Emergency mobile edge computing method and system based on space-air-ground integrated network

Abstract

The invention discloses an emergency mobile edge calculation method and system based on a space-air-ground integrated network, and the method comprises the steps: collecting the position information of vehicles and an unmanned plane based on an LEO satellite, and obtaining CHUs corresponding to all vehicles; in combination with the CHU corresponding to each vehicle, the unmanned aerial vehicle closest to the CHU is obtained, other unmanned aerial vehicles in the unmanned aerial vehicle cluster are determined, and a clustering strategy is obtained; in-cluster task splitting is executed, and a corresponding calculation task is unloaded to the in-cluster unmanned aerial vehicle for in-cluster cooperative calculation; a vehicle task unloading strategy is determined according to the position relation between a vehicle and a current CHU, a corresponding unmanned aerial vehicle cluster is switched or local calculation is used, aiming at the problem of insufficient coverage of unmanned aerial vehicles, the vehicle unloads a calculation task to a cluster head unmanned aerial vehicle CHU, task splitting is performed by the CHU, in order to ensure the continuity of calculation, a dynamic unloading strategy is designed, and the calculation efficiency is improved. And the vehicle switches the corresponding unmanned aerial vehicle cluster in the moving process and uses local calculation in good time, so that the sustainability of processing a calculation task by the vehicle is ensured.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communications, and in particular relates to an emergency mobile edge computing method and system based on an air-ground integrated network. Background Art

[0002] In the information age, people's daily lives are inseparable from the rapidly maturing Internet of Things (IoT) technology. In recent years, the deployment of a large number of IoT devices has improved people's quality of life and made the concept of the Internet of Everything possible. From the Internet of Things to the Internet of Vehicles, these IoT devices not only meet people's daily communication needs, but also provide an ideal platform for the development of intelligent connected vehicles.

[0003] As an important part of the vehicle-to-everything (V2X) network, intelligent connected vehicles have attracted much attention due to their rapid development in recent years. As the level of autonomous driving technology has risen to a new dimension, today's vehicles are usually equipped with multiple sensors and cameras to collect information about the road surface and surrounding environment. This information can not only help vehicles achieve highly autonomous autonomous driving, but also provide strong assistance for traffic flow control, post-disaster rescue and reconstruction in post-disaster areas. However, due to the limited computing power of the vehicle itself, it is usually difficult to process this complex information relying solely on the vehicle's own computing power. Therefore, Mobile Edge Computing (MEC) technology is considered to be an effective paradigm to solve this problem.

[0004] In existing V2X networks, base stations (BS) or roadside units (RSU) equipped with mobile edge computing servers are usually deployed on the ground, and vehicles can offload computing tasks to nearby mobile edge computing servers for calculation. However, these servers are often not suitable for high altitudes or post-disaster areas due to their long construction period and easy damage.

[0005] Fortunately, with the rapid development of next-generation mobile communication technology, satellite communications are rising rapidly. As an emerging communication tool, Low Earth Orbit (LEO) satellites have been deployed in large numbers in recent years due to their wide coverage, which not only reduces communication blind spots around the world, but also provides a new solution for emergency communications in disaster areas and areas where ground base stations are scarce. However, most satellite-based mobile edge computing methods only consider the wide coverage characteristics of low-orbit satellites, without considering the height of low-orbit satellites. When LEO satellites equipped with mobile edge computing servers provide services to ground users, high propagation delays are usually incurred due to their high altitude. Therefore, we need to find facilities closer to the ground to provide mobile edge computing services to reduce the total delay required to process tasks.

[0006] As one of the main aerial communication equipment, Unmanned Aerial Vehicle (UAV) is favored due to its rapid deployment. By deploying UAVs, emergency communication services can be provided for post-disaster search and rescue scenarios where ground BS is damaged and communication needs increase. With the rise of satellite communications, the Space-Air-Ground-Integrated-Network (SAGIN) that integrates UAVs into the satellite-ground network has set off another wave of research enthusiasm. However, most of the existing SAGIN-based mobile edge computing technologies focus on satellite participation in computing, and do not consider the significant impact of propagation delay on the total task processing delay. Therefore, in order to minimize this impact, the use of UAVs to provide emergency mobile edge computing services on the ground should be considered. However, existing technologies that use UAVs to provide mobile edge computing usually ignore the issues of UAV coverage and UAV utilization, which leads to a serious waste of communication resources. Summary of the invention

[0007] In order to solve the problems existing in the prior art, the present invention provides an emergency mobile edge computing method based on an integrated air-ground-space network, using UAVs to provide emergency mobile edge computing services for ground vehicles in scenarios where ground base stations are damaged. By jointly optimizing the UAV clustering and task splitting algorithms, the UAV utilization is maximized while minimizing the total system delay, and a dynamic unloading strategy is proposed to ensure the sustainability of vehicle computing services.

