An Edge-Cloud Cooperative Emergency Command and Dispatch System Based on an Enhanced Chaotic Map Marine Predators Algorithm

Abstract

The present invention provides a cloud-edge collaborative emergency command and dispatch system based on an enhanced chaotic mapping ocean predator algorithm. The system includes a device layer, an edge layer, and a cloud server layer. The device layer analyzes and processes the real-time task information generated by the mobile user device terminal and transmits it to the emergency command and dispatch controller. The edge layer includes edge servers, and each edge server contains several CPUs. The calculation and processing of tasks are mainly carried out in the edge layer. The emergency command and dispatch system and method proposed by the present invention can effectively solve the problem that trapped people cannot contact the outside world in the post-disaster emergency scenario. Moreover, the cloud-edge collaborative method proposed by the present invention can make an optimal scheduling strategy for different situations and can efficiently process the real-time tasks generated by mobile user devices.

Description

Technical Field

[0001] The present invention relates to a cloud-edge collaborative emergency command and dispatch system based on an enhanced chaotic mapping ocean predator algorithm. Background Art

[0002] In modern society, various emergencies occur frequently, such as natural disasters, accident disasters, public health events, etc. These events often have unpredictability and destructiveness, posing a serious threat to people's lives and property safety. Establishing an efficient emergency command and dispatch system is particularly important. The primary task of emergency command and dispatch is to quickly collect information such as resources, personnel, and materials involved in the event. Based on the collected relevant information, the command and dispatch system needs to make decisions according to the actual situation and issue corresponding command orders. Emergency command and dispatch need to coordinate various resources to ensure that these resources can be promptly and accurately put into emergency work.

[0003] Complex and unpredictable emergency rescue scenarios are typical applications of dispatch. Rescue personnel cannot enter the disaster area, which makes task scheduling invisible, intangible, and uncertain. Therefore, in an uncertain environment, how to dynamically reschedule tasks to respond to requests while making full use of limited resources remains a problem to be solved.

[0004] Existing methods for task scheduling mainly include optimization algorithms such as particle swarm optimization algorithm, genetic algorithm, ocean predator algorithm, etc. However, traditional optimization algorithms have the problem of being easily trapped in local optimal solutions. Summary of the Invention

[0005] Object of the Invention: The technical problem to be solved by the present invention is to provide a cloud-edge collaborative emergency command and dispatch system based on an enhanced chaotic mapping ocean predator algorithm in view of the deficiencies of the prior art, and establish an emergency scenario composed of K mobile user devices, E base stations, a cloud server, and an emergency command and dispatch controller;

[0006] Each BS base station is equipped with an edge server. Mobile user devices need to unload the tasks (real-time videos, data, etc.) they generate to the edge server for processing, and all BS base stations are linked through BS-BS 4G and 5G communication links. It is set that communication links cannot be established between some base stations damaged by the disaster area.

[0007] An emergency command and dispatch controller is deployed at the edge base station. The emergency command and dispatch controller is connected to all edge servers and the cloud server. The emergency command and dispatch controller has functions of resource allocation and task scheduling. The emergency command and dispatch controller collects server load information and task information tables of the network status, analyzes rescue requests and monitored information, and thus makes command and dispatch decisions;

[0008] The system includes a device layer, an edge layer, and a cloud server layer;

[0009] The device layer analyzes and processes the real-time task information generated by the mobile user device terminal and transmits it to the emergency command and dispatch controller;

[0010] The edge layer includes edge servers, and each edge server contains several CPUs. The calculation and processing of tasks are mainly carried out in the edge layer.

[0011] Cloud servers are deployed in the cloud server layer to provide storage resources and computing resources for tasks. The cloud servers also contain several CPUs to process real-time tasks; the cloud servers cover all edge servers, the cloud servers include all edge servers and mobile user devices, and the cloud servers establish connections with edge node base stations and mobile users through communication links.

