Cross-center home nursing personnel scheduling method and system considering newly-added demands of clients

Abstract

The invention provides a cross-center home nursing personnel scheduling method and system considering customer newly-added demands, and relates to the field of personnel scheduling, and the method comprises the steps: building a graph structure based on customer basic attributes and a spatial adjacency relation between customers, inputting the graph structure into a GNN model, and generating a node embedding representation; independently operating single center optimization, and generating an initial path and cost of each nursing center; integrating all customer points and path information; and allocating the saved cost based on a marginal contribution resource sharing MCRS method, solving through a multi-target collaborative optimization scheduling algorithm, and determining and outputting a multi-center collaborative scheduling scheme and a collaborative income allocation mechanism. According to the method provided by the invention, the saved cost is allocated according to the coordination degree through a marginal contribution resource sharing MCRS method, so that each nursing center can be effectively stimulated to actively participate in resource sharing, the utilization efficiency of nursing service resources is improved, and the problems of resource waste and idle manpower caused by a traditional single-center mode are avoided.

Description

Technical Field

[0001] This application relates to the field of personnel scheduling technology, specifically to a cross-center home care personnel scheduling method and system that takes into account new customer needs. Background Technology

[0002] With the rapid growth in demand for home-based elderly care, the increasing need for nursing care presents complex challenges to the management and operation of home care services. Traditional single-center home care dispatching models are no longer effective in addressing the imbalance of service resources and changing demands across centers, making it difficult to meet the increasingly diverse and dynamic nursing needs. For example, some care centers may experience an overabundance of orders and overworked caregivers, while others may have idle resources, leading to overall service inefficiency. Furthermore, the frequent emergence of unexpected new demands in home care services further exacerbates the difficulties of cross-center coordination and dispatch. Home care service orders can be divided into two categories: pre-booked orders and new orders. Pre-booked orders refer to routine demands where information collection and service details (such as time windows and geographical locations) have been completed before the caregiver's departure; while new orders are new demands generated in real time during the caregiver's duties, characterized by high unpredictability and typically high urgency.

[0003] Under the current service framework, there is a widespread problem of uneven resource allocation and delayed response to new demands. Firstly, due to the significant spatial unevenness of customer demand, the traditional single-center dispatch model often leads to a severe imbalance in order volume across different nursing centers. Furthermore, judging from order characteristics, residents' demand for home care workers exhibits a clear degree of urgency and unpredictability. Especially in the face of sudden health emergencies, the market's response to these dynamic demands is clearly insufficient, failing to allocate appropriate care resources in a timely manner.

[0004] While existing research has made positive progress in optimizing scheduling models, improving service efficiency, and reducing operating costs, several shortcomings remain: First, most current solutions focus on scheduling nursing resources within a single center or fixed area, lacking cross-center collaborative optimization mechanisms. This makes it difficult to effectively address the uneven distribution and load imbalance of nursing resources across different service centers. This results in low overall resource allocation efficiency, affecting service sustainability and fairness. Second, current scheduling methods are generally based on static demand, lacking real-time response capabilities to dynamically added nursing needs. Faced with emergencies or urgent orders, the system struggles to adjust service routes and allocate personnel in a timely manner, leading to service delays and decreased customer satisfaction. Third, some research models rely on overly simplistic assumptions, such as consistent nursing staff skill levels, constant travel speeds, and neglect of traffic conditions and route familiarity. While these assumptions simplify modeling, they also reduce the model's adaptability and reliability in practical applications. Fourth, in terms of algorithm design, while most optimization algorithms, such as particle swarm optimization, ant colony optimization, or simulated annealing, possess certain solution capabilities, they are still prone to slow convergence and getting trapped in local optima when facing large-scale, multi-constraint, and multi-objective optimization problems, affecting the practicality and stability of scheduling effectiveness. Fifth, current technologies still have weak systematic considerations regarding service fairness, differences in customer satisfaction, and the balance of nursing staff workload. They lack the ability to deeply explore and model the relationship between personalized needs, the complexity of nursing tasks, and nursing staff preferences, which can easily lead to rigid service arrangements and low resource matching efficiency. Sixth, in multi-center collaborative service scenarios, the lack of effective incentive mechanisms and benefit distribution strategies inhibits the enthusiasm of nursing centers for collaborative scheduling, hindering the formation of a resource-sharing and collaborative overall service network. In summary, the relevant technologies still have significant shortcomings in addressing dynamic new demands, achieving cross-center collaborative scheduling, and improving the overall system response efficiency and intelligence level. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a cross-center home care personnel scheduling method and system that takes into account new customer needs, solving the problems of significant deficiencies in responding to dynamic new needs, achieving cross-center collaborative scheduling, and improving the overall system response efficiency and intelligence level.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] In a first aspect, embodiments of this application provide a cross-center home care worker scheduling method that considers new customer needs. The method includes: constructing a graph structure based on customer basic attributes and spatial adjacency relationships between customers, and inputting it into a GNN model; generating node embedding representations through the trained graph neural network; independently running single-center optimization, constructing a three-objective optimization model for the customer subgraph of each center; during model optimization, using an fitness evaluation function based on explicit information including path and time, and incorporating node embedding representations to identify complex relationships between customers, generating initial paths and costs for each care center; and integrating all customer point and path information to re-execute cross-center scheduling. The system employs a joint scheduling approach, introducing a dynamic collaboration coefficient that integrates time window overlap, customer density distribution, and spatial proximity to measure the collaborative contribution of each center. Based on the Marginal Contribution Resource Sharing (MCRS) method, cost savings are allocated according to the degree of collaboration. This is then solved using a multi-objective collaborative optimization scheduling algorithm to determine and output a multi-center collaborative scheduling scheme and a collaborative benefit distribution mechanism. A dynamic insertion strategy is introduced: when a new request arrives, the feasibility of inserting the new request into the task sequence of each nursing staff member in the current path is evaluated, and the incremental cost is calculated. The system selects a location from all feasible insertion positions to complete the task insertion. If no feasible insertion scheme is available, a new service path is created to ensure timely task response.

