Emergency resource scheduling and path planning method

Abstract

The invention relates to the technical field of emergency response, in particular to an emergency resource scheduling and path planning method, which comprises the following steps that a dynamic weight evaluation system is constructed based on emergency resource scheduling factors, and the weight evaluation system can dynamically calculate and adjust factor weights so as to determine that the most critical resource is scheduled to the most required place under the emergency condition; and constructing a deep reinforcement learning model, and using historical emergency response data as a training set, so that the intelligent agent learns and optimizes a strategy of selecting an optimal path under different conditions in a simulated or historical emergency scene. Efficient emergency response: through comprehensive application of a dynamic weight evaluation system and a deep reinforcement learning model, the system can quickly identify and preferentially schedule the most critical resources to the most needed place, and meanwhile, the optimal path is planned, so that the emergency response speed and efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of emergency response technology, and in particular to an emergency resource scheduling and path planning method. Background Art

[0002] Traditional emergency resource scheduling and path planning methods often rely on manual decision-making or fixed algorithm models. These methods often find it difficult to achieve efficient, flexible and accurate scheduling when faced with complex and changing emergency scenarios.

[0003] First, the dispatch of emergency resources is affected by many factors, including time urgency, resource demand urgency, traffic congestion, weather impact, disaster severity, number of affected people, and road capacity. These factors change dynamically during the emergency response process, and the weights of different factors will also be adjusted as the situation develops. The traditional fixed weight evaluation system is difficult to accurately reflect this dynamic change, resulting in limited optimization of resource dispatch strategies.

[0004] Secondly, in terms of path planning, traditional algorithm models are often based on static road networks and fixed traffic conditions for planning, which makes it difficult to cope with unexpected situations such as traffic congestion and road closures that may occur in actual emergency response. This will result in the planned path being suboptimal and may even fail to reach the target location, thus affecting rescue efficiency.

[0005] Therefore, there is a need for an emergency resource scheduling and path planning method that can dynamically evaluate factor weights, intelligently plan the optimal path, effectively integrate multi-source heterogeneous data, provide real-time feedback on scheduling effects and continuously optimize scheduling strategies, as well as a system required to implement this method. Summary of the invention

[0006] The present invention aims at the technical problems existing in the prior art and provides an emergency resource scheduling and path planning method to solve the problem that the existing traditional emergency resource scheduling and path planning methods are often difficult to achieve efficient, flexible and accurate scheduling when faced with complex and changeable emergency scenarios.

[0007] The technical solution of the present invention to solve the above technical problems is as follows: an emergency resource scheduling and path planning method, comprising the following steps:

[0008] Building a dynamic weight evaluation system based on emergency resource dispatch factors, which can dynamically calculate and adjust factor weights to determine the dispatch of the most critical resources to the most needed locations in an emergency;

[0009] Build a deep reinforcement learning model, using historical emergency response data as a training set, so that the agent can learn and optimize the strategy of choosing the best path under different conditions in simulated or historical emergency scenarios;

[0010] Build a data preprocessing module to integrate multi-source heterogeneous data, extract key information, and provide real-time and accurate input for dynamic weight evaluation and reinforcement learning models;

[0011] Combining the output results of the dynamic weight evaluation system and the reinforcement learning model, an adaptive scheduling algorithm is designed, which is used to quickly adjust the resource allocation strategy according to the current situation.

[0012] Based on the above technical solution, the present invention can also be improved as follows.

[0013] Furthermore, the emergency resource scheduling factors include time urgency, resource demand urgency, traffic congestion, weather impact and disaster severity, number of affected people, and road capacity.

[0014] Furthermore, the dynamic weight evaluation system includes a weight adaptive adjustment mechanism, which is used to automatically adjust the weights of emergency resource scheduling factors according to real-time changes in emergency events to dynamically optimize resource scheduling strategies.

[0015] Furthermore, the construction of the dynamic weight evaluation system includes the following steps:

[0016] Define a set of key factors for emergency resource dispatch;

[0017] Set an initial weight for each key factor and obtain the latest data of key factors in real time;

[0018] Based on real-time data, the weight of each factor is dynamically adjusted using a specific algorithm;

[0019] Calculate the score of each scheduling plan based on the adjusted weights and real-time data, and select the plan with the highest score for execution;

[0020] Furthermore, the entropy weight method is used to calculate the basic weight, including the following process:

[0021] For each factor, its real-time data is standardized to the interval [0,1], denoted as X'i (i=1,2,...,7);

[0022] For each factor, calculate its information entropy Ei, the formula is:

[0023]

[0024] Where n is the number of scheduling options, p ij is the proportion of the standardized value of the i-th factor in the j-th solution to the total value of the factor;