[0012] Construct a problem model: Represent the mobile user devices (including trapped persons and sensor load terminals, etc.) in the emergency scenario as a set , where represents the th mobile user device, use the set to represent all edge servers in the emergency command and dispatch system, represents the th edge server, , represents the total number of edge servers, and there are edge servers in the entire emergency command and dispatch system; use to represent the cloud server; According to the information of different mobile user devices in the emergency scenario, it is divided into three situations, namely:

[0013] The first situation is that the mobile user device cannot contact the edge base station and cannot transmit the real-time information to the emergency command and dispatch controller through the edge server;

[0014] The second situation is that the mobile user device can establish a communication connection with the base station, but at this time, the task queue and the calculation volume of real-time data to be processed on the edge server where the mobile user is located exceed the computing power of the edge server, and it is impossible to analyze and process the real-time information transmitted by the user in time and transmit the information to the emergency command and dispatch controller;

[0015] The third situation is that the mobile user device can contact the base station, and the edge server device can process the real-time information of the mobile user device in the emergency scenario in time.

[0016] After a mobile user device generates real-time task information, the real-time task is processed accordingly according to the above three cases: for the first case, the real-time task information is transmitted to the cloud server for processing;

[0017] For the second case, the real-time task information is scheduled to other edge servers for processing;

[0018] For the third case, the real-time task information is processed on the local base station.

[0019] Construct the following mathematical model, assuming that each edge server and the cloud server can only process one real-time task at the same time:

[0020] Use the triple to represent the real-time task generated by the mobile device in time slot where represents the deadline of the real-time task, is the data transmission volume of the real-time task, represents the amount of computation of the real-time task;

[0021] When the task generated by the mobile user device is offloaded to the cloud server for processing, the completion delay of the task is expressed as:

[0022] ,

[0023] where represents the transmission delay of the task to the cloud server, The value of is expressed as: is the transmission rate between the mobile user device and the cloud server; represents the processing delay of the task on the cloud server, where is the processing rate of the cloud server; represents the delay of transmitting the task processing result from the cloud server to the mobile user device ;

[0024] When the task generated by the mobile user device needs to be scheduled from the edge server to other edge server sides (such as ) for computing and processing, the completion delay of the task is expressed as:

[0025] ,

[0026] where Indicates the transmission delay of a real-time task from a mobile user device to an edge server is expressed as: where is the transmission rate from the user device to the edge server to an edge server is the transmission rate; Indicates the scheduling delay of the task from the edge server to an edge server is expressed as: where is the transmission rate from the edge server to an edge server is the transmission rate; Indicates the processing delay of the real-time task on the edge server is where is the processing rate of the edge server ; Indicates the return delay of the task processing result from the edge server transmitted to the mobile user device is the return delay;

[0027] When the task is processed on the local edge server, the processing delay of the real-time task data is expressed as:

[0028] where

[0029] where indicates the computational processing delay of the task on the edge server is Indicates the return delay of the task processing result from the edge server transmitted to the mobile user device is the return delay;

[0030] The symbol indicates whether the real-time task is processed on the cloud server where indicates that the real-time task is not computationally processed on the cloud server indicates that the real-time task is processed on the cloud server side;

[0031] The symbol indicates whether the task is scheduled for processing where indicates that the real-time task does not require scheduling for processing indicates that the real-time task requires scheduling for processing.

[0032] The total completion delay of the real-time task is expressed as:

[0033] ,

[0034] Finally, the target optimization problem is modeled as:

[0035] ,

[0036] s.t. C1: ,

[0037] C2: ,

[0038] C3: ,

[0039] C4: ,

[0040] Among them, s.t. means subject to, C1 represents the constraint conditions of the edge server, C2 represents the constraint conditions of the mobile user device, and C3 and C4 represent the constraint conditions of the real-time task processing methods generated by the mobile user.

[0041] The improved ocean predator algorithm (ECMPA) is used to solve the target optimization problem:

[0042] The objective function fitness is expressed as:

[0043] ,

[0044] The position of each population is regarded as a solution to the task allocation in the emergency command and dispatch system, that is, the tasks generated by the mobile user are mapped to each edge server and cloud server through the solution , where the number of the server is , and it is set that the population position is represented by the formula . The tasks numbered {1, 2, 5, 7, 10} are assigned to the first edge server , the tasks numbered {3, 6, 9} are assigned to the second edge server , and the tasks numbered {4, 8} are assigned to the third edge server ;

[0045] Subsequently, the population is initialized, and the initial task allocation strategy is generated in the solution space through the following formula:

[0046] ,

[0047] Among them, is a random number in [0, 1], represents the population at the The dimensional position, and respectively represent the maximum and minimum values of the solution space boundary, represents the dimension of the randomly initialized population, represents the th task scheduling decision in the emergency command and dispatch system;

[0048] After initializing the population, the optimal solution is selected to construct the prey matrix and the elite matrix, and the fitness value of the scheduling scheme mapped by each population is calculated according to the objective function fitness. Subsequently, according to different iteration cycles, the population position is updated.