[0008] Secondly, embodiments of this application provide a cross-center home care worker scheduling system that takes into account new customer needs. The system includes: a construction and analysis module, a single-center optimization module, a center integration module, an allocation and solution module, and a dynamic insertion module.

[0009] Specifically, the analysis module constructs a graph structure based on basic customer attributes and spatial adjacency relationships between customers, inputs it into a GNN model, and generates node embedding representations through the trained graph neural network. The single-center optimization module independently runs single-center optimization, constructing a three-objective optimization model for each center's customer subgraph. During model optimization, an fitness evaluation function is used based on explicit information including path and time, incorporating node embedding representations to identify complex relationships between customers, generating initial paths and costs for each care center. The center integration module integrates all customer point and path information for re-coordinated cross-center scheduling, introducing dynamic coordination coefficients and... The system measures the collaborative contribution of each center by combining time window overlap, customer density distribution, and spatial proximity. The allocation solution module uses the marginal contribution resource sharing (MCRS) method to allocate cost savings according to the degree of collaboration, solving the problem through a multi-objective collaborative optimization scheduling algorithm to determine and output a multi-center collaborative scheduling scheme and a collaborative benefit allocation mechanism. The dynamic insertion module introduces a dynamic insertion strategy. When a new request arrives, it assesses the feasibility of inserting the new request into the current path's nursing staff task sequences and calculates the incremental cost. It selects a location from all feasible insertion positions to complete the task insertion. If no feasible insertion scheme exists, a new service path is created to ensure timely task response.

[0010] Thirdly, embodiments of this application provide an electronic device, which includes: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the cross-center home care worker scheduling method that takes into account new customer needs as described in the first aspect.

[0011] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program or instructions that, when executed by a processor, implement the cross-center home care worker scheduling method that takes into account new customer needs as described in the first aspect.

[0012] This application provides a method and system for dispatching cross-center home care workers, taking into account the new needs of customers. Compared with the prior art, it has the following advantages:

[0013] This application constructs a graph structure and inputs it into a GNN model, utilizing graph neural network technology to deeply explore the complex spatiotemporal relationships between customers, making scheduling decisions more accurate and efficient. The trained graph neural network generates node embedding representations, first independently running single-center optimization and incorporating these representations, then integrating all customer point and path information for cross-center joint scheduling. This incorporates time window overlap, customer density distribution, and spatial proximity measurements to measure the collaborative contribution of each center, significantly improving the utilization efficiency of nursing service resources and avoiding the resource waste and manpower idleness problems associated with the traditional single-center model. Furthermore, the dynamic insertion strategy enables real-time and efficient response to new and urgent nursing needs, significantly enhancing the system's rapid response capability to emergencies and improving user satisfaction and service flexibility. The marginal contribution resource sharing (MCRS) method allocates cost savings according to the degree of collaboration, effectively incentivizing each nursing center to actively participate in resource sharing, further improving the overall operational efficiency of the system. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating a cross-center home care worker scheduling method that takes into account new customer needs, as provided in an embodiment of this application.

[0016] Figure 2 This is an exemplary flowchart illustrating another cross-center home care worker scheduling method that considers new customer needs, provided in an embodiment of this application.

[0017] Figure 3 This application provides an embodiment of an undirected graph structure and adjacency matrix representation of client nodes.

