variable vector of the resource matching solution, P is the personnel path set, and E is the decision variable vector of the equipment scheduling solution;

[0041] S43. Within each cluster, based on the current service request simulation dataset, dung beetles adopt a scenario-aware behavior selection strategy and dynamically adjust the probabilities of autonomous search, excellent solution tracking, and local perturbation according to the current community scenario. The specific behavior selection probability is denoted as λ. search ,λ track ,λ disturb ,The behavior selection probability is updated in real time based on the scenario type and ,historical convergence;

[0042] S44. For all individuals in the cluster, perform corresponding transport, tracking or disturbance update operations according to the behavior selection results, and the individual position is updated to

[0043]

[0044] in, is the position of the individual in generation t, is the autonomous search vector, is the pheromone synergy vector, η and γ are the adjustment coefficients;

[0045] S45, introduce a multi-objective adaptive dynamic weight mechanism, and the fitness function is defined as f:

[0046]

[0047] in, is the dynamic weight of the tth generation, T resp is the average time from the initiation of service request to the completion of service response, U res The actual utilization level of retirement community service resources, R fin is the actual completion ratio of all service tasks in the scheduling process, B eq The degree of balance in resource allocation among different service objects and tasks;

[0048] S46. Periodically conduct cross-cluster pheromone communication, so that all clusters share the current optimal solution and excellent experience, broadcast the position and fitness of the global optimal individual to other clusters, and guide the individuals between clusters to converge to the global optimal solution;

[0049] S47. Calculate the fitness value of all individuals, retain excellent solutions and eliminate low-quality solutions in each cluster based on the fitness value, so as to maintain cluster vitality and population diversity;

[0050] S48, determine whether the termination condition is met, if the maximum number of iterations or the fitness convergence threshold is reached, stop the iteration, otherwise go to S43 to continue behavior selection, individual update and pheromone exchange;

[0051] S49, output the decision variable set corresponding to the high fitness individuals obtained by all clustering iterations as the candidate scheduling solution data set S cand ;

[0052] S410, candidate scheduling solution dataset S cand Conduct multiple rounds of evaluation and screening, and ultimately output a scheduling plan with the best comprehensive indicators.

[0053] Optionally, the S5 specifically includes:

[0054] S51. Import the candidate scheduling solution dataset into the digital twin model of the retirement community. Configure the corresponding service dispatch, resource matching, personnel path, and equipment scheduling parameters in the digital twin model for each candidate scheduling solution, and initialize the simulation environment to be consistent with the current state of the actual retirement community.

[0055] S52. Use the digital twin model of the elderly care community to virtually simulate each candidate scheduling plan. Drive the various entities, processes, and behavior nodes in the twin model according to the plan parameters, and dynamically reproduce the scheduling process of the plan in the entire community space, including service request response, personnel and equipment flow, dynamic resource allocation, and task completion.

[0056] S53. During the virtual simulation process, key performance data of each candidate scheduling solution is collected in real time, including service response time, resource utilization efficiency, task completion, and service balance, and abnormal events or bottleneck links are recorded;

[0057] S54. Perform a comprehensive comparison and multi-dimensional evaluation of the simulation results of all candidate scheduling schemes, and select the best performing scheduling scheme based on various performance indicators;

[0058] S55. Decompose and implement the optimal scheduling plan into various actual operational links of the retirement community, and implement service dispatch, resource allocation, personnel path and equipment scheduling in sequence according to the requirements of the simulation plan. During the execution process, continuously collect actual operation data, compare it with the twin model simulation results, and conduct dynamic feedback optimization.

[0059] The beneficial effects of the present invention are:

[0060] The present invention realizes the all-round digital and dynamic mapping of the multi-source perception data, physical space, service objects, service personnel, service equipment and service processes of the retirement community by introducing a hierarchical and nested digital twin model of the retirement community and an improved dung beetle optimization algorithm. Based on the digital twin model, the system can truly restore the real-time status and evolution process of each entity and process in the community, and support refined statistical analysis and multi-scenario simulation prediction of the elderly’s service needs, service processes and emergencies. By adopting the dung beetle optimization algorithm with a multi-objective adaptive dynamic weight mechanism, a scenario-aware behavior selection strategy, and a group clustering and cross-cluster pheromone communication mechanism, the global search capability and local response speed of service scheduling are improved, making service dispatch, resource matching, personnel path and equipment scheduling more intelligent and personalized.