[0049] The position of the predator represents the historical optimal solution, and the position of the prey represents the new solution. In the second stage of improving the marine predator algorithm, the present invention introduces the Guided Learning Strategy (GLS), and the Guided Learning Strategy GLS includes the feedback result and the parameter :

[0050] ,

[0051] ,

[0052] Among them, represents normalizing the feedback result, represents the learning experience, is the maximum value of the search space, is the minimum value of the search space, represents the normalization result to avoid the value being affected by the change of the boundary value;

[0053] In the guidance stage, the improved marine predator algorithm is guided to explore or exploit. When , it means that the current requirement of the improved marine predator algorithm is exploitation. When , it is detected that the current requirement of the improved marine predator algorithm is exploration, and the guidance algorithm will explore the global space.

[0054] An adaptive step size control factor is introduced in the update process of the prey to control the step size of the prey. The calculation formula of the adaptive step size control factor is:

[0055] ,

[0056] Among them, represents the current iteration number of the improved marine predator algorithm, is the maximum iteration number.

[0057] The update process of the improved marine predator algorithm includes the following three stages:

[0058] The first stage is the exploration stage, and the formula is described as:

[0059] When is

[0060] ,

[0061] ,

[0062] where represents the step size of the i-th predator individual, is a vector containing random numbers generated by a normal distribution, A is a constant, is a vector of uniformly generated random numbers; is the number of iterations of the algorithm; represents the prey matrix, represents the elite matrix; represents matrix multiplication in the prey matrix update process;

[0063] The second stage is a hybrid stage of exploration and exploitation. The population is updated using two movement methods. One is to update using Brownian motion, and the other is to update using Lévy flight;

[0064] The GLS strategy is introduced to determine whether the algorithm belongs to the exploitation or exploration stage for position update. When , it is the exploitation stage, and the predator updates its position according to the following formula:

[0065] ,

[0066] ,

[0067] where is a random vector representing Lévy flight;

[0068] When , it is the exploration stage, which will guide the predator to perform a global search. The update method of the population is:

[0069] ,

[0070] ,

[0071] ,

[0072] where CF is an adaptive parameter that controls the individual step size;

[0073] In the final stage of the optimization of the Enhanced Chaotic Mapping Marine Predation Algorithm (ECMPA), all populations are updated using Lévy flight:

[0074] When then , .

[0075] In addition to the above three update methods, vortices and fish aggregation effects will also be generated during the process of predators preying on prey. To address the impacts brought by vortices and fish aggregation effects, the following formula is used to update the population:

[0076] ,

[0077] where FADs is the local optimal solution in the exploration area of the predator, is a binary vector, ; and represent two random prey, and r represents a random number in;

[0078] After the population position information is updated, calculate the fitness value of each updated solution, and continuously iterate and update until the maximum number of iterations is reached or the algorithm converges. Finally, the emergency command and dispatch control center commands the mobile users for dispatch according to the optimal dispatch strategy.

[0079] Beneficial effects: The emergency command and dispatch system and method proposed by the present invention can effectively solve the problem that trapped people cannot contact the outside world in the post-disaster emergency scenario. Moreover, the cloud-edge collaboration method proposed by the present invention can make the optimal dispatch strategy for different situations and can efficiently process the real-time tasks generated by mobile user devices. In addition, the enhanced chaotic mapping marine predation algorithm proposed by the present invention can effectively find the dispatch strategy with the shortest task completion delay, thereby accelerating the task dispatch speed in the emergency scenario and realizing real-time rescue. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 is the system architecture diagram of the present invention.

[0081] Figure 2 is the flowchart of the cloud-edge collaboration dispatch scheme.

[0082] Figure 3 is the schematic diagram of task dispatch.

[0083] Figure 4 is the experimental result diagram.