[0008] In order to achieve the above objectives, in the first aspect, the present invention provides an emergency mobile edge computing method based on an integrated air-ground-space network, jointly optimizing UAV clustering and task splitting methods, maximizing UAV utilization while minimizing the total system delay, and adopting a dynamic unloading strategy to enable the vehicle to switch UAV clusters or use local computing during movement; specifically, the following steps are included: Collect the location information of vehicles and UAVs based on LEO satellites and obtain the Cluster-Head UAV (CHU) corresponding to all vehicles; Combine the CHU corresponding to each vehicle, obtain the drone closest to the CHU, determine the other drones in the drone cluster, and obtain the clustering strategy ,The clustering strategy includes the indexes of all CHUs and the UAV index in the cluster corresponding to each CHU, and the clustering results are sent to the corresponding CHU, which further processes the computing tasks according to the clustering results; Perform task splitting within the cluster and offload the corresponding computing tasks to the UAVs within the cluster for collaborative computing within the cluster; The vehicle task offloading strategy is determined according to the position relationship between the vehicle and the current CHU, and the corresponding drone cluster is switched or local computing is used.

[0009] Furthermore, based on the LEO satellite, the location information of vehicles and UAVs is collected to obtain the CHU corresponding to all vehicles, including: using the location information of all UAVs to establish a KD tree; for each vehicle, obtain its location information in the initial time slot and set the flag variable , to indicate whether the vehicle successfully finds a suitable CHU. Before obtaining a suitable CHU, it loops to query the UAVs closest to the vehicle, sorts the indexes in ascending order of distance, and returns the sorted indexes; In the obtained results, check whether each drone has been assigned to other clusters. If so, remove the drone from the waiting queue; calculate the elevation angle between the vehicle and the drone If the elevation angle is not less than the corresponding threshold, the drone is considered to be a qualified CHU of the current vehicle, the search for the current vehicle CHU is stopped, and the ; The above process is repeated until all vehicles have determined their corresponding CHUs.

[0010] Furthermore, by combining the CHU corresponding to each vehicle, the UAV closest to the CHU is obtained, and the other UAVs in the UAV cluster are determined to obtain the clustering strategy. The method includes: using the KD tree to obtain the distance of the drone closest to the current CHU, obtaining all possible drone indexes, filtering unselected drones from the obtained results, and excluding the index of the CHU. For drones that meet the conditions, calculating the distance between the drone and the current CHU, and checking whether the CHU closest to the drone is the current CHU; if so, treating it as a valid drone and adding it to the drone cluster corresponding to the CHU; Repeat the above process until all drones are divided into a drone cluster and the number of drones in the drone cluster does not exceed the maximum number, and the clustering strategy is obtained. .

[0011] Furthermore, CHU performs intra-cluster task splitting and offloads the corresponding computing tasks to the intra-cluster UAVs for intra-cluster collaborative computing, including: The optimal task split ratio within the UAV cluster to minimize the total delay of task processing is modeled as:

[0012]

[0013] Iteratively calculate the final task split ratio in the drone cluster based on the estimated computing delay index: When performing calculation tasks, the tasks are evenly split, the initial ECD of the iterative algorithm is calculated and iterated, and after each iteration, the latest task split ratio is determined according to the ECD of the current iteration cycle; the sum of the ECDs of the UAVs in the cluster is taken, and the ECD of each UAV is divided by the sum of the ECDs of the UAVs in the cluster to obtain the latest task split ratio of each UAV; the above process is repeated until the difference between the task split ratio of the new cycle and the task split ratio of the previous cycle is less than the threshold, and the final task split ratio in the cluster is obtained.

[0014] Furthermore, when the vehicle is dynamically unloaded, if there is an ongoing computing task in the current time slot, the current time slot is skipped; if there is no ongoing computing task in the current time slot, a new computing task is generated; When generating new computing tasks, if the vehicle is within the coverage of the current CHU, the vehicle will offload the computing tasks to the CHU. If the vehicle is beyond the coverage of the current CHU but within the coverage of the next CHU, the CHU corresponding to the vehicle will be switched to the next CHU, and the computing tasks will be offloaded to the drone cluster corresponding to the CHU for computing. If the vehicle is not within the coverage of any CHU, local calculations are performed.

[0015] Furthermore, the completion time of the task is calculated according to the task offloading strategy, and the time slots that can be skipped are determined. The time to complete the computing task includes the maximum value of the sum of the vehicle-to-UAV transmission delay, the intra-cluster UAV-to-UAV transmission delay, and the computing delay.

