A post-disaster wounded person rescue decision optimization method, device, equipment and storage medium

Abstract

This application relates to a method, apparatus, equipment, and storage medium for optimizing decision-making in post-disaster casualty rescue. The method includes: acquiring multidimensional physiological parameters and individual environmental factors of the casualties; using an improved START classification method and fuzzy comprehensive evaluation for initial injury quantification; and using a Markov state transition model to predict the injury evolution trend, generating a dynamic priority queue. A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed, aiming to maximize the expected number of survivors. A decomposition-coordination strategy is used to jointly solve the model, generating a globally optimized rescue plan. An event-driven mechanism is used to trigger rescheduling based on collected dynamic environmental data. Rolling time-domain optimization is performed based on the predictive control of the collaborative optimization model, and incremental updates are used to replan only the local plans affected by disturbances, outputting adaptively adjusted rescue instructions. This method can effectively improve the expected number of survivors and overall rescue efficiency.

Description

Technical Field

[0001] This application relates to the field of decision optimization technology for post-disaster casualty rescue, and in particular to a method, apparatus, equipment and storage medium for decision optimization of post-disaster casualty rescue. Background Technology

[0002] Following major natural disasters, the evacuation of the injured is a core component of emergency rescue, directly impacting the safety of affected people and the efficiency of post-disaster reconstruction. However, traditional evacuation models struggle to address the practical needs of managing evacuation order, routes, and methods in complex and dynamic environments. Statistics show that during an earthquake relief effort in one region, a temporary medical station received far more injured than its capacity within a short period, resulting in over 37% of the injured ultimately failing to receive treatment. In another region, during a rainstorm disaster, traffic paralysis caused approximately 45% of ambulances to arrive late due to route issues, severely impacting rescue efficiency. This reflects the following serious challenges that traditional evacuation models face in complex and dynamic environments: (1) The dynamic changes in the condition of the injured are ignored: Current rescue decisions mostly adopt static injury classification (i.e., START classification method), assuming that the condition of the injured remains unchanged during the rescue process. However, in actual rescue, the health condition of the injured will continue to deteriorate over time, and if they are not treated in time, their survival probability may drop sharply. These conditions of the injured are not dynamically tracked, which makes it impossible to adjust the treatment priority of many injured people who have "become more serious" and thus delays their treatment.

[0003] (2) Lack of coordination in the dispatch of rescue resources: Under the current circumstances, there is no dynamic matching mechanism between ambulances, helicopters and other transportation vehicles and medical facilities. This may lead to situations where some medical points are over-treated due to uneven distribution of the wounded, while other medical points are left without patients. This chaotic dispatch of rescue resources greatly affects the efficiency of the rescue of the wounded.

[0004] (3) Delayed route planning response: Traditional casualty evacuation models lack the ability to adjust in real time in response to disasters. Dynamic environmental factors such as road damage and traffic congestion after a disaster will render the pre-set routes ineffective. This affects both the choice of evacuation method and the accessibility of evacuation routes. This clearly exposes the drawbacks of the traditional static evacuation model.

[0005] In recent years, through extensive exploration and practice, domestic and international research has accumulated some theories and methods in the field of emergency medical rescue. However, with the increasing demand for intelligent disaster relief and the urgent need to improve rescue rates, collaborative optimization of rescue based on dynamic injury evolution has become a key means to improve rescue efficiency. Against this backdrop, current research still faces the following prominent challenges: (1) The prediction of injury is fixed and lacks accuracy.

[0006] Existing research mostly uses the START (Standardized Injury Level) assessment method, lacking precise quantitative modeling of the changes in the health status of emergency casualties over time. Although a few studies have introduced health status prediction functions, they still do not fully consider the coupled effects of individual differences and environmental factors, resulting in large prediction biases and failing to take effective measures to improve treatment priority when injuries suddenly worsen.

[0007] (2) The research on rescue order, route and method is fragmented and has not formed a collaborative framework.

[0008] Most existing literature only studies rescue priority ranking or route planning in isolation, failing to adequately integrate the three crucial factors of order, route, and method. The efficiency of casualty rescue is closely related to all three; such fragmented research makes it difficult to maximize overall rescue effectiveness and may even lead to locally optimal but globally suboptimal decisions. Such independent studies can severely impact emergency decision-making and rescue efficiency.

[0009] (3) The dynamic environment response capability is weak and cannot be rescheduled.