[0084] Figure 5 is the experimental result diagram. Detailed implementation manners

[0085] The following further specifically describes the present invention in conjunction with the accompanying drawings and specific implementation manners, and the above and / or other advantages of the present invention will become clearer.

[0086] As Figure 1 shown, an embodiment of the present invention provides a cloud-edge collaborative emergency command and dispatch system based on an enhanced chaotic mapping ocean predator algorithm, which includes a three-layer architecture of a device layer, an edge layer, and a cloud server layer.

[0087] The main task of the device layer is to analyze and process the real-time task information generated by the mobile user device terminal and transmit it to the emergency command and dispatch controller.

[0088] The edge layer includes edge servers, and each edge server includes several CPUs. The calculation and processing process of tasks mainly takes place in the edge layer.

[0089] In the cloud server layer, cloud servers with powerful computing capabilities are deployed, which can provide a large amount of storage resources and computing resources for tasks. The cloud servers also include several CPUs to process real-time tasks. The cloud server is located behind all edge nodes and includes all edge servers and mobile user devices. The cloud server establishes connections with edge node base stations and mobile users through communication links.

[0090] When real-time task data is generated by a mobile terminal, the real-time task is processed accordingly according to the above three different situations.

[0091] In the case where the mobile user device cannot contact the edge base station, the real-time task information is transmitted to the cloud server side for processing. When the edge server is overloaded and cannot process the real-time task information in time, the system framework will schedule the real-time data information to other edge servers for processing.

[0092] In the last case, that is, the mobile user can contact the base station, and the edge server device can process the real-time information of the user device in the emergency scenario in time, and the real-time task information is processed on the local base station.

[0093] Therefore, the following mathematical model can be constructed, assuming that each edge server and cloud server can only process one real-time task at the same time.

[0094] Use the triple to represent the real-time task generated by the mobile device in the time slot , where represents the deadline of the real-time task, is the size of the data transmission volume of the real-time task, Represents the computational workload of real-time tasks.

[0095] When the tasks generated by the mobile user device are offloaded to the cloud server for processing, the completion delay of the tasks Is expressed as:

[0096] ,

[0097] Where Represents the transmission delay of the task to the cloud server side. Among them, The value of can be expressed as the equation , Is the transmission rate between the mobile user device and the cloud server. Represents the processing delay of the task on the cloud server side, , where Is the processing rate of the cloud server. Represents the delay of the task processing result transmitted from the cloud server side to the mobile user device Of the delay;

[0098] When the mobile user device The tasks generated Need to be scheduled from the edge server To other edge server sides such as For computing and processing, the completion delay of the task Is expressed as:

[0099] ,

[0100] Where Represents the transmission delay of the real-time task from the mobile user device To the edge server Of the transmission delay, The value of can be expressed by the equation Indicates, Is the transmission rate from the user device to the edge server Of. Represents the scheduling delay of the task from the edge server To the edge server Is expressed as the equation , where Is the edge server To the edge server Of the transmission rate. Represents the processing delay of the real-time task on the edge server On, indicates the processing delay of the real-time task on the edge server On, Is the edge server Of the processing rate. Indicates the return latency of the task processing result from the edge server to the mobile user device ;

[0101] When the task is processed on the local edge server, the processing latency of the real-time task data is expressed as:

[0102] ,

[0103] wherein, represents the computing and processing latency of the task on the edge server ; represents the return latency of the task processing result from the edge server to the mobile user device ;

[0104] The symbol is used to indicate whether the real-time task is processed on the cloud server , , indicating that the real-time task is not computed and processed on the cloud server indicating that the real-time task is processed on the cloud server side. The symbol is used to indicate whether the task is scheduled for processing , , indicating that the real-time task does not require scheduling for processing indicating that the real-time task requires scheduling for processing

[0105] Therefore, the total completion latency of the real-time task is expressed as:

[0106] ,

[0107] Therefore, finally, we can model the target optimization problem as:

[0108] ,

[0109] s.t. C1: ,

[0110] C2: ,

[0111] C3: ,

[0112] C4: ,

[0113] Among them, C1 represents the constraint conditions of the edge server, C2 represents the constraint conditions of the mobile user device, and C3 and C4 represent the constraint conditions for the processing method of real-time tasks generated by the mobile user. This problem is a MINLP problem. Since the objective function is non-convex, when the problem scale is very large, it is almost impossible to find the exact solution of the problem within a reasonable time complexity. Therefore, the present invention proposes to use an improved MPA algorithm to solve the cloud-edge collaborative task scheduling model.

