Multi-agent-based method for diagnosis and feedback of community elderly care facility planning

Abstract

A multi-agent-based method for diagnosis and feedback of community elderly care facility planning. The method comprises five main steps of constructing a spatiotemporal data integration platform of a target community, constructing an elderly agent model of the target community, constructing a diagnosis agent for a community elderly care facility planning scheme, diagnosing a layout scheme of elderly care facilities in the target community, and providing decision feedback on the layout scheme of the elderly care facilities in the target community. The method aims at using population digital profiling technology to accurately analyze differentiated spatiotemporal behaviors of the elderly in a community, mine patterns, and construct an elderly agent of a target community; and determining the composition of key indicators for diagnosing a community elderly care facility planning scheme, and constructing a diagnosis agent, such that a real physical environment of the community and an elderly care facility planning scheme can be combined to perform facility layout scheme diagnosis and optimized decision feedback. The method can realize layout diagnosis of an elderly care facility planning scheme for a target community and output optimization suggestions.

Description

A Multi-Agent-Based Method for Planning, Diagnosis, and Feedback of Community-Based Elderly Care Facilities Technical Field

[0001] This invention relates to a multi-agent-based method for the diagnosis and feedback of community elderly care facility planning, belonging to the field of intelligent planning and decision support technology, and particularly to a method for diagnosing and optimizing community elderly care facility planning schemes using a multi-agent system. Background Technology

[0002] With the increasing aging of the population, the planning and layout of community-based elderly care facilities has become a focus of social attention. Because the needs of the elderly are often expressed passively, and these needs are multidimensional and internally diverse, current elderly care facility planning faces problems of insufficient supply, lagging supply, and uneven distribution, resulting in a mismatch between supply and demand. Reasonable planning of elderly care facilities not only affects the improvement of the quality of life for the elderly but also directly impacts the effective allocation of community resources and the efficiency of public services.

[0003] In practical applications, common planning methods include: one is to allocate elderly care facilities proportionally based on the total population of a region; the other is to construct an evaluation index system to analyze the supply and demand relationship and coupling coordination of urban living service facilities to achieve optimal allocation. These traditional planning methods mostly rely on static indicators for demand analysis, lacking dynamic adjustments to facility distribution and failing to effectively identify and adapt to the personalized needs of different elderly groups. These problems are particularly prominent when facing the rapid changes in the aging process of communities. Therefore, there is an urgent need for an intelligent planning tool that can comprehensively consider multiple factors, provide real-time feedback, and flexibly adjust to address the challenges brought about by aging. Summary of the Invention

[0004] Purpose of the Invention: The purpose of this invention is to provide a solution that can dynamically adjust and optimize the layout of community-based elderly care facilities based on a multi-agent intelligent planning and decision feedback mechanism. By introducing multi-agent technology, and considering different community characteristics, the needs of the elderly, and real-time data feedback, a multi-dimensional, real-time responsive elderly care facility planning and decision support system is constructed to address the limitations and inadequacies of existing planning methods in addressing the planning of community-based elderly care facilities.

[0005] The technical solution adopted in this invention is: a multi-agent-based method for planning, diagnosis, and feedback of community elderly care facilities. The system includes:

[0006] Step S1: Construct a spatiotemporal data integration platform for the target community.

[0007] Acquire high-precision LBS data, land use data, POI data, road network data, and administrative division spatial data of the target community and perform data preprocessing; import the preprocessed data into the geographic information platform to construct a spatiotemporal data integration platform for the target community;

[0008] Step S2: Construction of intelligent agent model for elderly people in the target community

[0009] Based on the spatiotemporal data integration platform described in step S1, people aged 60 and above are selected as the elderly population in the target community by age field. The points where the elderly stay and the points of behavior in the community are identified, the activity behavior chain of the elderly in the target community is constructed, and the label type combination and the community elderly profile genealogy are constructed by combining their individual ontology dimension and spatiotemporal behavior dimension. Based on the above elderly profile genealogy results, the target community elderly intelligent agent is constructed.

[0010] Step S3: Diagnostic Analysis and Intelligent Construction of Community Elderly Care Facility Planning Scheme

[0011] The key indicators for diagnosing community elderly care facility planning schemes are determined. The integrity, utilization rate, accessibility, and fairness of the community elderly care facility layout plan are used as constraints. The thresholds of the above-mentioned diagnostic indicators for the layout of community elderly care facilities are determined by combining the current planning standards and the minimum standard method. A diagnostic intelligent agent agent model for the layout of community elderly care facilities is then constructed.

[0012] Step S4: Diagnosis of the current status of elderly care facility layout in the target community

[0013] Using the spatiotemporal data integration platform of the target community constructed in step S1, import the spatial environment data of the target community and the layout plan of the community elderly care facilities. Combine the target community elderly intelligent agent described in step S2 and the community elderly care facility layout diagnostic intelligent agent described in step S3 to construct and run a multi-agent model of the target community, and perform layout diagnosis through the diagnostic intelligent agent.