[0010] Most models assume fixed parameters such as road capacity and medical resource capacity, making them unsuitable for adapting to real-time changes in post-disaster environmental conditions. For example, in the event of sudden situations such as road damage, aftershocks and landslides, traffic congestion, or saturation of medical facilities, existing systems lack rapid rescheduling capabilities, requiring manual intervention for replanning. This is insufficient to meet minute-level response demands and far exceeds the tolerance limits of the golden window for treatment. Furthermore, robust design is neglected; if the preset plan is significantly affected by parameter disturbances, the overall scheduling scheme fails, requiring a complete recalculation. Summary of the Invention

[0011] Based on this, it is necessary to provide a method, device, equipment, and storage medium for optimizing post-disaster casualty rescue decisions that can achieve second-level rescheduling response within the golden rescue window, significantly reduce the computational load caused by global recalculation, and ultimately effectively improve the expected number of survivors and overall rescue efficiency.

[0012] A method for optimizing decision-making in the rescue of injured people after a disaster, the method comprising: We acquire multidimensional physiological parameters and individual environmental factors of the injured, use an improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use a Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate.

[0013] A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem jointly and generate a globally optimized rescue plan.

[0014] Real-time environmental dynamic data is collected through multi-source sensing, and rescheduling is triggered by an event-driven mechanism.

[0015] Rolling time-domain optimization is performed based on predictive control of the collaborative optimization model. Incremental updates are used to replan only the local schemes affected by disturbances, and adaptively adjusted rescue commands are output.

[0016] A disaster relief and casualty evacuation decision optimization device, the device comprising: The priority queue generation module is used to obtain the multidimensional physiological parameters and individual environmental factors of the wounded, use the improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use the Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate.

[0017] The rescue plan generation module is used to construct a three-in-one collaborative optimization model of rescue order, rescue route and rescue method. With the goal of maximizing the expected number of survivors, it adopts a decomposition-coordination strategy to solve the problem together and generate a globally optimized rescue plan.

[0018] The rescheduling trigger module is used to collect dynamic environmental data in real time through multi-source sensing and trigger rescheduling using an event-driven mechanism.

[0019] The rescue command output module is used to perform rolling time-domain optimization based on the predictive control of the collaborative optimization model. It replans only the local schemes affected by disturbances in an incremental update manner and outputs adaptively adjusted rescue commands.

[0020] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program performing the following steps: We acquire multidimensional physiological parameters and individual environmental factors of the injured, use an improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use a Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate.

[0021] A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem jointly and generate a globally optimized rescue plan.

[0022] Real-time environmental dynamic data is collected through multi-source sensing, and rescheduling is triggered by an event-driven mechanism.

[0023] Rolling time-domain optimization is performed based on predictive control of the collaborative optimization model. Incremental updates are used to replan only the local schemes affected by disturbances, and adaptively adjusted rescue commands are output.

[0024] A computer-readable storage medium having a computer program stored thereon, the computer program performing the following steps when executed by a processor: We acquire multidimensional physiological parameters and individual environmental factors of the injured, use an improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use a Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate.

[0025] A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem jointly and generate a globally optimized rescue plan.

[0026] Real-time environmental dynamic data is collected through multi-source sensing, and rescheduling is triggered by an event-driven mechanism.

[0027] Rolling time-domain optimization is performed based on predictive control of the collaborative optimization model. Incremental updates are used to replan only the local schemes affected by disturbances, and adaptively adjusted rescue commands are output.