[0061] The present invention adaptively adjusts the scheduling target weights according to the actual scenario requirements to achieve a dynamic balance of service response speed, resource utilization, task completion and service balance. In the face of complex working conditions such as resource conflicts, high loads and emergencies, the real-time simulation of the digital twin model and the diversity of the optimization algorithm are coordinated to quickly provide the optimal or near-optimal scheduling plan, and the scheduling effect is continuously optimized through closed-loop feedback. Compared with the existing technology, the present invention has achieved significant improvements in model expression ability, scheduling intelligence level, system robustness and the ability to adapt to complex dynamic scenarios. It effectively overcomes the shortcomings of traditional methods in refined modeling, intelligent scheduling and global dynamic optimization, and improves the overall quality and operational efficiency of elderly care community services. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0063] Figure 1 This is a flow chart of the intelligent service linkage scheduling method for elderly care communities based on the digital twin model proposed in the present invention;

[0064] Figure 2 This is an algorithm flow chart for multi-objective scheduling optimization using the dung beetle optimization algorithm for the intelligent service linkage scheduling method for the elderly care community based on the digital twin model proposed in the present invention. DETAILED DESCRIPTION

[0065] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0066] refer to Figure 1 and Figure 2,The intelligent service linkage scheduling method for elderly care communities based on the digital twin model includes the following steps:

[0067] S1. Collect multi-source perception data of the elderly care community through IoT sensing devices, pre-process the multi-source perception data of the elderly care community, and generate a standard data set;

[0068] S2. Based on the standard data set, a digital twin model of the elderly care community is constructed. The physical entities and service processes of the community are digitally modeled, and a real-time correspondence between the physical entities and the virtual twins is established.

[0069] S3. Based on the digital twin model of the elderly care community, conduct data statistics and scenario analysis on the elderly’s service needs, service processes, and emergencies, generate service request sequences and scheduling scenario simulation samples, and form a service request simulation dataset;

[0070] S4. Based on the service request simulation dataset, the dung beetle optimization algorithm is used to globally optimize service dispatch, resource matching, personnel path, and equipment scheduling, generating a candidate scheduling solution dataset. The candidate scheduling solution dataset is evaluated and evolved according to the fitness function.

[0071] S5. Use the digital twin model of the retirement community to conduct virtual simulation of the candidate scheduling plan data set to obtain scheduling effects and service execution results, select the optimal candidate scheduling plan, and issue it to the retirement community for actual implementation;

[0072] S6. Real-time collection of operational data during the actual scheduling execution process is performed and fed back to the digital twin model dataset to achieve continuous updating and optimization of the twin state, forming a closed-loop linkage between the physical and virtual worlds.

[0073] In this embodiment, the multi-source perception data of the elderly care community specifically includes health parameters of the elderly, service requests, personnel location information, service equipment status and community environment data.

[0074] In this embodiment, preprocessing the multi-source perception data of the retirement community specifically includes denoising, standardizing, filling missing values ​​and unifying the format of the multi-source perception data of the retirement community.

[0075] In this embodiment, S2 specifically includes:

[0076] S21. Classify and organize the physical space, service objects, service personnel, service equipment, and service process information of the retirement community in the standard dataset, and mark them as physical layer, logical layer, and behavioral layer input datasets respectively;

[0077] S22. At the physical layer, a 3D modeling method is used to digitally represent the buildings, rooms, and hardware facilities in the retirement community, and to construct a physical node set V phy, each node space coordinate is p i =(x i ,y i ,z i ) and append the entity type tag, where x i is the X-axis coordinate, y i is the Y-axis coordinate, z i is the Z-axis coordinate;

[0078] S23. At the logic layer, based on the service process information, define the process node set S and the directed connection relationship R. The service process is modeled with a directed graph P = (S, R). All standard processes are modeled with module objects F. j Form packaging, ready for pluggable operation;

[0079] S24. At the behavior layer, a behavior node set B is constructed for all service objects, including the elderly, caregivers, and equipment, and a multi-channel attribute tensor X is established for each behavior node. i , where each channel corresponds to health parameters, location information, status labels, and interaction history;

[0080] S25. Establish a virtual-real mapping function φ, link the physical layer nodes, logical layer process nodes, and behavioral layer subject nodes through mapping relationships. The mapping from physical entities to virtual twins is φ:V phy →V virt , where V phy is the physical layer node set of the retirement community, V virt A collection of virtual twin nodes for the retirement community;

