Unmanned aerial vehicle task calculation flight strategy optimization method and system

Abstract

The invention discloses an unmanned aerial vehicle task calculation flight strategy optimization method and system, and is used for solving the technical problem that the processing cost of a calculation task is too high due to an existing unmanned aerial vehicle task calculation flight strategy optimization technology. The method comprises the steps of determining a task transmission unloading rate based on acquired user coordinate data and unmanned aerial vehicle coordinate data; based on preset unloading decision constraints, the obtained unmanned aerial vehicle state data, the task transmission unloading rate and the obtained user calculation task data, target task completion time delay and target task completion energy consumption are determined, and the target task completion time delay and the target task completion energy consumption are determined in combination with the preset unloading decision constraints, the preset unmanned aerial vehicle constraints, the user unmanned aerial vehicle unloading data and the unmanned aerial vehicle coordinate data. Constructing a joint optimization problem; converting the joint optimization problem, and determining a multi-agent Markov decision process model; and performing optimization solution on the multi-agent Markov decision process model by adopting a multi-agent deep reinforcement learning algorithm to generate a target unmanned aerial vehicle task calculation flight strategy.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a method and system for optimizing a flight strategy for mission calculation of an unmanned aerial vehicle. Background Art

[0002] In recent years, UAV technology has developed rapidly. Due to its small size, flexibility, easy deployment, high adaptability, and high probability line-of-sight link, it is not only widely used in the military field, but also has received extensive attention in the civilian field. The application of UAVs in disaster relief, communication assistance, network coverage, etc. has made significant progress.

[0003] Drones can hover and fly. As high-altitude mobile platforms, they can carry mobile edge computing servers, solve the problems of traditional mobile edge computing deployment, and provide more flexible and efficient computing and communication services. Especially in remote areas and disaster areas, drones can be quickly deployed as MEC (Multi-access Edge Computing Node) nodes to ensure the continuity of network connection and data processing, significantly improve emergency response capabilities and network service quality, and thus provide timely and reliable communication and computing services.

[0004] Most of the existing UAV mission calculation flight strategy optimization technologies consider that a task can be calculated locally by the user or offloaded to the UAV for calculation. The user side considers how to offload the subtask. However, the user side will consume a lot of energy to offload and calculate at the same time, resulting in excessively high processing costs for the computing tasks. Summary of the invention

[0005] The present invention provides a method and system for optimizing flight strategy for unmanned aerial vehicle mission calculation, which are used to solve the technical problem that the processing cost of computing tasks is too high due to the existing unmanned aerial vehicle mission calculation flight strategy optimization technology.

[0006] A first aspect of the present invention provides a method for optimizing a UAV mission calculation flight strategy, comprising:

[0007] Obtain user computing task data, user coordinate data, drone status data, and drone coordinate data;

[0008] Determine user-drone distance data based on the user coordinate data and the drone coordinate data, and determine a task transmission offloading rate according to the user-drone distance data and drone height data in the drone coordinate data;

[0009] Based on the preset unloading decision constraints, determine the user drone unloading data, and determine the target task completion delay and the target task completion energy consumption according to the user drone unloading data, the drone status data, the task transmission unloading rate, and the computing task complexity and computing task size in the user computing task data;

[0010] A joint optimization problem is constructed using the preset unloading decision constraints, the target task completion delay, the target task completion energy consumption, the preset drone constraints, the user drone unloading data and the drone coordinate data;

[0011] Transforming the joint optimization problem to determine a multi-agent Markov decision process model;

[0012] A multi-agent deep reinforcement learning algorithm is used to optimize and solve the multi-agent Markov decision process model to generate a target UAV mission calculation flight strategy.

[0013] Optionally, the drone coordinate data includes three-dimensional cluster drone coordinates and three-dimensional calculation drone coordinates; the user drone distance data includes distance data between the user and the cluster drone, distance data between the user and the calculation drone, and distance data between drones; determining the user drone distance data based on the user coordinate data and the drone coordinate data includes:

[0014] The three-dimensional user coordinates in the user coordinate data are used to perform Euclidean distance calculations with the three-dimensional cluster drone coordinates and the three-dimensional calculation drone coordinates, and the distance data between the user and the cluster drone and the distance data between the user and the calculation drone are output;

[0015] The distance data between the drones is determined according to the three-dimensional cluster drone coordinates and the three-dimensional calculated drone coordinates.

[0016] Optionally, the task transmission offloading rate includes a user-drone transmission rate and a drone-drone task offloading rate; and determining the task transmission offloading rate according to the user-drone distance data and the drone height data in the drone coordinate data includes:

[0017] Determine the connection probability between the user and the drone based on the distance data between the user and the cluster drone, the distance data between the user and the calculation drone, and the drone height data in the drone coordinate data;

[0018] Determine the link channel power gain between the user and the drone according to the connection probability between the user and the drone, the distance data between the user and the cluster drone, and the distance data between the user and the computing drone;

[0019] The preset Shannon formula is used to determine the transmission rate between the user and the UAV according to the link channel power gain between the user and the UAV;

[0020] Based on the distance data between the drones, a task offloading rate between the drones is determined.

[0021] Optionally, the computing task complexity includes computing UAV computing task complexity and cluster UAV computing task complexity; the computing task size includes computing UAV computing task size and cluster UAV computing task size; the user-UAV transmission rate includes user-computing UAV transmission rate and user-cluster UAV transmission rate; the determining the target task completion delay and the target task completion energy consumption according to the user UAV unloading data, the UAV status data, the task transmission unloading rate and the computing task complexity and computing task size in the user computing task data includes:

[0022] According to the computing task complexity of the computing drone, the computing task size of the computing drone, and the transmission rate between the user and the computing drone, the completion delay of the user unloading the computing drone task and the energy consumption of the user unloading the computing drone task are determined;

[0023] Based on the user drone unloading data, the complexity of the cluster drone computing task, the size of the cluster drone computing task, and the transmission rate between the user and the cluster drone, determine the completion delay of the user unloading cluster drone task and the energy consumption of the user unloading cluster drone task;

[0024] Determine the target task completion delay according to the drone status data, the user drone uninstallation data, the user uninstallation calculation drone task completion delay and the fixed user uninstallation cluster drone task completion delay;

[0025] The target task completion energy consumption is determined based on the drone status data, the user drone unloading data, the user unloading calculation drone task completion energy consumption and the user unloading cluster drone task completion energy consumption.

[0026] Optionally, the using of the preset unloading decision constraint, the target task completion delay, the target task completion energy consumption, the preset drone constraint, the user drone unloading data and the drone coordinate data to construct a joint optimization problem includes:

[0027] Determining a task completion cost based on the target task completion delay and the target task completion energy consumption;

[0028] A joint optimization problem is constructed by using the preset unloading decision constraints, the task completion cost, the preset drone constraints, the user drone unloading data and the drone coordinate data.