[0114] Figure 2 The specific process of the scheduling method proposed by the present invention is as follows. In the present invention, the goal of task scheduling is to reduce the completion time of emergency command and dispatch and achieve efficient rescue. Therefore, the objective function is expressed as:

[0115] ,

[0116] Regarding the position of each population as a solution to task allocation in the mobile edge computing network, that is, the tasks generated by the mobile user are mapped to each edge server and cloud server through the solution , where the number of the server is . Assuming that the population position is represented by Equation , then this solution means that the tasks numbered {1, 2, 5, 7, 10} are assigned to the server , the tasks numbered {3, 6, 9} are assigned to , and the tasks numbered {4, 8} are assigned to , as shown in Figure 2 .

[0117] Subsequently, the population is initialized. The improved MPA algorithm proposed by the present invention follows the general initialization strategy of typical meta-heuristic algorithms. At the beginning of the algorithm, the following formula is used to generate the initial task allocation strategy in the solution space:

[0118] ,

[0119] Among them, is a random number in [0, 1], represents the position of the population in the th dimension, and represent the maximum and minimum values of the solution space boundary respectively, represents the dimension of the randomly initialized population, represents the rd task scheduling decision in the emergency command and dispatch system.

[0120] After initializing the population, the optimal solution is selected to construct the prey matrix and the elite matrix, and the fitness value of the scheduling scheme mapped by each population is calculated according to the above objective function. Subsequently, according to different iteration cycles, the population position is updated, which also corresponds to the scheduling of real-time task information.

[0121] The position of the predator represents the historical optimal solution, and the position of the prey represents the new solution. Different from the traditional MPA algorithm, in order to solve the balance problem between exploration and exploitation in the second stage of the traditional MPA algorithm, a guiding learning strategy (GLS) is introduced on this basis to guide the exploration and exploitation of the algorithm. GLS is a mechanism for analyzing and guiding algorithms, which includes defining feedback results and parameters . GLS mainly includes two stages: feedback and guidance. In the feedback stage, the current demand of the algorithm is evaluated by calculating the dispersion of historical individuals (learning experience). A high dispersion indicates that the algorithm is exploring, while concentration indicates exploitation. Finally, the feedback results are standardized so that the same set of evaluation criteria can be used in different situations.

[0122] ,

[0123] ,

[0124] Among them, represents standardizing the feedback results. represents the learning experience, is the maximum value of the search space, is the minimum value of the search space, represents the standardized result, avoiding the value being affected by the change of boundary values.

[0125] In the guidance stage, the algorithm is guided to explore or exploit. When , the current demand of the algorithm is exploitation. When , it is detected that the current demand of the algorithm is exploration, and the algorithm will be guided to explore the global space.

[0126] In addition, in order to improve the convergence speed and accuracy of the algorithm, an adaptive step size control factor is introduced in the update step of the prey to control the step size of the prey. After introducing the adaptive step size control factor, as the number of iterations increases, the step size of population update also increases, thus accelerating the convergence speed of the ECMPA algorithm. Among them, the definition of the adaptive step size control factor in the present invention is:

[0127] ,

[0128] Among them, represents the current iteration number of the enhanced marine predator algorithm proposed by the present invention, is the maximum number of iterations set for the algorithm.

[0129] Therefore, the update process of the improved MPA algorithm proposed in the present invention can be divided into the following three stages. In the first stage, that is, the exploration stage, the predators try to move faster until they find their prey. This is the essence of all metaheuristic algorithms at the beginning of the optimization process, and the formula is described as:

[0130] When is true,

[0131] ,

[0132] ,

[0133] where represents the step size of the i-th predator individual, is a vector containing random numbers generated from a normal distribution, used to represent Brownian motion. A is a constant, usually set to 0.5. is a vector of uniformly generated random numbers. is the iteration number of the algorithm. represents the prey matrix, represents the elite matrix.

