Emergency resource collaborative scheduling optimization method based on dynamic planning

Abstract

The invention discloses a first-aid resource collaborative scheduling optimization method based on dynamic planning, and belongs to the technical field of resource scheduling, and the method comprises the steps: obtaining first-aid resource multi-source data; performing preprocessing and fusion on the first-aid resource multi-source data accessed in real time to generate first-aid resource fusion data containing time, space, demand and resource states; constructing a dynamic planning model according to multi-objective optimization and constraints, solving an optimal plan, and generating a first-aid resource collaborative scheduling optimization dynamic planning scheme; and carrying out cooperative scheduling optimization on the first-aid resources according to the first-aid resource cooperative scheduling optimization dynamic planning scheme. According to the method, the problems that cooperative scheduling optimization cannot be carried out on emergency resources based on dynamic planning in the prior art, a resource allocation strategy cannot be adjusted in real time, and the emergency efficiency and the patient survival rate are reduced are solved. Cooperative scheduling optimization can be performed on first-aid resources based on dynamic planning, a resource allocation strategy can be adjusted in real time, and the first-aid efficiency and the patient survival rate are improved.

Description

Technical Field

[0001] This invention relates to the field of resource scheduling technology, specifically to a dynamic programming-based method for optimizing collaborative scheduling of emergency medical resources. Background Technology

[0002] While rapid social development brings numerous opportunities, it also poses severe challenges to the level of emergency rescue technology. How to fully and rationally utilize existing resources and technologies, and effectively dispatch emergency rescue forces within disaster-stricken areas to ensure stable operation during emergencies, has become an urgent issue. In particular, during public health emergencies or major accidents, the demand for emergency resources (ambulances, medical personnel, emergency medicines, hospital beds, etc.) fluctuates dramatically in time and space, and resources are limited. Therefore, how to achieve coordinated and optimized dispatch of emergency resources is a problem that urgently needs to be solved.

[0003] Existing technologies cannot optimize the coordinated scheduling of emergency resources based on dynamic programming, nor can they adjust resource allocation strategies in real time, thus reducing emergency response efficiency and patient survival rates. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic programming-based method for collaborative scheduling and optimization of emergency medical resources. This method can optimize the collaborative scheduling of emergency medical resources based on dynamic programming, adjust resource allocation strategies in real time, improve emergency medical efficiency and patient survival rate, and solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A dynamic programming-based method for optimizing the collaborative scheduling of emergency medical resources includes: Real-time data acquisition devices based on the Internet of Things and edge computing acquire multi-source data on emergency medical resources; Based on API gateway, message queue and stream processing entry, real-time access to multi-source emergency medical resource data is performed, and the real-time access to multi-source emergency medical resource data is preprocessed and fused to generate emergency medical resource fusion data containing time, space, demand and resource status. Based on multi-objective optimization and constraints, a dynamic programming model is constructed to solve the optimal program and generate an optimized dynamic programming scheme for the coordinated scheduling of emergency resources. The emergency medical resources are optimized through a dynamic programming scheme based on collaborative scheduling, thereby achieving the global optimum of collaborative scheduling.

[0006] Preferably, real-time data acquisition devices based on the Internet of Things and edge computing acquire multi-source data on emergency medical resources and perform the following operations: Monitor 120 emergency calls, one-click emergency calls via mobile app, automatic alarms from wearable devices, and information reported by medical institutions, and collect real-time request data; Monitor the GPS location, vehicle status, onboard equipment status, personnel configuration, vehicle type, vehicle loading capacity, and the station to which the vehicle belongs, and collect ambulance resource data; The system monitors real-time bed occupancy, queue length, specialty capability matrix, estimated waiting time, equipment configuration, operating room availability, and blood bank inventory in hospitals, collecting hospital resource data. Monitor the current location, current task status, cumulative working hours, and skill tags of medical staff to collect personnel resource data; The system monitors real-time road congestion index, traffic accident locations, road construction control information, estimated travel time matrix, meteorological and geographic information, and collects environmental resource data. Based on real-time request data, ambulance resource data, hospital resource data, personnel resource data, and environmental resource data, multi-source emergency resource data with time, space, demand, and resource characteristics is formed.

[0007] Preferably, multi-source emergency medical resource data is accessed in real time via API gateway, message queue, and stream processing entry point, and the real-time accessed multi-source emergency medical resource data is preprocessed by performing the following operations: Clean the multi-source data of emergency resources, correct GPS drift and remove data noise based on Kalman filtering, identify invalid requests and remove malicious calls and repeated alarms, perform equipment anomaly detection, and provide timely warnings and repairs when offline equipment and data anomalies occur. Standardize multi-source emergency medical resource data, convert it into a unified format, unify coordinate system and address, and perform time alignment based on timestamps to form standardized multi-source emergency medical resource data.

[0008] Preferably, multi-source emergency medical resource data is fused to generate fused emergency medical resource data containing time, space, demand, and resource status, and the following operations are performed: Spatiotemporal alignment of multi-source emergency medical resource data is performed; multi-source reports of the same event are merged based on spatiotemporal window association; vehicle GPS and traffic camera trajectories are associated based on trajectory matching; and requests from similar times and locations are grouped into the same event based on event clustering. Entity association is performed on multi-source emergency resource data to establish the relationship between requests and resources to determine the response between requests and vehicles, the relationship between patients and medical records is established to retrieve historical medical records, and the relationship between vehicles and hospitals is established to determine the matching of transport destinations. Feature extraction is performed on multi-source emergency medical resource data. Feature vectors related to the collaborative scheduling optimization of emergency medical resources are extracted from the multi-source emergency medical resource data. The extracted feature vectors are then fused to form emergency medical resource fusion data containing time, space, demand and resource status. The extracted feature vectors include spatiotemporal features, resource features, patient features, and environmental features; Spatiotemporal characteristics include time period and geographic location clustering; resource characteristics include vehicle type matching degree and medical care skill matching degree; patient characteristics include age group, symptom combination and risk of deterioration; environmental characteristics include weather influence coefficient and traffic congestion index.