[0081] S26, standard service process unit F j Register in the logic layer in the form of modular objects to form a pluggable process unit pool, and configure the logic layer process management engine to support dynamic loading, replacement and reconstruction of service process units. The process execution sequence is recorded as

[0082] S27. Embed a lifecycle state set L = {create, activate, run, sleep, deregister} for each twin entity and define a lifecycle state transition function in is the state of entity i at time t, For transfer conditions;

[0083] S28. Real-time monitoring of multi-channel attribute changes of behavior nodes, detection of lifecycle state transition events, and automatic triggering of dynamic adjustment of logic layer service processes and process unit switching based on state transition results;

[0084] S29. Integrate all nodes, attributes, processes, and mapping relationships of the physical layer, logical layer, and behavioral layer into a unified data structure;

[0085] S210. Finally, a digital twin model of the retirement community is formed that is hierarchically nested, multi-channel in attributes, pluggable in processes, and lifecycle-driven.

[0086] In this embodiment, S3 specifically includes:

[0087] S31. Obtain the behavior layer node set B in the digital twin model of the retirement community, and calculate the multi-channel attribute tensor X of each behavior node. i Conduct time series collection to form a time series attribute data set;

[0088] S32, the behavior layer node set B and the physical layer node set V phy The multi-channel time series attribute data is processed by using an adaptive weighted dynamic aggregation method, and the weight coefficient α of each attribute channel is set. j , integrating health parameters, location information, status tags and interaction history to build multi-dimensional service demand statistics

[0089]

[0090] in, is the value of attribute channel j at time s, α j is the adaptive weight of attribute channel j, d is the number of attribute channels, and w is the sliding window length;

[0091] S33. Based on the process node set S and directed connection relationship R of the logical layer service, the occurrence frequency of various service processes is counted to form the service process statistical matrix F ij ;

[0092] S34. Mark the emergencies in the digital twin model dataset of the retirement community and establish the emergency event set E event , record the behavior node, physical location, state parameters and timestamp when each event occurs;

[0093] S35. Using a scenario analysis algorithm, perform cluster analysis and scenario division on the time series attribute data, service demand statistics, service process statistical matrix, and emergency event data, and map common service scenarios and emergency situations in the elderly care community into a scenario category set C;

[0094] S36, combine the multi-channel attributes of behavior nodes and service process nodes to j The typical service request, resource distribution, process status and event characteristics under the scenario are encoded to generate the scenario feature vector z j ;

[0095] S37, based on the scene category set C and the scene feature vector z j, construct service request sequence RQ={rq1,rq2,...,rq h}, where rq i Represents a service request, including the request body, request content, time, location and priority information, where h is the total number of service requests;

[0096] S38. For each service request sequence RQ and scenario category C, simulate different scheduling response paths, record the process node activation sequence and service execution status under each path, and form a scheduling scenario simulation sample data set S sim ;

[0097] S39, compare the service request sequence RQ with the scheduling scenario simulation sample data set S sim Integration, output is service request simulation dataset S RQsim ={(rq i ,sim j )}.

[0098] In this embodiment, the S4 specifically includes:

[0099] S41, the service request simulation data set S RQsim As optimization input, service assignment, resource matching, personnel path and equipment scheduling are set as decision variables, and the dung beetle population is initialized with an individual size of N;

[0100] S42, divide the dung beetle population into several clusters, and each cluster independently initializes the decision variable set (A g ,M g ,P g ,E g ), where g is the cluster number, A is the decision variable vector of the service dispatching solution, M is the decision variable vector of the resource matching solution, P is the personnel path set, and E is the decision variable vector of the equipment scheduling solution;

[0101] S43. Within each cluster, based on the current service request simulation dataset, dung beetles adopt a scenario-aware behavior selection strategy and dynamically adjust the probabilities of autonomous search, excellent solution tracking, and local perturbation according to the current community scenario. The specific behavior selection probability is denoted as λ. search ,λ track ,λ disturb ,The behavior selection probability is updated in real time based on the scenario type and ,historical convergence;

[0102] S44. For all individuals in the cluster, perform corresponding transport, tracking or disturbance update operations according to the behavior selection results, and the individual position is updated to

[0103]