[0028] The aforementioned method, apparatus, equipment, and storage medium for optimizing post-disaster casualty rescue decisions firstly combines an improved START classification method with fuzzy comprehensive evaluation and introduces a Markov state transition model to predict injury evolution trends, generating a dynamic priority queue that includes survival probability and deterioration rate, thus solving the problem of dynamic quantification of injuries. Secondly, this dynamic priority is used as the core input and integrated into a collaborative optimization model that combines rescue order, route, and method, aiming to maximize the expected number of survivors. A decomposition-coordination strategy is used for joint solution, breaking the fragmented state of traditional independent optimization of order, route, and method. Finally, rescheduling is driven by multi-source sensing trigger events, and rolling time-domain optimization is performed using predictive control of the collaborative optimization model. Incremental updates only replan the scheme for local disturbances, achieving a unity of global optimization and real-time response. In summary, dynamic injury prediction provides accurate priority basis for collaborative optimization, the global scheme output by collaborative optimization provides a baseline trajectory for dynamic rescheduling, and rolling time-domain optimization and incremental updates ensure the scheme's rapid adaptability under sudden environmental changes. This significantly improves the timeliness of priority adjustment for critically injured patients, avoids resource misallocation and path failure, achieves second-level rescheduling response within the golden rescue window, and greatly reduces the computational load caused by global recalculation, ultimately effectively increasing the expected number of survivors and overall rescue efficiency. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the technical architecture of an emergency medical rescue system based on multi-source perception and intelligent decision-making in one embodiment; Figure 2This is a flowchart illustrating a method for optimizing decision-making in the rescue of injured people after a disaster, as shown in one embodiment. Figure 3 This is a schematic diagram of the MPC rolling optimization and real-time control process from a micro-control perspective in one embodiment. Figure 4 This is a flowchart of a hierarchical rescue optimization algorithm from the perspective of a scheduling algorithm in one embodiment; Figure 5 This is a detailed flowchart of injury assessment from the perspective of an assessment algorithm in one embodiment. Figure 6 This is a structural block diagram of a disaster patient rescue decision optimization device in one embodiment; Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0031] This application provides an optimization method for disaster relief and evacuation decisions, which can be applied to, for example... Figure 1 The technical architecture of the emergency medical rescue system based on multi-source perception and intelligent decision-making is shown below. This system architecture includes: a multi-source data acquisition layer, three core decision engine modules, a solution output and execution layer, and a reinforcement learning offline training closed loop. It covers the entire process from injury assessment, resource scheduling, route planning to real-time dynamic adjustment and system self-optimization. Because it relies on multi-source data such as physiological sensors (blood pressure and heart rate), historical knowledge bases (case road networks), and environmental monitoring (drone GPS), three core engine modules were designed, and execution logs are recorded. Reinforcement learning is used for training, continuously iterating model parameters to achieve system self-optimization. Specifically: The multi-source data acquisition layer includes: physiological sensors for real-time acquisition of multi-dimensional physiological parameters such as blood pressure, heart rate, and respiratory rate of the injured; a historical knowledge base for storing past medical records, road network topology information, and medical point capacity records; and environmental monitoring equipment, including drones, GPS, and meteorological sensors, for real-time acquisition of dynamic environmental data such as road traffic status, aftershock and landslide information, and temperature and humidity.

[0032] The core decision engine module includes: Module 1 (Injury Assessment): Using fuzzy evaluation and Markov prediction, the raw data is transformed into a "priority queue" and a "survival probability curve" to solve the "who to save" problem.

[0033] Module 2 (Rescue Optimization): Utilizing multi-objective programming and improved genetic algorithms, the module formulates the "optimal allocation plan" and "global path planning" based on the priority of injuries, solving the problems of "how to dispatch vehicles and which routes to take".

[0034] Module 3 (Dynamic Response): Utilizing Kalman filtering and MPC rolling optimization, it handles sudden environmental changes and outputs "rescheduling instructions" to solve the problem of "what to do when encountering emergencies".

[0035] The solution output and execution layer includes: a rescue command terminal, used to receive reschedule commands and issue specific action commands to rescue vehicles and personnel; and an execution log recorder, used to record the actual execution results and environmental feedback data for each dispatch.

[0036] Feedback optimization: The reinforcement learning offline training closed loop uses a reinforcement learning trainer whose input is connected to an execution log recorder and whose output is connected to the parameter update interface of the injury assessment module, rescue optimization module, and dynamic response module. This is used to continuously iterate and optimize the internal weight coefficients and prediction models of each module using historical execution logs, so as to achieve the system's self-continuous evolution.

[0037] Data flow: The multi-source data acquisition layer inputs the collected data into the injury assessment module and the dynamic response module respectively; the output of the injury assessment module serves as the input of the rescue optimization module; the output of the rescue optimization module serves as the initial baseline scheme of the dynamic response module; the rescheduling instruction output by the dynamic response module is sent to the rescue instruction terminal; the execution log recorder sends the feedback data to the reinforcement learning trainer; the reinforcement learning trainer sends the updated model parameters back to the three core decision engine modules, forming a closed loop.

[0038] In one embodiment, such as Figure 2 As shown, a method for optimizing decision-making in the rescue of injured people after a disaster is provided, and this method is applied to... Figure 1 Taking the system architecture in [the document] as an example, the following steps are included: Step 202: Obtain the multidimensional physiological parameters and individual environmental factors of the injured, use the improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use the Markov state transition model to predict the trend of injury evolution, and generate a dynamic priority queue containing survival probability and deterioration rate.

[0039] Step 204: Construct a three-in-one collaborative optimization model of rescue order, rescue route and rescue method. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem together and generate a globally optimized rescue plan.

[0040] Step 206: Collect dynamic environmental data in real time through multi-source sensing, and trigger rescheduling using an event-driven mechanism.

[0041] Step 208: Perform rolling time-domain optimization based on the predictive control of the collaborative optimization model, and replan only the local schemes affected by disturbances in an incremental update manner, and output the adaptively adjusted rescue command.