[0029] Optionally, the user drone unloading data includes first unloading data, second unloading data, third unloading data and fourth unloading data; the preset unloading decision constraints include first unloading decision constraints, second unloading decision constraints, third unloading decision constraints, fourth unloading decision constraints and fifth unloading decision constraints; the preset drone constraints include first drone constraints, second drone constraints, third drone constraints, fourth drone constraints, fifth drone constraints, sixth drone constraints, seventh drone constraints and eighth drone constraints; the drone coordinate data also includes two-dimensional cluster drone coordinates; the joint optimization problem is specifically:

[0030]

[0031] Among them, P is a joint optimization problem; is the two-dimensional cluster UAV coordinate, representing the optimization variable of the cluster UAV trajectory, , is the horizontal coordinate of the mth swarm UAV at time slot t, is the ordinate of the mth cluster UAV at time slot t; B is the user UAV unloading data, which represents the unloading decision variables of the user and cluster UAV, , is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the uth user offloads the task to the nth computing drone for calculation at time slot t. The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t. The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t. The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local calculation on the task at time slot t, where M is the set of cluster drones; is the task completion cost at time slot t; T is the time slot set; C1 is the first offloading decision constraint, indicating that the task of each ground user (user) can only be offloaded to a cluster drone or a computing drone, and U is the set of users; C2 is the second offloading decision constraint, indicating the constraint of the offloading decision variables calculated locally by the cluster drone; C3 is the third offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to the computing drone, and N is the set of computing drones; C4 is the fourth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to another cluster drone; C5 is the fifth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone; C6 is the first drone constraint, indicating the mobility constraint of the service range of each cluster drone, The maximum length of the service area divided by the mth cluster drone, is the maximum width of the service area divided by the mth swarm drone; C7 is the second drone constraint, indicating the coverage constraint that ground users can offload tasks to swarm drones. is the two-dimensional user coordinate of the u-th user at time slot t, is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal coverage of the mth cluster UAV; C8 is the constraint of the third UAV, which means the constraint to avoid collision between cluster UAVs. is the three-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional coordinates of a cluster of drones at time slot t, is the minimum safety distance; C9 is the fourth drone constraint, which indicates the constraint to avoid overlapping coverage between cluster drones. is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional cluster UAV coordinates of the cluster UAV at time slot t; C10 is the fifth UAV constraint, which represents the constraint of the horizontal flight angle of the cluster UAV. is the horizontal flight angle of the mth swarm UAV at time slot t; C11 is the sixth UAV constraint, which indicates the constraint on the horizontal flight distance of the swarm UAV. is the flight distance of the mth swarm drone at time slot t, is the maximum flight distance allowed by the cluster drone; C12 is the seventh drone constraint, indicating the state of the cluster drone. is the drone status data of the mth cluster drone at time slot t. When , it means that the mth cluster drone is in a fault state at time slot t. , indicating that the mth cluster drone is in a normal state at time slot t; C13 is the eighth drone constraint, indicating that the computing task in the tth time slot needs to be completed before the maximum tolerable delay. is the target task completion delay, is the maximum tolerable delay time.

[0032] A second aspect of the present invention provides a UAV mission calculation flight strategy optimization system, comprising:

[0033] The acquisition module is used to obtain user computing task data, user coordinate data, drone status data and drone coordinate data;

[0034] A determination module, configured to determine user-drone distance data based on the user coordinate data and the drone coordinate data, and determine a task transmission offloading rate according to the user-drone distance data and drone height data in the drone coordinate data;

[0035] The delay energy consumption determination module is used to determine the user drone unloading data based on the preset unloading decision constraints, and determine the target task completion delay and the target task completion energy consumption according to the user drone unloading data, the drone status data, the task transmission unloading rate and the computing task complexity and computing task size in the user computing task data;

[0036] A construction module, for constructing a joint optimization problem by using the preset unloading decision constraints, the target task completion delay, the target task completion energy consumption, the preset drone constraints, the user drone unloading data and the drone coordinate data;

[0037] A conversion module, used for converting the joint optimization problem and determining a multi-agent Markov decision process model;

[0038] The solution module is used to optimize and solve the multi-agent Markov decision process model using a multi-agent deep reinforcement learning algorithm to generate a target UAV mission calculation flight strategy.

[0039] A third aspect of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for optimizing the flight strategy for calculating a drone mission as described in any one of the above items.

[0040] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the steps of the method for optimizing the flight strategy for calculating a UAV mission as described in any one of the above items.

[0041] A fifth aspect of the present invention provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the steps of the method for optimizing the flight strategy for calculating a UAV mission as described in any one of the above items.

[0042] It can be seen from the above technical solutions that the present invention has the following advantages:

[0043] The above technical scheme of the present invention provides a method for optimizing a UAV task calculation flight strategy. First, user calculation task data, user coordinate data, UAV status data and UAV coordinate data are obtained; then, based on the user coordinate data and the UAV coordinate data, the user UAV distance data is determined, and the task transmission unloading rate is determined according to the user UAV distance data and the UAV height data in the UAV coordinate data; based on the preset unloading decision constraints, the user UAV unloading data is determined, and the target task completion delay and the target task completion energy consumption are determined according to the user UAV unloading data, the UAV status data, the task transmission unloading rate and the calculation task complexity and the calculation task size in the user calculation task data; the preset unloading decision constraints, the target task completion delay, the target task completion energy consumption, the preset UAV constraints, the user UAV unloading data and the UAV coordinate data are used to construct a joint optimization problem; the joint optimization problem is transformed to determine the multi-agent Markov decision process model; finally, a multi-agent deep reinforcement learning algorithm is used to optimize and solve the multi-agent Markov decision process model to generate a target UAV task calculation flight strategy; based on the above scheme, the acquired user calculation task data, user coordinate data, and UAV coordinate data are processed to obtain the target task completion delay and target task completion energy consumption, and then combined with the preset unloading decision constraints, preset UAV constraints, user UAV unloading data, and UAV status data to obtain a multi-agent Markov decision process model, and the model is solved by a multi-agent deep reinforcement learning algorithm to generate a process of calculating the target UAV task flight strategy. The present invention comprehensively considers the delay and energy consumption, so that the same order of magnitude of changes are produced in the optimization process. At the same time, the present invention controls the UAV to calculate the user task according to the obtained target UAV task calculation flight strategy, and does not require the user end to unload and calculate at the same time, which can reduce energy consumption and thus reduce the processing cost of the computing task. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 A flowchart of the steps of a method for optimizing a UAV mission calculation flight strategy provided in Embodiment 1 of the present invention;

[0046] Figure 2 A schematic diagram of the network architecture of a system model corresponding to the method for optimizing the flight strategy for calculating a UAV mission provided in the first embodiment of the present invention;

[0047] Figure 3 This is a structural block diagram of a UAV mission calculation flight strategy optimization system provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0048] The embodiments of the present invention provide a method and system for optimizing flight strategy for UAV mission calculation, which are used to solve the technical problem that the processing cost of computing tasks is too high due to the existing UAV mission calculation flight strategy optimization technology.