[0016] Furthermore, the UAV clustering and task splitting methods are jointly optimized to maximize the UAV utilization while minimizing the total system delay. Based on the modeling of the vehicle-to-UAV link, the low-orbit satellite-to-UAV link, and the inter-UAV link, the satellite-to-UAV communication delay, the vehicle-to-UAV communication delay, the UAV-to-UAV communication delay, and the computational delay of the UAVs within the cluster are modeled; and then the total system delay is obtained based on the delay modeling; The optimization goal is to obtain the optimal UAV clustering and unloading decision, while reducing the total system delay and improving the UAV utilization rate. The optimization problem is modeled as:

[0017]

[0018] in, Indicates the clustering of drones. If , representing the drone has been clustered, otherwise ; Assume that each drone can only be assigned to one drone cluster; and Represents the weight coefficient.

[0019] In a second aspect, the present invention provides an emergency mobile edge computing system based on an air-ground integrated network, including a vehicle corresponding CHU acquisition module, a clustering strategy acquisition module, a splitting module, and an execution module; The vehicle-corresponding CHU acquisition module collects the location information of vehicles and drones based on LEO satellites and obtains the CHUs corresponding to all vehicles; The clustering strategy acquisition module is used to combine the CHU corresponding to each vehicle, obtain the UAV closest to the CHU, determine other UAVs in the UAV cluster, and obtain the clustering strategy. The clustering strategy contains the indexes of all CHUs and the UAV index in the cluster corresponding to each CHU, and sends the clustering results to the corresponding CHU. The CHU further processes the computing task according to the clustering results. The splitting module is used to perform intra-cluster task splitting and offload the corresponding computing tasks to the UAVs in the cluster for intra-cluster collaborative computing; The execution module is used to determine the vehicle task unloading strategy according to the position relationship between the vehicle and the current CHU, switch the corresponding UAV cluster or use local calculation.

[0020] In the third aspect, the present invention provides an emergency mobile communication system based on an integrated air-space-ground network, including a vehicle, a LEO satellite and a drone equipped with a MEC server. The LEO satellite collects the initial position information of the drone and the initial position information of the vehicle. The drone can process the computing tasks generated by the vehicle. When the vehicle is driving on the road, the vehicle and the drone, the low-orbit satellite and the drone, and the drones can communicate with each other. When the system is running, the computing tasks are processed based on the above-mentioned emergency mobile edge computing method based on the integrated air-space-ground network.

[0021] The present invention can also provide a computer device, including a processor and a memory, the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and when the processor executes the computer executable program, it can implement the above-mentioned emergency mobile edge computing method based on the integrated air-space-ground network.

[0022] Compared with the prior art, the present invention has at least the following beneficial effects: For scenarios such as remote areas or post-disaster areas where ground MEC servers are unable to provide computing services, the vehicle's own processing capabilities alone cannot meet the large amount of data processing needs, and the existing technology of using satellites to perform calculations in this scenario will lead to large propagation delays, thereby increasing the total task processing delay; most MEC frameworks that use an integrated air-ground-space network to assist in computing have the problem of centralized use of satellite computing; MEC methods that use drones to assist have the problem of ignoring drone coverage and drone utilization; to address the above problems, the present invention uses multiple drones in an integrated air-ground-space network to collaboratively provide computing services, thereby optimizing the propagation delay caused by satellite computing, thereby reducing the computing delay of ground vehicles and enhancing the vehicle's task processing capabilities. In order to improve the utilization rate of drones, the wide coverage characteristics of low-orbit satellites are used to cluster drones. Each vehicle unloads the task to the corresponding drone cluster. In view of the poor effect of traditional clustering methods on drone clustering, a distance-based drone clustering algorithm is designed. For the collaborative computing scenario of drones within a cluster, a new metric is designed, which comprehensively considers the two key factors affecting the total task processing delay: the computing power of the drone and the distance between the cluster head drone and other drones in the cluster. On this basis, an iterative algorithm is designed to determine the task splitting ratio within the cluster. In order to make this method closer to the actual situation, a vehicle-to-drone elevation angle model is established to represent the coverage of the drone. In view of the problem of insufficient drone coverage, the vehicle offloads the computing task to the cluster head drone CHU, which splits the task. In order to ensure the continuity of computing, a dynamic unloading strategy is designed to enable the vehicle to switch the corresponding drone cluster during movement and use local computing in a timely manner, thereby ensuring the sustainability of vehicle processing computing tasks.

[0023] Comparison of the system utility values ​​with the other two existing MEC methods shows that the method of the present invention is more effective and feasible. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 One of the application scenarios of this invention is the search and rescue scenario in post-disaster areas, which provides mobile edge computing services for search and rescue vehicles.

[0025] Figure 2 This is the UAV coverage model, and the UAV coverage is determined by the elevation angle from the vehicle to the UAV.

[0026] Figure 3 The flowchart of the method proposed in the present invention is shown in FIG.

[0027] Figure 4 This is a comparison chart of the utility values ​​of the method described in the present invention and other mobile edge computing methods.

[0028] Figure 5 Comparison of the average computational delay of each computational task at different vehicle speeds between the method proposed in the present invention and other algorithm combinations.