[0134] The second stage of the MPA algorithm is a hybrid stage of exploration and exploitation. This population is updated using two types of movements, one using Brownian motion for update and the other using Lévy flight for update. We introduce the GLS strategy to determine whether the algorithm belongs to the exploitation or exploration stage for position update.

[0135] When , it is the exploitation stage, and the predators update their positions according to the following formula:

[0136] ,

[0137] ,

[0138] where, is a random vector representing Lévy flight.

[0139] When , it is the exploration stage, which will guide the predators to conduct a global search, and the update method of its population is:

[0140] ,

[0141] ,

[0142] ,

[0143] Among them, CF is an adaptive parameter that controls the individual step size.

[0144] In the last stage of the MPA algorithm optimization, all populations are updated using Lévy flight:

[0145] When ,

[0146] ,

[0147] ,

[0148] In addition to the above three update methods, the predator will also produce the FADs effect during the process of preying on the prey. To solve the problems brought by FADs, we use the following formula to update the population.

[0149] ,

[0150] Among them, is a binary vector, , , represents two randomly selected prey.

[0151] After the population position information is updated, calculate the fitness value of each updated solution, and continuously iterate and update until the maximum number of iterations is reached or the algorithm converges. Finally, the command and dispatch control center commands the mobile users according to the optimal scheduling strategy for scheduling.

[0152] In the present invention, in order to adapt to emergency scenarios and achieve real-time task scheduling, the real-time tasks generated by all user devices in each time slot are scheduled to ensure the minimum task completion delay. Figure 3 As shown in the task scheduling schematic diagram, in order to find the solution with the least time consumption, the improved MPA algorithm will update the position of the prey, which corresponds to the scheduling of some tasks. For example, when is updated to , it means that the task numbered 2 is scheduled from the first server to the third server , the task numbered 4 is scheduled from the third server to the first server , the task numbered 6 is scheduled from the second server to the third server , the task numbered 10 is scheduled from the first server to the third server .

[0153] To verify the performance of the cloud-edge collaborative emergency command and dispatch system proposed by the present invention and the dispatch method based on the enhanced chaotic mapping marine predator algorithm proposed by the present invention, simulation experiments were conducted on mobile user devices in the post-disaster emergency scenario. The simulation was carried out in a circular area with a radius of 120m. The simulated experimental scenario has 36 devices, 12 edges, and 1 cloud. The edge servers and mobile user devices are randomly distributed in the circular area. In each time slot, it is set that some random user devices cannot communicate with the edge servers. The transmission rate between the edge and the cloud is [20, 30] Mbps. The network topology between the edges is randomly generated, and the transmission rate is [150, 300] Mbps. In the simulated edge environment, the computing power of the nodes (VMs) is set to [1, 2.5] GHz. Each edge server has [4, 6] processors. The length of each task is randomly generated, and the length range of the random tasks is [125, 350] GHz. The data transmission volume of the tasks is [250, 500] MB. It is set that the energy consumption of the virtual machine during computing is 100W, and the energy consumption during idle is 10% of the computing energy consumption. In addition, it is assumed that the number of the population is 30.

[0154] It can be seen from Figure 4 and Figure 5 that the cloud-edge collaborative dispatch method proposed by the present invention can effectively reduce the task completion delay of the dispatch system, which indicates that the cloud-edge collaborative dispatch system proposed by the present invention can effectively adapt to the emergency scenario and achieve efficient rescue. Moreover, it can be seen from the experimental results that the enhanced chaotic mapping marine predator algorithm proposed by the present invention has better performance and can effectively reduce the task completion delay. In Figure 4 and Figure 5 , MPA represents the traditional marine predator algorithm, WOA represents the whale optimization algorithm, MRFOSSA represents the hybrid intelligent optimization algorithm, IPSO represents the improved particle swarm optimization algorithm, and ECMPA represents the improved marine predator algorithm proposed by the present invention.

[0155] The present invention provides a cloud-edge collaborative emergency command and dispatch system based on the enhanced chaotic mapping marine predator algorithm. There are many methods and ways to specifically implement this technical solution. The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Each component not clearly defined in this embodiment can be implemented by the existing technology.