[0009] Preferably, a dynamic programming model is constructed based on multi-objective optimization and constraints, and the following operations are performed: Discretize the time axis into multiple decision stages, each stage representing a fixed time window; Design state variables to determine resource status, demand status, hospital status, and environmental status; Resource status includes the location, status, estimated idle time, and stock of critical medical supplies for each ambulance; demand status includes the location, priority, waiting time, and estimated probability of worsening condition for each incomplete request; hospital status includes the remaining capacity of the emergency department, specialist treatment capacity, and estimated waiting time for each receiving hospital; environmental status includes the congestion index of key road sections. Under the conditions of resource status, demand status, hospital status, and environment, determine the decision variables, including which ambulance to dispatch for a new emergency request, which hospital the ambulance should transfer the patient to after handling the situation on-site, and whether to relocate idle vehicles. With constraints such as vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking, and with the goals of maximizing treatment success rate, minimizing total scheduling cost, maximizing fairness, and satisfying time window and capacity constraints, a globally optimal dynamic programming model is established to achieve coordinated scheduling of emergency resources.

[0010] Preferably, the optimal planning is solved to generate a dynamic programming scheme for the coordinated scheduling of emergency resources, and the following operations are performed: The integrated emergency medical resource data is input into a dynamic programming model. The model is then used to analyze the integrated emergency medical resource data. Constraints include vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking. The objectives are to maximize the success rate of treatment, minimize the total scheduling cost, maximize fairness, and meet time window and capacity constraints. Resource availability, regional load, response time prediction, and hospital saturation are calculated in real time to solve for the optimal plan and determine the dynamic programming scheme for collaborative scheduling of emergency medical resources.

[0011] Preferably, the emergency resources are coordinated and optimized according to the dynamic programming scheme for coordinated dispatch, and the following operations are performed: Based on the dynamic planning scheme for collaborative scheduling of emergency resources, dispatch instructions are distributed to vehicles and hospitals. Based on multi-party collaborative interfaces, the system works with hospital emergency departments, blood banks, and transportation departments to generate dispatch orders, route navigation, personnel grouping, and suggestions for the selection and expansion of temporary treatment points. The system also connects with the command center, vehicle terminals, and hospital systems.

[0012] Preferably, it also includes real-time monitoring of the dynamic programming scheme for collaborative scheduling of emergency resources, and adjusting the dynamic programming model based on the monitoring feedback, so that the parameters of the dynamic programming model are adaptively optimized and the model is updated, triggering state updates and rescheduling, forming an adaptive loop.

[0013] Preferably, feature extraction is performed on multi-source emergency medical resource data. Feature vectors related to the collaborative scheduling and optimization of emergency medical resources are extracted from the multi-source emergency medical resource data, and the extracted feature vectors are fused to form fused emergency medical resource data containing time, space, demand, and resource status, including: Multi-source emergency medical resource data is classified into three types: precise data, semi-fuzzy data, and highly fuzzy data. Preprocessing is then performed on each of the precise, semi-fuzzy, and highly fuzzy data to obtain preprocessed data. The preprocessed data is subjected to basic feature vector extraction to obtain spatiotemporal feature vector, resource feature vector, patient feature vector and environmental feature vector; among them, spatiotemporal feature vector is extracted based on LSTM model and improved K-means algorithm, resource feature vector is extracted based on vehicle adaptation matrix and weighted calculation method, patient feature vector is extracted based on one-hot encoding, CNN model and medical history correction rule, and environmental feature vector is extracted based on meteorological department quantization value and LSTM prediction model. The four types of basic feature vectors are subjected to feature cross-mapping to obtain cross feature vectors. The cross feature vectors are then concatenated with the four types of basic feature vectors to form an initial fusion feature matrix. The feature cross-mapping includes spatiotemporal-environment cross-mapping, resource-patient cross-mapping, spatiotemporal-patient cross-mapping, and resource-environment cross-mapping, which generate corresponding cross feature vectors respectively. Construct a weighting influence factor, and calculate dynamic weights based on the weighting influence factor to obtain the target weights for each feature dimension; The initial fusion feature matrix is ​​fused based on the target weights to obtain emergency medical resource fusion data.

[0014] Preferably, before standardizing the multi-source data of emergency medical resources, the method further includes conducting a data quality assessment of the multi-source data of emergency medical resources, and standardizing the multi-source data of emergency medical resources when the assessment results meet the requirements. The data quality assessment of multi-source emergency medical resource data includes: The association features of multi-source emergency medical resource data are extracted. The association features include real-time features, accuracy features, correlation features, completeness features and availability features. Each dimension feature is integrated with single-source data features and cross-source correlation features. The features of each dimension are calculated, normalized and fused to obtain the association feature set. A temporal-association dual-stream attention network is constructed as a dynamic evaluation model. The associated feature set is input into the temporal stream branch and the associated stream branch of the dynamic evaluation model. The temporal feature vector and the associated feature vector are dynamically fused by a dual-stream fusion layer to obtain a fused feature matrix. The fused feature matrix is ​​input into a piecewise quality evaluation function to output a data quality evaluation value. The temporal stream branch uses an LSTM-Transformer hybrid structure to extract temporal dependency features and output a temporal feature vector. The associated stream branch uses a GCN-attention hybrid structure to extract association constraint features and output an association feature vector. The data quality assessment value is compared with a preset data quality assessment threshold. When the data quality assessment value is determined to be greater than or equal to the preset data quality assessment threshold, the multi-source data of emergency medical resources is standardized.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention utilizes real-time data acquisition devices based on the Internet of Things and edge computing to acquire multi-source data on emergency medical resources. This data is then preprocessed and fused to generate integrated emergency medical resource data containing time, space, demand, and resource status. Constrained by vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking, and aiming to maximize treatment success rate, minimize total scheduling cost, maximize fairness, and meet time window and capacity constraints, a dynamic programming model is established and the optimal programming solution is obtained. This generates a dynamic programming scheme for collaborative scheduling optimization of emergency medical resources. Based on this scheme, collaborative scheduling optimization of emergency medical resources is performed, achieving global optimization of collaborative scheduling. This system allows for collaborative scheduling optimization of emergency medical resources based on dynamic programming, enabling real-time adjustment of resource allocation strategies to improve emergency medical efficiency and patient survival rates. Attached Figure Description