[0025] Redundancy di = 1-Ei, which represents the information utility value of the factor;

[0026] Calculate the basic weight W' of each factor according to the redundancy i , the formula is:

[0027]

[0028] Define the feedback indicator F according to the actual situation of emergency response;

[0029] According to the comparison between the feedback index F and the preset target value, the weight adjustment coefficient α is calculated, and the formula is:

[0030]

[0031] Among them, F min and F max are the minimum and maximum values ​​of the feedback indicators respectively;

[0032] According to the weight adjustment coefficient α and the basic weight W' i , calculate the adjusted weight W" i , the formula is

[0033] W″ i =αW' i +(1-α)W i

[0034] Among them, W i is the initial weight;

[0035] According to the adjusted weight W" i and real-time data X' j , calculate the score S of each scheduling solution i , the formula is:

[0036]

[0037] The solution with the highest score is selected as the final scheduling decision;

[0038] T is time urgency, which indicates the time pressure to respond as quickly as possible after an emergency occurs;

[0039] R is the urgency of resource demand, which indicates the urgency of the resource demand of the emergency event;

[0040] C is the traffic congestion condition, which indicates the traffic obstacles that may be encountered during the emergency response process;

[0041] W is weather impact, which indicates the potential impact of weather conditions on emergency response;

[0042] D is the severity of the disaster, which indicates the destructive power and impact range of the emergency event;

[0043] P is the number of affected population, indicating the number of people directly affected by the emergency event;

[0044] H is the road capacity, which represents the traffic efficiency and reliability of the road during the emergency response process.

[0045] Furthermore, the deep reinforcement learning model uses historical emergency response data as a training set, allowing the agent to learn and optimize the selection of the optimal path under different conditions in simulated or historical emergency scenarios.

[0046] The deep reinforcement learning model selects the optimal path including the following steps:

[0047] Identify the various states in the emergency response process and all possible actions that can be taken;

[0048] Construct a reward function to evaluate the immediate reward after the agent takes an action in a given state;

[0049] Build a deep neural network as the agent's policy network to predict the best action based on the current state;

[0050] Use trained deep neural networks for path planning in simulated or historical emergency scenarios;

[0051] Calculate the efficiency and effectiveness of emergency response based on the path and resource allocation plan selected by the agent;

[0052] Based on the simulation results, the deep neural network is fine-tuned to further optimize the path selection strategy;

[0053] In actual emergency response, the deep neural network is input according to the current status to obtain the optimal path and resource allocation plan;

[0054] Execute the selected path and resource allocation plan to respond to emergencies.

[0055] Furthermore, the data preprocessing module also includes a data fusion and verification unit, and the data fusion and verification unit is used to integrate data from different sources.

[0056] Further, the adaptive scheduling algorithm includes:

[0057] Based on the output of the dynamic weight evaluation system, the resource scheduling strategy is initialized to determine the preliminary resource allocation and path planning;

[0058] Monitor key data in the emergency response process in real time, and dynamically update relevant parameters in the resource scheduling strategy based on real-time data;

[0059] Adopt an adaptive adjustment mechanism to dynamically adjust resource scheduling strategies based on real-time data and historical experience;

[0060] Define scheduling effect evaluation indicators, and calculate the effect score of the current scheduling strategy based on the evaluation indicators;

[0061] If the effect score of the current scheduling strategy is lower than the preset threshold, the strategy is iterated and optimized;

[0062] Outputting the final optimized resource scheduling strategy, wherein the resource scheduling strategy includes resource allocation and path planning;

[0063] The resource allocation adjustment formula is:

[0064] R alloc (t) = R base +ΔR(t)

[0065] Among them, R alloc (t) represents the resource allocation at time t;

[0066] R base Indicates the basic resource allocation amount;

[0067] ΔR(t) represents the resource allocation increment adjusted according to real-time data;

[0068] The path planning adjustment formula is:

[0069]

[0070] Among them, Path opt (t) represents the optimal path at time t;

[0071] PathSct represents the set of all possible paths;

[0072] Time(P,t) represents the rescue time required along path P at time t;

[0073] Cost(P) represents the cost of path P;

[0074] λ is the weight coefficient.

[0075] Furthermore, it also includes an emergency response effect evaluation and feedback mechanism, which is used to compare actual scheduling results with expected goals, to evaluate the effectiveness of resource scheduling and path planning, and to feed back the evaluation results to the dynamic weight evaluation system and reinforcement learning model.

[0076] Furthermore, the emergency response effect evaluation and feedback mechanism also includes a user feedback interface, which allows emergency response personnel or decision makers to provide direct feedback on the scheduling results.