[0049] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only 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.

[0050] See also Figure 1 , Figure 1 A flowchart of the steps of a method for optimizing a UAV mission calculation flight strategy provided in Embodiment 1 of the present invention.

[0051] The present invention provides a method for optimizing a UAV mission calculation flight strategy, comprising:

[0052] Step 101: Obtain user computing task data, user coordinate data, drone status data, and drone coordinate data.

[0053] User computing task data includes computing task complexity, computing task size, and maximum tolerable delay time.

[0054] The drone coordinate data includes three-dimensional cluster drone coordinates, three-dimensional calculated drone coordinates, two-dimensional cluster drone coordinates, and two-dimensional calculated drone coordinates.

[0055] The user coordinate data includes three-dimensional user coordinates and two-dimensional user coordinates.

[0056] Please note that Figure 2 The system model corresponding to the UAV task calculation flight strategy optimization method proposed in the present invention is a three-layer network architecture of multi-UAV collaborative assisted mobile edge computing composed of ground terminal users U, cluster UAVs M and computing UAVs N, wherein ground terminal users U={1,2,...,u}, cluster UAVs M={1,2,...,m}, computing UAVs N={1,2,...,n}, n=1, that is, the present invention only considers one computing UAV (N=1), and ground users can offload tasks to cluster UAVs or computing UAVs (considering that ground users who are not within the coverage of cluster UAVs are offloaded to computing UAVs), wherein, if the computing resources of the cluster UAV are limited and cannot meet its computing needs, the cluster UAV can be offloaded to other cluster UAVs or computing UAVs for computing, and the cluster UAVs can process tasks in parallel.

[0057] Furthermore, the present invention divides time into multiple time slots t, and uses T={1,2,...,t} to represent the time slot set. Each user in each time slot t will generate corresponding user computing task data, and the user computing task data generated by each ground user in the tth time slot is represented by a triple ,in, The size of the computing task generated by the u-th user (ground user) in the t-th time slot. The computing task can be offloaded to the cluster drone or computing drone for execution; The computational complexity of the task generated for the u-th user (ground user) in the t-th time slot, i.e., the number of CPU cycles required to compute 1 bit of input data, usually depends on the type of application; is the maximum tolerable delay time, that is, the maximum time to complete the computing task in the tth time slot. It is meaningless to complete the computing task beyond this time.

[0058] Furthermore, in order to reduce the impact of the change in flight altitude on the energy consumption of the swarm UAVs, the swarm UAVs m are set to fly at a fixed altitude h m The three-dimensional coordinates of the flight, that is, the swarm drone (three-dimensional swarm drone coordinates) can be expressed as: , the two-dimensional coordinates of the cluster UAV can be expressed as: By setting the fifth and sixth drone constraints, the horizontal flight angle of the cluster drone is limited to , the flight distance is ,in, is the three-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal coordinate of the mth swarm UAV at time slot t, is the ordinate of the mth swarm drone at time slot t, is the drone height data of the mth cluster drone, is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal flight angle of the mth swarm UAV at time slot t, is the flight distance of the mth swarm drone at time slot t, The maximum flight distance allowed for swarm drones.

[0059] Based on the above foundation, the position update calculation formula of the cluster drone can be obtained, which is:

[0060]

[0061] in, is the horizontal coordinate of the mth swarm UAV at time slot t+1; is the ordinate of the mth swarm UAV at time slot t+1; is the horizontal coordinate of the mth swarm UAV at time slot t; is the ordinate of the mth swarm UAV at time slot t; is the horizontal flight angle of the mth swarm UAV at time slot t; is the flight distance of the mth swarm drone at time slot t.

[0062] Furthermore, in order to ensure that the swarm drone m moves within the corresponding service range, the two-dimensional swarm drone coordinates of the swarm drone m need to satisfy the set first drone constraint: ,in, is the horizontal coordinate of the mth swarm UAV at time slot t, is the ordinate of the mth swarm drone at time slot t, The maximum length of the service area divided by the mth cluster drone, The maximum width of the service area divided by the m-th cluster drone.

[0063] Furthermore, the three-dimensional user coordinates are , the two-dimensional user coordinates are , based on the set second drone constraint, when the ground user wants to offload the task to the cluster drone m, he needs to be within the horizontal coverage of the cluster drone m. The second drone constraint is specifically: ,in, is the two-dimensional user coordinate of the u-th user at time slot t, is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal coverage of the mth swarm drone.

[0064] Furthermore, in order to avoid collision between any two swarm drones and to avoid overlapping coverage, the distance between swarm drones should not be less than the minimum safety distance d according to the set third drone constraint and fourth drone constraint. min , the coverage between clustered drones should be no less than twice the horizontal coverage of clustered drones, where the third drone constraint can be expressed as: ; The fourth drone constraint can be expressed as: , is the three-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional coordinates of a cluster of drones at time slot t, is the minimum safe distance, A collection of swarm drones. is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The two-dimensional coordinates of a cluster of drones at time slot t.

[0065] Step 102: Determine the user-UAV distance data based on the user coordinate data and the UAV coordinate data, and determine the task transmission offloading rate according to the user-UAV distance data and the UAV height data in the UAV coordinate data.

[0066] The user drone distance data includes the distance data between the user and the cluster drone, the distance data between the user and the computing drone, and the distance data between drones; among them, the distance data between drones includes the distance between cluster drones (i.e., the distance between a cluster drone m and another cluster drone m'), and the distance between a cluster drone and the computing drone.

[0067] The task transmission offloading rate includes the transmission rate between users and drones and the task offloading rate between drones.

[0068] Specifically, the process of determining the user-drone distance data based on the user coordinate data and the drone coordinate data can be achieved by executing the following steps S11 to S12:

[0069] Step S11, using the three-dimensional user coordinates in the user coordinate data to perform Euclidean distance calculations with the three-dimensional cluster drone coordinates and the three-dimensional calculation drone coordinates, and outputting the distance data between the user and the cluster drone and the distance data between the user and the calculation drone;

[0070] Step S12: Determine the distance data between the drones based on the three-dimensional cluster drone coordinates and the three-dimensional calculated drone coordinates.

[0071] It should be noted that, similar to the existing research on UAV edge computing, the system model proposed in this invention is based on a quasi-static scenario, that is, during the data transmission process, the network topology of the cluster UAV and the ground user remains fixed. In the Cartesian coordinate system, the distance between the ground user u and the cluster UAV m or computing UAV n and the distance between the cluster UAV m and the cluster UAV m' or computing UAV n in time slot t can be obtained according to the Euclidean distance calculation formula. The processing process can be expressed as:

[0072]

[0073] in, is the distance between the u-th user and the m-th cluster drone at time slot t; is the three-dimensional user coordinate of the u-th user at time slot t; is the three-dimensional cluster drone coordinates of the mth cluster drone at time slot t; Calculate the distance between the u-th user and the n-th drone at time slot t; The three-dimensional calculated drone coordinates of the nth calculated drone. The calculated drone position is fixed, that is, the three-dimensional calculated drone coordinates remain unchanged. , is the horizontal coordinate of the nth calculated drone, Calculate the vertical coordinate of the nth drone, Calculate the drone altitude data for the nth drone; is the distance between the mth swarm drone and the m'th swarm drone at time slot t; is the three-dimensional cluster drone coordinates of the m'th cluster drone at time slot t; is the distance between the mth cluster UAV and the nth calculation UAV at time slot t; M is the set of cluster UAVs.