[0016] Figure 1 This is a flowchart of the dynamic programming-based emergency resource collaborative scheduling optimization method of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0018] To address the current limitations of dynamic programming in optimizing the coordinated scheduling of emergency resources and incorporating real-time adjustments to resource allocation strategies, which consequently reduces emergency response efficiency and patient survival rates, please refer to [link to relevant documentation]. Figure 1 This embodiment provides the following technical solution: A dynamic programming-based method for optimizing the collaborative scheduling of emergency medical resources includes: Real-time data acquisition devices based on the Internet of Things and edge computing acquire multi-source data on emergency medical resources; Based on API gateway, message queue and stream processing entry, real-time access to multi-source emergency medical resource data is performed, and the real-time access to multi-source emergency medical resource data is preprocessed and fused to generate emergency medical resource fusion data containing time, space, demand and resource status. Based on multi-objective optimization and constraints, a dynamic programming model is constructed to solve the optimal program and generate an optimized dynamic programming scheme for the coordinated scheduling of emergency resources. The emergency medical resources are optimized through a dynamic programming scheme based on collaborative scheduling, thereby achieving the global optimum of collaborative scheduling.

[0019] Specifically, by breaking down regional, hierarchical, and type barriers, and by unifying, dynamically assigning, and coordinating the allocation of diverse resources such as ambulances, emergency personnel, and hospital beds, emergency response time can be shortened, treatment success rates can be improved, and resource utilization can be maximized.

[0020] In this embodiment, a real-time data acquisition device based on the Internet of Things and edge computing acquires multi-source data on emergency medical resources and performs the following operations: Monitor 120 emergency calls, one-click emergency calls via mobile app, automatic alarms from wearable devices, and information reported by medical institutions, and collect real-time request data; The system monitors the GPS location, vehicle status (driving, on standby, refueling / charging, maintenance), onboard equipment status (defibrillator, ventilator, medication inventory), personnel configuration (doctor, nurse and driver information, qualifications, continuous working hours), vehicle type (general type, BLS type, critical care type, negative pressure isolation type), vehicle loading capacity (number of stretchers, equipment list), and vehicle station (parking location, service radius) of ambulances, collecting ambulance resource data. The system monitors real-time bed occupancy (resuscitation beds, observation beds, ICU beds), queue length (triage waiting queue length), specialty capability matrix (chest pain center, stroke center, trauma center, pediatrics, burns), estimated waiting time (time from arrival to doctor's consultation), equipment configuration (CT and MRI availability), operating room vacancy status, and blood bank inventory, collecting hospital resource data. The system monitors the current location (in vehicle, at hospital, in field), current task status, cumulative working hours, and skill tags (endotracheal intubation, cardiopulmonary resuscitation, pediatric emergency care) of medical staff, and collects personnel resource data. The system monitors real-time road congestion index (from Gaode / Baidu API), traffic accident locations (from traffic police platform), road construction control information, estimated travel time matrix, weather (rainfall, snow, fog, visibility, temperature / humidity, air quality index), and geographic information (3D building models, elevator availability) to collect environmental resource data. Based on real-time request data, ambulance resource data, hospital resource data, personnel resource data, and environmental resource data, multi-source emergency resource data with time, space, demand, and resource characteristics is formed.

[0021] In this embodiment, multi-source emergency medical resource data is accessed in real time via an API gateway, message queue, and stream processing entry point. The real-time accessed multi-source emergency medical resource data is then preprocessed, and the following operations are performed: Clean the multi-source data of emergency resources, correct GPS drift and remove data noise based on Kalman filtering, identify invalid requests and remove malicious calls and repeated alarms, perform equipment anomaly detection, and provide timely warnings and repairs when offline equipment and data anomalies occur. Standardize multi-source emergency medical resource data, convert it into a unified format, unify coordinate system and address, and perform time alignment based on timestamps to form standardized multi-source emergency medical resource data.

[0022] In this embodiment, multi-source emergency medical resource data is fused to generate fused emergency medical resource data containing time, space, demand, and resource status, and the following operations are performed: Spatiotemporal alignment of multi-source emergency medical resource data is performed; multi-source reports of the same event are merged based on spatiotemporal window association; vehicle GPS and traffic camera trajectories are associated based on trajectory matching; and requests from similar times and locations are grouped into the same event based on event clustering. Entity association is performed on multi-source emergency resource data to establish the relationship between requests and resources to determine the response between requests and vehicles, the relationship between patients and medical records is established to retrieve historical medical records, and the relationship between vehicles and hospitals is established to determine the matching of transport destinations. Feature extraction is performed on multi-source emergency medical resource data. Feature vectors related to the collaborative scheduling optimization of emergency medical resources are extracted from the multi-source emergency medical resource data. The extracted feature vectors are then fused to form emergency medical resource fusion data containing time, space, demand and resource status. The extracted feature vectors include spatiotemporal features, resource features, patient features, and environmental features; Spatiotemporal characteristics include time period and geographic location clustering; resource characteristics include vehicle type matching degree and medical care skill matching degree; patient characteristics include age group, symptom combination and deterioration risk; environmental characteristics include weather influence coefficient and traffic congestion index.