[0014] Step S5: Decision feedback on the layout plan of elderly care facilities in the target community

[0015] The operation and diagnostic results of the multi-agent model described in step S4 are optimized and evaluated to generate a set of optimization types for the target community elderly care facility planning scheme and corresponding optimization suggestions. Then, the layout scheme of the target community elderly care facility, the diagnostic report results and optimization suggestions are visualized and interactively displayed through urban three-dimensional digital holographic sand table equipment and virtual reality equipment, and output through printing equipment.

[0016] Furthermore, the preprocessing method for LBS data in step S1 is as follows:

[0017] Step S1-11: Stop Point Identification; Using a workstation equipped with a multi-core CPU and GPU, the LBS data is processed and analyzed. Trajectory points in the acquired LBS data that have not moved within 10 minutes or have moved within a 100-meter range are clustered, and their average center is identified as the stop point. Information about the stop point is generated (including user ID, age, latitude and longitude of the stop point, duration of stay, and start and end times of stay). The average center refers to a point returned for each stop, which is used to represent the average center of the stay in terms of time and distance. The generated elements will have time-type intervals.

[0018] Steps S1-12: Data cleaning and effective user screening; Effective users are those whose number of days with a stay point is greater than or equal to half of the total number of days in the observation data; The specific data cleaning and effective user screening method is as follows: Summarize the data table of all stay points within the selected study period, summarize the stay days of each user by ID, screen out effective users, and associate all original stay point data by effective user ID. The data processed in step S1-11 is called effective user stay data; Use a distributed storage system to store the original data and calculation results;

[0019] The spatiotemporal data integration platform in step S1 includes basic spatial data and spatiotemporal trajectory data. The basic spatial data is obtained from the Open Street Map website and specifically includes land use data, business type POI data, urban road network vector data, and administrative division spatial data. The user spatiotemporal trajectory data includes the valid user dwell data obtained in steps S1-12. The specific attribute descriptions of each type of data are shown in the table below:

[0020] Drones equipped with high-precision cameras and LiDAR (Light Detection and Ranging) are used to conduct aerial photography and 3D modeling of specific areas, acquiring high-resolution spatial data to verify and supplement land use and road network data in Open Street Map. Internet of Things (IoT) sensors are used to monitor pedestrian and vehicle flow dynamics in real time, supplementing user spatiotemporal trajectory data.

[0021] Furthermore, the specific process for constructing the community elderly profile genealogy in step S2 is as follows:

[0022] Step S2-11: Construction of the spatiotemporal behavior chain of the elderly; Based on the spatiotemporal data integration platform described in S1, spatiotemporal data of the elderly population aged 60 and above are extracted. The LBS data of the elderly in the target community is spatially connected with the 20M buffer area of ​​land use data and POI data to identify their main activity areas and activity types, specifically including residential areas, commercial areas, medical areas, cultural activity areas, and green spaces. Based on time series analysis, the spatiotemporal behavior chain of the elderly is constructed to describe their daily behavioral activity patterns.

[0023] Step S2-12: Label group combination; Based on the ontological dimensions (gender, age) and spatiotemporal behavioral dimensions (activity type, frequency, duration) of the elderly, the K-means clustering algorithm is used to divide the elderly into groups and generate label groups; The specific formula is as follows: K-means(X,k)

[0024] Where X is the feature vector of the elderly (including ontology dimension and spatiotemporal behavior dimension), k is the number of clusters, and the optimal value of k is determined by the elbow rule;

[0025] Step S2-13: Construction of Elderly Profile Lineage; Based on the label group results, an elderly profile lineage is constructed. The area where the elderly spend the most time during the day (7:00 AM to 7:00 PM) excluding their residence is used as the basis for profile classification. The decision tree algorithm is used to classify the elderly population and generate elderly profile labels, specifically including "Healthy and Active," "Home-based Dependent," and "Medical Needs-based" (among which, profiles whose activity location is only their residence are classified as "Home-based Dependent"). The comparison table of activity location and profile classification is shown below:

[0026] The decision tree algorithm used is the CART algorithm, and the specific formula is as follows: CART(D,A)

[0027] Where D is the dataset of elderly people's behavioral activities, and A is the set of feature attributes.