[0074] Further, the process of determining the task transmission offloading rate according to the user drone distance data and the drone height data in the drone coordinate data can be implemented by executing the following steps S21 to S24:

[0075] Step S21, determining the connection probability between the user and the drone based on the distance data between the user and the cluster drone, the distance data between the user and the calculation drone, and the drone height data in the drone coordinate data;

[0076] The user-UAV connection probability includes the LoS connection probability between the ground user u and the cluster UAV m or the computing UAV n at time slot t, and the NLoS connection probability between the ground user u and the cluster UAV m or the computing UAV n at time slot t.

[0077] It should be noted that since the communication channel between the ground user u and the cluster UAV m or computing UAV n is affected by the altitude, elevation angle and propagation environment type, its communication link model is established as a probabilistic path loss model containing LoS (Line of Sight) and NLoS (Non Line of Sight) with different occurrence probabilities. Therefore, the calculation process of the LoS connection probability between the ground user u and the cluster UAV m or computing UAV n in time slot t can be expressed as:

[0078]

[0079] in, Calculate the LoS connection probability between the u-th ground user and the m-th clustered UAV or the n-th UAV at time slot t; is the first environmental parameter (e.g. rural or urban); is the second environmental parameter; is the altitude data (drone altitude data in drone coordinate data), including the drone altitude data of the calculation drone and the drone altitude data of the cluster drone. , is the drone height data of the mth cluster drone, Calculate the drone altitude data for the nth drone; For distance data (user drone distance data), including, , is the distance between the u-th user and the m-th cluster drone at time slot t, Calculate the distance between the u-th user and the n-th drone at time slot t.

[0080] Furthermore, according to the LoS connection probability between the ground user u and the cluster UAV m or the computing UAV n in the time slot t, the NLoS connection probability between the ground user u and the cluster UAV m or the computing UAV n in the time slot t is determined. The process can be expressed as:

[0081]

[0082] in, Calculate the NLoS connection probability between the u-th ground user and the m-th swarm UAV or the n-th UAV at time slot t; Calculate the LoS connection probability between the u-th ground user and the m-th clustered UAV or the n-th UAV at time slot t.

[0083] Step S22: Determine the link channel power gain between the user and the drone according to the connection probability between the user and the drone, the distance data between the user and the cluster drone, and the distance data between the user and the calculation drone;

[0084] The link channel power gain between user and UAV is the link channel power gain from user u to cluster UAV m or to computing UAV n.

[0085] Specifically, according to the distance data between the user and the cluster drone and the distance data between the user and the computing drone, the path loss of the LoS from the ground user u to the cluster drone m or computing drone n, the path loss of the NLoS from the ground user u to the cluster drone m or computing drone n, and the free space path consumption are determined. The calculation process can be expressed as:

[0086]

[0087] in, Calculate the path loss of LoS between the u-th ground user and the m-th swarm UAV or the n-th UAV; is the free space path consumption; is the additional loss of LoS; Calculate the NLoS path loss for the u-th ground user and the m-th swarm UAV or the n-th UAV; is the additional loss of NLoS; f is the carrier frequency; D is the distance data (user drone distance data), including, , is the distance between the u-th user and the m-th cluster drone at time slot t, is the distance between the u-th user and the n-th drone at time slot t; c is the speed of light.

[0088] Furthermore, based on the free space path consumption obtained in the above steps, the LoS connection probability between the ground user u and the clustered UAV m or the calculated UAV n in time slot t, the LoS path loss between the ground user u and the clustered UAV m or the calculated UAV n, the NLoS connection probability between the ground user u and the clustered UAV m or the calculated UAV n in time slot t, and the NLoS path loss between the ground user u and the clustered UAV m or the calculated UAV n, the power gain of the link channel between the user and the UAV is determined. The process can be expressed as:

[0089]

[0090] in, Calculate the path loss of LoS between the u-th ground user and the m-th swarm UAV or the n-th UAV; is the free space path consumption; Calculate the LoS connection probability between the u-th ground user and the m-th clustered UAV or the n-th UAV at time slot t; Calculate the NLoS connection probability between the u-th ground user and the m-th swarm UAV or the n-th UAV at time slot t; Calculate the NLoS path loss for the u-th ground user and the m-th swarm UAV or the n-th UAV; Calculate the link channel power gain for the u-th user to the m-th clustered drone or to the n-th drone at time slot t.

[0091] Step S23: using the preset Shannon formula to determine the transmission rate between the user and the drone according to the link channel power gain between the user and the drone;

[0092] The user-to-UAV transmission rate is the transmission rate at which the task of ground user u is offloaded to cluster UAV m or to computing UAV n.

[0093] It should be noted that the process of determining the transmission rate between the user and the UAV by using the preset Shannon formula according to the link channel power increase between the user and the UAV can be expressed as:

[0094]

[0095] in, The transmission rate of the task offloading from the u-th ground user to the m-th cluster UAV or to the n-th computing UAV at time slot t; The channel bandwidth between the u-th user and the m-th cluster drone or the n-th computing drone; is the transmission power of the u-th user; Calculate the link channel power gain for the u-th user to the m-th clustered drone or to the n-th drone; is the additive white Gaussian noise power.

[0096] Step S24: Determine the task offloading rate between the UAVs based on the distance data between the UAVs.

[0097] The task offloading rate between drones is the task offloading rate between cluster drone m and cluster drone m' or between cluster drone n and computing drone n.

[0098] It should be noted that, since the altitude of the cluster drones and the computing drones is relatively high, the communication between them is considered to be LoS (Line of Sight) communication. Specifically, according to the distance data between drones, that is, according to the distance between cluster drones (that is, the distance between cluster drone m and another cluster drone m') or the distance between cluster drones and computing drones, the channel power gain between cluster drone m and cluster drone m' or computing drone n is determined; this process can be expressed as:

[0099]

[0100] in, Calculate the channel power gain of the mth cluster UAV and the m'th cluster UAV or the nth UAV at time slot t; The distance between the mth cluster UAV and the m'th cluster UAV or the nth calculation UAV at time slot t; is the channel power gain at a reference distance of 1 m; M is the set of clustered drones; N is the set of computing drones.