[0077] Furthermore, it also includes a visual report generation module, which is used to automatically generate a visual report containing one or more key information such as scheduling strategy, resource allocation, path planning, response time, and rescue effect.

[0078] The beneficial effects of the present invention are:

[0079] Efficient emergency response: Through the comprehensive application of dynamic weight evaluation system and deep reinforcement learning model, the system can quickly identify and prioritize the most critical resources to the most needed locations, while planning the optimal path, significantly improving the speed and efficiency of emergency response.

[0080] Flexibility and pertinence: The dynamic weight evaluation system can automatically adjust the weight of factors according to changes in the actual emergency situation, ensuring the flexibility and pertinence of resource scheduling. This dynamism enables the system to cope with various complex and changing emergency environments and achieve effective resource allocation.

[0081] Data-driven decision-making: The data preprocessing module integrates multi-source heterogeneous data and ensures the accuracy and consistency of the data through data fusion and verification units, providing high-quality input for dynamic weight evaluation and reinforcement learning models. This makes the system's decisions more data-based, scientific and reliable.

[0082] Adaptive scheduling capability: Adaptive scheduling algorithms can dynamically adjust resource scheduling strategies based on real-time data and historical experience to ensure that resources can be dispatched to where they are most needed in the most efficient way. This adaptability enhances the robustness and adaptability of the system, enabling it to flexibly respond to various emergencies.

[0083] Continuous optimization and evolution: The emergency response effect evaluation and feedback mechanism comprehensively evaluates the effectiveness of resource scheduling and path planning by comparing the actual scheduling results with the expected goals, and feeds the evaluation results back to the dynamic weight evaluation system and reinforcement learning model to continuously optimize the scheduling strategy. This continuous feedback and optimization cycle promotes the continuous evolution of the entire system and improves its ability to respond to various emergency situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 It is a schematic diagram of the principle of the emergency resource scheduling and path planning method of the present invention. DETAILED DESCRIPTION

[0085] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0086] like Figure 1 As shown, a method for emergency resource scheduling and path planning includes the following steps:

[0087] Building a dynamic weight evaluation system based on emergency resource dispatch factors, which can dynamically calculate and adjust factor weights to determine the dispatch of the most critical resources to the most needed locations in an emergency;

[0088] Build a deep reinforcement learning model, using historical emergency response data as a training set, so that the agent can learn and optimize the strategy of choosing the best path under different conditions in simulated or historical emergency scenarios;

[0089] Build a data preprocessing module to integrate multi-source heterogeneous data, extract key information, and provide real-time and accurate input for dynamic weight evaluation and reinforcement learning models;

[0090] Combining the output results of the dynamic weight evaluation system and the reinforcement learning model, an adaptive scheduling algorithm is designed, which is used to quickly adjust the resource allocation strategy according to the current situation.

[0091] This method comprehensively applies a dynamic weight evaluation system, a deep reinforcement learning model, a data preprocessing module, and an adaptive scheduling algorithm. Its core principles can be summarized as follows:

[0092] Dynamic Weight Assessment System: This approach first establishes an assessment system that can dynamically calculate and adjust the weights of factors. This system can identify and assess which resources are most critical at a specific time and place based on the specific needs of an emergency. This dynamism ensures flexibility and targeting of resource scheduling, allowing resources to be most effectively allocated to where they are most needed.

[0093] Deep reinforcement learning model: Using historical emergency response data as a training set, this method builds a deep reinforcement learning model. This model simulates or reviews historical emergency scenarios to allow the agent to learn and optimize the strategy of choosing the best path under different conditions. In this way, the model can gradually master the ability to make the best decisions in complex and changing emergency environments.

[0094] Data preprocessing module: To support the effective operation of dynamic weight evaluation and reinforcement learning models, this method also builds a data preprocessing module. This module is responsible for integrating heterogeneous data from multiple different sources and extracting key information from them. In this way, it can provide real-time and accurate input data for the above two core components, ensuring that they can make decisions based on the latest and most comprehensive information.

[0095] Adaptive Scheduling Algorithm: Finally, the method combines the output of the dynamic weight evaluation system and the reinforcement learning model to design an adaptive scheduling algorithm. This algorithm can quickly adjust the resource allocation strategy according to the current situation to ensure that resources can be dispatched to where they are most needed in the most efficient way. This adaptability enables the method to flexibly respond to various emergencies and ensure the timeliness and effectiveness of emergency response.

[0096] As an implementation method, the emergency resource scheduling factors include time urgency, resource demand urgency, traffic congestion, weather impact and disaster severity, number of affected people, and road capacity.