[0042] In the aforementioned method for optimizing post-disaster casualty rescue decisions, firstly, an improved START classification method is combined with fuzzy comprehensive evaluation, and a Markov state transition model is introduced to predict the evolution trend of injuries, generating a dynamic priority queue that includes survival probability and deterioration rate, thus solving the problem of dynamic quantification of injuries. Secondly, this dynamic priority is used as the core input and integrated into a collaborative optimization model that combines rescue order, route, and method, aiming to maximize the expected number of survivors. A decomposition-coordination strategy is used to solve the problem jointly, breaking the fragmented state of traditional independent optimization of order, route, and method. Finally, rescheduling is driven by multi-source sensing trigger events, and rolling time-domain optimization is performed using predictive control of the collaborative optimization model. Incremental updates are used to replan only the scheme for local disturbances, achieving a unity of global optimization and real-time response. In summary, dynamic injury prediction provides accurate priority basis for collaborative optimization, the global scheme output by collaborative optimization provides a baseline trajectory for dynamic rescheduling, and rolling time-domain optimization and incremental updates ensure the rapid adaptability of the scheme under sudden environmental changes. This significantly improves the timeliness of priority adjustment for critically injured patients, avoids resource misallocation and path failure, achieves second-level rescheduling response within the golden rescue window, and greatly reduces the computational load caused by global recalculation, ultimately effectively increasing the expected number of survivors and overall rescue efficiency.

[0043] In one embodiment, multidimensional physiological parameters of the injured person's respiratory, circulatory, and consciousness states, as well as individual environmental factors such as ambient temperature and humidity, are acquired. Fuzzy comprehensive evaluation is used to map nonlinear physiological indicators to an initial injury quantification score. This initial injury quantification score is used as the initial state vector of a Markov state transition model. A dynamic transition matrix is ​​constructed by combining environmental factors and time decay factors to infer the survival probability curves and deterioration rates for multiple future time windows. Based on the survival probability curves and the deterioration rates, a priority scoring function is constructed. ; in, Rate the priority. , , These are the weighting coefficients, To predict the probability of survival, To predict the rate of deterioration, The waiting time is calculated. Based on the priority scoring function, the predicted injury evolution results corresponding to each injury quantification score are prioritized to obtain a dynamic priority queue.

[0044] In one embodiment, the collaborative optimization model includes a casualty-vehicle allocation model and a vehicle routing model. The casualty survival probability, vehicle capacity, and road network travel time are incorporated into a 0-1 integer programming model, with the objective function of maximizing the expected number of survivors. The ε-constraint method is used to solve this model, constructing the casualty-vehicle allocation model. Multiple Pareto-optimal casualty-vehicle matching schemes are generated through this model. Based on the casualty-vehicle matching schemes, a vehicle routing model with time windows is established. With the objective of minimizing the sum of rescue time and penalty cost, an improved genetic algorithm is used to plan the globally optimal path sequence. The fitness function of the improved genetic algorithm is as follows: ; in, For the fitness function, Total travel time This is a penalty term for time windows and capacity constraints. The penalty coefficient is used. A multi-agent collaborative mechanism is introduced, treating vehicles using different rescue methods as independent agents. Task broadcasting, bidding, and evaluation are conducted through a contract network protocol. The target execution instructions for multi-capacity collaboration are determined based on simulation.

[0045] In one embodiment, real-time data on road network travel time, remaining capacity of medical points, and vehicle status are collected through multi-source sensing. Kalman filtering is used to fuse and estimate the state of the raw data to estimate the actual travel time and remaining capacity. A rule engine is constructed, and event triggering conditions are set to trigger rescheduling using an event-driven mechanism: a rescheduling event is triggered when the increase in the actual travel time from a preset baseline value exceeds a first threshold; or when the remaining capacity is lower than a second threshold; or when a vehicle fault signal is received.

[0046] In one embodiment, a rolling time-domain window length is set. After the rescheduling event is triggered, a multi-objective function is constructed to minimize time delay, mortality risk, and scheduling oscillations. A quadratic programming solver is used for online optimization within the rolling time domain to generate a preliminary adjustment plan. Vehicles and path nodes affected by the disturbance are identified, and an affected local subgraph is constructed. Incremental rescheduling is performed only on the paths and allocation relationships within the local subgraph, keeping the plan unchanged in the undisturbed areas. The adaptively adjusted rescue command is then output.

[0047] In one embodiment, such as Figure 3 As shown, the real-time control module details the system's mechanisms for responding to real-time changes and unexpected events: Perception and Fusion: GPS and sensor data are processed by Kalman filtering to estimate actual travel time and remaining capacity.