Claims

1. An edge-cloud collaborative emergency command and dispatch system based on an enhanced chaotic mapping ocean predator algorithm, characterized in that, An emergency scenario is established, which consists of K mobile user devices, E base stations, a cloud server, and an emergency command and dispatch controller. Each BS base station is equipped with an edge server. Mobile user devices need to offload the tasks they generate to the edge server for processing. All BS base stations are linked through BS-BS 4G and 5G communication links. It is assumed that communication links cannot be established between the base stations damaged by the disaster area. The emergency command and dispatch controller is deployed at the edge base station. The emergency command and dispatch controller is connected to all edge servers and the cloud server. The emergency command and dispatch controller has functions of resource allocation and task scheduling. The emergency command and dispatch controller collects server load information and task information tables of the network status, analyzes rescue requests and monitored information, and thus makes command and dispatch decisions. The system includes a device layer, an edge layer, and a cloud server layer. The device layer analyzes and processes the real-time task information generated by the mobile user device side and transmits it to the emergency command and dispatch controller. The edge layer includes edge servers. Each edge server contains a CPU, and the calculation and processing process of tasks is carried out in the edge layer. A cloud server is deployed in the cloud server layer, which is used to provide storage resources and computing resources for tasks. The cloud server contains a CPU to process real-time tasks. The cloud server covers all edge servers. The cloud server includes all edge servers and mobile user devices. The cloud server establishes connections with edge node base stations and mobile users through communication links. Construct a problem model: Represent mobile user devices in an emergency scenario as a set , where represents the th mobile user device, and the set represents all edge servers in the emergency command and dispatch system, represents the th edge server, , represents the total number of edge servers. There are edge servers in the entire emergency command and dispatch system; use to represent the cloud server; According to the information of different mobile user devices in the emergency scenario, it is divided into three cases, namely: In the first case, the mobile user device cannot contact the edge base station and cannot transmit real-time information to the emergency command and dispatch controller through the edge server. In the second case, the mobile user device can establish a communication connection with the base station. However, at this time, the task queue and the amount of real-time data to be processed on the edge server where the mobile user is located exceed the computing power of the edge server, and it is impossible to analyze and process the real-time information transmitted by the user in time and transmit the information to the emergency command and dispatch controller. In the third case, the mobile user device can contact the base station, and the edge server device can process the real-time information of the mobile user device in the emergency scenario in time. When a mobile user device generates real-time task information, according to the above three cases, the real-time task is processed accordingly: for the first case, the real-time task information is transmitted to the cloud server side for processing. For the second case, the real-time task information is scheduled to other edge servers for processing. For the third case, the real-time task information is processed on the local base station.

2. The system according to claim 1, wherein The following mathematical model is constructed, assuming that each edge server and cloud server can only process one real-time task at the same time: Using a triple to represent a mobile device at a time slot generates a real-time task, where represents the deadline of the real-time task, is the data transfer volume size of the real-time task, represents the computational amount of the real-time task; When the tasks generated by the mobile user device are offloaded to the cloud server for processing, the completion delay of the tasks is expressed as: ， Among them, represents the transmission delay of the task to the cloud server side, The value of is expressed as: , is the transmission rate between the mobile user device and the cloud server; represents the processing delay of the task on the cloud server side, , where is the processing rate of the cloud server; represents the delay of the task processing result transmitted from the cloud server side to the mobile user device ; When a mobile user device generates a task that needs to be scheduled from an edge server to another edge server for computing and processing, the completion delay of the task is expressed as: ， Among them, represents the transmission delay of a real-time task from a mobile user device to an edge server The value of which is expressed as: where , is the transmission rate from the user device to the edge server ; represents the scheduling delay of a task from an edge server to an edge server which is expressed as: where is the transmission rate from edge server to edge server ; represents the processing delay of a real-time task on an edge server which is , where is the processing rate of the edge server represents the return delay of the task processing result from the edge server transmitted to the mobile user device ; When the task is processed on the local edge server, the processing latency of the real-time task data is expressed as: ， Among them, represents the computing and processing delay of the task on the edge server and represents the return delay for the task processing result to be transmitted from the edge server to the mobile user device . Use the symbol to indicate whether the real-time task is processed on the cloud server, , indicating that the real-time task is not calculated and processed on the cloud server, indicating that the real-time task is processed on the cloud server side; Use the symbol to indicate whether the task is scheduled for processing, , indicating that the real-time task does not need to be scheduled for processing, indicating that the real-time task needs to be scheduled for processing; Total Completion Delay of Real-Time Tasks It is expressed as: ， Finally, the target optimization problem is modeled as: ， s.t. C1: , C2: ， C3: ， C4: ， Among them, s.t. means subject to, C1 represents the constraint conditions of the edge server, C2 is the constraint conditions of the mobile user device, and C3 and C4 are the constraint conditions of the processing methods of the real-time tasks generated by the mobile user.