[0028] Furthermore, the specific process for constructing the intelligent agent for the elderly in the target community in step S2 is as follows:

[0029] Step S2-21: Initialize the elderly agent model; Based on the elderly profile genealogy constructed above, extract behavioral features for each type of elderly person, specifically including activity type, frequency, and duration as the initial attributes of the elderly agent; Use the Mesa framework in Python to create an agent instance and define the agent's attribute set Aa: Aa = {activity type, frequency, duration}

[0030] Step S2-22: Defining the behavioral rules of the elderly agent; Based on the behavioral patterns of the elderly, the behavioral rules of the agent are defined using a Markov Decision Process (MDP) to simulate the daily activity paths and usage preferences of elderly care facilities; The specific formula is as follows: MDP(S m A m ,P m ,R m )

[0031] Among them, S m Let A be the state space. m For the behavior space, P m Let R be the state transition probability matrix.m This is the reward function.

[0032] Furthermore, the key indicators for the diagnosis of the community elderly care facility layout scheme in step S3 are specifically as follows:

[0033] Furthermore, the specific construction process of the diagnostic agent proxy model in step S3 is as follows:

[0034] Step S3-21: Determination of the baseline values ​​for the diagnostic indicators of the intelligent agent; the baseline values ​​for the diagnostic indicators of the community elderly care facility layout scheme in step S3 are determined based on the current planning standards and the minimum standard method, specifically as follows:

[0035] Step S3-22: Construction of diagnostic agent indicators; Combining the baseline values ​​of the diagnostic agents described in Step S3-21, these are used as constraints for constructing the diagnostic agent for community elderly care facility planning. Community elderly care facility planning schemes that meet the baseline values ​​are defined as passing the diagnosis, while those that do not are considered failing. Specific diagnostic results are output, including the detailed calculation results of the aforementioned key indicators. The specific rule set for the diagnostic agent for community elderly care facility planning is as follows: Completeness rule: If C1 = 1, the facility types are complete; otherwise, the facility types are incomplete. Utilization rate rule: If U ≥ 60%, the facility utilization rate meets the standard; otherwise, the utilization rate is insufficient. Coverage rate rule: If C ≥ 90%, the coverage rate meets the standard; otherwise, the coverage rate is insufficient. Accessibility rule: If A ≤ 15min, the accessibility meets the standard; otherwise, the accessibility is insufficient. Fairness rule: If F ≤ 0.3, the fairness meets the standard; otherwise, the fairness is insufficient.

[0036] Furthermore, the specific construction and operation process of the target community multi-agent model in step S4 is as follows:

[0037] Step S4-1: Construction of the target community multi-agent model framework; Based on the target community spatiotemporal data integration platform constructed in Step S1, import the spatial environment data of the corresponding community, specifically including land use data, business type POI data, urban road network vector data, and community elderly care facility layout plan data; Use a workstation equipped with a multi-core CPU and GPU to run the Mesa framework in Python to build a multi-agent model; Define the spatial environment framework of the model based on the aforementioned spatial data; Based on the elderly agent model constructed in Step S2 and the diagnostic agent model constructed in Step S3, initialize the attribute set of the corresponding agents;

[0038] Step S4-2: Run the multi-agent model of the target community; import the distribution data of the elderly population in the target community, assign initial residence and specific activity preferences (including home-dependent, health-active, and medical-needs types) to the elderly agents; simulate their daily activity paths and use of elderly care facilities according to the behavior rules of the corresponding elderly agents based on the Markov decision process within each time step, and monitor the activity data of the elderly agents in real time through the diagnostic agent and calculate the actual values ​​of various diagnostic indicators of the elderly care facility plan; the diagnostic agent judges whether the elderly care facility layout plan of the target community has passed the diagnosis by comparing the actual values ​​with the benchmark values, generates diagnostic results, and outputs a diagnostic report.

[0039] Furthermore, regarding the optimization suggestions for the target community elderly care facility planning scheme in step S5, if the diagnostic indicators pass, the output is "No optimization required"; if the diagnostic indicators fail, the specific optimization type determination suggestion is as follows:

[0040] Furthermore, in step S5, the target community elderly care facility layout plan, diagnostic report results, and optimization suggestions are visualized and interactively displayed using urban 3D digital holographic sand table equipment and virtual reality equipment. Specifically, healthy individuals interact using panoramic virtual glasses equipped with binocular resolution sensors (3664x1920 pixels) and a 98° field of view, and a display. Elderly individuals interact using an intelligent guide robot equipped with a 3664x1920 pixel camera, text-to-speech conversion, over 90% accurate speech recognition, second-level facial recognition technology, meter-level positioning accuracy, and a high-fidelity voice output device, enabling real-world interaction through image and text-to-speech conversion. Then, high-precision 3D printing equipment is used to print the target community elderly care facility layout plan and the corresponding diagnostic results and optimization suggestions. Beneficial effects:

[0041] Compared with the prior art, the advantages of this invention are as follows:

[0042] 1. This invention acquires high-precision LBS data, land use data, POI data, road network data, and administrative division spatial data of the target community, and performs data preprocessing and integration to construct a unified spatiotemporal data integration platform. This platform then accurately screens the elderly population aged 60 and above, identifies their resting points and behavioral points, and generates an elderly profile system by combining individual ontology dimensions and spatiotemporal behavioral dimensions, and constructs an intelligent agent model for the elderly. This process achieves precise characterization and behavioral analysis of the elderly population, providing a scientific and reliable data foundation and decision support for the planning of community elderly care facilities, significantly improving the accuracy and relevance of planning schemes.