[0104] in, is the position of the individual in generation t, is the autonomous search vector, is the pheromone synergy vector, η and γ are the adjustment coefficients;

[0105] S45, introduce a multi-objective adaptive dynamic weight mechanism, and the fitness function is defined as f:

[0106]

[0107] in, is the dynamic weight of the tth generation, T resp is the average time from the initiation of service request to the completion of service response, U res The actual utilization level of retirement community service resources, R fin is the actual completion ratio of all service tasks in the scheduling process, B eq The degree of balance in resource allocation among different service objects and tasks;

[0108] S46. Periodically conduct cross-cluster pheromone communication, so that all clusters share the current optimal solution and excellent experience, broadcast the position and fitness of the global optimal individual to other clusters, and guide the individuals between clusters to converge to the global optimal solution;

[0109] S47. Calculate the fitness value of all individuals, retain excellent solutions and eliminate low-quality solutions in each cluster based on the fitness value, so as to maintain cluster vitality and population diversity;

[0110] S48, determine whether the termination condition is met, if the maximum number of iterations or the fitness convergence threshold is reached, stop the iteration, otherwise go to S43 to continue behavior selection, individual update and pheromone exchange;

[0111] S49, output the decision variable set corresponding to the high fitness individuals obtained by all clustering iterations as the candidate scheduling solution data set S cand ;

[0112] S410, candidate scheduling solution dataset S cand Conduct multiple rounds of evaluation and screening, and ultimately output a scheduling plan with the best comprehensive indicators.

[0113] In this embodiment, the S5 specifically includes:

[0114] S51. Import the candidate scheduling solution dataset into the digital twin model of the retirement community. Configure the corresponding service dispatch, resource matching, personnel path, and equipment scheduling parameters in the digital twin model for each candidate scheduling solution, and initialize the simulation environment to be consistent with the current state of the actual retirement community.

[0115] S52. Use the digital twin model of the elderly care community to virtually simulate each candidate scheduling plan. Drive the various entities, processes, and behavior nodes in the twin model according to the plan parameters, and dynamically reproduce the scheduling process of the plan in the entire community space, including service request response, personnel and equipment flow, dynamic resource allocation, and task completion.

[0116] S53. During the virtual simulation process, key performance data of each candidate scheduling solution is collected in real time, including service response time, resource utilization efficiency, task completion, and service balance, and abnormal events or bottleneck links are recorded;

[0117] S54. Perform a comprehensive comparison and multi-dimensional evaluation of the simulation results of all candidate scheduling schemes, and select the best performing scheduling scheme based on various performance indicators;

[0118] S55. Decompose and implement the optimal scheduling plan into various actual operational links of the retirement community, and implement service dispatch, resource allocation, personnel path and equipment scheduling in sequence according to the requirements of the simulation plan. During the execution process, continuously collect actual operation data, compare it with the twin model simulation results, and conduct dynamic feedback optimization.

[0119] Example 1:

[0120] To verify the feasibility of this invention, it was applied to a smart senior care community with a building area of ​​approximately 38,000 square meters. It currently houses 410 registered seniors and is staffed by 82 caregivers and service personnel. Key functional areas include living quarters, medical care stations, rehabilitation training areas, a dining and activity center, and an intelligent security system. The community has deployed 360 health monitoring devices, 14 mobile care robots, and 80 sets of environmental sensors. Previously, the community used traditional manual scheduling and zoning management to allocate service resources, making it difficult to efficiently coordinate multi-source demand and respond to emergencies. Data shows that from February to March 2024, the average daily service request from seniors was approximately 1,010. Peak hours for tasks such as health checks, food delivery, daily care, and security patrols were particularly congested, with some elderly residents often waiting for service for over 25 minutes. Some caregivers performed 37 services per day, while others only performed 12. For nighttime fall alerts or emergencies, the average response time from traditional scheduling was over 11 minutes, equipment utilization was only 62%, and resident service satisfaction hovered around 83%.