[0101] Furthermore, the task offloading rate between drones is determined based on the channel power gain of the cluster drone m and the cluster drone m' or the calculated drone n. The process can be expressed as:

[0102]

[0103] in, is the task offloading rate of the mth cluster UAV and the m'th cluster UAV or the nth computing UAV at time slot t; is the channel bandwidth between the mth cluster UAV and the nth computing UAV or the m'th cluster UAV; Calculate the channel power gain of the mth cluster UAV and the m'th cluster UAV or the nth UAV at time slot t; is the transmission power of the mth cluster UAV; is the additive white Gaussian noise power.

[0104] Step 103: Based on the preset unloading decision constraints, determine the user drone unloading data, and determine the target task completion delay and target task completion energy consumption according to the user drone unloading data, drone status data, task transmission unloading rate, and computing task complexity and computing task size in the user computing task data.

[0105] The user drone unloading data includes first unloading data, second unloading data, third unloading data and fourth unloading data.

[0106] The computational task complexity includes the computational task complexity of a single UAV and the computational task complexity of a cluster of UAVs.

[0107] The computing task size includes the computing task size of the UAV and the computing task size of the cluster UAV.

[0108] The user-UAV transmission rate includes the user-computing UAV transmission rate and the user-swarm UAV transmission rate.

[0109] It should be noted that the ground user u generates a computing task in time slot t , considering that the ground user has no local computing power, user u can only offload the task to the cluster drone m or computing drone n, and in order to avoid interference, at any time the user can only offload the task he generates to one drone (including cluster drones and computing drones). The offloading decision (the first offloading decision constraint) can be expressed as:

[0110]

[0111] in, is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the u-th user offloads the task to the n-th computing drone for calculation at time slot t; U is the set of time slots; M is the set of cluster drones; N is the set of computing drones.

[0112] Furthermore, after receiving the task offloaded by the user, the swarm drone can choose other swarm drones or computing drones for local computing or offloading. Its offloading decision (second offloading decision constraint, third offloading decision constraint, fourth offloading decision constraint, fifth offloading decision constraint) is described as:

[0113]

[0114] in, The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t; The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t; The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local computation on the task at time slot t, where M is the set of cluster drones.

[0115] It is worth mentioning that the present invention takes into account the flight status of the cluster UAV m (i.e., whether it is faulty), that is, the UAV status data of each cluster UAV is obtained. The UAV status data can be expressed as: , among which, when When , it means that the mth cluster drone is in a fault state at time slot t. , it means that the mth cluster drone is in a normal state (no fault) at time slot t.

[0116] Further, according to the user drone unloading data, drone status data, task transmission unloading rate and the computing task complexity and computing task size in the user computing task data, the process of determining the target task completion delay and the target task completion energy consumption can be achieved by executing the following steps S31 to S34:

[0117] Step S31, determining the completion delay of the user unloading the computing drone task and the energy consumption of the user unloading the computing drone task according to the computing drone computing task complexity, the computing drone computing task size, and the transmission rate between the user and the computing drone;

[0118] The task completion delay of user offloading computing UAV is the task completion delay of ground user offloading to computing UAV in time slot t.

[0119] The energy consumption of user offloading computing UAV task completion is the computing energy consumption of the ground user offloading to the computing UAV task completion in time slot t.

[0120] It should be noted that when the ground user is not within the coverage of the swarm drone, the ground user can offload the tasks generated by the ground user to the computing drone. That is, according to the computing task complexity of the computing drone, the computing task size of the computing drone, and the transmission rate between the user and the computing drone, the task completion delay of the ground user unloading to the computing drone in time slot t and the computing energy consumption of the task completed by the ground user unloading to the computing drone in time slot t can be determined. The process can be expressed as:

[0121]

[0122] in, is the task completion delay of the u-th ground user offloading to the n-th computing drone at time slot t; The computing UAV computing task size offloaded to the nth computing UAV by the uth ground user at time slot t; The transmission rate of the computing UAV is calculated for offloading the task of the u-th ground user at time slot t to the n-th; The computational complexity of the computational UAV task offloaded to the nth computational UAV by the uth ground user at time slot t; To calculate the CPU cycle frequency of the drone; is the computational energy consumption offloaded from the u-th ground user to the n-th computing UAV at time slot t; is the transmission power of the u-th ground user; The effective capacitance for calculating drones depends on the chip architecture.

[0123] Step S32, based on the user drone unloading data, the complexity of the cluster drone computing task, the size of the cluster drone computing task, and the transmission rate between the user and the cluster drone, determine the completion delay of the user unloading cluster drone task and the energy consumption of the user unloading cluster drone task;

[0124] The completion delay of the task unloading from the cluster drone by the user is the completion delay of the task unloaded from the user to the cluster drone.

[0125] The energy consumption of completing tasks unloaded by users to cluster drones is the energy consumption of completing tasks unloaded by users to cluster drones.

[0126] It should be noted that in time slot t, when ground user u unloads to cluster drone m, cluster drone m can perform local calculation or unload to other cluster drones m' or computing drone n for auxiliary calculation. According to the user drone unloading data, cluster drone computing task complexity, cluster drone computing task size, and user and cluster drone transmission rate, the user unloading cluster drone task completion delay and user unloading cluster drone task completion energy consumption can be determined. The process can be expressed as:

[0127]

[0128]

[0129] in, is the task completion delay of the u-th user offloading to the m-th cluster drone at time slot t; The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t; is the time delay for the mth swarm UAV to complete the calculation locally at time slot t; The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local calculation on the task at time slot t; is the mth cluster drone unloaded to the The task completion delay of a swarm of drones; The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t; is the task completion delay of the mth cluster UAV unloaded to the nth computing UAV at time slot t; Calculate the task size for the swarm UAV that offloads from the u-th ground user to the m-th swarm UAV at time slot t; is the transmission rate of the task offloading from the u-th ground user to the m-th cluster UAV at time slot t; Calculate the task complexity of the swarm UAVs for the uth ground user offloading to the mth swarm UAV at time slot t; is the CPU cycle frequency of the mth cluster drone; To calculate the CPU cycle frequency of the drone; The energy consumption for completing the task offloading from the u-th user to the m-th swarm drone at time slot t; is the energy consumption calculated locally by the mth cluster UAV at time slot t; is the mth cluster drone unloaded to the Energy consumption of a swarm of drones to complete their mission; is the energy consumption of the task completed by the m-th cluster UAV offloaded to the n-th computing UAV at time slot t; is the transmission power of the u-th user; is the effective capacitance of the chip architecture of the mth swarm drone; is the transmission power of the mth cluster UAV; is the task offloading rate of the mth cluster UAV and the m'th cluster UAV at time slot t; is the CPU cycle frequency of the m'th cluster drone; The transmission rate of the computing UAV is calculated for offloading the task of the u-th ground user at time slot t to the n-th; is the effective capacitance of the chip architecture of the m'th swarm drone; Calculate the effective capacitance of the chip architecture of the drone for the nth time.

[0130] Step S33, determining the target task completion delay according to the drone status data, the user drone unloading data, the user unloading calculation drone task completion delay and the user unloading cluster drone task completion delay;

[0131] The target task completion delay is the task completion delay of user u in time slot t.