[0043] 2. This invention, based on an elderly profile system, utilizes Markov Decision Processes (MDPs) to define behavioral rules for elderly agents, simulating their daily activity paths and preferences for using elderly care facilities. By combining key indicators of community-based elderly care facility layout, a diagnostic agent model is constructed to achieve intelligent diagnosis and optimization of facility layout. This process addresses the problems of irrational layout and low resource utilization in traditional planning, not only improving the scientific rigor and rationality of planning schemes but also significantly enhancing the utilization rate and service equity of elderly care facilities through multi-agent collaborative optimization.

[0044] 3. This invention constructs a diagnostic intelligent agent agent model, which, based on current planning standards and the minimum standard method, clarifies the benchmark values ​​for the integrity, utilization rate, coverage rate, accessibility, and fairness of community-based elderly care facilities, and uses these as constraints for the diagnostic intelligent agent. Through a set of rules (such as integrity rules, utilization rate rules, coverage rate rules, accessibility rules, and fairness rules), the planning scheme is intelligently diagnosed, and specific diagnostic results are output. This method solves the problems of lax standard enforcement and unbalanced resource allocation in traditional planning, ensuring the scientific, standardized, and operable nature of elderly care facility planning.

[0045] 4. This patent uses a diagnostic intelligence agent to accurately evaluate various indicators (completeness, utilization rate, coverage, accessibility, and fairness) of the layout plan for community elderly care facilities, and outputs targeted optimization suggestions based on the diagnostic results. This solves the problems of vague optimization direction and unreasonable resource allocation in traditional planning, providing a scientific basis for precise adjustments to the planning scheme.

[0046] 5. This patent utilizes a three-dimensional digital holographic sand table, virtual reality equipment, and an intelligent guide robot to visualize and interactively display the layout plan, diagnostic report results, and optimization suggestions for elderly care facilities in the target community in a high-precision, multimodal format, further enhancing the intuitiveness and operability of the planning results. This method solves the problems of difficulty in public understanding and low participation in traditional planning, significantly improving the transparency of planning schemes and public participation, and providing decision-makers and community residents with an efficient and intuitive communication and feedback channel. Attached Figure Description

[0047] Figure 1 is a flowchart of a community elderly care facility planning diagnosis and feedback method based on multi-agent system according to the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0049] The technical solution of this invention will be explained in detail using the diagnosis and decision-making process of a planning scheme for elderly care facilities in a community in Nanjing, Jiangsu Province:

[0050] As shown in Figure 1, a multi-agent-based method for the planning, diagnosis, and feedback of community elderly care facilities includes the following steps:

[0051] Step S1: Construct a spatiotemporal data integration platform for the target community.

[0052] Acquire high-precision LBS data, land use data, POI data, road network data, and administrative division spatial data of the target community and perform data preprocessing; import the preprocessed data into a geographic information platform to construct a spatiotemporal data integration platform for the target community. Specifically, this includes the following steps:

[0053] S1-11 Stop Point Identification. Using a workstation equipped with a multi-core CPU and GPU, LBS data is processed and analyzed. Trajectory points that have not moved within 10 minutes or have moved within a 100-meter range are clustered in the acquired LBS data. Their average center is identified as the stop point, and stop point information is generated (including user ID, age, stop point latitude and longitude, stop duration, and stop start and end times). The average center refers to a point returned for each stop, representing the average center of the stop in terms of time and distance. The generated features will have time-type intervals.

[0054] S1-12 Data Cleaning and Effective User Screening. Effective users are defined as those whose number of days with a stay point is greater than or equal to half of the total number of days in the observation data. The specific data cleaning and effective user screening method is as follows: Summarize the data table of all stay points within the selected study period, summarize the stay days for each user by ID, screen out effective users, and associate all original stay point data by effective user ID. The data processed in step S1-11 is called effective user stay data. The original data and calculation results are stored using the distributed storage system Hadoop HDFS.

[0055] S1-21 Constructs a spatiotemporal data integration platform. This platform includes basic spatial data and spatiotemporal trajectory data. The basic spatial data is obtained from the Open Street Map website and specifically includes land use data, business type POI data, urban road network vector data, and administrative division spatial data. User spatiotemporal trajectory data includes the valid user dwell time data obtained in step S1-11. The specific attribute descriptions of each type of data are shown in the table below:

[0056] Drones equipped with high-precision cameras and LiDAR (Light Detection and Ranging) are used to conduct aerial photography and 3D modeling of specific areas, acquiring high-resolution spatial data to verify and supplement land use and road network data in Open Street Map. Internet of Things (IoT) sensors are used to monitor pedestrian and vehicle flow dynamics in real time, supplementing user spatiotemporal trajectory data.