[0121] This paper introduces a digital twin-based intelligent service linkage scheduling method for senior care communities. The community fully integrates real-time IoT sensor data from buildings, rooms, equipment, and personnel. All collected data undergoes standardized preprocessing to drive the continuous dynamic updating of the senior care community's digital twin model. The model comprises a multi-layered structure consisting of physical space, service processes, and behavioral objects. This model not only maps the space and status of each entity but also reflects, in real time, changes in the health of the elderly, service requests, caregiver distribution, and equipment operation. For example, during the daily peak period of April 15, 2024, 213 service requests were received between 9:00 AM and 11:00 AM for services such as health checks, meal deliveries, and inspections. The system automatically generated eight scheduling plans for simulation and performance comparison. Using virtual simulation in a digital twin environment, the system evaluated each plan in real time based on metrics such as service response speed, personnel and equipment utilization, and task completion rate. Using an improved dung beetle optimization algorithm, the system adaptively adjusted target weights and clustering coordination strategies, screened the optimal scheduling results, and automatically pushed them to the community for execution.

[0122] In actual operation, caregivers and robots automatically assign personnel, adjust routes, and dispatch equipment based on the optimal solution. During peak hours that day, the average response time for all service requests dropped to 5.2 minutes, with the longest single wait time reduced from 28 minutes to 10 minutes. The utilization rate of care robots reached 92.4%, and the average service load per caregiver became more balanced, with the maximum to minimum workload ratio dropping from 2.5 to 1.3. In response to a sudden elderly person fall alarm at 10:08 PM that evening, the system quickly located the target using a digital twin model, dispatching the nearest night shift caregiver and emergency robot for coordinated response. The actual arrival time was 3.6 minutes, and the incident was fully resolved within 8 minutes, ensuring the safety of the elderly person. Monthly statistics show that in April 2024, there were 31,248 community service requests, with the average response time dropping to 4.8 minutes from 9.1 minutes the previous month. Equipment utilization increased to 91.7%, and satisfaction rose to 96.8%.

[0123] Table 1 Comparison of service performance of elderly care communities before and after digital twin intelligent scheduling

[0124]

[0125] The data in Table 1 demonstrates that the digital twin intelligent scheduling approach has significantly improved service management in senior care communities. The total number of service requests and the average daily number of service requests remained largely unchanged before and after optimization, reflecting the relatively stable demand for community services. However, after implementing intelligent scheduling optimization, the average response time was significantly reduced from 9.1 minutes using traditional manual scheduling to 4.8 minutes, nearly doubling service efficiency. After initiating a service request, seniors can now receive care and assistance more quickly, effectively alleviating the pressure of waiting during peak hours.

[0126] The maximum wait time for services has dropped from 28.3 minutes to 10.0 minutes, improving the extreme wait times previously experienced during peak hours or resource constraints. This not only enhances the service experience for seniors but also helps prevent health risks associated with service delays. Equipment utilization has increased significantly, from 62.0% to 91.7%. This demonstrates that optimized scheduling has enabled more efficient and balanced use of various devices within the community, reducing idle equipment and wasted resources.

[0127] Service satisfaction increased from 83.0% before the optimization to 96.8%, demonstrating high user recognition of overall service quality. The maximum to minimum workload ratio for caregivers decreased from 2.5 to 1.3, demonstrating a significant improvement in staffing balance. This reduced overwork and underwork for some caregivers, and strengthened team cohesion and work enthusiasm.

[0128] The average response time for emergencies dropped from 11.4 minutes to 3.8 minutes, improving the community's ability to respond quickly to emergencies and ensuring the life safety and health needs of the elderly. The data in Table 1 fully demonstrates that the digital twin and intelligent optimization scheduling solution have significant practical results in improving service response efficiency, resource utilization, emergency response capabilities and resident satisfaction, providing solid data support for the intelligent management and high-quality services of retirement communities.

[0129] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for intelligent service linkage scheduling in elderly care communities based on a digital twin model, characterized in that: The steps include: S1. Collect multi-source perception data of the elderly care community through IoT sensing devices, pre-process the multi-source perception data of the elderly care community, and generate a standard data set; S2. Based on the standard data set, a digital twin model of the elderly care community is constructed. The physical entities and service processes of the community are digitally modeled, and a real-time correspondence between the physical entities and the virtual twins is established. S3. Based on the digital twin model of the elderly care community, conduct data statistics and scenario analysis on the elderly’s service needs, service processes, and emergencies, generate service request sequences and scheduling scenario simulation samples, and form a service request simulation dataset; S4. Based on the service request simulation dataset, the dung beetle optimization algorithm is used to globally optimize service dispatch, resource matching, personnel path, and equipment scheduling, generating a candidate scheduling solution dataset. The candidate scheduling solution dataset is evaluated and evolved according to the fitness function. S5. Use the digital twin model of the retirement community to conduct virtual simulation of the candidate scheduling plan data set to obtain scheduling effects and service execution results, select the optimal candidate scheduling plan, and issue it to the retirement community for actual implementation; S6. Real-time collection of operational data during the actual scheduling execution process is performed and fed back to the digital twin model dataset to achieve continuous updating and optimization of the twin state, forming a closed-loop linkage between the physical and virtual worlds.