[0029] Figure 6 This is a comparison of the average computing delay of each computing task under different average task sizes between the method proposed in the present invention and other algorithm combinations. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] like Figure 1 As shown, the present invention considers an emergency mobile edge computing scenario for post-disaster search and rescue, where some rescue vehicles are traveling counterclockwise along the street. The rescue vehicles need to collect environmental information and survivor information during movement, generating a large number of computing tasks. In disaster scenarios, ground MEC servers are usually severely damaged and cannot be restored in a short time. Due to the limited local computing power of the vehicle, relying solely on the local computing of the vehicle will result in a large computing delay. In order to avoid this problem, the present invention proposes an emergency mobile edge computing method based on an integrated air-space-ground network to assist vehicle computing.

[0032] The method proposed in the present invention involves three layers: satellite layer, drone layer and ground layer: At the satellite level, there is a LEO satellite deployed above the area, using To represent the 3D coordinates of the LEO satellite. Before the vehicle starts to drive, the LEO satellite collects the initial position information of the UAV and the initial position information of the vehicle. The UAVs are clustered based on the initial position information of the vehicle and the initial position information of the UAV to determine the appropriate Cluster-Head UAV (CHU) and the UAVs in each cluster. The clustering results containing the information of the UAVs in each cluster are then returned to the corresponding CHU, and each CHU corresponds to a vehicle.

[0033] At the drone level, there are Drones equipped with MEC servers are evenly distributed over the area to process The computing tasks generated by each vehicle. The 3D coordinates of To indicate that , the initial position set of all UAVs is expressed as , assuming that the position of each drone is fixed.

[0034] On the ground floor, there are The car driving on the road at a speed of When the vehicle The 3D coordinates of ,in, , , is the total number of time slots. During the driving process of the vehicle, the computing tasks generated by the vehicle are sent to their corresponding CHUs.

[0035] After receiving the task, CHU splits it and distributes it to the drones in the drone cluster for calculation according to the determined task split ratio. Each vehicle is associated with only one CHU. Therefore, define Indicates vehicle The corresponding CHU is used Represents its 3D coordinates.

[0036] Step 1: Modeling channels between layers (1) Vehicle-to-UAV link: The link between the ground vehicle and the UAV is considered as a line-of-sight link. Therefore, the channel gain of the vehicle-to-UAV link can be modeled as ,in Indicates the reference distance The channel gain at For vehicles with CHU The distance between the vehicles can therefore be calculated To CHU Transfer rate:

[0037] in , Respectively represent vehicles and CHU The bandwidth and channel noise power between them. For vehicles of transmission power.

[0038] (2) Low-orbit satellite to UAV link: When the low-orbit satellite sends information about the UAVs in the cluster to the CHU, the effect of rainfall attenuation on the signal is considered. The rainfall attenuation between is modeled as a random process following the Weibull distribution. Therefore, the interference from the LEO satellite to the UAV link can be calculated using the following formula:

[0039] in , CHU and the antenna gain of the LEO satellite, It's CHU The distance to the LEO satellite, is the wavelength. Rain attenuation Weibull distribution modeling is used. Based on this, the satellite and CHU (S2U) transmission rate:

[0040] in Represents CHU The channel bandwidth of the satellite communication link, represents the transmission power of the LEO satellite, represents the power of the background noise, Rain attenuation for the link from LEO satellite to UAV.

[0041] (3) UAV-to-UAV link: Unlike ground communication, communication between UAVs usually has a line-of-sight path. Therefore, in this model, the UAV-to-UAV (U2U) channel can be modeled as a free space path loss model. With other UAVs in the same cluster The channel gain is modeled as follows:

[0042] in, The reference distance The channel gain at For CHU With other drones in the cluster The distance between is the free space path loss coefficient, obey The standard normal distribution of represents the small-scale fading component with zero mean and unity variance.

[0043] Since the link between UAVs is usually a line-of-sight link, the interference between UAVs is usually greater than other types of interference. Therefore, the present invention only considers the interference of this type of factors. The present invention only models the interference between UAVs, which can be applied to a variety of scenarios in the communication scenario. Based on this, CHU can be calculated With other drones in the cluster The transmission rates between are as follows:

[0044] in represents the channel bandwidth of the UAV-to-UAV communication link, Represents CHU The transmission power, is the background noise power. , are the transmit power and channel gain of other UAVs respectively, It means the sum of the signals of other UAVs except the current UAV. For example, there are 3 UAVs in a UAV cluster, numbered 1, 2, and 3. If the rate of UAV No. 1 needs to be calculated, the signals of UAV No. 2 and UAV No. 3 need to be regarded as interference. UAV No. 2 and UAV No. 3 are the other UAVs mentioned above. The sum is to find the sum of the signals of UAV No. 2 and UAV No. 3 in addition to the signal of UAV No. 1.