[0132] It should be noted that based on the above derivation process, the target task completion delay is determined according to the drone status data, the user drone unloading data, the user unloading calculation drone task completion delay and the fixed user unloading cluster drone task completion delay. The processing process can be expressed as:

[0133]

[0134] in, The target task completion delay; is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the u-th user offloads the task to the n-th computing drone for calculation at time slot t; is the drone status data of the mth cluster drone at time slot t. When , it means that the mth cluster drone is in a fault state at time slot t. , it means that the mth cluster drone is in a normal state (no fault) at time slot t; is the task completion delay of the u-th user offloading to the m-th cluster drone at time slot t; is the task completion delay of the u-th ground user offloading to the n-th computing UAV at time slot t.

[0135] Step S34: determine the target task completion energy consumption based on the drone status data, the user drone unloading data, the user unloading calculated drone task completion energy consumption, and the user unloading cluster drone task completion energy consumption.

[0136] The target task completion energy consumption is the energy consumption of the system task completed in time slot t.

[0137] It should be noted that based on the above derivation process, the target task completion energy consumption is determined according to the drone status data, user drone unloading data, user unloading calculation drone task completion energy consumption and user unloading cluster drone task completion energy consumption. The processing process can be expressed as:

[0138]

[0139] in, Complete energy consumption for target tasks; is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the u-th user offloads the task to the n-th computing drone for calculation at time slot t; is the drone status data of the mth cluster drone at time slot t. When , it means that the mth cluster drone is in a fault state at time slot t. , it means that the mth cluster drone is in a normal state (no fault) at time slot t; The energy consumption for completing the task offloading from the u-th user to the m-th swarm drone at time slot t; is the computing energy consumption offloaded from the u-th ground user to the n-th computing UAV at time slot t.

[0140] Step 104: construct a joint optimization problem using preset unloading decision constraints, target task completion delay, target task completion energy consumption, preset drone constraints, user drone unloading data, and drone coordinate data.

[0141] Further, step 104 may include the following sub-steps S41-S42:

[0142] Step S41, determining the task completion cost based on the target task completion delay and the target task completion energy consumption;

[0143] It should be noted that the cost of the system in time slot t (task completion cost) is determined according to the target task completion delay and target task completion energy consumption. The process can be expressed as:

[0144]

[0145] in, Cost of completing the task; is the energy consumption weight; Complete energy consumption for target tasks; is the delay weight, satisfying ; The target task completion delay.

[0146] Step S42: construct a joint optimization problem using preset unloading decision constraints, task completion costs, preset drone constraints, user drone unloading data, and drone coordinate data.

[0147] The user drone unloading data includes first unloading data, second unloading data, third unloading data and fourth unloading data.

[0148] The preset unloading decision constraints include a first unloading decision constraint, a second unloading decision constraint, a third unloading decision constraint, a fourth unloading decision constraint and a fifth unloading decision constraint.

[0149] The preset drone constraints include a first drone constraint, a second drone constraint, a third drone constraint, a fourth drone constraint, a fifth drone constraint, a sixth drone constraint, a seventh drone constraint and an eighth drone constraint.

[0150] The drone coordinate data also includes two-dimensional cluster drone coordinates.

[0151] It should be noted that the optimization goal of the present invention is to establish the problem of minimizing system energy consumption and minimizing user task completion delay by combining the flight trajectory and task allocation of the UAV. Therefore, based on the above derivation process, according to the task completion cost, user UAV unloading data and UAV coordinate data, as well as the preset unloading decision constraints and preset UAV constraints, the constructed joint optimization problem can be described as:

[0152]

[0153] Among them, P is a joint optimization problem; is the two-dimensional cluster UAV coordinate, representing the optimization variable of the cluster UAV trajectory, , is the horizontal coordinate of the mth swarm UAV at time slot t, is the ordinate of the mth cluster UAV at time slot t; B is the user UAV unloading data, which represents the unloading decision variables of the user and cluster UAV, , is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the uth user offloads the task to the nth computing drone for calculation at time slot t. The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t. The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t. The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local calculation on the task at time slot t, where M is the set of cluster drones; is the task completion cost at time slot t; T is the time slot set; C1 is the first offloading decision constraint, indicating that the task of each ground user (user) can only be offloaded to a cluster drone or a computing drone, and U is the set of users; C2 is the second offloading decision constraint, indicating the constraint of the offloading decision variables calculated locally by the cluster drone; C3 is the third offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to the computing drone, and N is the set of computing drones; C4 is the fourth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to another cluster drone; C5 is the fifth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone; C6 is the first drone constraint, indicating the mobility constraint of the service range of each cluster drone, The maximum length of the service area divided by the mth cluster drone, is the maximum width of the service area divided by the mth swarm drone; C7 is the second drone constraint, indicating the coverage constraint that ground users can offload tasks to swarm drones. is the two-dimensional user coordinate of the u-th user at time slot t, is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal coverage of the mth cluster UAV; C8 is the constraint of the third UAV, which means the constraint to avoid collision between cluster UAVs. is the three-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional coordinates of a cluster of drones at time slot t, is the minimum safety distance; C9 is the fourth drone constraint, which indicates the constraint to avoid overlapping coverage between cluster drones. is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional cluster UAV coordinates of the cluster UAV at time slot t; C10 is the fifth UAV constraint, which represents the constraint of the horizontal flight angle of the cluster UAV. is the horizontal flight angle of the mth swarm UAV at time slot t; C11 is the sixth UAV constraint, which indicates the constraint on the horizontal flight distance of the swarm UAV. is the flight distance of the mth swarm drone at time slot t, is the maximum flight distance allowed by the cluster drone; C12 is the seventh drone constraint, indicating the state of the cluster drone. is the drone status data of the mth cluster drone at time slot t. When , it means that the mth cluster drone is in a fault state at time slot t. , indicating that the mth cluster drone is in a normal state at time slot t; C13 is the eighth drone constraint, indicating that the computing task in the tth time slot needs to be completed before the maximum tolerable delay. is the target task completion delay, is the maximum tolerable delay time.

[0154] Step 105: transform the joint optimization problem and determine the multi-agent Markov decision process model.

[0155] It should be noted that since the discrete variable B and the continuous variable As well as non-convex objective functions and constraints, problem P is a mixed integer nonlinear programming (MINLP) problem. Traditional heuristic algorithms are difficult to guarantee high-quality feasible solutions in a dynamic and uncertain environment. Therefore, it can be solved by using a multi-agent deep reinforcement learning algorithm.