[0097] As an implementation method, the dynamic weight evaluation system includes a weight adaptive adjustment mechanism, which is used to automatically adjust the weights of emergency resource scheduling factors according to real-time changes in emergency events (specifically, such as the speed of disaster spread, newly affected areas, etc.) to dynamically optimize resource scheduling strategies.

[0098] Specifically, the construction of the dynamic weight evaluation system includes the following steps:

[0099] Define a set of key factors for emergency resource dispatch; (including time urgency (T), resource demand urgency (R), traffic congestion (C), weather impact (W), disaster severity (D), number of affected people (P) and road capacity (H))

[0100] Set an initial weight for each key factor and obtain the latest data of key factors in real time;

[0101] Based on real-time data, the weight of each factor is dynamically adjusted using a specific algorithm;

[0102] Calculate the score of each scheduling plan based on the adjusted weights and real-time data, and select the plan with the highest score for execution;

[0103] Furthermore, the entropy weight method is used to calculate the basic weight, including the following process:

[0104] For each factor, its real-time data is standardized to the interval [0,1], denoted as X'i (i=1,2,...,7);

[0105] For each factor, calculate its information entropy Ei, the formula is:

[0106]

[0107] Where n is the number of scheduling solutions, p ij is the proportion of the standardized value of the i-th factor in the j-th solution to the total value of the factor;

[0108] Redundancy di = 1-Ei, which represents the information utility value of the factor;

[0109] Calculate the basic weight W' of each factor according to the redundancy i , the formula is:

[0110]

[0111] According to the actual situation of emergency response, define the feedback index F; (such as rescue efficiency, reduction ratio of affected population, etc.);

[0112] According to the comparison between the feedback index F and the preset target value, the weight adjustment coefficient α is calculated, and the formula is:

[0113]

[0114] Among them, F min and F max are the minimum and maximum values ​​of the feedback indicators respectively;

[0115] According to the weight adjustment coefficient α and the basic weight W' i , calculate the adjusted weight W" i , the formula is

[0116] W″ i =αW' i +(1-α)W i

[0117] Among them, W i is the initial weight;

[0118] According to the adjusted weight W" i and real-time data X' j , calculate the score S of each scheduling solution i , the formula is:

[0119]

[0120] The solution with the highest score is selected as the final scheduling decision;

[0121] T is time urgency, which indicates the time pressure to respond as quickly as possible after an emergency occurs;

[0122] R is the urgency of resource demand, which indicates the urgency of the resource demand of the emergency event;

[0123] C is the traffic congestion condition, which indicates the traffic obstacles that may be encountered during the emergency response process;

[0124] W is weather impact, which indicates the potential impact of weather conditions on emergency response;

[0125] D is the severity of the disaster, which indicates the destructive power and impact range of the emergency event;

[0126] P is the number of affected population, indicating the number of people directly affected by the emergency event;

[0127] H is the road capacity, which represents the traffic efficiency and reliability of the road during the emergency response process.

[0128] As an implementation method, the deep reinforcement learning model uses historical emergency response data as a training set, allowing the agent to learn and optimize the selection of the optimal path under different conditions in simulated or historical emergency scenarios.

[0129] Specifically, the deep reinforcement learning model selects the optimal path including the following steps:

[0130] Identify the various states in the emergency response process and all possible actions that can be taken (i.e., choose different paths or resource allocation plans)

[0131] Construct a reward function to evaluate the immediate reward after the agent takes an action in a given state;

[0132] The reward function should take into account factors such as the efficiency of emergency response (such as rescue time, resource utilization efficiency), the rationality of resource use (such as avoiding resource waste and ensuring sufficient resources), and the degree of reduction in the affected population (such as the number of rescued people and the reduction in casualties).

[0133] Build a deep neural network as the agent's policy network to predict the best action based on the current state;

[0134] Using historical emergency response data as a training set, the neural network is trained through deep reinforcement learning algorithms (such as Actor-Critic, PPO, etc.) to enable it to learn and optimize the strategy of selecting the optimal path under different conditions.

[0135] Use trained deep neural networks for path planning in simulated or historical emergency scenarios;

[0136] Calculate the efficiency and effectiveness of emergency response based on the path and resource allocation plan selected by the agent;

[0137] Based on the simulation results, the deep neural network is fine-tuned to further optimize the path selection strategy;

[0138] In actual emergency response, the deep neural network is input according to the current status to obtain the optimal path and resource allocation plan;

[0139] Execute the selected path and resource allocation plan to respond to emergencies.

[0140] The examples are as follows:

[0141] In order to describe in detail the actual operation process of the deep reinforcement learning model selecting the optimal path, the present invention also supplements the following embodiments:

[0142] This embodiment uses a deep reinforcement learning model, combined with historical emergency response data, to optimize the strategy of selecting the optimal path under different conditions.