[0018] Figure 4 This is a diagram illustrating the node embedding process of a GCN module and a GAT module provided in an embodiment of this application;

[0019] Figure 5 This is an exemplary flowchart of a dynamic insertion strategy provided in an embodiment of this application;

[0020] Figure 6 This is a schematic diagram of the structure of a cross-center home care worker dispatch system that takes into account new customer needs, provided in an embodiment of this application;

[0021] Figure 7This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0024] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0025] The following section first introduces a cross-center home care worker scheduling method that takes into account new customer needs, provided by an embodiment of this application.

[0026] This application provides a flowchart illustrating a cross-center home care worker scheduling method that considers new customer needs, as shown in the embodiments below. Figure 1 As shown, the cross-center home care worker scheduling method that takes into account new customer needs may include the following steps S110-S150.

[0027] S110. Based on the basic attributes of customers and the spatial adjacency relationships between customers, construct a graph structure and input it into the GNN model. Generate node embedding representations through the trained graph neural network.

[0028] S120. Independently run single-center optimization, construct a three-objective optimization model for the customer subgraph of each center. During the model optimization process, the fitness evaluation function is used to identify the complex relationships between customers based on explicit information including path and time, and node embedding representation is incorporated to generate the initial path and cost for each care center.

[0029] S130. Integrate all customer point and route information to re-perform cross-center joint scheduling, introduce dynamic collaboration coefficients, and combine time window overlap, customer density distribution and spatial proximity to measure the collaborative contribution of each center.

[0030] S140. The marginal contribution resource sharing MCRS method allocates cost savings according to the degree of collaboration, and solves the problem through a multi-objective collaborative optimization scheduling algorithm to determine and output a multi-center collaborative scheduling scheme and a collaborative benefit distribution mechanism.

[0031] S150. Introduce a dynamic insertion strategy; when a new request arrives, assess the feasibility of inserting the new request into the current path of each nursing staff task sequence and calculate the incremental cost; select a location from all feasible insertion locations to complete the task insertion; if there is no feasible insertion solution, create a new service path to ensure timely task response.

[0032] The above are specific implementation methods of the cross-center home care personnel scheduling method that takes into account the new needs of customers provided in the embodiments of this application. It can be understood that this study focuses on the dynamic scheduling optimization problem in the cross-center home care service system, aiming to improve resource utilization efficiency and service response speed to meet the ever-growing home care needs of residents.

[0033] Please refer to the above as well. Figure 1 and Figure 2 This application constructs a graph structure and inputs it into a GNN model, utilizing graph neural network technology to deeply explore the complex spatiotemporal relationships between customers, making scheduling decisions more accurate and efficient. The trained graph neural network generates node embedding representations, first independently running single-center optimization and incorporating these node embedding representations to generate initial paths and costs for each nursing center. This obtains the baseline performance of each center in a non-cooperative state and provides a foundation for subsequent cross-center collaborative optimization. Then, based on single-center path optimization, a cross-center collaborative optimization mechanism is introduced, integrating all customer point and path information to re-perform cross-center joint scheduling. This incorporates time window overlap, customer density distribution, and spatial proximity measurements to measure the collaborative contribution of each center, addressing the dual challenges of cross-center resource coordination and dynamic new demand response in home care scenarios. This significantly improves the utilization efficiency of nursing service resources and avoids the resource waste and manpower idleness problems associated with the traditional single-center model.

[0034] Furthermore, this application, through a dynamic insertion strategy, can respond to new and urgent nursing needs in real time and efficiently, significantly enhancing the system's rapid response capability to emergencies and improving user satisfaction and service flexibility. By allocating cost savings according to the degree of collaboration through the Marginal Contribution Resource Sharing (MCRS) method, it effectively incentivizes nursing centers to actively participate in resource sharing, ensuring the long-term stable operation of the overall scheduling system and further improving the overall operational efficiency of the system.

[0035] In some embodiments, before constructing a graph structure based on basic customer attributes and spatial adjacency relationships between customers and inputting it into a GNN model, and generating node embedding representations through the trained graph neural network, the cross-center home care worker scheduling method considering new customer needs further includes: preprocessing the acquired customer data and dividing customer locations into different areas using the K-Means algorithm.

[0036] In some alternative embodiments, the objective function of the three-objective optimization model considers total path length, time window breach penalty, and nurse load balancing; total path length is the total distance traveled by all nurses in completing the task; time window breach penalty is the penalty cost incurred when a nurse fails to complete the service within the time window specified by the client; nurse load balancing is used to measure whether the task allocation among nurses is balanced to avoid overloading of individual nurses.

[0037] In the embodiments of this application, it can be understood that the scheduling problem of this application is a three-objective optimization problem. By simultaneously considering the service path length, time window limit, and workload of nursing staff, the allocation of nursing staff's work tasks is more scientific and reasonable, avoiding the overwork of individual nursing staff, ensuring the stability and reliability of nursing service quality, and achieving a balanced improvement in resource utilization efficiency and service quality.