[0045] Step 2: Modeling the communication delay of each layer (1) Satellite-to-UAV communication delay LEO Satellite and CHU The transmission delay between is expressed as:

[0046] in, For the The amount of clustering information data of each drone cluster. Since the LEO satellite is located far away from the drone layer, its propagation delay is larger than other delays and cannot be ignored. Based on this, the LEO satellite and CHU The propagation delay between can be calculated as:

[0047] in, It's CHU The distance to the LEO satellite, It's the speed of light.

[0048] (2) Communication delay between vehicle and drone vehicle with CHU The transmission delay between

[0049] in, is the size of the task data.

[0050] (3) UAV-to-UAV communication delay CHU With other drones in the cluster The transmission delay between can be expressed as:

[0051] in, For CHU Offloading tasks to UAVs task splitting ratio.

[0052] (4) Calculation delay As an example, in the proposed method, the vehicle has two ways to perform computation: local computation and offloading the task to the drone cluster. , To represent vehicles Adopt local computing and offload computing to drone clusters.

[0053] when When the vehicle Offloading computational tasks to CHU ,otherwise The task splitting is performed by the CHU and offloaded to the corresponding UAVs in the cluster. The computational delay can be calculated as follows:

[0054] in, , UAV The computing power and computing resources required.

[0055] Using binary indicators To indicate CHU Whether to offload computing tasks to drones in the cluster ,if , which means CHU Offloading computational tasks to UAVs ,otherwise . Similarly, if =1, vehicle Perform local computation, otherwise The computational delay can be calculated as follows:

[0056] in For vehicles computing power.

[0057] Step 3: Modeling the drone coverage The coverage of the drone modeled by the present invention is as follows: Figure 2 As shown, in the time slot , the vehicle can be calculated using the following formula with CHU Horizontal distance between:

[0058] exist The elevation angle at the moment can be calculated as follows:

[0059] In time slot ,if ,vehicle Located at CHU If the elevation angle is within the coverage area, the computing task can be offloaded to the CHU. Below threshold , the vehicle is beyond the coverage of the current CHU and the computing task cannot be offloaded to the CHU. It should be noted that the elevation angle calculated by the formula is in degrees.

[0060] Step 4: Modeling the optimization problem When modeling the optimization problem, the present invention considers a continuous optimization process during the driving process of a vehicle. The vehicle generates computing tasks during the driving process. When one task is completed, the next task is generated. Therefore, the total delay of the system It can be calculated as:

[0061] in , For vehicles The corresponding number of drones in the drone cluster and the number of computing tasks completed during the entire driving process.

[0062] When the vehicle performs local computation, the drone does not participate in the computation. Therefore, when calculating the drone utilization, only the case where the vehicle offloads the task to the drone for computation is considered. The sum of the drone utilization in each time slot is It can be calculated as:

[0063] The optimization goal is to improve the utilization rate of drones while reducing the total system delay by obtaining the optimal drone clustering and unloading decision. Therefore, the optimization problem is modeled as follows:

[0064]

[0065] in, Indicates the clustering of drones. , representing the drone has been clustered, otherwise ; The present invention assumes that each drone can only be assigned to one drone cluster; and Represents the weight coefficient, which can be flexibly adjusted according to the needs of the scene.

[0066] Due to constraints and constraints is non-convex, and is nonlinear, so this optimization problem is a mixed integer nonlinear programming problem (MINLP), which cannot be solved by traditional methods. To solve this complex MINLP problem, it is decomposed into three sub-problems: UAV clustering, task splitting ratio determination, and unloading strategy determination. In particular, the optimization function value is used as the system utility value.

[0067] Step 5: Drone Clustering Aiming at the UAV clustering scenario, a distance-based UAV Cluster (DUC) algorithm assisted by low-orbit satellite is proposed. Before the DUC algorithm is executed, the LEO satellite collects the location information of vehicles and UAVs as the input of the DUC algorithm. The DUC algorithm can be divided into two parts: initial CHU determination and UAV determination within the cluster.

[0068] When the algorithm starts executing, a KD (K-Dimensional Tree) is built using the location information of all UAVs to quickly query the nearest UAV. The position information of the time slot (initial time slot) sets a flag variable , to indicate whether the vehicle successfully finds the appropriate CHU. Before finding a suitable CHU, loop to query the drones closest to the vehicle, sort the indexes in ascending order of distance and return the sorted indexes. In the query results, first check whether each drone has been assigned to other clusters, that is, ,if , then the drone is removed from the waiting queue. Then, the elevation angle between the vehicle and the drone is calculated. ,if , the drone is considered to be a qualified CHU of the current vehicle, the search for the current vehicle CHU is stopped, and the The above process is repeated until all vehicles have determined their corresponding CHUs.