[0023] In this embodiment, a dynamic programming model is constructed based on multi-objective optimization and constraints, and the following operations are performed: Discretize the time axis into multiple decision stages, each stage representing a fixed time window; Design state variables to determine resource status, demand status, hospital status, and environmental status; Resource status includes the location, status, estimated idle time, and stock of critical medical supplies for each ambulance; demand status includes the location, priority, waiting time, and estimated probability of worsening condition for each incomplete request; hospital status includes the remaining capacity of the emergency department, specialist treatment capacity, and estimated waiting time for each receiving hospital; environmental status includes the congestion index of key road sections. Under the conditions of resource status, demand status, hospital status, and environment, determine the decision variables, including which ambulance to dispatch for a new emergency request, which hospital the ambulance should transfer the patient to after handling the situation on-site, and whether to relocate idle vehicles. With constraints such as vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking, and with the goals of maximizing treatment success rate, minimizing total scheduling cost, maximizing fairness, and satisfying time window and capacity constraints, a globally optimal dynamic programming model is established to achieve coordinated scheduling of emergency resources.

[0024] In this embodiment, the optimal planning is solved to generate an optimized dynamic programming scheme for the coordinated scheduling of emergency resources, and the following operations are performed: The integrated emergency medical resource data is input into a dynamic programming model. The model is then used to analyze the integrated emergency medical resource data. Constraints include vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking. The objectives are to maximize the success rate of treatment, minimize the total scheduling cost, maximize fairness, and meet time window and capacity constraints. Resource availability, regional load, response time prediction, and hospital saturation are calculated in real time to solve for the optimal plan and determine the dynamic programming scheme for collaborative scheduling of emergency medical resources.

[0025] In this embodiment, the emergency resources are optimized for coordinated scheduling based on the dynamic programming scheme, and the following operations are performed: Based on the dynamic planning scheme for collaborative scheduling of emergency resources, dispatch instructions are distributed to vehicles and hospitals. Based on multi-party collaborative interfaces, the system works with hospital emergency departments, blood banks, and transportation departments to generate dispatch orders, route navigation, personnel grouping, and suggestions for the selection and expansion of temporary treatment points. The system also connects with the command center, vehicle terminals, and hospital systems.

[0026] In this embodiment, the method also includes real-time monitoring of the dynamic planning scheme for collaborative scheduling of emergency resources, and adjusting the dynamic planning model based on the monitoring results, so that the parameters of the dynamic planning model are adaptively optimized and the model is updated, triggering state updates and rescheduling, forming an adaptive loop.

[0027] In summary, by acquiring multi-source emergency medical resource data and performing preprocessing and fusion, fused emergency medical resource data containing time, space, demand, and resource status is generated. Constrained by vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking, and with the objectives of maximizing treatment success rate, minimizing total scheduling cost, maximizing fairness, and satisfying time window and capacity constraints, a dynamic programming model is established and the optimal plan is solved. Furthermore, an optimized dynamic programming scheme for collaborative scheduling of emergency medical resources is generated, thereby optimizing the collaborative scheduling of emergency medical resources and achieving global optimization of collaborative scheduling, which can improve emergency medical efficiency and patient survival rate.

[0028] Feature extraction is performed on multi-source emergency medical resource data. Feature vectors related to the collaborative scheduling and optimization of emergency medical resources are extracted from the multi-source data. These extracted feature vectors are then fused to form fused emergency medical resource data containing time, space, demand, and resource status information, including: Multi-source emergency medical resource data is classified into three types: precise data, semi-fuzzy data, and highly fuzzy data. Preprocessing is then performed on each of the precise, semi-fuzzy, and highly fuzzy data to obtain preprocessed data. The preprocessed data is subjected to basic feature vector extraction to obtain spatiotemporal feature vector, resource feature vector, patient feature vector and environmental feature vector; among them, spatiotemporal feature vector is extracted based on LSTM model and improved K-means algorithm, resource feature vector is extracted based on vehicle adaptation matrix and weighted calculation method, patient feature vector is extracted based on one-hot encoding, CNN model and medical history correction rule, and environmental feature vector is extracted based on meteorological department quantization value and LSTM prediction model. The four types of basic feature vectors are subjected to feature cross-mapping to obtain cross feature vectors. The cross feature vectors are then concatenated with the four types of basic feature vectors to form an initial fusion feature matrix. The feature cross-mapping includes spatiotemporal-environment cross-mapping, resource-patient cross-mapping, spatiotemporal-patient cross-mapping, and resource-environment cross-mapping, which generate corresponding cross feature vectors respectively. Construct a weighting influence factor, and calculate dynamic weights based on the weighting influence factor to obtain the target weights for each feature dimension; The initial fusion feature matrix is ​​fused based on the target weights to obtain emergency medical resource fusion data.

[0029] In this embodiment, the precise data includes: objective data measured by instruments and automatically recorded by the system, including traffic congestion index, real-time vehicle location, patient age, remaining number of emergency beds, and weather impact coefficient.

[0030] In this embodiment, semi-fuzzy data refers to data that contains some subjective evaluation but can be quantified, including the matching degree of medical and nursing skills, the vehicle failure risk coefficient, and the road traffic efficiency prediction.

[0031] In this embodiment, the highly fuzzy data is unstructured data based on subjective descriptions and experience-based judgments, including combinations of patient symptoms, risk of deterioration, and complexity of the on-site environment.

[0032] In this embodiment, precise data preprocessing includes wavelet transform denoising and Z-score normalization.

[0033] In this embodiment, semi-fuzzy data preprocessing includes fuzziness reduction based on fuzzy C-means clustering, such as clustering the continuous scores of medical and nursing skill matching degree into three levels: high fit, medium fit, and low fit, and then mapping them to quantized values ​​of [0.8, 0.5, 0.2].