[0038] It should be noted that the objective optimization satisfies the following expression:

[0039]

[0040] minF3=minβ(M k -minM k (3)

[0041]

[0042] Tp>T k ,P∈P D ,k∈K (11)

[0043]

[0044]

[0045] Where: F1 represents the total travel distance, F2 represents the total time penalty, F3 represents the difference between the maximum and minimum workload, Equation (1) represents the sum of the distance costs along the travel routes of all nursing staff to minimize the total travel distance; Equation (2) represents the sum of the late arrival time for each customer point exceeding the latest allowed service time multiplied by the penalty coefficient to minimize the total time penalty; Equation (3) measures the difference between the maximum and minimum workload among all nursing staff and minimizes this difference to achieve load balancing; Equation (4) represents that each demand point is served only once by one nursing staff; Equation (5) represents that the medical staff providing door-to-door service start from their own nursing center and eventually return to their original nursing center; Equation (6) restricts the nursing staff... The number of times a staff member cycles within the subset must not form an independent loop; Equation (7) ensures that the medical staff providing door-to-door services have the corresponding qualifications; Equation (8) states that the service start time of the nursing staff at the customer's location must not be earlier than the earliest allowed service time at that location; Equation (9) limits the maximum working time of the nursing staff; Equation (10) ensures that the number of nursing staff dispatched by the nursing center on the same day does not exceed the upper limit of the available resources of the center; Equation (11) ensures that the submission time of a new order must be later than the time when the assigned nursing staff leaves their center; Equation (12) limits the total number of nursing staff dispatched by each nursing center to not exceed the available resources of the center when scheduling across centers; Equations (13)-(14) are the range of variable values.

[0046] It should also be noted that the meanings of the parameters involved in the aforementioned formulas are shown in Table 1:

[0047] Table 1: Definitions and Explanations of Model Symbols

[0048]

[0049]

[0050] In some embodiments, during the construction of the graph structure, each customer is considered a graph node, and the spatial adjacency relationships between customers constitute the edges of the graph. The input information includes node features, edge features, and graph structure type. Node features include the customer's geographic coordinates, start and end times of the time window, service duration, and demand level; edge features are derived from corresponding values ​​in the distance matrix to characterize the spatial distance between customers; the graph structure type is a constructed undirected graph, in which the existence of edges is determined by the distance between customers being less than a specified threshold, and adjacency matrix construction is supported; after the graph structure is constructed, an input graph G = (V, E) is formed, where V is the set of customer nodes and E is the set of edges between customers.

[0051] In the embodiments of this application, it is understood that in the cross-center home care scheduling problem, the spatial distribution of task data, time window constraints, and dependencies between customers are quite complex; traditional optimization methods cannot effectively capture these complex spatiotemporal correlation characteristics. Therefore, this application introduces a Graph Neural Network (GNN) to perform structured modeling of task data and generate node embedding vectors for scheduling optimization, thereby improving the scheduling model's ability to perceive potential correlations between customers. Please refer to... Figure 3 This application employs graph neural network technology to capture the complex spatial and temporal constraints between clients, generating high-quality node embedding vectors. This enhances the scheduling model's ability to perceive complex structures and relationships, making scheduling decisions more accurate and effective. It overcomes the shortcomings of traditional single-center scheduling models in terms of uneven resource allocation and response delays, achieving efficient collaborative utilization of nursing resources.

[0052] In one example, please refer to Figure 4 The GNN model includes two types of graph neural network models: the GCN module and the GAT module. These are used for node embedding training to extract spatiotemporal feature relationships in the graph structure. The GCN module uses the classic graph convolutional layer GCN-Conv to achieve feature aggregation. Through joint learning of the adjacency matrix and the feature matrix, it obtains the aggregated representation of each node in the spatial structure. The GAT module introduces the attention mechanism GAT-Conv to perform weighted fusion of neighboring nodes, enabling the model to adaptively learn the influence weights of different neighbors on the central node, thereby improving the representation ability. After training, the graph neural network generates a fixed-length embedding vector for each client node. The fixed-length embedding vector contains the client's own features and structural context information, which are used as input features for the scheduling optimization algorithm.

[0053] In another example, the loss module design in this application employs a contrastive learning-based graph neural network (GNN), which maps 5-dimensional node features (center index, delivery point identifier, normalized time window, node centrality) to a 16-dimensional embedding space through a three-layer graph convolutional network. The model uses a marginal loss function to optimize the node representation. The core idea is to maximize the inner product of the embedding vectors of edge-connected node pairs while minimizing the inner product of random node pairs. The loss function is defined as:

[0054]

[0055] in, It is the score of positive sample pairs (node ​​pairs with edges); It is the score of negative sample pairs (random node pairs); is the embedding vector of node i; m = 5 is the marginal value; N = 1000 is the number of samples sampled.