[0143] Specific implementation steps

[0144] State definition and feature extraction

[0145] During the emergency response process, the following state variables are defined:

[0146] Current Location: The current coordinates of the emergency vehicle or resource.

[0147] Target: The coordinates of the location where rescue or resource allocation is required.

[0148] Time urgency (T): Remaining rescue time window, unit: hours.

[0149] Resource requirement urgency (R): The urgency of resource requirements at the target location, range: [0,1].

[0150] Traffic congestion condition (C): The current road congestion level, range: [0,1], 0 means unobstructed, 1 means severe congestion.

[0151] Weather impact (W): The impact of the current weather on emergency response, range: [0,1], 0 means no impact, 1 means severe impact.

[0152] Disaster severity (D): the degree of damage caused by the disaster, range: [0,1].

[0153] Affected population (P): The number of people affected by the target location.

[0154] Road capacity (H): The current road capacity, unit: vehicles / hour.

[0155] For non-numerical data (such as weather impact and disaster severity), one-hot encoding or normalization is used to convert them into numerical features.

[0156] Action Space Construction

[0157] The action space is defined as all possible paths and resource allocation schemes that the agent can take. Each action is represented by a path selection (Path) and a resource allocation (Resource), that is:

[0158] Path: The path that the emergency vehicle or resource takes from its current location to its target location.

[0159] Resource: The amount or type of resources allocated to the target location.

[0160] Reward function design

[0161] The reward function comprehensively considers the efficiency of emergency response, the rationality of resource use, and the reduction of the affected population, and is defined as follows:

[0162]

[0163] T rescue time: the time required for the agent to reach the target location and complete the rescue after taking action a from the current state s.

[0164] Ruse: The amount of resources allocated by the agent in action a.

[0165] RTotal demand: The total resource demand at the target location.

[0166] P rescued: The number of people that the agent successfully rescued in action a.

[0167] P affected: The total number of people affected at the target location.

[0168] α, β, γ: are the weight coefficients of time, resources and population factors respectively, and α, β, γ>0.

[0169] Deep neural network construction and training

[0170] Build a deep neural network as the agent's policy network, with the input layer receiving state variables and the output layer predicting the best action. Use historical emergency response data as a training set and train the neural network using deep reinforcement learning algorithms (such as Actor-Critic, PPO, etc.).

[0171] During the training process, the agent continuously tries different paths and resource allocation schemes in simulated or historical emergency scenarios and evaluates their effects based on the reward function. The weights of the neural network are adjusted through the back-propagation algorithm so that it gradually learns the strategy of selecting the optimal path under different conditions.

[0172] Simulation and Optimization

[0173] In simulated or historical emergency scenarios, the trained deep neural network is used for path planning. Based on the path and resource allocation plan selected by the agent, the efficiency and effectiveness of the emergency response, such as rescue time, resource utilization efficiency, and number of rescued people, are calculated.

[0174] Based on the simulation results, the deep neural network is fine-tuned to further optimize the path selection strategy. During the fine-tuning process, the neural network architecture, learning rate, batch size and other hyperparameters can be adjusted to improve the generalization ability and accuracy of the model.

[0175] Actual emergency response

[0176] In actual emergency response, the deep neural network is input according to the current state to obtain the optimal path and resource allocation plan. The selected path and resource allocation plan are executed to carry out emergency response. At the same time, data from the emergency response process is collected in real time for subsequent model updating and optimization.

[0177] As an embodiment, the data preprocessing module also includes a data fusion and verification unit, which is used to integrate data from different sources (such as satellite images, social media feedback, sensor data, etc., and ensure the accuracy and consistency of the data through a data verification algorithm to provide high-quality input for dynamic weight evaluation and reinforcement learning models).

[0178] As a key component of the data preprocessing module, the data fusion and verification unit has the core function of integrating data resources from different sources, such as satellite images, social media feedback, and sensor data. These data sources each have unique values ​​and perspectives, but there may also be problems such as format differences, redundancy, or inconsistency. Therefore, the data fusion and verification unit first needs to effectively integrate these data to ensure that they can form a unified and comprehensive data set.

[0179] During the data integration process, the data fusion and verification unit will use a series of data verification algorithms. The purpose of these algorithms is to verify the accuracy and consistency of the data and eliminate erroneous or invalid information, thereby ensuring that the data provided to the dynamic weight evaluation and reinforcement learning model is high-quality and reliable. Through this process, the data preprocessing module can provide a solid foundation for subsequent decision support and improve the effectiveness and accuracy of emergency resource scheduling and path planning.