[0034] In this embodiment, the preprocessing of highly fuzzy data includes: cleaning with improved DS evidence theory, constructing an emergency scene recognition framework {low risk, medium risk, high risk, extremely high risk}, dividing the data into groups using an effective label set coverage algorithm, i.e., label set = symptom urgency, medical history correlation, and scene conditions, inputting the data into a classifier trained based on Bert+DS evidence theory, outputting the category probability, and then performing a pignistic transformation to obtain a quantified initial deterioration risk value; the initial deterioration risk value is a continuous value in the interval [0,1].

[0035] In this embodiment, the time periodic feature is extracted using an LSTM model to extract the temporal pattern. The input is the number of emergency requests in the past 7 days, holiday markers, and commuting peak markers. The output is the time urgency coefficient. The time series is decomposed by Fourier transform, and the amplitude and phase of the first 5 main frequency components are extracted as periodic features. The spatiotemporal feature vector includes timestamp, time urgency coefficient, periodic features, cluster unit ID, spatial demand density, and emergency response radius.

[0036] In this embodiment, the geographic location clustering adopts an improved K-means algorithm, which divides the service area into 100×100 meter grid clustering units and calculates the spatial demand density and emergency response radius of each clustering unit; the spatial demand density is the number of requests in the past hour / unit area; the emergency response radius is the straight-line distance from the unit center to the nearest emergency station / road traffic coefficient.

[0037] In this embodiment, resource feature extraction includes vehicle type matching degree and medical skills matching degree; vehicle type matching degree is obtained by constructing a vehicle adaptation matrix, assigning weights to matrix elements based on historical scheduling success rates, and outputting matching degree values; wherein the rows of the vehicle adaptation matrix = vehicle type: general ambulance / critical care ambulance / maternal and infant ambulance; columns = patient symptom type: trauma / cardiovascular / pediatrics / obstetrics and gynecology.

[0038] In this embodiment, the matching degree of medical and nursing skills is calculated by weighting qualification scores and experience scores, and the output matching degree value is (0-1). The qualification score is quantified by the nurse's license / physician's license level (basic = 0.5, intermediate = 0.7, advanced = 1.0); the experience score is obtained by the number of patients with similar symptoms treated in the past year / patient satisfaction, calculated using the following formula: ; Experience rating for healthcare professionals in managing specific symptoms.

[0039] In this embodiment, the resource feature vector includes emergency vehicle ID, vehicle type, vehicle matching degree, medical team ID, medical skill matching degree, vehicle remaining range, and current load status.

[0040] In this embodiment, patient feature extraction includes age group, symptom combination, and risk of deterioration; age groups are divided into children (0-14 years), youth (15-44 years), middle-aged (45-64 years), and elderly (65+ years), mapped to 4-dimensional one-hot encoding; symptom combinations are based on an emergency symptom dictionary, using a CNN model to extract symptom feature vectors (64 dimensions), and then weighted according to physician experience rules; risk of deterioration is based on the results of high-fuzzy data processing and medical history correction. ,in Does the patient have a similar history of illness? (Yes = 1, No = 0) Quantized values ​​output from preprocessing highly fuzzy data; This is the revised risk value for patient deterioration.

[0041] In this embodiment, the patient feature vector includes patient ID, age group code, symptom feature vector, deterioration risk value, and demand priority.

[0042] In this embodiment, environmental feature extraction includes weather impact coefficient and traffic congestion index; the weather impact coefficient adopts the quantitative value of the meteorological department, combined with regional terrain correction, with a range of [0,1]: sunny day = 0.1, light rain / light snow = 0.3, moderate rain / moderate snow = 0.5, heavy rain / heavy snow = 0.7, rainstorm / snowstorm = 0.9; the traffic congestion index adopts real-time data from the transportation department, and predicts the index 15 minutes later through the LSTM model, outputting the current congestion index and the predicted congestion index.

[0043] In this embodiment, the environmental feature vector includes weather impact coefficient, current congestion index, predicted congestion index, road closure signage, and terrain complexity coefficient.

[0044] In this embodiment, spatiotemporal and environmental aspects intersect: ; The environmental risk level is classified based on the weather impact coefficient.

[0045] In this embodiment, resource-patient crossover occurs: ; An emergency adaptation flag is set to 1 if and only if the requirement priority is greater than 0.7 and the matching degree is greater than 0.6.

[0046] In this embodiment, spatiotemporal-patient crossover occurs: , Patient clustering in the same area = number of requests in the current unit / average number of requests in the past hour.

[0047] In this embodiment, resource-environment interaction occurs: ; Vehicle traffic restriction sign = 1 when there is a bridge height restriction sign or a narrow road sign.

[0048] In this embodiment, before concatenating the four types of cross feature vectors (each with 2 dimensions) with the four types of basic feature vectors, the high-dimensional features (symptom feature vectors) are reduced to 16 dimensions using PCA, and the low-dimensional features are copied and expanded to 16 dimensions to form an initial fusion feature matrix with consistent dimensions.

[0049] In this embodiment, the weighting factors include data quality factor, collection stability factor, demand urgency factor, resource scarcity factor, feature correlation factor, and scenario adaptation factor.

[0050] In this embodiment, the data quality factor (Q) is calculated based on the data error rate: accurate data Semi-fuzzy data Highly fuzzy data .

[0051] In this embodiment, the acquisition stability factor (S) is calculated based on the success rate and latency fluctuation of the last 10 acquisitions: (Delay fluctuation coefficient = delay standard deviation / average delay, take 1.0 when average delay = 0).

[0052] In this embodiment, the urgency factor (E) is directly adopted as the patient's demand priority (0-1), and E=1.0 when the risk of deterioration is ≥0.8.