[0048] Event trigger: Establish a rule engine; when the passage time increases dramatically ( Insufficient capacity (remaining capacity) When a vehicle malfunctions, a replanning process is triggered.

[0049] MPC Core: A rolling time domain of 30-60 minutes is defined, constructing a multi-objective function that includes minimizing time delay, mortality risk, and rescheduling oscillations. A QP / SQP solver is employed, and an incremental update strategy is introduced to recalculate only the affected vehicle paths to improve efficiency.

[0050] Human-computer interaction and learning: The system allows for human intervention (locking specific variables) and records the interaction data for offline reinforcement learning to update the weights and prediction model of MPC.

[0051] In one embodiment, such as Figure 4 As shown, the optimized scheduling module breaks down the rescue process into three specific algorithmic levels, reflecting the process from macro-level matching to micro-level coordination. Specifically, these include: First layer (casualty-vehicle matching): A 0-1 integer programming problem is established to maximize the expected number of survivors, considering constraints such as capacity and infectious disease isolation. - The constraint method generates the Pareto optimal solution set.

[0052] The second layer (single-entity optimal path planning): employs an improved genetic algorithm. It searches for the optimal path sequence through a fitness function (time + penalty), selection, crossover (OX sequential crossover), mutation, and local search (2-Opt operator).

[0053] The third layer (multi-capacity collaboration): Introduces a multi-agent mechanism. Ambulances, helicopters, and civilian vehicles act as different agents. Through task broadcasting, bidding (calculating ETA (Estimated Time of Arrival) and Cost (execution costs, such as fuel consumption, time, and risks)), and contract network protocol evaluation, tasks are assigned to the most suitable agent. Finally, collaborative scheduling strategies are simulated in a multi-agent simulation environment, and various performance indicators, such as total rescue time and survival rate, are comprehensively evaluated to output the final instruction set to guide actual rescue operations.

[0054] In one embodiment, such as Figure 5 As shown, the injury assessment module demonstrates in detail how a dynamic priority score is derived from raw physiological data.

[0055] Phase 1 (Initial Quantification): Through fuzzy comprehensive evaluation, physiological parameters (R, C, M) are mapped to standardized scores of 0-25. Combined with expert scoring weights, an initial comprehensive score is obtained. .

[0056] Phase Two (Evolutionary Prediction): Predicting injury evolution using Markov Chains (SPM). Considering environmental factors (temperature and humidity), time decay factors, and individual factors, a dynamic transition matrix is ​​generated to extrapolate future survival probabilities. and deterioration rate .

[0057] Phase 3 (Dynamic Priority Adjustment): Construct a priority objective function, combining survival probability, deterioration rate, and waiting time to calculate the final score. If the score exceeds the threshold, a top-level alert is triggered; otherwise, the current priority is maintained (i.e., the order remains unchanged), and the final output updates the priority queue in real time.

[0058] It is worth noting that, based on Figures 3 to 5 The technical approach of this system can be summarized as a complete closed loop of "data-driven perception - hierarchical intelligent decision-making - real-time rolling control - closed-loop continuous evolution": (1) Multi-source sensing and data fusion layer The system first establishes a comprehensive data sensing network, using IoT technology to collect physiological data of the injured (blood pressure, heart rate), and combines this with drones and environmental sensors to acquire road condition and environmental data. Kalman filtering and particle filtering algorithms are then used to clean and estimate the state of the raw data, eliminating noise and accurately estimating the actual travel time. ) and remaining resource capacity ( This provides reliable input for decision-making.

[0059] (2) Intelligent injury assessment and classification (Module 1) After acquiring the data, the system's primary task is to determine the urgency of the rescue efforts. Module 1 employs a fuzzy comprehensive evaluation method to map nonlinear physiological indicators into quantified injury scores for initial assessment. Then, a Markov Chain prediction model is introduced, combining environmental factors (temperature and humidity) and individual differences to dynamically predict the survival probability curve and deterioration rate of the injured within future time windows. Finally, a dynamically updated priority queue is generated to ensure that critically injured patients receive resources first.