[0069] After determining the CHU corresponding to each vehicle, we start to query the UAV closest to the CHU to determine other UAVs in the UAV cluster. First, we use the KD tree to query the distance of the UAV closest to the current CHU to obtain all possible UAV indexes, and then filter out invalid UAVs: First, we filter out the unselected UAVs from the query results, i.e. , and exclude the index of CHU to ensure that CHU will not be added to other drone clusters repeatedly. After that, for the qualified drones, calculate the distance between it and the current CHU, and check whether the CHU closest to the drone is the current CHU; if so, it is regarded as a valid drone and added to the drone cluster corresponding to the CHU. Repeat the above process until all drones are divided into a drone cluster. And ensure that the number of drones in the drone cluster does not exceed the maximum number .

[0070] After the algorithm is executed, the clustering strategy can be obtained , which contains the indexes of all CHUs and the UAV index in the cluster corresponding to each CHU. The LEO satellite sends the clustering results to the corresponding CHU, and the CHU further processes the computing tasks based on the clustering results.

[0071] Step 6: Intra-cluster task splitting When the CHU receives the clustering results from the LEO satellite, the vehicle can offload the computing tasks to the corresponding CHU. The CHU performs intra-cluster task splitting and offloads the corresponding computing tasks to the UAVs in the cluster for intra-cluster collaborative computing.

[0072] The total processing delay of the computing task is composed of the maximum value of the sum of the vehicle-to-UAV transmission delay, the UAV-to-UAV transmission delay, and the task processing delay. In the previous step, the total system delay was modeled. Since the satellite communication delay is fixed, the vehicle will offload the computing task to the CHU, so the vehicle-to-UAV transmission delay is not within the optimization range of this step. Therefore, the goal is to determine the optimal task splitting ratio within the UAV cluster to minimize the total task processing delay. Therefore, this subproblem can be modeled as:

[0073]

[0074] Through the modeling of the sub-problems, it can be seen that the transmission delay and computing delay are mainly related to the distance and the computing power of the drone. To this end, the present invention proposes an indicator called estimated computing delay (ECD), and designs an iterative algorithm based on this indicator to determine the final task split ratio in the drone cluster. ECD is the split ratio determined in the current iteration cycle. The normalized vector of the distance from the drone to the CHU and the calculated delay. it In the UAV The ECD can be calculated as:

[0075] in, and are the weight coefficients of distance and current computing delay respectively. In the computing scenario within the drone cluster, a larger distance will lead to a larger transmission delay, and a smaller drone computing power will lead to a larger computing delay. Therefore, after normalizing the distance and computing delay, the reciprocal is taken to ensure that the smaller the distance and the greater the computing power, the larger the ECD of the drone, which is convenient for determining the task splitting ratio in subsequent iterations. The weight coefficient is used to adjust the weights of the two, so as to achieve a better task splitting ratio determination effect.

[0076] When CHU From the vehicle When computing tasks, the tasks are first split evenly, and the initial ECD of the iterative algorithm is calculated based on this to iterate; after each iteration, the latest task split ratio is determined based on the ECD of the current iteration cycle. Take the sum of the ECDs of the drones in the cluster, divide the ECD of each drone by the sum of the ECDs of the drones in the cluster, and get the latest task split ratio for each drone. Repeat the above process until the difference between the task split ratio of the new cycle and the task split ratio of the previous cycle is less than the threshold tol , at this point the algorithm converges and determines the final task splitting rate within the cluster .

[0077] Step 7: Determine the vehicle dynamic unloading strategy The present invention uses com Indicator to indicate whether there is a computing task in progress in the current time slot. , it means that a computing task is in progress and the current time slot will be skipped. Otherwise, if , a new computing task will be generated and further processed. After generating a new computing task, first determine whether the vehicle is within the coverage of the current CHU, that is, calculate the vehicle To CHU Elevation .if , then set it to , the vehicle will offload the computing task to the CHU If the vehicle If the vehicle is beyond the coverage of the current CHU but within the coverage of the next CHU, the CHU corresponding to the vehicle will be switched to the next CHU and the computing task will be offloaded to the drone cluster corresponding to the CHU for computing. If it is not within the coverage area of ​​any CHU, it performs local calculation and sets .

[0078] After determining the task offloading strategy, the completion time of the task is calculated to determine which time slots can be skipped. Assuming that the task return time is ignored, the time to complete this computing task is composed of the maximum value of the sum of the vehicle-to-UAV transmission delay, the intra-cluster UAV-to-UAV transmission delay, and the computing delay. Therefore, the total task processing delay can be modeled as follows:

[0079] The above are all the steps of the emergency mobile edge computing method based on air-ground integration proposed by us. In order to have a clearer understanding of all the steps of our method, the overall process of the method is designed, such as Figure 3 shown.

[0080] Figure 4 The optimization function values ​​of the optimization problem modeled in step 4 are shown with different numbers of time slots, corresponding to the utility values ​​on the y-axis. More time slots means longer vehicle travel time. As can be seen from the figure, our method has the highest utility value, which means that our solution improves the utilization rate of drones while reducing the total system latency. And as the vehicle's operating time increases, the advantage of our method becomes more obvious. Figure 5 and Figure 6 The figure shows the comparison of the average computational latency of each task between our proposed method and other methods at different vehicle speeds and task data sizes. In the mobile edge scenario, the smaller the task computational latency, the faster the task computation, and the better the method effect. As can be seen from the figure, our solution always maintains the lowest computational latency, and can show good optimization effects for different vehicle speeds and task data sizes.