2. The method for intelligent service linkage scheduling in a retirement community based on a digital twin model according to claim 2 is characterized in that: The multi-source perception data of the retirement community specifically includes elderly health parameters, service requests, personnel location information, service equipment status and community environment data.

3. The method for intelligent service linkage scheduling in a retirement community based on a digital twin model according to claim 2 is characterized in that: The preprocessing of multi-source perception data of retirement communities specifically includes denoising, standardization, missing value filling and format unification of multi-source perception data of retirement communities.

4. The method for intelligent service linkage scheduling in a retirement community based on a digital twin model according to claim 1 is characterized in that: The S2 specifically includes: S21. Classify and organize the physical space, service objects, service personnel, service equipment, and service process information of the retirement community in the standard dataset, and mark them as physical layer, logical layer, and behavioral layer input datasets respectively; S22. At the physical layer, a 3D modeling method is used to digitally represent the buildings, rooms, and hardware facilities in the retirement community, and to construct a physical node set V phy , each node space coordinate is p i =(x i ,y i ,z i ) and append the entity type tag, where x i is the X-axis coordinate, y i is the Y-axis coordinate, z i is the Z-axis coordinate; S23. At the logic layer, based on the service process information, define the process node set S and the directed connection relationship R. The service process is modeled with a directed graph P = (S, R). All standard processes are modeled with module objects F. j Form packaging, ready for pluggable operation; S24. At the behavior layer, a behavior node set B is constructed for all service objects, including the elderly, caregivers, and equipment, and a multi-channel attribute tensor X is established for each behavior node. i , where each channel corresponds to health parameters, location information, status labels, and interaction history; S25. Establish a virtual-real mapping function φ, link the physical layer nodes, logical layer process nodes, and behavioral layer subject nodes through mapping relationships. The mapping from physical entities to virtual twins is φ:V phy →V virt , where V phy is the physical layer node set of the retirement community, V virt It is a collection of virtual twin nodes for the retirement community; S26, standard service process unit F j Register in the logic layer in the form of modular objects to form a pluggable process unit pool, and configure the logic layer process management engine to support dynamic loading, replacement and reconstruction of service process units. The process execution sequence is recorded as S27. Embed a lifecycle state set L = {create, activate, run, sleep, deregister} for each twin entity and define a lifecycle state transition function in is the state of entity i at time t, For transfer conditions; S28. Real-time monitoring of multi-channel attribute changes of behavior nodes, detection of lifecycle state transition events, and automatic triggering of dynamic adjustment of logic layer service processes and process unit switching based on state transition results; S29. Integrate all nodes, attributes, processes, and mapping relationships of the physical layer, logical layer, and behavioral layer into a unified data structure; S210. Finally, a digital twin model of the retirement community is formed that is hierarchically nested, multi-channel in attributes, pluggable in processes, and lifecycle-driven.

5. The method for intelligent service linkage scheduling in a retirement community based on a digital twin model according to claim 1 is characterized in that: The S3 specifically includes: S31. Obtain the behavior layer node set B in the digital twin model of the retirement community, and calculate the multi-channel attribute tensor X of each behavior node. i Conduct time series collection to form a time series attribute data set; S32, the behavior layer node set B and the physical layer node set V phy The multi-channel time series attribute data is processed by using an adaptive weighted dynamic aggregation method, and the weight coefficient α of each attribute channel is set. j , integrate health parameters, location information, status tags and interaction history to build multi-dimensional service demand statistics D i (t) ; S33. Based on the process node set S and directed connection relationship R of the logical layer service, the occurrence frequency of various service processes is counted to form the service process statistical matrix F ij ; S34. Mark the emergencies in the digital twin model dataset of the retirement community and establish the emergency event set E event , record the behavior node, physical location, state parameters and timestamp when each event occurs; S35. Using a scenario analysis algorithm, perform cluster analysis and scenario division on the time series attribute data, service demand statistics, service process statistical matrix, and emergency event data, and map common service scenarios and emergency situations in the elderly care community into a scenario category set C; S36, combine the multi-channel attributes of behavior nodes and service process nodes to j The typical service request, resource distribution, process status and event characteristics under the scenario are encoded to generate the scenario feature vector z j ; S37, based on the scene category set C and the scene feature vector z j , construct service request sequence RQ={rq1,rq2,...,rq h }, where rq i Represents a service request, including the request body, request content, time, location and priority information, where h is the total number of service requests; S38. For each service request sequence RQ and scenario category C, simulate different scheduling response paths, record the process node activation sequence and service execution status under each path, and form a scheduling scenario simulation sample data set S sim ; S39, compare the service request sequence RQ with the scheduling scenario simulation sample data set S sim Integration, output is service request simulation dataset S RQsim ={(rq i ,sim j )}.