[0180] As an implementation method, the adaptive scheduling algorithm includes:

[0181] Based on the output of the dynamic weight evaluation system, the resource scheduling strategy is initialized to determine the preliminary resource allocation and path planning;

[0182] Real-time monitoring of key data in the emergency response process, including time urgency, resource demand urgency, traffic congestion, weather impact, disaster severity, number of affected people, and road capacity, etc., and dynamically updating relevant parameters in the resource scheduling strategy based on real-time data;

[0183] Adopt an adaptive adjustment mechanism to dynamically adjust resource scheduling strategies based on real-time data and historical experience;

[0184] Specific adjustment strategies include: when time urgency increases, prioritize resources closest to the target location; when resource demand urgency increases, increase the amount of resources allocated to the target location; when traffic congestion worsens, choose a detour route, etc.

[0185] Define dispatch effect evaluation indicators, such as rescue time, resource utilization efficiency, number of rescuers, etc.; calculate the effect score of the current dispatch strategy based on the evaluation indicators;

[0186] If the effect score of the current scheduling strategy is lower than the preset threshold, the strategy is iterated and optimized;

[0187] Adopt reinforcement learning algorithms to fine-tune scheduling strategies based on historical data and real-time feedback to improve their adaptability and accuracy;

[0188] Outputting the final optimized resource scheduling strategy, wherein the resource scheduling strategy includes resource allocation and path planning;

[0189] The resource allocation adjustment formula is:

[0190] R alloc (t) = R base +ΔR(t)

[0191] Among them, R alloc (t) represents the resource allocation at time t;

[0192] R base Indicates the basic resource allocation amount;

[0193] ΔR(t) represents the resource allocation increment adjusted according to real-time data;

[0194] The path planning adjustment formula is:

[0195]

[0196] Among them, Path opt (t) represents the optimal path at time t;

[0197] PathSct represents the set of all possible paths;

[0198] Time(P,t) represents the rescue time required along path P at time t;

[0199] Cost(P) represents the cost of path P;

[0200] λ is the weight coefficient.

[0201] The present invention is further supplemented with the following embodiments:

[0202] This embodiment describes in detail the specific operation process of the adaptive scheduling algorithm.

[0203] Step a: Scheduling strategy initialization

[0204] According to the output of the dynamic weight evaluation system, the preliminary resource allocation and path planning are determined. For example, if the time urgency is high, the resources closest to the target location are prioritized; if the resource demand urgency is high, the amount of resources allocated to the target location is increased.

[0205] Step b: Real-time data monitoring and updating

[0206] Real-time monitoring of key data in the emergency response process, such as time urgency, resource demand urgency, etc., and dynamic update of relevant parameters in the resource scheduling strategy based on real-time data. For example, when time urgency increases, timely adjustment of resource scheduling strategy, giving priority to scheduling resources closer to the target location.

[0207] Step c: Strategy adjustment and optimization

[0208] Adopt an adaptive adjustment mechanism to dynamically adjust resource scheduling strategies based on real-time data and historical experience. For example, when traffic congestion worsens, a detour route is selected to shorten rescue time; when the urgency of resource demand increases, the amount of resources allocated to the target location is increased to ensure rescue effectiveness.

[0209] Step d: Scheduling effect evaluation

[0210] Define dispatch effect evaluation indicators, such as rescue time, resource utilization efficiency, number of rescuers, etc., and calculate the effect score of the current dispatch strategy based on the evaluation indicators. For example, if the rescue time is short and the resource utilization efficiency is high, the effect score of the current dispatch strategy is high.

[0211] Step e: Strategy iteration and optimization

[0212] If the effectiveness score of the current scheduling strategy is lower than the preset threshold, the strategy is iterated and optimized. The reinforcement learning algorithm is used to fine-tune the scheduling strategy based on historical data and real-time feedback to improve its adaptability and accuracy. For example, by adjusting the parameters in the resource allocation adjustment formula and the path planning adjustment formula, the resource allocation and path planning strategies are optimized.

[0213] Step f: Scheduling decision output

[0214] Output the final optimized resource scheduling strategy, including resource allocation and path planning. Send the optimized scheduling strategy to emergency responders or decision makers so that they can respond quickly and optimize resource scheduling and path planning. At the same time, collect data from the emergency response process in real time for subsequent model updates and optimization.

[0215] As an implementation method, it also includes an emergency response effect evaluation and feedback mechanism, which is used to compare the actual scheduling results with the expected goals, to evaluate the effectiveness of resource scheduling and path planning, and to feed back the evaluation results to the dynamic weight evaluation system and reinforcement learning model to continuously optimize the scheduling strategy.

[0216] This mechanism aims to comprehensively evaluate the effectiveness of resource scheduling and path planning by comparing actual scheduling results with expected goals. This evaluation process not only focuses on whether the scheduling results have achieved the expected emergency response effect, but also deeply analyzes the performance of the scheduling strategy in actual execution.