[0056] In some embodiments, the solution process of the aforementioned multi-objective cooperative optimization scheduling algorithm includes:

[0057] S210. In the initialization phase, the population size and reference point set are set first. The population size determines the number of individuals participating in evolution in each generation. The reference points are used to guide the diversity distribution of the solution set in the target space, so as to avoid local optima and improve the multi-objective balanced coverage capability.

[0058] S220. Based on the standard procedure of the NSGA-III algorithm, non-dominated sorting is performed, and the parent and child generations are merged and grouped according to the level until the population is full; when the complete solution set in the last level may exceed the population capacity, the improved survival selection stage is entered.

[0059] S230. First, map each candidate solution to the nearest reference direction and count the current representativeness of each reference line. Prioritize the selection of solutions from reference directions with insufficient representativeness, and select the solution with the smallest achievement scalarization function (ASF) value among the selected reference directions by calculating the achievement scalarization function (ASF) value.

[0060] S240. Before performing cross-center joint scheduling, the starting position of the number in the global matrix is ​​determined for each nursing center by calculating the customer index offset to avoid duplicate customer numbers.

[0061] S250. Merge the internal distance matrices of each center into a global distance matrix, and supplement the distance values ​​between customers across regions based on the distance relationship between centers;

[0062] S260. Integrate the service time window information of all customers to generate a unified time window list, so as to unify and integrate the customer information and scheduling parameters of each nursing center, and build a cross-center scheduling basic data structure to provide a basis for subsequent path generation and feasibility judgment.

[0063] S270. Treat all nursing center customer points and nursing staff resources as a unified whole and perform global scheduling to achieve overall path optimization and resource integration.

[0064] In this embodiment, it is understood that before performing cross-center joint scheduling, it is necessary to unify and integrate the customer information and scheduling parameters of each nursing center to construct a cross-center scheduling basic data structure: First, by calculating the customer index offset, the starting position of the number in the global matrix is ​​determined for each nursing center to avoid duplicate customer numbers; then, the internal distance matrices of each center are merged into a global distance matrix, and the distance values ​​between cross-regional customers are supplemented according to the distance relationship between centers; then, the service time window information of all customers is integrated to generate a unified time window list to provide a basis for path generation and feasibility judgment. Finally, in order to achieve overall path optimization and effective resource integration, the customer points and nursing staff resources of all nursing centers are regarded as a unified whole for global scheduling.

[0065] In one example, the aforementioned treats all nursing center customer points and nursing staff resources as a unified whole, and performs global scheduling to achieve overall path optimization and resource integration. Specifically, S270 may include the following steps:

[0066] S310. Integrate information from various customer centers and nursing staff to build a unified multi-objective optimization model, including objectives such as path cost, time window constraints, and load balancing.

[0067] S320. The improved reference point non-dominated sorting RNSGA-III algorithm is used for path planning and generation. Through non-dominated sorting, the trade-offs of each objective are balanced, and multiple path schemes are generated.

[0068] S330. Perform constraint checks on the generated paths to ensure that each path meets all customer time window constraints while taking into account the skill matching and working time constraints of nursing staff.

[0069] In the embodiments of this application, it can be understood that the aforementioned S270 mainly includes three steps: resource and customer merging modeling, path solving and scheduling generation, and feasibility verification and constraint checking. The solution algorithm of this application adopts the improved reference point non-dominated sorting RNSGA-III algorithm as a framework. This algorithm is a multi-objective evolutionary algorithm developed on the basis of NSGA-III, which is particularly suitable for optimization problems with high objective dimension and complex solution space. Its core idea is to introduce a set of uniformly distributed reference points in the evolution process, and perform association operations by calculating the vertical distance between individuals and reference points in the objective space, thereby ensuring that the solution set has good distribution and diversity in the objective space.

[0070] Furthermore, the generated solution vectors are encoded in real-number form. The algorithm uses Float-Random-Sampling for initial sampling. Float-Random-Sampling is a sampling method in NumPy used to generate floating-point random samples, and is an implementation of a random number generator. Its core is to generate pseudo-random numbers through bitwise operations, making it suitable for scenarios requiring a large number of floating-point samples. This application searches in the continuous solution space using standard genetic operators (tournament selection, crossover, mutation). The fitness of individuals is evaluated through non-dominated sorting and a reference point-based association strategy. The algorithm associates solutions with reference directions through vertical distance calculation and maintains the uniformity of the solution set distribution through ASF value comparison, further improving the solution quality and diversity. This application uses the improved reference point non-dominated sorting RNSGA-III algorithm for multi-objective solving, effectively coordinating the contradictions between multiple optimization objectives and achieving a balanced and comprehensive solution.