[0156] Furthermore, the Markov decision process (MDP) can effectively handle the uncertainty and dynamics in the decision-making process, and provides an effective mathematical framework for drones to make real-time and optimal task offloading decisions in time-varying environments. The Markov decision process model is based on elements such as state, action, reward, and transition probability, and is used to describe how an agent achieves its goal by taking a series of actions within a given time frame. Therefore, in multi-drone swarm-assisted mobile edge computing, cluster drones determine their positions and offloading decisions to obtain the minimum cost of completing the system task. Considering that the movement of drones affects the state of the environment, the total cost of the system is determined by the current state of the system environment and the joint actions of all drones. Moreover, the previous state and the previous action jointly trigger the system environment to enter a new random state. In this case, the optimization problem P of task offloading can be expressed as a multi-agent Markov decision process model, which is specifically:

[0157]

[0158] Among them, M is the set of agents, and agents represent swarm drones; S is the state set of all agents, that is, the state space; A m is the action space of agent m; P is the state transition probability; R m is the reward function of agent m; is the discount factor, .

[0159] Furthermore, based on the optimization problem P, the state set S of each agent at time slot t is t The two-dimensional swarm drone coordinates of the swarm drone , UAV status data of swarm drones , the user's two-dimensional user coordinates , the size of tasks generated by users The composition can be described as:

[0160]

[0161] Furthermore, the action of each swarm drone (agent) includes the flight trajectory (horizontal flight distance and angle) and unloading decision, so the action space of the swarm drone can be expressed as:

[0162]

[0163] in, is the action of the mth swarm drone at time slot t; is the flight distance of the mth swarm UAV at time slot t; is the horizontal flight angle of the mth swarm UAV at time slot t; The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t; The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t; The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local calculation on the task at time slot t; is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the u-th user offloads the task to the n-th computing drone for computation at time slot t.

[0164] It is worth mentioning that according to the constraints of the optimization problem P, we can get The value range of each element in is: , , , , , .

[0165] Furthermore, the M agents cooperate to minimize the total cost of the system task completion, while satisfying relevant constraints, such as coverage overlap and collision constraints between drones. When all constraints are met, the reward function of the cluster drone m is defined as the system cost C t A negative number; when some constraints are not met, the reward function will be punished accordingly. The reward function of agent m in time slot t is expressed as:

[0166]

[0167] in, is the reward value at time slot t; Cost of completing the task; is the penalty corresponding to the overlapping constraint; is the penalty corresponding to the collision constraint; is the penalty corresponding to the coverage constraint.

[0168] Furthermore, in a dynamic and time-varying environment, the transition probability plays a key role in the unloading decision of the UAV in the current state. The transition probability can be defined as P(S t+1 |S t ,A t ), where S t is the current state of the agent, A t is the action currently selected by the agent, S t+1 is the state of the agent in the t+1 time slot.

[0169] Step 106: Use a multi-agent deep reinforcement learning algorithm to optimize and solve the multi-agent Markov decision process model to generate a target UAV mission calculation flight strategy.

[0170] It should be noted that, based on the Markov decision process model established above, the algorithm of multi-agent deep reinforcement learning can be used to solve the optimal solution, so as to obtain the target drone task calculation flight strategy, and control the drone to calculate the task according to the target drone task calculation flight strategy, so as to obtain the optimal task calculation result, minimize the system energy consumption and minimize the user task completion delay, thereby reducing the processing cost of the calculation task. At the same time, the present invention comprehensively considers the delay and energy consumption, and introduces a parameter between the delay and energy consumption to adjust two different dimensions, so that the same order of magnitude changes are produced in the optimization process, and will not be ignored because the order of magnitude is too small; the present invention also considers designing layered drones, first dividing the service area for the cluster drones, if the ground user is within the coverage area of ​​the cluster drone, the task can be unloaded to the cluster drone, when the cluster drone computing resources are insufficient, the task component will be unloaded again to the same layer drone or the upper layer drone server for calculation through the server. This can greatly reduce the delay and energy consumption caused by wireless communication; in addition, the present invention considers that in the scenario of multiple drones assisting mobile edge computing, when a drone fails, other drones can re-plan the path and re-divide the service area, and cover the user equipment covered by the original faulty drone within the coverage range of the normal drone, so as to provide computing services for it and complete the remaining computing tasks, thereby ensuring the user's service quality, and achieving the optimal overall task delay and system energy consumption by optimizing the drone trajectory and offloading decisions.

[0171] In this embodiment, the present invention is based on a three-layer network architecture computing offloading model of ground users and cluster drones. In the multi-drone collaborative assisted mobile edge computing offloading scenario, the status of drones in the drone cluster (whether it is faulty) is considered, and after the drone failure, the service area is redivided by adjusting the drone trajectory to ensure the service quality of users in the area covered by the faulty drone; at the same time, the present invention combines the flight trajectory and task allocation of the drone to minimize the system energy consumption and the user task completion delay.

[0172] As a comparison of technical effects, we can refer to existing technologies. In recent years, drone technology has developed rapidly. Due to its small size, mobility, flexibility, easy deployment, high adaptability, and high probability line-of-sight links, it is not only widely used in the military field, but also has received widespread attention in the civilian field. Drones have made significant progress in the application of disaster relief, communication assistance, network coverage, etc. Drones can hover and fly. As high-altitude mobile platforms, they can carry mobile edge computing servers to solve the problems of traditional mobile edge computing deployment and provide more flexible and efficient computing and communication services. Especially in remote areas and disaster areas, drones can be quickly deployed as MEC nodes to ensure the continuity of network connection and data processing, and significantly improve emergency response capabilities and network service quality.

[0173] With the development of autonomous UAV technology, multi-UAV cooperative control has become an important research direction. Through cooperative communication, multi-UAVs show high task execution efficiency, excellent coordination, intelligence and autonomy. In order to ensure the cooperative control performance of multi-UAV systems during task execution, the flight status of UAVs has become the focus of research. When multiple UAVs cooperate to perform tasks such as environmental monitoring, fire monitoring and cooperative search, if one or more UAVs encounter component failure, there is a risk of loss of control. In severe cases, the faulty UAV may collide with surrounding UAVs, causing the entire flight formation to lose control. Therefore, research on coping with UAV failures and fault-tolerant cooperative control issues is of great theoretical significance and practical necessity for the safe conduct of monitoring tasks and the safe control of multi-UAV systems.

[0174] It is worth noting that with the increase of working time, complexity and severity of the environment, actuators and sensors may wear and age. In addition, in the cooperative formation flight of multiple drones, the number of system components in the whole system increases significantly. The multi-drone system involves communication network connection, which gives the faulty drone the opportunity to send its error status information to nearby drones, greatly increasing the probability of collision and mission failure.

[0175] In summary, the existing UAV task calculation flight strategy optimization technology considers the task to be calculated locally by the user and by the UAV, but does not consider the situation that the UAV has limited computing resources, resulting in the inability to process the tasks unloaded by the user, and only considers optimizing energy consumption, but does not consider optimizing latency. In addition, the existing UAV task calculation flight strategy optimization technology considers that a task can be calculated locally or unloaded to the UAV for calculation, and the user side considers how to choose how to unload the subtask, but the user side will consume a lot of energy by unloading and calculating at the same time. At the same time, the existing UAV task calculation flight strategy optimization technology does not take into account whether the UAV is faulty, and cannot guarantee that the tasks unloaded by the user can be covered by the faulty UAV calculation, and does not consider how to handle the tasks of users within its coverage area after the UAV fails.