[0081] Based on the above conception, the present invention provides an emergency mobile edge computing system based on an air-ground integrated network, including a vehicle corresponding CHU acquisition module, a clustering strategy acquisition module, a splitting module and an execution module; The vehicle-corresponding CHU acquisition module collects the location information of vehicles and drones based on LEO satellites and obtains the CHUs corresponding to all vehicles; The clustering strategy acquisition module is used to combine the CHU corresponding to each vehicle, obtain the UAV closest to the CHU, determine other UAVs in the UAV cluster, and obtain the clustering strategy. The clustering strategy contains the indexes of all CHUs and the UAV index in the cluster corresponding to each CHU, and sends the clustering results to the corresponding CHU. The CHU further processes the computing task according to the clustering results. The splitting module is used to perform intra-cluster task splitting and offload the corresponding computing tasks to the UAVs in the cluster for intra-cluster collaborative computing; The execution module is used to determine the vehicle task unloading strategy according to the position relationship between the vehicle and the current CHU, switch the corresponding UAV cluster or use local calculation.

[0082] In the third aspect, the present invention provides an emergency mobile communication system based on an integrated air-space-ground network, including a vehicle, a LEO satellite and a drone equipped with a MEC server. The LEO satellite collects the initial position information of the drone and the initial position information of the vehicle. The drone can process the computing tasks generated by the vehicle. When the vehicle is driving on the road, the vehicle and the drone, the low-orbit satellite and the drone, and the drones can communicate with each other. When the system is running, the computing tasks are processed based on the above-mentioned emergency mobile edge computing method based on the integrated air-space-ground network.

[0083] The present invention may also provide a computer device, including a processor and a memory, the memory is used to store a computer executable program, the processor reads part or all of the computer executable program from the memory and executes it, and when the processor executes part or all of the computer executable program, it can implement the emergency mobile edge computing method based on the integrated air-space-ground network of the present invention. The computer device may be a laptop computer, a desktop computer or a workstation. The processor of the present invention may be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC) or a readily available field programmable gate array (FPGA).

[0084] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. An emergency mobile edge computing method based on an air-ground integrated network is characterized in that: The UAV clustering and task splitting methods are jointly optimized to maximize the UAV utilization while minimizing the total system latency. A dynamic offloading strategy is adopted to enable the vehicle to switch UAV clusters or use local computing during movement. The specific steps include: Collect the location information of vehicles and drones based on LEO satellites and obtain the CHU corresponding to all vehicles; Combine the CHU corresponding to each vehicle, obtain the drone closest to the CHU, determine the other drones in the drone cluster, and obtain the clustering strategy ,The clustering strategy includes the indexes of all CHUs and the UAV index in the cluster corresponding to each CHU, and the clustering results are sent to the corresponding CHU, which further processes the computing tasks according to the clustering results; Perform task splitting within the cluster and offload the corresponding computing tasks to the UAVs within the cluster for collaborative computing within the cluster; The vehicle task offloading strategy is determined according to the position relationship between the vehicle and the current CHU, and the corresponding drone cluster is switched or local computing is used.

2. The emergency mobile edge computing method based on the air-ground integrated network according to claim 1 is characterized in that: Based on the LEO satellite, the location information of vehicles and UAVs is collected to obtain the CHU corresponding to all vehicles, including: using the location information of all UAVs to establish a KD tree; for each vehicle, obtain its location information in the initial time slot and set the flag variable , to indicate whether the vehicle successfully finds a suitable CHU. Before obtaining a suitable CHU, it loops to query the UAVs closest to the vehicle, sorts the indexes in ascending order of distance, and returns the sorted indexes; In the obtained results, check whether each drone has been assigned to other clusters. If so, remove the drone from the waiting queue; calculate the elevation angle between the vehicle and the drone If the elevation angle is not less than the corresponding threshold, the drone is considered to be a qualified CHU of the current vehicle, the search for the current vehicle CHU is stopped, and the ; The above process is repeated until all vehicles have determined their corresponding CHUs.

3. The emergency mobile edge computing method based on the air-ground integrated network according to claim 2 is characterized in that: Combine the CHU corresponding to each vehicle, obtain the drone closest to the CHU, determine the other drones in the drone cluster, and obtain the clustering strategy The method includes: using the KD tree to obtain the distance of the drone closest to the current CHU, obtaining all possible drone indexes, filtering unselected drones from the obtained results, and excluding the index of the CHU. For drones that meet the conditions, calculating the distance between the drone and the current CHU, and checking whether the CHU closest to the drone is the current CHU; if so, treating it as a valid drone and adding it to the drone cluster corresponding to the CHU; Repeat the above process until all drones are divided into a drone cluster (and ensure that the number of drones in the drone cluster does not exceed the maximum number, and obtain the clustering strategy .