[0071] In some embodiments, the dynamic insertion strategy considers the increase in path distance, the satisfaction of time window constraints, and the cost of delay penalty; the increase in path distance is used to characterize the change in the total path distance after the insertion of a new customer; the satisfaction of time window constraints is used to characterize whether the service time window requirements of all customers can still be met after insertion; the cost of delay penalty is used to calculate the service delay penalty as part of the insertion cost when the path violates the time window constraint.

[0072] For example, the dynamic insertion process corresponding to the aforementioned dynamic insertion strategy at the execution level may include the following steps:

[0073] S410. Traverse the service paths of all nursing staff and record the current path structure;

[0074] S420. Try inserting the new customer at each possible insertion position for each path, and evaluate the changes in path distance and time window cost after insertion.

[0075] S430. Calculate the insertion cost for each feasible solution, and select the insertion position that minimizes the incremental cost to complete the task arrangement.

[0076] S440. If none of the existing paths meet the insertion conditions, a new service path will be created for the customer to ensure that its service requests can be responded to in a timely manner and included in the overall scheduling system.

[0077] In the embodiments of this application, please refer to Figure 5It is understandable that in cross-center home care scheduling systems, new customer requests are often sudden and urgent. Relying on global path recalculation to handle these requests would not only be computationally expensive but would also disrupt the established scheduling arrangements. The dynamic insertion strategy designed in this application aims to address the real-time response needs of new orders in home care services, ensuring the flexibility and timeliness of the scheduling system. During algorithm execution, when the system receives a new customer order, it will determine the optimal insertion position for the order in real time based on the current scheduling path information and service constraints.

[0078] Based on the overall optimization algorithm, this application introduces a dynamic insertion strategy to quickly respond to new orders in a locally optimal and low-interference manner. This dynamic insertion strategy traverses existing paths and evaluates insertion costs, embedding new customers into appropriate positions while ensuring time window constraints and path feasibility, minimizing system reordering costs and realizing a "run-while-scheduling" response mechanism for sudden demands. This module, together with the previously mentioned GNN embedding modeling, RNSGA-III path optimization, and collaborative allocation mechanism, constitutes a complete dynamic multi-center scheduling framework, enhancing the system's real-time performance and adaptability.

[0079] In some embodiments, this application provides a cross-center home care worker dispatch system 500 that takes into account new customer needs, such as... Figure 6 As shown, the cross-center home care worker dispatch system 500, which takes into account new customer needs, may include the following modules:

[0080] The analysis module 510 is used to construct a graph structure based on the basic attributes of customers and the spatial adjacency relationships between customers, and input it into the GNN model to generate node embedding representations through the trained graph neural network.

[0081] The single-center optimization module 520 is used to run single-center optimization independently. It constructs a three-objective optimization model for the customer subgraph of each center. During the model optimization process, the fitness evaluation function is used to identify the complex relationships between customers based on explicit information including path and time, and incorporates node embedding representation to generate the initial path and cost for each care center.

[0082] The central integration module 530 is used to integrate all customer point and path information to re-perform cross-center joint scheduling, introduce dynamic collaboration coefficients, and integrate time window overlap, customer density distribution and spatial proximity to measure the collaborative contribution of each center.

[0083] The allocation solution module 540 is used to allocate the cost savings according to the degree of collaboration based on the marginal contribution resource sharing MCRS method, so as to solve the problem through the multi-objective collaborative optimization scheduling algorithm, determine and output the multi-center collaborative scheduling scheme and the collaborative benefit distribution mechanism;

[0084] The dynamic insertion module 550 is used to introduce a dynamic insertion strategy. When a new request arrives, it evaluates the feasibility of inserting the new request into the current path of each nursing staff task sequence and calculates the incremental cost. It selects a location from all feasible insertion locations to complete the task insertion. If there is no feasible insertion solution, it creates a new service path to ensure timely task response.

[0085] According to embodiments of this application, any and multiple modules among the construction analysis module 510, single-center optimization module 520, center integration module 530, allocation solution module 540, and dynamic insertion module 550 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module.

[0086] Figure 6 Each module in the system shown has the function of implementing each step in the aforementioned cross-center home care worker scheduling method that takes into account new customer needs, and can achieve the corresponding technical effect. For the sake of brevity, it will not be elaborated here.

[0087] In some embodiments, this application provides an electronic device, the structural schematic of which is shown below. Figure 7 As shown.

[0088] The electronic device may include a processor 610 and a memory 620 storing computer program instructions.

[0089] Specifically, the processor 610 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0090] Memory 620 may include mass storage for data or instructions. For example, and not limitingly, memory 620 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 620 may include removable or non-removable (or fixed) media. Where appropriate, memory 620 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 620 is non-volatile solid-state memory.