[0057] Step S2: Construction of intelligent agent model for elderly people in the target community

[0058] Based on the spatiotemporal data integration platform described in step S1, individuals aged 60 and above are selected as the target elderly population in the community using the age field. The points of stay and behavior of the elderly in the community are identified, constructing the activity behavior chain of the elderly in the target community. Combined with their individual ontology dimension and spatiotemporal behavior dimension, tag types are combined, and a community elderly profile genealogy is constructed. Based on the above elderly profile genealogy results, an intelligent agent for the elderly in the target community is constructed. Specifically, the following steps are included:

[0059] S2-11 Construction of the Spatiotemporal Behavior Chain for the Elderly. Based on the spatiotemporal data integration platform described in step S1, spatiotemporal data of the elderly population aged 60 and above are extracted. The LBS data of the elderly in the target community is spatially connected with the 20M buffer area of ​​land use data and POI data to identify their main activity areas and activity types, specifically including residential areas, commercial areas, medical areas, cultural activity areas, and green spaces. Based on time series analysis, a spatiotemporal behavior chain for the elderly is constructed to describe their daily behavioral activity patterns.

[0060] S2-12 Label Cluster Combination. Based on the ontological dimensions (gender, age) and spatiotemporal behavioral dimensions (activity type, frequency, duration) of the elderly, the K-means clustering algorithm is used to divide the elderly into groups and generate label clusters. The specific formula is as follows: K-means(X,k)

[0061] Where X is the feature vector of the elderly (including ontology dimension and spatiotemporal behavior dimension), k is the number of clusters, and the optimal value of k is determined by the elbow rule.

[0062] S2-13 Construction of Elderly Profile Lineage. Based on the label grouping results, an elderly profile lineage was constructed. The area where the elderly spend the most time during the day (7:00 AM to 7:00 PM) excluding their residence was used as the basis for profile classification. A decision tree algorithm was used to classify the elderly population and generate elderly profile labels, specifically including "Healthy and Active," "Home-based Dependent," and "Medical Needs-based" (among which, profiles whose activity location is only their residence are classified as "Home-based Dependent"). The comparison table of activity location and profile classification is shown below:

[0063] The decision tree algorithm used is the CART algorithm, and the specific formula is as follows: CART(D,A)

[0064] Where D is the dataset of elderly people's behavioral activities, and A is the set of feature attributes.

[0065] S2-21: Initialization of the Elderly Agent Model. Based on the elderly profile system constructed in S2-1, behavioral features of each elderly category are extracted, specifically including activity type, frequency, and duration as initial attributes of the elderly agent. Agent instances are created using the Mesa framework in Python, and the agent's attribute set Aa is defined: Aa = {activity type, frequency, duration}.

[0066] S2-22: Definition of Behavioral Rules for Elderly Agents. Based on the behavioral patterns of the elderly, Markov Decision Process (MDP) is used to define the behavioral rules of the agents, simulating the daily activity paths and usage preferences of elderly care facilities. The specific formula is as follows: MDP(S m A m ,P m ,R m )

[0067] Among them, S m Let A be the state space. m For the behavior space, P m Let R be the state transition probability matrix. m This is the reward function.

[0068] Step S3: Diagnostic Analysis and Intelligent Construction of Community Elderly Care Facility Planning Scheme

[0069] The key indicators for diagnosing community-based elderly care facility planning schemes are determined. These indicators are then used as constraints, considering the completeness, utilization rate, accessibility, and equity of the community-based elderly care facility layout plan. Thresholds for these diagnostic indicators are determined using current planning standards and the minimum standard method. A diagnostic agent model for community-based elderly care facility layout is then constructed. The specific steps include:

[0070] S3-11: Define key indicators for diagnosing community-based elderly care facility layout plans. Details are as follows:

[0071] S3-21: Determination of the baseline values ​​for the diagnostic indicators of the intelligent agent. The baseline values ​​for the diagnostic indicators of the community elderly care facility layout scheme in step S3 are determined based on current planning standards and the minimum standard method, specifically as follows:

[0072] S3-22: Construction of Diagnostic Agent Indicators. Combining the baseline values ​​of the diagnostic agents described in step S3-21, these are used as constraints for constructing the diagnostic agent for community elderly care facility planning. Community elderly care facility planning schemes that meet the baseline values ​​are defined as passing the diagnosis; otherwise, they fail. Specific diagnostic results are output, including the detailed calculation results of the aforementioned key indicators. The specific rule set for the diagnostic agent for community elderly care facility planning is as follows: Completeness Rule: If C1 = 1, the facility types are complete; otherwise, the facility types are incomplete. Utilization Rate Rule: If U ≥ 60%, the facility utilization rate meets the standard; otherwise, the utilization rate is insufficient. Coverage Rate Rule: If C ≥ 90%, the coverage rate meets the standard; otherwise, the coverage rate is insufficient. Accessibility Rule: If A ≤ 15min, the accessibility meets the standard; otherwise, the accessibility is insufficient. Fairness Rule: If F ≤ 0.3, the fairness meets the standard; otherwise, the fairness is insufficient.