4. The emergency mobile edge computing method based on the air-ground integrated network according to claim 1 is characterized in that: CHU performs intra-cluster task splitting and offloads the corresponding computing tasks to the UAVs in the cluster for intra-cluster collaborative computing, including: The optimal task split ratio within the UAV cluster to minimize the total delay of task processing is modeled as: Iteratively calculate the final task split ratio in the drone cluster based on the estimated computing delay index: When performing calculation tasks, the tasks are evenly split, the initial ECD of the iterative algorithm is calculated and iterated, and after each iteration, the latest task split ratio is determined according to the ECD of the current iteration cycle; the sum of the ECDs of the UAVs in the cluster is taken, and the ECD of each UAV is divided by the sum of the ECDs of the UAVs in the cluster to obtain the latest task split ratio of each UAV; the above process is repeated until the difference between the task split ratio of the new cycle and the task split ratio of the previous cycle is less than the threshold, and the final task split ratio in the cluster is obtained.

5. The emergency mobile edge computing method based on the air-ground integrated network according to claim 1 is characterized in that: When the vehicle is dynamically unloaded, if there is an ongoing computing task in the current time slot, the current time slot is skipped; if there is no ongoing computing task in the current time slot, a new computing task is generated; When generating new computing tasks, if the vehicle is within the coverage of the current CHU, the vehicle will offload the computing tasks to the CHU. If the vehicle is beyond the coverage of the current CHU but within the coverage of the next CHU, the CHU corresponding to the vehicle will be switched to the next CHU, and the computing tasks will be offloaded to the drone cluster corresponding to the CHU for computing. If the vehicle is not within the coverage of any CHU, local calculations are performed.

6. The emergency mobile edge computing method based on the air-ground integrated network according to claim 1 is characterized in that: The completion time of the task is calculated according to the task offloading strategy, and the time slots that can be skipped are determined. The time to complete the computing task includes the maximum value of the sum of the vehicle-to-UAV transmission delay, the intra-cluster UAV-to-UAV transmission delay, and the computing delay.

7. The emergency mobile edge computing method based on the air-ground integrated network according to claim 1 is characterized in that: Jointly optimize UAV clustering and task splitting methods to maximize UAV utilization while minimizing the total system delay. Based on the modeling of vehicle-to-UAV links, low-orbit satellite-to-UAV links, and inter-UAV links, the satellite-to-UAV communication delay, vehicle-to-UAV communication delay, UAV-to-UAV communication delay, and the computational delay of UAVs within a cluster are modeled; Then, the total system delay is obtained according to the delay modeling; The optimization goal is to obtain the optimal UAV clustering and unloading decision, while reducing the total system delay and improving the UAV utilization rate. The optimization problem is modeled as: in, Indicates the clustering of drones. If , representing the drone has been clustered, otherwise ; Assume that each drone can only be assigned to one drone cluster; and Represents the weight coefficient.

8. An emergency mobile edge computing system based on an air-ground integrated network, characterized in that: It includes a vehicle corresponding CHU acquisition module, a clustering strategy acquisition module, a splitting module and an execution module; The vehicle-corresponding CHU acquisition module collects the location information of vehicles and drones based on LEO satellites and obtains the CHUs corresponding to all vehicles; The clustering strategy acquisition module is used to combine the CHU corresponding to each vehicle, obtain the UAV closest to the CHU, determine other UAVs in the UAV cluster, and obtain the clustering strategy. The clustering strategy contains the indexes of all CHUs and the UAV index in the cluster corresponding to each CHU, and sends the clustering results to the corresponding CHU. The CHU further processes the computing task according to the clustering results. The splitting module is used to perform intra-cluster task splitting and offload the corresponding computing tasks to the UAVs in the cluster for intra-cluster collaborative computing; The execution module is used to determine the vehicle task unloading strategy according to the position relationship between the vehicle and the current CHU, switch the corresponding UAV cluster or use local calculation.

9. The emergency mobile communication system based on the integrated air-ground-space network is characterized by: The invention comprises a vehicle, a LEO satellite and a drone equipped with a MEC server. The LEO satellite collects the initial position information of the drone and the initial position information of the vehicle. The drone can process the computing tasks generated by the vehicle. When the vehicle is driving on the road, the vehicle and the drone, the low-orbit satellite and the drone, and the drones themselves can communicate. When the system is running, the computing tasks are processed based on the emergency mobile edge computing method based on the integrated air-space-ground network as described in any one of claims 1-7.

10. A computer device, characterized in that: It includes a processor and a memory, the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and when the processor executes the computer executable program, it can implement the emergency mobile edge computing method based on the air-space-ground integrated network as described in any one of claims 1-7.

Images