[0217] After the evaluation is completed, the mechanism will provide detailed feedback to the dynamic weight evaluation system and reinforcement learning model. For the dynamic weight evaluation system, feedback helps it adjust the weights of various factors according to the actual scheduling effect, ensuring that the most critical resources can be more accurately identified and prioritized in subsequent scheduling. For the reinforcement learning model, feedback serves as an important basis for its learning and optimization, helping it learn lessons from actual scheduling experience and continuously revise and improve the strategy of selecting the optimal path under different conditions.

[0218] Through this continuous feedback and optimization cycle, the emergency response effectiveness evaluation and feedback mechanism can drive the continuous evolution of the entire resource scheduling and path planning system, enhance its ability to respond to various emergencies, and ensure that resources can be efficiently and accurately dispatched to where they are most needed in the shortest time.

[0219] As a supplement, the emergency response effect evaluation and feedback mechanism also includes a user feedback interface, which allows emergency responders or decision makers to provide direct feedback on the scheduling results to further refine the optimization direction of the scheduling strategy.

[0220] As an implementation method, it also includes a visualization report generation module, which is used to automatically generate a visualization report containing one or more key information including scheduling strategy, resource allocation, path planning, response time, and rescue effect.

[0221] The present invention also relates to an emergency resource scheduling and path planning system, which includes the emergency resource scheduling and path planning method, as well as the computing equipment, data storage equipment and communication interface required to implement the method, and is used to quickly respond and optimize resource scheduling and path planning in emergency situations.

[0222] The system is a comprehensive solution designed to respond to emergencies and ensure that resources are dispatched quickly and efficiently to where they are needed. The core of the system is its internal integrated emergency resource scheduling and routing method, which is carefully designed to quickly analyze, evaluate and develop the optimal resource scheduling and routing strategy in emergency situations.

[0223] In order to implement this method, the system is equipped with necessary hardware and software resources. Among them, the computing equipment is responsible for performing complex computing tasks, such as dynamic weight evaluation, training and reasoning of deep reinforcement learning models, etc., to ensure the efficient operation of the method. The data storage device is used to store a large amount of historical emergency response data, real-time data, and intermediate data and result data generated during the operation of the method, providing data support for the continuous optimization and decision-making of the method.

[0224] In addition, the communication interface is an important part of the system. It is responsible for communicating with other systems, equipment or personnel to ensure the real-time transmission and sharing of information. In an emergency, the communication interface can quickly receive and process requests and information from all parties, providing timely and accurate data support for resource scheduling and path planning.

[0225] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An emergency resource scheduling and path planning method, characterized in that: The following steps are involved: Building a dynamic weight evaluation system based on emergency resource dispatch factors, which can dynamically calculate and adjust factor weights to determine the dispatch of the most critical resources to the most needed locations in an emergency; Build a deep reinforcement learning model, using historical emergency response data as a training set, so that the agent can learn and optimize the strategy of choosing the best path under different conditions in simulated or historical emergency scenarios; Build a data preprocessing module to integrate multi-source heterogeneous data, extract key information, and provide real-time and accurate input for dynamic weight evaluation and reinforcement learning models; Combining the output results of the dynamic weight evaluation system and the reinforcement learning model, an adaptive scheduling algorithm is designed, which is used to quickly adjust the resource allocation strategy according to the current situation.

2. The method for emergency resource scheduling and path planning according to claim 1, characterized in that: The emergency resource dispatching factors include time urgency, urgency of resource demand, traffic congestion, weather impact and disaster severity, number of affected people, and road capacity.

3. The method for emergency resource scheduling and path planning according to claim 2, characterized in that: The dynamic weight evaluation system includes a weight adaptive adjustment mechanism, which is used to automatically adjust the weights of emergency resource scheduling factors according to real-time changes in emergency events to dynamically optimize resource scheduling strategies.

4. The method for emergency resource scheduling and path planning according to claim 3, characterized in that: The construction of the dynamic weight evaluation system includes the following steps: Define a set of key factors for emergency resource dispatch; Set an initial weight for each key factor and obtain the latest data of key factors in real time; Based on real-time data, the weight of each factor is dynamically adjusted using a specific algorithm; Based on the adjusted weights and real-time data, the score of each scheduling plan is calculated, and the plan with the highest score is selected for execution.