[0060] (3) Multi-level rescue resource optimization scheduling (Module 2) Module Two is the core decision-making layer of the system. For complex NP-hard problems like optimizing and scheduling rescue resources, a "layered solution" strategy is often adopted. Specifically, it includes: Layer One: Matching optimization (0-1 integer programming), solving the "who rides which vehicle" problem. With the goal of maximizing the survival rate, it considers hard constraints such as vehicle capacity and infectious disease isolation, utilizing... - The first layer uses a constraint method to find the Pareto optimal solution; the second layer uses a single-entity optimal path planning method (improved genetic algorithm) to solve the "how the vehicle goes" problem. For a given matching scheme, an improved genetic algorithm (incorporating 2-Opt local search) is used to plan the globally optimal path, avoiding congestion; the third layer uses multi-agent collaboration to solve the "how multiple forces cooperate" problem. Using the ContractNet Protocol, ambulances, helicopters, and other vehicles act as agents to bid for and negotiate tasks, achieving efficient integration of social transportation capacity.

[0061] (4) Dynamic response and real-time rolling control (Module 3) To address the rapidly changing conditions at rescue sites, Module 3 introduces model predictive control. The system performs real-time monitoring, immediately triggering replanning upon detecting drastic changes in road conditions or vehicle malfunctions. It employs rolling time-domain optimization, meaning it doesn't aim for a one-time global optimum, but rather performs online optimization within a 30-60 minute rolling time domain, aiming to minimize delays, mortality risks, and scheduling oscillations. Incremental computation ensures real-time performance (<30 seconds), with the system recalculating only the subgraph affected by the event, while other vehicles maintain their original plans, significantly reducing computational load.

[0062] To address two extreme problems: first, some lightly and moderately injured patients at high risk of deterioration rapidly develop into seriously injured individuals due to not being prioritized for transfer; second, some critically injured patients with extremely low survival rates consume a large amount of rescue resources, leading to a decline in overall rescue efficiency. Taking into account dynamic constraints such as the rate of deterioration of the injured, available waiting time for treatment, rescue distance, and hospital capacity, this study breaks through the limitations of fixed-priority scheduling to address how to dynamically adjust the order of injured patient transport based on real-time injury evolution, rescue resource status, and road network conditions, and establishes an adaptive priority decision-making mechanism that links "injury-time-resources".

[0063] Therefore, this invention breaks through the current common three-level static injury classification standard of light, moderate and severe in the field of emergency rescue. It comprehensively considers the continuous deterioration of injuries and the dynamic decline in survival probability caused by time delays, harsh environments and lack of basic medical treatment during the process of waiting for treatment, being stranded at the scene, and long-distance transfer. By constructing a dynamic quantitative model of injury that integrates time dimension, treatment delay and environmental factors, it can accurately characterize the rate of deterioration and the probability of state transition of patients with different injuries, shifting from "static classification" to "dynamic real-time assessment" and providing core data support for subsequent rescue. Secondly, it focuses on the dispatch of multiple rescue tools. By comprehensively dispatching professional ambulances, social volunteer vehicles, rescue helicopters and other rescue tools, and combining the differences of various tools in terms of capacity, speed, applicable scenarios and dispatch authority, it supplements the collaborative division of labor rules, dynamic capacity allocation and task allocation logic of multiple types of rescue forces. It solves the core problem of collaborative dispatch of multiple rescue methods, clarifies the applicable scenarios, dispatch priorities and task boundaries of different rescue tools, and alleviates the real contradiction of single resource capacity shortage and low overall resource utilization. Finally, by constructing a dynamic route optimization mechanism that takes into account the time window (golden treatment period), route accessibility, and resource saturation, the traditional route method based on static road network data and fixed resource information is optimized. Dynamic planning and iterative adjustment of the rescue route are carried out to achieve timely response to emergencies such as real-time traffic congestion, temporary road damage, and hospital resource saturation, so as to avoid the failure of the transfer route and the delay of treatment, thereby ensuring the timeliness and feasibility of the rescue route.

[0064] It should be understood that, although Figures 2-5 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figures 2-5 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0065] In one embodiment, such as Figure 6 As shown, a post-disaster casualty rescue decision optimization device is provided, including: a priority queue generation module 602, a rescue plan generation module 604, a rescheduling trigger module 606, and a rescue command output module 608, wherein: The priority queue generation module 602 is used to obtain the multidimensional physiological parameters and individual environmental factors of the injured, use the improved START classification method and fuzzy comprehensive evaluation to perform initial injury quantification, and use the Markov state transition model to predict the injury evolution trend to generate a dynamic priority queue containing survival probability and deterioration rate.

[0066] The rescue plan generation module 604 is used to construct a three-in-one collaborative optimization model of rescue order, rescue route and rescue method. With the goal of maximizing the expected number of survivors, it adopts a decomposition-coordination strategy to solve the problem together and generate a globally optimized rescue plan.

[0067] The rescheduling trigger module 606 is used to collect dynamic environmental data in real time through multi-source sensing and trigger rescheduling using an event-driven mechanism.