[0091] Memory 620 may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory 620 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any of the above embodiments of the cross-center home care worker scheduling method that considers new customer needs.

[0092] The processor 610 reads and executes computer program instructions stored in the memory 620 to implement any of the cross-center home care worker scheduling methods in the above embodiments that take into account new customer needs.

[0093] In one example, the electronic device may also include a communication interface 630 and a bus 600. Wherein, such as Figure 7 As shown, the processor 610, memory 620, and communication interface 630 are connected via bus 600 and communicate with each other.

[0094] The communication interface 630 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0095] Bus 600 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 600 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0096] Furthermore, in conjunction with the cross-center home care worker scheduling method considering new customer needs in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the cross-center home care worker scheduling methods considering new customer needs in the above embodiments.

[0097] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0098] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0099] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0100] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0101] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for scheduling cross-center home care workers that considers new customer needs, characterized in that, include: Based on the basic attributes of customers and the spatial adjacency relationships between customers, a graph structure is constructed and input into the GNN model, and node embedding representations are generated through the trained graph neural network. The single-center optimization is run independently. A three-objective optimization model is constructed for the customer subgraph of each center. During the model optimization process, the fitness evaluation function is used to identify the complex relationships between customers based on explicit information including path and time, and the node embedding representation is incorporated to generate the initial path and cost for each care center. Integrate all customer point and route information to re-perform cross-center joint scheduling, introduce dynamic collaboration coefficients, and combine time window overlap, customer density distribution and spatial proximity to measure the collaborative contribution of each center; Based on the marginal contribution resource sharing (MCRS) method, the cost savings are allocated according to the degree of collaboration, and solved by a multi-objective collaborative optimization scheduling algorithm to determine and output a multi-center collaborative scheduling scheme and a collaborative benefit distribution mechanism. A dynamic insertion strategy is introduced; when a new request arrives, the feasibility of inserting the new request into each nurse's task sequence in the current path is evaluated, and the incremental cost is calculated; the task insertion is completed by selecting a position from all feasible insertion positions; If no feasible insertion solution is found, a new service path will be created to ensure timely task response.

2. The cross-center home care worker scheduling method considering new customer needs as described in claim 1, characterized in that, Before constructing a graph structure based on basic customer attributes and spatial adjacency relationships between customers and inputting it into a GNN model, and generating node embedding representations through the trained graph neural network, the cross-center home care worker scheduling method that considers new customer needs further includes: preprocessing the acquired customer data and using the K-Means algorithm to divide customer locations into different areas; The objective function of the three-objective optimization model considers the total path length, time window penalty for breach of contract, and the load balancing of nursing staff. The total path length is the total distance traveled by all caregivers during the completion of their tasks; the time window breach penalty is the penalty cost incurred when a caregiver fails to complete the service within the time window specified by the client; the caregiver load balancing is used to measure whether the task allocation among caregivers is balanced, so as to avoid overloading of individual personnel.

3. The cross-center home care worker scheduling method considering new customer needs as described in claim 1, characterized in that, In the process of constructing the graph structure, each customer is regarded as a graph node, the spatial adjacency relationship between customers constitutes the edge of the graph, and the input information includes node features, edge features and graph structure type; The node features include the customer's geographic coordinates, time window start and end times, service duration, and demand level; the edge features are derived from corresponding values ​​in the distance matrix to characterize the spatial distance between customers; the graph structure type is a constructed undirected graph, in which the existence of edges is determined by the distance between customers being less than a specified threshold, and adjacency matrix construction is supported; after the graph structure is constructed, an input graph G = (V, E) is formed, where V is the set of customer nodes and E is the set of edges between customers.

4. The method for dispatching cross-center home care workers considering new customer needs as described in any one of claims 1-3, characterized in that, The GNN model includes two graph neural network models, namely the GCN module and the GAT module, which are used for node embedding training to extract spatiotemporal feature relationships in graph structures. The GCN module uses the classic graph convolutional layer GCN-Conv to achieve feature aggregation. Through joint learning of the adjacency matrix and the feature matrix, it obtains the aggregated representation of each node in the spatial structure. The GAT module introduces the attention mechanism GAT-Conv to perform weighted fusion of neighboring nodes, enabling the model to adaptively learn the influence weights of different neighbors on the central node and improve representation ability. The trained graph neural network generates a fixed-length embedding vector for each customer node. The fixed-length embedding vector contains the customer's own features and structural context information, which are used as input features for the scheduling optimization algorithm.