[0053] In this embodiment, the resource stress factor (R) is: Weight boosting is triggered when R ≤ 0.5.

[0054] In this embodiment, the feature correlation factor (C) is used to calculate the mutual information between the feature and the scheduling optimization target: C=1.0 when mutual information ≥0.6, C=0.7 when mutual information is 0.3-0.6, and C=0.3 when mutual information is <0.3.

[0055] In this embodiment, the scenario adaptation factor (A) is adjusted based on the current scenario. For major accident scenarios, A = 1.2, and for normal scenarios, A = 1.0.

[0056] In this embodiment, the target weight is calculated using a first adjustment coefficient and a second adjustment coefficient. The first adjustment coefficient reflects the reliability of the data and the second adjustment coefficient reflects the priority of emergency response.

[0057] In this embodiment, the first adjustment coefficient : ; In this embodiment, the second adjustment coefficient : ; In this embodiment, ; in, The target weight; The baseline weight.

[0058] The working principle and beneficial effects of the above technical solution are as follows: Multi-source emergency medical resource data is processed hierarchically and preprocessed separately to make the data more standardized and orderly, facilitating the extraction of effective features and improving data quality; basic feature vectors are extracted from multiple dimensions such as time, space, resources, patients, and environment to comprehensively reflect information related to the coordinated scheduling of emergency medical resources, enabling a more accurate grasp of the scheduling scenario; an initial fusion feature matrix is ​​formed through feature cross-mapping and feature splicing, and then fused with dynamic weights, integrating multiple features to make the fused emergency medical resource data contain richer and more accurate information; the fused emergency medical resource data, which integrates key information such as time, space, demand, and resource status, provides a more reliable basis for optimizing the coordinated scheduling of emergency medical resources, helping to allocate emergency medical resources more accurately and improve the rationality of the scheduling plan; the construction of dynamic weights for weighted influencing factors can adapt to the dynamic changes of various factors in the emergency scenario, enabling the scheduling plan to be flexibly adjusted according to the actual situation, enhancing the adaptability and effectiveness of scheduling.

[0059] Before standardizing multi-source data on emergency medical resources, the process also includes conducting a data quality assessment of the multi-source data on emergency medical resources, and standardizing the multi-source data on emergency medical resources when the assessment results meet the requirements. The data quality assessment of multi-source emergency medical resource data includes: The association features of multi-source emergency medical resource data are extracted. The association features include real-time features, accuracy features, correlation features, completeness features and availability features. Each dimension feature is integrated with single-source data features and cross-source correlation features. The features of each dimension are calculated, normalized and fused to obtain the association feature set. A temporal-association dual-stream attention network is constructed as a dynamic evaluation model. The associated feature set is input into the temporal stream branch and the associated stream branch of the dynamic evaluation model. The temporal feature vector and the associated feature vector are dynamically fused by a dual-stream fusion layer to obtain a fused feature matrix. The fused feature matrix is ​​input into a piecewise quality evaluation function to output a data quality evaluation value. The temporal stream branch uses an LSTM-Transformer hybrid structure to extract temporal dependency features and output a temporal feature vector. The associated stream branch uses a GCN-attention hybrid structure to extract association constraint features and output an association feature vector. The data quality assessment value is compared with a preset data quality assessment threshold. When the data quality assessment value is determined to be greater than or equal to the preset data quality assessment threshold, the multi-source data of emergency medical resources is standardized.

[0060] In this embodiment, the real-time characteristic is: transmission delay. Update frequency Spatiotemporal synchronization ( Second; For spatial data timestamps; (Time stamp for time data).

[0061] In this embodiment, the accuracy feature is: static consistency. ( Let k be the static data of system i. This is the k-th static data item of system j; (where K is the reference value for the k-th static data item; K is the number of static data items); Dynamic precision. ( (Relative error between dynamic data and true values); rule compliance .

[0062] In this embodiment, the correlation feature is: spatiotemporal matching consistency. Business link correlation Cross-source coupling ( Let n be the correlation coefficient between the i-th and j-th data types, and m be the number of data types.

[0063] In this embodiment, the integrity feature is the non-null rate of the key field. Record integrity Association integrity .

[0064] In this embodiment, usability feature: data validity Scene adaptability Redundancy percentage .

[0065] In this embodiment, a cross-dimensional association matrix is ​​constructed, and the association strength coefficient is calculated. Determine the interaction between dimension i and dimension j: ; ; in, Let be the Pearson correlation coefficient between dimensions i and j; Let i be the feature covariance of dimension i and j. Let i be the standard deviation of the feature in dimension i. The business weight for dimension i; The business weight for dimension j.

[0066] In this embodiment, single-dimensional features and cross-dimensional correlation features are fused to generate a correlation feature set: ;in, Let i be the feature value of dimension i; Let be the correlation strength coefficient between dimensions i and j; The business weight for dimension i; For device operation;

[0067] In this embodiment, the temporal flow branch adopts an LSTM-Transformer hybrid structure: Layer structure: Input layer (feature dimension = 25) → LSTM layer (hidden units = 128, number of layers = 3, dropout = 0.25) → Transformer encoder (attention heads = 8, number of layers = 2) → fully connected layer (output dimension = 64), outputting temporal feature vectors. .

[0068] In this embodiment, the association flow branch adopts a GCN-attention hybrid structure: Graph structure construction: using 6 types of data sources as nodes (calls, vehicles, personnel, organizations, road conditions, and supplies), with edge weights as cross-source coupling matrices, a dynamic association graph is constructed; Layer structure: Input layer (node ​​feature dimension = 25) → GCN layer (hidden unit number = 128, number of layers = 3) → Graph attention layer (attention head number = 8) → Fully connected layer (output dimension = 64), outputting the association feature vector. .