[0068] The rescue command output module 608 is used to perform rolling time-domain optimization based on the predictive control of the collaborative optimization model, and to replan only the local scheme affected by the disturbance in an incremental update manner, and output the adaptively adjusted rescue command.

[0069] In one embodiment, the rescue plan generation module 604 is further configured to incorporate the survival probability of the injured, vehicle capacity, and road network travel time into a 0-1 integer programming model, using the maximization of the expected number of survivors as the objective function, and solving it using the ε-constraint method to construct an injured-vehicle allocation model. Multiple Pareto-optimal injured-vehicle matching schemes are generated through this model. Based on the injured-vehicle matching schemes, a vehicle path planning model with a time window is established, aiming to minimize the sum of rescue time and penalty cost. An improved genetic algorithm is used to plan the globally optimal path sequence. The fitness function of the improved genetic algorithm is: ; in, For the fitness function, Total travel time This is a penalty term for time windows and capacity constraints. As a penalty coefficient, a multi-agent collaborative mechanism is introduced, treating vehicles with different rescue methods as independent intelligent agents. Through the contract network protocol, tasks are broadcast, bid, and evaluated. The target execution instructions for multi-capacity collaboration are determined based on simulation.

[0070] For specific limitations regarding the post-disaster casualty rescue decision optimization device, please refer to the limitations of the post-disaster casualty rescue decision optimization method described above, which will not be repeated here. Each module in the aforementioned post-disaster casualty rescue decision optimization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0071] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for optimizing disaster relief and rescue decisions. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0072] Those skilled in the art will understand that Figures 6-7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0073] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to perform the following steps: We acquire multidimensional physiological parameters and individual environmental factors of the injured, use an improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use a Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate.

[0074] A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem jointly and generate a globally optimized rescue plan.

[0075] Real-time environmental dynamic data is collected through multi-source sensing, and rescheduling is triggered by an event-driven mechanism.

[0076] Rolling time-domain optimization is performed based on predictive control of the collaborative optimization model. Incremental updates are used to replan only the local schemes affected by disturbances, and adaptively adjusted rescue commands are output.

[0077] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: We acquire multidimensional physiological parameters and individual environmental factors of the injured, use an improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use a Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate.

[0078] A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem jointly and generate a globally optimized rescue plan.

[0079] Real-time environmental dynamic data is collected through multi-source sensing, and rescheduling is triggered by an event-driven mechanism.

[0080] Rolling time-domain optimization is performed based on predictive control of the collaborative optimization model. Incremental updates are used to replan only the local schemes affected by disturbances, and adaptively adjusted rescue commands are output.

[0081] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchlink, DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0082] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0083] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for optimizing decision-making in the rescue of injured people after a disaster, characterized in that, The method includes: We acquire multidimensional physiological parameters and individual environmental factors of the wounded, use an improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use a Markov state transition model to predict the trend of injury evolution, generating a dynamic priority queue that includes survival probability and deterioration rate. A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed. With the goal of maximizing the expected number of survivors, a decomposition-coordination strategy is used to solve the problem jointly and generate a globally optimized rescue plan. Real-time environmental dynamic data is collected through multi-source sensing, and rescheduling is triggered by an event-driven mechanism. Based on the predictive control of the collaborative optimization model, rolling time-domain optimization is performed, and incremental updates are used to replan only the local schemes affected by disturbances, outputting adaptively adjusted rescue commands.

2. The method according to claim 1, characterized in that, Multidimensional physiological parameters and individual environmental factors of the injured were obtained. An improved START classification method and fuzzy comprehensive evaluation were used for initial injury quantification. A Markov state transition model was used to predict the evolution trend of the injury, and a dynamic priority queue containing survival probability and deterioration rate was generated, including: The study obtains multidimensional physiological parameters of the patient's respiratory, circulatory and consciousness status, as well as individual environmental factors such as ambient temperature and humidity. Through fuzzy comprehensive evaluation, nonlinear physiological indicators are mapped into an initial quantitative score of injury. The initial injury quantification score is used as the initial state vector of the Markov state transition model. A dynamic transition matrix is ​​constructed by combining environmental factors and time decay factors to infer the survival probability curve and deterioration rate for multiple future time windows. Based on the survival probability curve and the deterioration rate, a priority scoring function is constructed: in, Rate the priority. , , These are the weighting coefficients, To predict the probability of survival, To predict the rate of deterioration, Waiting time; The predicted injury evolution results corresponding to each injury quantification score are prioritized and sorted according to the priority scoring function to obtain a dynamic priority queue.