5. The method for emergency resource scheduling and path planning according to claim 4, characterized in that: The entropy weight method is used to calculate the basic weight, which includes the following process: For each factor, its real-time data is standardized to the interval [0,1], denoted as X'i (i=1,2,...,7); For each factor, calculate its information entropy Ei, the formula is: Where n is the number of scheduling options, p ij is the proportion of the standardized value of the i-th factor in the j-th solution to the total value of the factor; Redundancy di = 1-Ei, which represents the information utility value of the factor; Calculate the basic weight W' of each factor according to the redundancy i , the formula is: Define the feedback indicator F according to the actual situation of emergency response; According to the comparison between the feedback index F and the preset target value, the weight adjustment coefficient α is calculated, and the formula is: Among them, F min and F max are the minimum and maximum values ​​of the feedback indicators respectively; According to the weight adjustment coefficient α and the basic weight W' i , calculate the adjusted weight W" i , the formula is "W" i =αW' i +(1-α)W i Among them, W i is the initial weight; According to the adjusted weight W" i and real-time data X' j , calculate the score S of each scheduling solution i , the formula is: The solution with the highest score is selected as the final scheduling decision; T is time urgency, which indicates the time pressure to respond as quickly as possible after an emergency occurs; R is the urgency of resource demand, which indicates the urgency of the resource demand of the emergency event; C is the traffic congestion condition, which indicates the traffic obstacles that may be encountered during the emergency response process; W is weather impact, which indicates the potential impact of weather conditions on emergency response; D is the severity of the disaster, which indicates the destructive power and impact range of the emergency event; P is the number of affected population, indicating the number of people directly affected by the emergency event; H is the road capacity, which represents the traffic efficiency and reliability of the road during the emergency response process.

6. The method for emergency resource scheduling and path planning according to claim 1, characterized in that: The deep reinforcement learning model uses historical emergency response data as a training set, allowing the agent to learn and optimize the selection of the best path under different conditions in simulated or historical emergency scenarios. The deep reinforcement learning model selects the optimal path including the following steps: Identify the various states in the emergency response process and all possible actions that can be taken; Construct a reward function to evaluate the immediate reward after the agent takes an action in a given state; Build a deep neural network as the agent's policy network to predict the best action based on the current state; Use trained deep neural networks for path planning in simulated or historical emergency scenarios; Calculate the efficiency and effectiveness of emergency response based on the path and resource allocation plan selected by the agent; Based on the simulation results, the deep neural network is fine-tuned to further optimize the path selection strategy; In actual emergency response, the deep neural network is input according to the current status to obtain the optimal path and resource allocation plan; Execute the selected path and resource allocation plan to respond to emergencies.

7. The method for emergency resource scheduling and path planning according to claim 1, characterized in that: The data preprocessing module further includes a data fusion and verification unit, which is used to integrate data from different sources.

8. The method for emergency resource scheduling and path planning according to claim 1, characterized in that: The adaptive scheduling algorithm includes: Based on the output of the dynamic weight evaluation system, the resource scheduling strategy is initialized to determine the preliminary resource allocation and path planning; Monitor key data in the emergency response process in real time, and dynamically update relevant parameters in the resource scheduling strategy based on real-time data; Adopt an adaptive adjustment mechanism to dynamically adjust resource scheduling strategies based on real-time data and historical experience; Define scheduling effect evaluation indicators, and calculate the effect score of the current scheduling strategy based on the evaluation indicators; If the effect score of the current scheduling strategy is lower than the preset threshold, the strategy is iterated and optimized; Outputting the final optimized resource scheduling strategy, wherein the resource scheduling strategy includes resource allocation and path planning; The resource allocation adjustment formula is: R alloc (t)=R base +ΔR(t) Among them, R alloc (t) represents the resource allocation at time t; R base Indicates the basic resource allocation amount; ΔR(t) represents the resource allocation increment adjusted according to real-time data; The path planning adjustment formula is: Among them, Path opt (t) represents the optimal path at time t; PathSct represents the set of all possible paths; Time(P,t) represents the rescue time required along path P at time t; Cost(P) represents the cost of path P; λ is the weight coefficient.

9. The method for emergency resource scheduling and path planning according to claim 1, characterized in that: It also includes an emergency response effect evaluation and feedback mechanism, which is used to compare actual scheduling results with expected goals, to evaluate the effectiveness of resource scheduling and path planning, and to feed back the evaluation results to the dynamic weight evaluation system and reinforcement learning model.

10. The method for emergency resource scheduling and path planning according to claim 9, characterized in that: The emergency response effect evaluation and feedback mechanism also includes a user feedback interface, which allows emergency response personnel or decision makers to provide direct feedback on the dispatch results.

11. The method for emergency resource scheduling and path planning according to claim 1, characterized in that: It also includes a visual report generation module, which is used to automatically generate a visual report containing one or more key information such as scheduling strategy, resource allocation, path planning, response time, and rescue effect.

Images