[0069] In this embodiment, the dual-stream fusion layer is a priority vector for emergency rescue scenarios. (Corresponding to real-time performance, accuracy, relevance, completeness, and availability); Time-series weights ; This is a time-series feature vector; This is the transpose of the priority vector; Priority vector for emergency rescue scenarios; associated weights The physical constraint feature vector R is [emergency response time constraint, resource capacity constraint], calculated using domain rules; The fused feature matrix is: .

[0070] In this embodiment, the quality assessment function is: the urgency level of the emergency rescue scenario. (0 = normal, 1 = major accident); ; in, The weight matrix obtained during training, For bias terms; This is a data quality assessment value.

[0071] In this embodiment, the loss function for dynamically evaluating the model is: ;in, Cross-entropy loss; Quality labels marked manually; These are the actual physical constraint values; Total loss; This is the physical constraint feature vector output by the model.

[0072] The working principle and beneficial effects of the above technical solution are as follows: It extracts correlation features from multiple dimensions such as real-time performance, accuracy, relevance, completeness, and availability, and integrates single-source and cross-source features, enabling a comprehensive and detailed measurement of the quality of multi-source emergency medical resource data, avoiding the one-sidedness of single-dimensional evaluation; it constructs a time-series-relevance dual-stream attention network as a dynamic evaluation model, dynamically weighting and fusing time-series and correlation feature vectors through a dual-stream fusion layer, fully considering the temporal dependence and correlation constraints of the data, and more accurately reflecting the dynamic changes in data quality, resulting in more reliable evaluation results; it performs data quality evaluation first, and only performs standardization processing when the evaluation value reaches a preset threshold, avoiding useless standardization operations on low-quality data and saving computational resources and time costs; it ensures that the data quality meets the standards entering the standardization stage, enabling standardization processing to play a more effective role, providing a high-quality data foundation for subsequent emergency medical resource collaborative scheduling optimization based on this data, thereby improving the accuracy and effectiveness of the scheduling scheme.

[0073] It should be noted that, in this document, relational terms such as "first" and "second" are used only 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 process, method, article, or apparatus.

[0074] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic programming-based method for optimizing the collaborative scheduling of emergency medical resources, characterized in that, include: Real-time data acquisition devices based on the Internet of Things and edge computing acquire multi-source data on emergency medical resources; Based on API gateway, message queue and stream processing entry, real-time access to multi-source emergency medical resource data is performed, and the real-time access to multi-source emergency medical resource data is preprocessed and fused to generate emergency medical resource fusion data containing time, space, demand and resource status. Based on multi-objective optimization and constraints, a dynamic programming model is constructed to solve the optimal program and generate an optimized dynamic programming scheme for the coordinated scheduling of emergency resources. The emergency medical resources are optimized through a dynamic programming scheme based on collaborative scheduling, thereby achieving the global optimum of collaborative scheduling.

2. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 1, characterized in that, Real-time data acquisition devices based on the Internet of Things and edge computing acquire multi-source data on emergency medical resources and perform the following operations: Monitor 120 emergency calls, one-click emergency calls via mobile app, automatic alarms from wearable devices, and information reported by medical institutions, and collect real-time request data; Monitor the GPS location, vehicle status, onboard equipment status, personnel configuration, vehicle type, vehicle loading capacity, and the station to which the vehicle belongs, and collect ambulance resource data; The system monitors real-time bed occupancy, queue length, specialty capability matrix, estimated waiting time, equipment configuration, operating room availability, and blood bank inventory in hospitals, collecting hospital resource data. Monitor the current location, current task status, cumulative working hours, and skill tags of medical staff to collect personnel resource data; The system monitors real-time road congestion index, traffic accident locations, road construction control information, estimated travel time matrix, meteorological and geographic information, and collects environmental resource data. Based on real-time request data, ambulance resource data, hospital resource data, personnel resource data, and environmental resource data, multi-source emergency resource data with time, space, demand, and resource characteristics is formed.

3. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 2, characterized in that, Real-time access to multi-source emergency medical resource data is achieved through API gateway, message queue, and stream processing entry point. The real-time access to this multi-source emergency medical resource data is then preprocessed, and the following operations are performed: Clean the multi-source data of emergency resources, correct GPS drift and remove data noise based on Kalman filtering, identify invalid requests and remove malicious calls and repeated alarms, perform equipment anomaly detection, and provide timely warnings and repairs when offline equipment and data anomalies occur. Standardize multi-source emergency medical resource data, convert it into a unified format, unify coordinate system and address, and align it based on timestamps to form standardized multi-source emergency medical resource data.

4. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 3, characterized in that, The multi-source emergency medical resource data is fused to generate fused emergency medical resource data containing time, space, demand, and resource status. The following operations are performed: Spatiotemporal alignment of multi-source emergency medical resource data is performed; multi-source reports of the same event are merged based on spatiotemporal window association; vehicle GPS and traffic camera trajectories are associated based on trajectory matching; and requests from similar times and locations are grouped into the same event based on event clustering. Entity association is performed on multi-source emergency resource data to establish the relationship between requests and resources to determine the response between requests and vehicles, the relationship between patients and medical records is established to retrieve historical medical records, and the relationship between vehicles and hospitals is established to determine the matching of transport destinations. Feature extraction is performed on multi-source emergency medical resource data. Feature vectors related to the collaborative scheduling optimization of emergency medical resources are extracted from the multi-source emergency medical resource data. The extracted feature vectors are then fused to form fused emergency medical resource data containing time, space, demand and resource status. The extracted feature vectors include spatiotemporal features, resource features, patient features, and environmental features; Spatiotemporal characteristics include time period and geographic location clustering; resource characteristics include vehicle type matching degree and medical care skill matching degree; patient characteristics include age group, symptom combination and deterioration risk; environmental characteristics include weather influence coefficient and traffic congestion index.

5. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 4, characterized in that, Construct a dynamic programming model based on multi-objective optimization and constraints, and perform the following operations: Discretize the time axis into multiple decision stages, each stage representing a fixed time window; Design state variables to determine resource status, demand status, hospital status, and environmental status; Resource status includes the location, status, estimated idle time, and stock of critical medical supplies for each ambulance; demand status includes the location, priority, waiting time, and estimated probability of worsening condition for each incomplete request; hospital status includes the remaining capacity of the emergency department, specialist treatment capacity, and estimated waiting time for each receiving hospital; environmental status includes the congestion index of key road sections. Under the conditions of resource status, demand status, hospital status, and environment, determine the decision variables, including which ambulance to dispatch for a new emergency request, which hospital the ambulance should transfer the patient to after handling the situation on-site, and whether to relocate idle vehicles. With constraints such as vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking, and with the goals of maximizing treatment success rate, minimizing total scheduling cost, maximizing fairness, and satisfying time window and capacity constraints, a globally optimal dynamic programming model is established to achieve coordinated scheduling of emergency resources.

6. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 5, characterized in that, Solve for the optimal planning, generate an optimized dynamic programming scheme for the coordinated scheduling of emergency resources, and perform the following operations: The integrated emergency medical resource data is input into a dynamic programming model. The model is then used to analyze the integrated emergency medical resource data. Constraints include vehicle capacity, personnel qualifications, road accessibility, hospital capacity limits, and priority ranking. The objectives are to maximize the success rate of treatment, minimize the total scheduling cost, maximize fairness, and meet time window and capacity constraints. Resource availability, regional load, response time prediction, and hospital saturation are calculated in real time to solve for the optimal plan and determine the dynamic programming scheme for collaborative scheduling of emergency medical resources.

7. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 6, characterized in that, Based on the dynamic programming scheme for collaborative scheduling of emergency medical resources, the following operations are performed to optimize the collaborative scheduling of emergency medical resources: Based on the dynamic planning scheme for collaborative scheduling of emergency resources, dispatch instructions are distributed to vehicles and hospitals. Based on multi-party collaborative interfaces, the system works with hospital emergency departments, blood banks, and transportation departments to generate dispatch orders, route navigation, personnel grouping, and suggestions for the selection and expansion of temporary treatment points. The system also connects with the command center, vehicle terminals, and hospital systems.

8. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 7, characterized in that, It also includes real-time monitoring of the dynamic programming scheme for the coordinated scheduling of emergency resources, and adjusting the dynamic programming model based on the monitoring feedback, so that the parameters of the dynamic programming model are adaptively optimized and the model is updated, triggering state updates and rescheduling, forming an adaptive loop.

9. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 4, characterized in that, Feature extraction is performed on multi-source emergency medical resource data. Feature vectors related to the collaborative scheduling and optimization of emergency medical resources are extracted from the multi-source data. These extracted feature vectors are then fused to form fused emergency medical resource data containing time, space, demand, and resource status information, including: Multi-source emergency medical resource data is classified into three types: precise data, semi-fuzzy data, and highly fuzzy data. Preprocessing is then performed on each of the precise, semi-fuzzy, and highly fuzzy data to obtain preprocessed data. The preprocessed data is subjected to basic feature vector extraction to obtain spatiotemporal feature vector, resource feature vector, patient feature vector and environmental feature vector; among them, spatiotemporal feature vector is extracted based on LSTM model and improved K-means algorithm, resource feature vector is extracted based on vehicle adaptation matrix and weighted calculation method, patient feature vector is extracted based on one-hot encoding, CNN model and medical history correction rule, and environmental feature vector is extracted based on meteorological department quantization value and LSTM prediction model. The four types of basic feature vectors are subjected to feature cross-mapping to obtain cross feature vectors. The cross feature vectors are then concatenated with the four types of basic feature vectors to form an initial fusion feature matrix. The feature cross-mapping includes spatiotemporal-environment cross-mapping, resource-patient cross-mapping, spatiotemporal-patient cross-mapping, and resource-environment cross-mapping, which generate corresponding cross feature vectors respectively. Construct a weighting influence factor, and calculate dynamic weights based on the weighting influence factor to obtain the target weights for each feature dimension; The initial fusion feature matrix is ​​fused based on the target weights to obtain emergency medical resource fusion data.

10. The emergency medical resource collaborative scheduling optimization method based on dynamic programming according to claim 3, characterized in that, Before standardizing multi-source data on emergency medical resources, the process also includes conducting a data quality assessment of the multi-source data on emergency medical resources, and standardizing the multi-source data on emergency medical resources when the assessment results meet the requirements. The data quality assessment of multi-source emergency medical resource data includes: The association features of multi-source emergency medical resource data are extracted. The association features include real-time features, accuracy features, correlation features, completeness features and availability features. Each dimension feature is integrated with single-source data features and cross-source correlation features. The features of each dimension are calculated, normalized and fused to obtain the association feature set. A temporal-association dual-stream attention network is constructed as a dynamic evaluation model. The associated feature set is input into the temporal stream branch and the associated stream branch of the dynamic evaluation model. The temporal feature vector and the associated feature vector are dynamically fused by a dual-stream fusion layer to obtain a fused feature matrix. The fused feature matrix is ​​input into a piecewise quality evaluation function to output a data quality evaluation value. The temporal stream branch uses an LSTM-Transformer hybrid structure to extract temporal dependency features and output a temporal feature vector. The associated stream branch uses a GCN-attention hybrid structure to extract association constraint features and output an association feature vector. The data quality assessment value is compared with a preset data quality assessment threshold. When the data quality assessment value is determined to be greater than or equal to the preset data quality assessment threshold, the multi-source data of emergency medical resources is standardized.

Images