3. The method according to claim 1, characterized in that, The collaborative optimization model includes: a casualty-vehicle allocation model and a vehicle routing model; Construct a three-in-one collaborative optimization model integrating rescue order, rescue route, and rescue method, including: The survival probability of the wounded, vehicle capacity, and road network travel time are incorporated into a 0-1 integer programming model. The objective function is to maximize the expected number of survivors. The ε-constraint method is used to solve the problem and construct the wounded-vehicle allocation model. Multiple Pareto optimal wounded-vehicle matching schemes are generated through the wounded-vehicle allocation model. Based on the aforementioned casualty-vehicle matching scheme, a vehicle path planning model with a time window is established. With the goal of minimizing the sum of rescue time and penalty cost, an improved genetic algorithm is used to plan the globally optimal path sequence. The fitness function of the improved genetic algorithm is: in, For the fitness function, Total travel time This is a penalty term for time windows and capacity constraints. This is the penalty coefficient; A multi-agent collaborative mechanism is introduced, treating vehicles with different rescue methods as independent intelligent agents. Task broadcasting, bidding, and evaluation are conducted through the contract network protocol, and the target execution instructions for multi-capacity collaboration are determined based on simulation.

4. The method according to claim 3, characterized in that, Real-time collection of dynamic environmental data through multi-source sensing, and triggering rescheduling using an event-driven mechanism, including: By collecting real-time data on road network travel time, remaining capacity of medical points and vehicle status through multi-source sensing, and using Kalman filtering to fuse the raw data and estimate the status, the actual travel time and remaining capacity are estimated. A rule engine is built, event triggering conditions are set, and an event-driven mechanism is used to trigger rescheduling: when the increase of the actual passage time from the preset baseline value exceeds the first threshold, a rescheduling event is triggered; or when the remaining capacity is lower than the second threshold, a rescheduling event is triggered; or when a vehicle fault signal is received, a rescheduling event is triggered.

5. The method according to claim 4, characterized in that, Based on the predictive control of the collaborative optimization model, rolling time-domain optimization is performed, and incremental updates are used to replan only the local schemes affected by disturbances, outputting adaptively adjusted rescue commands, including: Set the length of the rolling time domain window. After the rescheduling event is triggered, construct a multi-objective function with the goal of minimizing time delay, death risk and scheduling oscillation. Use a quadratic programming solver to optimize it online in the rolling time domain and generate a preliminary adjustment plan. Identify vehicles and path nodes affected by disturbances, construct affected local subgraphs, perform incremental replanning only on paths and allocation relationships within the local subgraphs, keep the schemes of undisturbed areas unchanged, and output adaptively adjusted rescue instructions.

6. A decision optimization device for post-disaster casualty rescue, characterized in that, The device includes: The priority queue generation module is used to obtain the multidimensional physiological parameters and individual environmental factors of the wounded, use the improved START classification method and fuzzy comprehensive evaluation to quantify the initial injury, and use the Markov state transition model to predict the trend of injury evolution, and generate a dynamic priority queue containing survival probability and deterioration rate. The rescue plan generation module is used to construct a three-in-one collaborative optimization model of rescue order, rescue route and rescue method. With the goal of maximizing the expected number of survivors, it adopts a decomposition-coordination strategy to solve the problem together and generate a globally optimized rescue plan. The rescheduling trigger module is used to collect dynamic environmental data in real time through multi-source sensing and trigger rescheduling using an event-driven mechanism. The rescue command output module is used to perform rolling time-domain optimization based on the predictive control of the collaborative optimization model, and to replan only the local scheme affected by the disturbance in an incremental update manner, and output the adaptively adjusted rescue command.

7. The apparatus according to claim 6, characterized in that, The rescue plan generation module is also used to incorporate the survival probability of the injured, vehicle capacity, and road network travel time into a 0-1 integer programming model, with the objective function of maximizing the expected number of survivors. The ε-constraint method is used to solve the problem and construct an injured-vehicle allocation model. Multiple Pareto-optimal injured-vehicle matching schemes are generated through the injured-vehicle allocation model. Based on the injured-vehicle matching schemes, a vehicle path planning model with time windows is established. With the objective of minimizing the sum of rescue time and penalty cost, an improved genetic algorithm is used to plan the globally optimal path sequence. The fitness function of the improved genetic algorithm is: in, For the fitness function, Total travel time This is a penalty term for time windows and capacity constraints. As a penalty coefficient, a multi-agent collaborative mechanism is introduced, treating vehicles with different rescue methods as independent intelligent agents. Through the contract network protocol, tasks are broadcast, bid, and evaluated. The target execution instructions for multi-capacity collaboration are determined based on simulation.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images