6. The method for intelligent service linkage scheduling in a retirement community based on a digital twin model according to claim 1 is characterized in that: The S4 specifically includes: S41, the service request simulation data set S RQsim As optimization input, service assignment, resource matching, personnel path and equipment scheduling are set as decision variables, and the dung beetle population is initialized with an individual size of N; S42, divide the dung beetle population into several clusters, and each cluster independently initializes the decision variable set (A g ,M g ,P g ,E g ), where g is the cluster number, A is the decision variable vector of the service dispatching solution, M is the decision variable vector of the resource matching solution, P is the personnel path set, and E is the decision variable vector of the equipment scheduling solution; S43. Within each cluster, based on the current service request simulation dataset, dung beetles adopt a scenario-aware behavior selection strategy and dynamically adjust the probabilities of autonomous search, excellent solution tracking, and local perturbation according to the current community scenario. The specific behavior selection probability is denoted as λ. search ,λ track ,λ disturb ,The behavior selection probability is updated in real time based on the scenario type and ,historical convergence; S44. For all individuals in the cluster, perform corresponding transport, tracking or disturbance update operations according to the behavior selection results, and the individual position is updated to S45, introduce a multi-objective adaptive dynamic weight mechanism, and define the fitness function as f; S46. Periodically conduct cross-cluster pheromone communication, so that all clusters share the current optimal solution and excellent experience, broadcast the position and fitness of the global optimal individual to other clusters, and guide the individuals between clusters to converge to the global optimal solution; S47. Calculate the fitness value of all individuals, retain excellent solutions and eliminate low-quality solutions in each cluster based on the fitness value, so as to maintain cluster vitality and population diversity; S48, determine whether the termination condition is met, if the maximum number of iterations or the fitness convergence threshold is reached, stop the iteration, otherwise go to S43 to continue behavior selection, individual update and pheromone exchange; S49, output the decision variable set corresponding to the high fitness individuals obtained by all clustering iterations as the candidate scheduling solution data set S cand ; S410, candidate scheduling solution dataset S cand Conduct multiple rounds of evaluation and screening, and ultimately output a scheduling plan with the best comprehensive indicators.

7. The method for intelligent service linkage scheduling in a retirement community based on a digital twin model according to claim 1 is characterized in that: The S5 specifically includes: S51. Import the candidate scheduling solution dataset into the digital twin model of the retirement community. Configure the corresponding service dispatch, resource matching, personnel path, and equipment scheduling parameters in the digital twin model for each candidate scheduling solution, and initialize the simulation environment to be consistent with the current state of the actual retirement community. S52. Use the digital twin model of the elderly care community to virtually simulate each candidate scheduling plan. Drive the various entities, processes, and behavior nodes in the twin model according to the plan parameters, and dynamically reproduce the scheduling process of the plan in the entire community space, including service request response, personnel and equipment flow, dynamic resource allocation, and task completion. S53. During the virtual simulation process, key performance data of each candidate scheduling solution is collected in real time, including service response time, resource utilization efficiency, task completion, and service balance, and abnormal events or bottleneck links are recorded; S54. Perform a comprehensive comparison and multi-dimensional evaluation of the simulation results of all candidate scheduling schemes, and select the best performing scheduling scheme based on various performance indicators; S55. Decompose and implement the optimal scheduling plan into various actual operational links of the retirement community, and implement service dispatch, resource allocation, personnel path and equipment scheduling in sequence according to the requirements of the simulation plan. During the execution process, continuously collect actual operation data, compare it with the twin model simulation results, and conduct dynamic feedback optimization.

Images