[0073] Step S4: Diagnosis of the current status of elderly care facility layout in the target community

[0074] Using the spatiotemporal data integration platform for the target community constructed in step S1, the spatial environment data and layout plan of the community's elderly care facilities are imported. A multi-agent model of the target community is constructed and run, combining the elderly intelligent agent described in step S2 and the community elderly care facility layout diagnostic intelligent agent described in step S3. Layout diagnosis is then performed through the diagnostic intelligent agent. Specifically, the following steps are included:

[0075] S4-1: Construction of the Multi-Agent Model Framework for the Target Community. Based on the spatiotemporal data integration platform for the target community built in step S1, import the spatial environment data of the corresponding community, specifically including land use data, POI data, road network data, and community elderly care facility layout plan data; use a workstation equipped with a multi-core CPU and GPU to run the Mesa framework in Python to build a multi-agent model. Define the spatial environment framework of the model based on the aforementioned spatial data; based on the elderly agent model built in step S2 and the diagnostic agent model built in step S3, initialize the attribute sets of the corresponding agents.

[0076] S4-2: Multi-agent model operation in the target community. Import the distribution data of the elderly population in the target community, assign initial residences and specific activity preferences (including home-dependent, health-active, and medical-needs-oriented) to the elderly agents; within each time step, simulate their daily activity paths and elderly care facility usage behavior according to the corresponding elderly agent's behavioral rules using a Markov decision process; monitor the activity data of the elderly agents in real time through a diagnostic agent and calculate the actual values ​​of various diagnostic indicators for the elderly care facility plan; the diagnostic agent compares the actual values ​​with benchmark values ​​to determine whether the elderly care facility layout plan for the target community has passed the diagnosis, generates diagnostic results, and outputs a diagnostic report.

[0077] Step S5: Decision feedback on the layout plan of elderly care facilities in the target community

[0078] The operation and diagnostic results of the multi-agent model described in step S4 are optimized and evaluated to generate a set of optimization types for the target community's elderly care facility planning scheme and corresponding optimization suggestions. Then, the target community's elderly care facility layout scheme, diagnostic report results, and optimization suggestions are visualized and interactively displayed using urban 3D digital holographic sand table equipment and virtual reality equipment, and output via printing equipment. Specifically, the following steps are included:

[0079] S5-1: Generation of optimization suggestions for the planning scheme of elderly care facilities in the target community. If the diagnostic indicators pass, the output is "No optimization required"; if the diagnostic indicators fail, the specific optimization type determination suggestion is as follows:

[0080] S5-2: Visualization and Interactive Display of the Solution. The layout plan for the target community's elderly care facilities, diagnostic report results, and optimization suggestions will be visualized and interactively displayed using a city 3D digital holographic sand table and virtual reality equipment. For healthy individuals, panoramic virtual glasses equipped with binocular sensors and a display with a resolution of 3664x1920 pixels and a 98° field of view will be used for interaction. For the elderly, an intelligent guide robot with a 3664x1920 pixel camera, text-to-speech conversion, over 90% accurate speech recognition, second-level facial recognition technology, meter-level positioning accuracy, and high-fidelity voice output will be used for real-world image and text-to-speech interaction. Finally, high-precision 3D printing equipment will be used to print the layout plan for the target community's elderly care facilities, along with the corresponding diagnostic results and optimization suggestions.