3. The system according to claim 2, characterized in that, An improved ocean predator algorithm is used to solve the target optimization problem: The objective function fitness is expressed as: ， Regarding the position of each population as a solution to task allocation in the emergency command and dispatch system, that is, the tasks generated by mobile users are mapped to each edge server and cloud server through the solution , where the server numbers are . It is set that the population position is represented by Equation . The tasks numbered {1, 2, 5, 7, 10} are assigned to the first edge server , the tasks numbered {3, 6, 9} are assigned to the second edge server , and the tasks numbered {4, 8} are assigned to the third edge server . Subsequently, the population is initialized, and an initial task allocation strategy is generated in the solution space through the following formula: ， Among them, is a random number in [0, 1], represents the population at the dimensional position, and represent the maximum and minimum values of the solution space boundary respectively, represents the dimension of the randomly initialized population, represents the th task scheduling decision in the emergency command and dispatch system; After initializing the population, the optimal solution is selected to construct the prey matrix and the elite matrix, and the fitness value of the scheduling scheme mapped by each population is calculated according to the objective function fitness. Subsequently, according to different iteration cycles, the population position is updated.

4. The system according to claim 3, wherein Use the position of the predator to represent the historical optimal solution, and the position of the prey to represent the new solution. Introduce a guided learning strategy, and the guided learning strategy includes feedback results and parameters : ， ， Among them, represents standardizing the feedback result, represents learning experience, is the maximum value of the search space, is the minimum value of the search space, represents the standardized result; In the guidance stage, the improved marine predator algorithm is guided to explore or exploit. When it means that the current requirement of the improved marine predator algorithm is exploitation. When it is detected that the current requirement of the improved marine predator algorithm is exploration, and the guidance algorithm will explore the global space.

5. The system according to claim 4, characterized in that, Introduce an adaptive step size control factor during the update of the prey to control the step size of the prey. The adaptive step size control factor is calculated as follows: ， Among them, represents the current iteration number of the improved marine predator algorithm, is the maximum number of iterations.

6. The system according to claim 5, wherein The update process of the improved marine predator algorithm includes the following three stages: The first stage is the exploration stage, which is described by the formula: When then ， ， Among them represents the step size of the $i$-th predator individual, is a vector containing random numbers generated by a normal distribution, and $A$ is a constant, is a vector of uniformly generated random numbers; is the number of iterations of the algorithm; represents the prey matrix, represents the elite matrix; represents matrix multiplication during the update process of the prey matrix The second stage is a mixed stage of exploration and exploitation. The population is updated by two movement methods. One is to update by Brownian motion, and the other is to update by Lévy flight; Introduce the GLS strategy judgment algorithm to determine whether it belongs to the development or exploration stage, and perform position updates. When , it is the development stage, and the predator updates its position according to the following formula: ， ， Among them, is a random vector representing Lévy flight; When , it is the exploration stage, which will guide the predator to conduct a global search. The update method of the population is as follows: ， ， ， Among them, CF is an adaptive parameter that controls the individual step size; In the final stage of the MPA algorithm optimization, all populations are updated by Lévy flight: When , , .

7. The system according to claim 6, characterized in that, During the process of predators preying on prey, eddies and fish aggregation effects will also be generated. To solve the influence brought by the eddies and fish aggregation effects, the following formula is used to update the population: ， Among them, FADs are the local optimal solutions in the predator exploration area, is a binary vector, ; and represent two randomly selected prey, represents a random number in; When the population position information is updated, the fitness value of each updated solution is calculated, and the iteration is continuously updated until the maximum iteration number is reached or the algorithm converges. Finally, the emergency command and dispatch control center commands the mobile users to conduct scheduling according to the optimal scheduling strategy.

Images