5. The method for dispatching cross-center home care workers considering new customer needs as described in any one of claims 1-3, characterized in that, The solution process of the multi-objective cooperative optimization scheduling algorithm includes: In the initialization phase, the population size and reference point set are first set. The population size determines the number of individuals participating in evolution in each generation, and the reference points are used to guide the diversity distribution of the solution set in the target space, so as to avoid local optima and improve the multi-objective balanced coverage capability. Based on the standard procedure of the NSGA-III algorithm, non-dominated sorting is performed, and the parent and child generations are merged and grouped according to level until the population is full; when the complete solution set in the last level may exceed the population capacity, the improved survival selection stage is entered. First, map each candidate solution to the nearest reference direction and count the current representativeness of each reference line; prioritize the selection of solutions from reference directions with insufficient representativeness, and select the solution with the smallest achievement scalarization function (ASF) value among the selected reference directions by calculating the achievement scalarization function (ASF) value. Before performing cross-center joint scheduling, the starting position of the number in the global matrix for each nursing center is determined by calculating the customer index offset to avoid duplicate customer numbers. The internal distance matrices of each center are merged into a global distance matrix, and the distance values ​​between customers across regions are supplemented based on the distance relationships between centers; Integrate the service time window information of all customers to generate a unified time window list, so as to unify and integrate customer information and scheduling parameters of various nursing centers, and build a cross-center scheduling basic data structure to provide a basis for subsequent path generation and feasibility judgment; Treat all nursing center client locations and nursing staff resources as a unified whole, and conduct global scheduling to achieve overall path optimization and resource integration.

6. The cross-center home care worker scheduling method considering new customer needs as described in claim 5, characterized in that, The concept of treating all nursing center client locations and nursing staff resources as a unified whole and performing global scheduling to achieve overall path optimization and resource integration includes: By integrating information from various customer centers and nursing staff, a unified multi-objective optimization model is constructed, which includes objectives such as path cost, time window constraints, and load balancing. The improved reference point non-dominated sorting RNSGA-III algorithm is used for path planning and generation. Through non-dominated sorting, the trade-offs of each objective are balanced, and multiple path schemes are generated. Constraint checks are performed on the generated paths to ensure that each path meets all customer time window constraints while taking into account the skill matching and working time constraints of caregivers.

7. The method for cross-center home care worker scheduling considering new customer needs as described in any one of claims 1-3, characterized in that, The dynamic insertion strategy considers the increase in path distance, the satisfaction of time window constraints, and the cost of delay penalty. The increase in path distance is used to characterize the change in the total path distance after a new customer is inserted; The satisfaction of the time window constraint is used to characterize whether the service time window requirements of all customers can still be met after insertion; The delay penalty cost is used to calculate the service delay penalty as part of the insertion cost when the path violates the time window limit; The dynamic insertion strategy, at the execution level, corresponds to the following dynamic insertion process: Traverse all nursing staff service paths and record the current path structure; For each path, try inserting the new customer at every possible insertion point and evaluate the changes in path distance and time window cost after insertion; Calculate the insertion cost for each feasible solution, and select the insertion position that minimizes the incremental cost to complete the task arrangement. If none of the existing paths meet the insertion criteria, a new service path will be created for the customer to ensure that their service requests are responded to in a timely manner and included in the overall scheduling system.

8. A cross-center home care worker dispatch system that considers new customer needs, characterized in that, include: The analysis module is used to construct a graph structure based on the basic attributes of customers and the spatial adjacency relationships between customers, and input it into the GNN model. The trained graph neural network then generates node embedding representations. The single-center optimization module is used to run single-center optimization independently. It constructs a three-objective optimization model for the customer subgraph of each center. During the model optimization process, the fitness evaluation function is used to identify complex relationships between customers based on explicit information including path and time, and incorporates the node embedding representation to generate the initial path and cost for each care center. The central integration module is used to integrate all customer point and path information to re-perform cross-center joint scheduling, introduce dynamic collaboration coefficients, and integrate time window overlap, customer density distribution and spatial proximity to measure the collaborative contribution of each center. The allocation solution module is used to allocate cost savings according to the degree of collaboration based on the marginal contribution resource sharing MCRS method. It is solved by a multi-objective collaborative optimization scheduling algorithm to determine and output a multi-center collaborative scheduling scheme and a collaborative benefit distribution mechanism. The dynamic insertion module is used to introduce dynamic insertion strategies; when a new request arrives, it evaluates the feasibility of inserting the new request into the current path of each nurse's task sequence and calculates the incremental cost; it then selects a position from all feasible insertion positions to complete the task insertion. If no feasible insertion solution is found, a new service path will be created to ensure timely task response.

9. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the cross-center home care worker scheduling method as described in any one of claims 1 to 7, taking into account new customer needs.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions that, when executed by a processor, implement the cross-center home care worker scheduling method as described in any one of claims 1 to 7, taking into account new customer needs.

Images