[0081] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A multi-agent-based community nursing home planning diagnosis and feedback method, characterized in that, The method includes the following steps: Step S1: Construct a spatiotemporal data integration platform for the target community Acquire high-precision LBS data, land use data, POI data, road network data, and administrative division spatial data of the target community and perform data preprocessing; import the preprocessed data into the geographic information platform to construct a spatiotemporal data integration platform for the target community; Step S2: Construction of intelligent agent model for elderly people in the target community Based on the spatiotemporal data integration platform described in step S1, people aged 60 and above are selected as the elderly population in the target community by age field. The points where the elderly stay and the points of behavior in the community are identified, the activity behavior chain of the elderly in the target community is constructed, and the label type combination and the community elderly profile genealogy are constructed by combining their individual ontology dimension and spatiotemporal behavior dimension. Based on the above elderly profile genealogy results, the target community elderly intelligent agent is constructed. Step S3: Diagnostic Analysis and Intelligent Construction of Community Elderly Care Facility Planning Scheme The key indicators for diagnosing community elderly care facility planning schemes are determined. The integrity, utilization rate, accessibility, and fairness of the community elderly care facility layout plan are used as constraints. The thresholds of the above-mentioned diagnostic indicators for the layout of community elderly care facilities are determined by combining the current planning standards and the minimum standard method. A diagnostic intelligent agent agent model for the layout of community elderly care facilities is then constructed. Step S4: Diagnosis of the current status of elderly care facility layout in the target community Using the spatiotemporal data integration platform of the target community constructed in step S1, import the spatial environment data of the target community and the layout plan of the community elderly care facilities. Combine the target community elderly intelligent agent described in step S2 and the community elderly care facility layout diagnostic intelligent agent described in step S3 to construct and run a multi-agent model of the target community, and perform layout diagnosis through the diagnostic intelligent agent. Step S5: Decision feedback on the layout plan of elderly care facilities in the target community The operation and diagnostic results of the multi-agent model described in step S4 are optimized and evaluated to generate a set of optimization types for the target community elderly care facility planning scheme and corresponding optimization suggestions. Then, the layout scheme of the target community elderly care facility, the diagnostic report results and optimization suggestions are visualized and interactively displayed through urban three-dimensional digital holographic sand table equipment and virtual reality equipment, and output through printing equipment.

2. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 1, characterized in that, The preprocessing method for LBS data in step S1 is as follows: Step S1-11 Stop point identification: Using a workstation equipped with a multi-core CPU and GPU, the LBS data is processed and analyzed. Trajectory points in the acquired LBS data that have not moved within 10 minutes or have moved within a 100-meter range are clustered, and their average centers are identified as stop points, generating stop point information. The average center refers to a point returned for each stop, used to represent the average center of the stop in terms of time and distance. The generated features will have time-type intervals. Steps S1-12: Data cleaning and effective user screening; the effective users are those whose number of days with a stay point is greater than or equal to half of the total number of days in the observation data; the specific data cleaning and effective user screening method is as follows: summarize the data table of all stay points within the selected study period, summarize and count the number of stay days for each user by ID, screen out the effective users, and associate all the original stay point data by effective user ID. The data processed in step S1-11 is called the effective user stay data; use a distributed storage system to store the original data and calculation results.

3. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 2, characterized in that, The spatiotemporal data integration platform in step S1 includes basic spatial data and spatiotemporal trajectory data. The basic spatial data is obtained from the Open Street Map website and specifically includes land use data, business type POI data, urban road network vector data, and administrative division spatial data. The user spatiotemporal trajectory data includes the valid user dwell data obtained in steps S1-12. The specific attribute descriptions of each type of data are shown in the table below: By using drones equipped with high-precision cameras and lidar, aerial photography and 3D modeling of specific areas are conducted to obtain high-resolution spatial data, which is used to verify and supplement land use and road network data in Open Street Map. Using IoT sensors to monitor the dynamics of pedestrian and vehicle traffic in real time, supplementing users' spatiotemporal trajectory data.

4. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 3, characterized in that, The specific process for constructing the community elderly profile genealogy in step S2 is as follows: Step S2-11: Construction of the spatiotemporal behavior chain of the elderly; Based on the spatiotemporal data integration platform described in S1, spatiotemporal data of the elderly population aged 60 and above are extracted. The LBS data of the elderly in the target community is spatially connected with the 20M buffer area of ​​land use data and POI data to identify their main activity areas and activity types, specifically including residential areas, commercial areas, medical areas, cultural activity areas, and green spaces. Based on time series analysis, the spatiotemporal behavior chain of the elderly is constructed to describe their daily behavioral activity patterns. Step S2-12: Label group combination; Based on the ontology and spatiotemporal behavior dimensions of the elderly, the K-means clustering algorithm is used to divide the elderly population into groups and generate label groups; the specific formula is as follows: K-means(X,k) Where X is the feature vector of the elderly, k is the number of clusters, and the optimal value of k is determined by the elbow rule; Step S2-13: Construction of Elderly Profile Lineage; Based on the label group results, an elderly profile lineage is constructed. The area where the elderly spend the most time during the day (excluding their residence) is used as the basis for profile classification. The decision tree algorithm is used to classify the elderly population and generate elderly profile labels, specifically including "healthy and active," "home-based dependent," and "medical-needing." The activity location and profile classification correspondence table is shown below: The decision tree algorithm used is the CART algorithm, and the specific formula is as follows: CART(D,A) Where D is the dataset of elderly people's behavioral activities, and A is the set of feature attributes.

5. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 4, characterized in that, The specific process for constructing the intelligent agent for the elderly in the target community in step S2 is as follows: Step S2-21: Initialize the elderly agent model; Based on the elderly profile genealogy constructed above, extract behavioral features for each type of elderly person, specifically including activity type, frequency, and duration as initial attributes of the elderly agent; Use the Mesa framework in Python to create agent instances and define the agent's attribute set Aa: Aa = {Activity type, frequency, duration} Step S2-22: Defining the behavioral rules of the elderly agent; Based on the behavioral patterns of the elderly, the behavioral rules of the agent are defined using a Markov decision process to simulate the daily activity paths and usage preferences of elderly care facilities; the specific formula is as follows: MDP(S m ,A m ,P m ,R m ) Among them, S m Let A be the state space. m For the behavior space, P m Let R be the state transition probability matrix. m This is the reward function.

6. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 5, characterized in that, The key indicators for the diagnosis of the community elderly care facility layout plan in step S3 are specifically as follows:

7. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 1, characterized in that, The specific construction process of the diagnostic agent proxy model in step S3 is as follows: Step S3-21: Determination of the baseline values ​​for the diagnostic indicators of the intelligent agent; the baseline values ​​for the diagnostic indicators of the community elderly care facility layout scheme in step S3 are determined based on the current planning standards and the minimum standard method, specifically as follows: Step S3-22: Construction of diagnostic agent indicators; Combining the baseline values ​​of the diagnostic agents described in Step S3-21, these are used as constraints for constructing the diagnostic agent for community elderly care facility planning. Community elderly care facility planning schemes that meet the baseline values ​​are defined as passing the diagnosis, while those that do not are considered failing. Specific diagnostic results are output, including the detailed calculation results of the aforementioned key indicators. The specific rule set for the diagnostic agent for community elderly care facility planning is as follows: Completeness rule: If C1 = 1, the facility types are complete; otherwise, the facility types are incomplete. Utilization rate rule: If U ≥ 60%, the facility utilization rate meets the standard; otherwise, the utilization rate is insufficient. Coverage rate rule: If C ≥ 90%, the coverage rate meets the standard; otherwise, the coverage rate is insufficient. Accessibility rule: If A ≤ 15min, the accessibility meets the standard; otherwise, the accessibility is insufficient. Fairness rule: If F ≤ 0.3, the fairness meets the standard; otherwise, the fairness is insufficient.

8. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 7, characterized in that, The specific construction and operation process of the target community multi-agent model in step S4 is as follows: Step S4-1: Construction of the target community multi-agent model framework; Based on the target community spatiotemporal data integration platform constructed in Step S1, import the spatial environment data of the corresponding community, specifically including land use data, business type POI data, urban road network vector data, and community elderly care facility layout plan data; Use a workstation equipped with a multi-core CPU and GPU to run the Mesa framework in Python to build a multi-agent model; Define the spatial environment framework of the model based on the aforementioned spatial data; Based on the elderly agent model constructed in Step S2 and the diagnostic agent model constructed in Step S3, initialize the attribute set of the corresponding agents; Step S4-2: Run the multi-agent model of the target community; import the distribution data of the elderly population in the target community, assign initial residence and specific activity preferences to the elderly agents; simulate their daily activity paths and use of elderly care facilities according to the behavior rules of the corresponding elderly agents based on the Markov decision process within each time step, and monitor the activity data of the elderly agents in real time through the diagnostic agent and calculate the actual values ​​of various diagnostic indicators of the elderly care facility plan; the diagnostic agent judges whether the elderly care facility layout plan of the target community has passed the diagnosis by comparing the actual values ​​with the benchmark values, generates diagnostic results and outputs a diagnostic report.

9. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 8, characterized in that, If the diagnostic indicators for the optimization proposal of the target community elderly care facility planning scheme in step S5 pass, the output will be "No optimization required"; if the diagnostic indicators fail, the specific optimization type determination suggestion will be as follows:

10. The method for planning, diagnosis, and feedback of community elderly care facilities based on multi-agent systems according to claim 9, characterized in that, In step S5, the layout plan of the target community elderly care facilities, the diagnostic report results, and the optimization suggestions are visualized and interactively displayed using urban 3D digital holographic sand table equipment and virtual reality equipment. For healthy individuals, panoramic virtual glasses equipped with binocular resolution sensors (3664x1920 pixels) and a 98° field of view are used for interaction. For the elderly, an intelligent guide robot with a 3664x1920 pixel camera, text-to-speech conversion, over 90% accurate speech recognition, second-level facial recognition technology, meter-level positioning accuracy, and high-fidelity voice output equipment is used for real-world interaction involving images and text-to-speech. Then, high-precision 3D printing equipment is used to print the layout plan of the target community elderly care facilities, along with the corresponding diagnostic results and optimization suggestions.

Images