[0176] In view of the above problems, the present invention proposes a method for optimizing the flight strategy of UAV task calculation. In the scenario of multiple UAVs assisting mobile edge computing, when a certain UAV fails, other UAVs can re-plan the path and re-divide the service area, and cover the user equipment covered by the original faulty UAV within their own coverage to provide computing services for it and complete the remaining computing tasks, thereby ensuring the service quality of the user and minimizing the task completion delay and system energy consumption. Specifically, the present invention comprehensively considers the delay and energy consumption, introduces a parameter between the delay and energy consumption to adjust two different dimensions, and thus makes the same order of magnitude change in the optimization process, and will not be ignored because the order of magnitude is too small; at the same time, the present invention considers designing layered UAVs, first dividing the service area for the cluster UAV, if the ground user is within the coverage of the cluster UAV, the task can be unloaded to the cluster UAV, when the computing resources of the cluster UAV are insufficient, the task component will be unloaded again to the same layer UAV or the upper layer UAV server for calculation through the server. This can greatly reduce the delay and energy consumption caused by wireless communication; in addition, the present invention considers that in the scenario of multiple drones assisting mobile edge computing, when a drone fails, other drones can re-plan the path and re-divide the service area, and cover the user equipment covered by the original faulty drone within the coverage range of the normal drone, so as to provide computing services for it and complete the remaining computing tasks, thereby ensuring the user's service quality, and achieving the optimal overall task delay and system energy consumption by optimizing the drone trajectory and offloading decisions.

[0177] In an embodiment of the present invention, the present invention provides a method for optimizing a UAV task calculation flight strategy. First, user calculation task data, user coordinate data, UAV state data and UAV coordinate data are obtained; then, based on the user coordinate data and the UAV coordinate data, the user UAV distance data is determined, and the task transmission unloading rate is determined according to the user UAV distance data and the UAV height data in the UAV coordinate data; based on a preset unloading decision constraint, the user UAV unloading data is determined, and according to the user UAV unloading data, the UAV state data, the task transmission unloading rate and the calculation task complexity and the calculation task size in the user calculation task data, the target task completion delay and the target task completion energy consumption are determined; the preset unloading decision constraint, the target task completion delay, the target task completion energy consumption, the preset UAV constraint, the user UAV unloading data and the UAV coordinate data are used to construct a joint optimization problem; the joint optimization problem is transformed to determine the multi-agent Markov decision policy process model; finally, a multi-agent deep reinforcement learning algorithm is used to optimize and solve the multi-agent Markov decision process model to generate a target UAV task calculation flight strategy; based on the above scheme, the acquired user calculation task data, user coordinate data, and UAV coordinate data are processed to obtain the target task completion delay and the target task completion energy consumption, and then combined with the preset unloading decision constraints, preset UAV constraints, user UAV unloading data, and UAV status data, a multi-agent Markov decision process model is obtained, and the model is solved by a multi-agent deep reinforcement learning algorithm to generate a process of calculating the flight strategy of the target UAV task. The present invention comprehensively considers the delay and energy consumption, so that the same order of magnitude of changes are produced in the optimization process. At the same time, the present invention controls the UAV to calculate the user task according to the obtained target UAV task calculation flight strategy, and does not require the user end to unload and calculate at the same time, which can reduce energy consumption and thus reduce the processing cost of the computing task.

[0178] See also Figure 3 , Figure 3 A structural block diagram of a UAV mission calculation flight strategy optimization system provided in Embodiment 2 of the present invention.

[0179] The present invention provides a UAV mission calculation flight strategy optimization system, comprising:

[0180] The acquisition module 301 is used to acquire user computing task data, user coordinate data, drone status data and drone coordinate data;

[0181] A determination module 302 is used to determine user-drone distance data based on the user coordinate data and the drone coordinate data, and determine a task transmission offloading rate according to the user-drone distance data and the drone height data in the drone coordinate data;

[0182] The delay energy consumption determination module 303 is used to determine the user drone unloading data based on the preset unloading decision constraints, and determine the target task completion delay and the target task completion energy consumption according to the user drone unloading data, drone status data, task transmission unloading rate and the computing task complexity and computing task size in the user computing task data;

[0183] A construction module 304 is used to construct a joint optimization problem using preset unloading decision constraints, target task completion delay, target task completion energy consumption, preset drone constraints, user drone unloading data, and drone coordinate data;

[0184] A conversion module 305 is used to convert the joint optimization problem and determine a multi-agent Markov decision process model;

[0185] The solution module 306 is used to optimize and solve the multi-agent Markov decision process model using a multi-agent deep reinforcement learning algorithm to generate a target UAV mission calculation flight strategy.

[0186] Furthermore, the drone coordinate data includes three-dimensional cluster drone coordinates and three-dimensional calculation drone coordinates; the user drone distance data includes user-cluster drone distance data, user-calculation drone distance data, and drone-to-drone distance data; the determination module 302 includes:

[0187] The first submodule is used to use the three-dimensional user coordinates in the user coordinate data to perform Euclidean distance calculations with the three-dimensional cluster drone coordinates and the three-dimensional calculation drone coordinates, and output the distance data between the user and the cluster drone and the distance data between the user and the calculation drone;

[0188] The second submodule is used to determine the distance data between drones based on the three-dimensional cluster drone coordinates and the three-dimensional calculated drone coordinates.

[0189] Furthermore, the task transmission offloading rate includes the transmission rate between the user and the drone and the task offloading rate between the drones; the determination module 302 further includes:

[0190] The third submodule is used to determine the connection probability between the user and the drone based on the distance data between the user and the cluster drone, the distance data between the user and the calculation drone, and the drone height data in the drone coordinate data;

[0191] The fourth submodule is used to determine the link channel power gain between the user and the drone according to the connection probability between the user and the drone, the distance data between the user and the cluster drone, and the distance data between the user and the computing drone;

[0192] The fifth submodule is used to determine the transmission rate between the user and the UAV according to the power gain of the link channel between the user and the UAV by using a preset Shannon formula;

[0193] The sixth submodule is used to determine the task offloading rate between UAVs based on the distance data between UAVs.

[0194] Further, the computing task complexity includes the computing task complexity of the computing drone and the computing task complexity of the cluster drone; the computing task size includes the computing task size of the computing drone and the computing task size of the cluster drone; the user-drone transmission rate includes the user-computing drone transmission rate and the user-swarm drone transmission rate; the delay energy consumption determination module 303 is specifically used to:

[0195] According to the computing task complexity of the computing drone, the computing task size of the computing drone, and the transmission rate between the user and the computing drone, the completion delay of the user unloading the computing drone task and the energy consumption of the user unloading the computing drone task are determined;

[0196] Based on the user drone unloading data, the complexity of the cluster drone computing task, the size of the cluster drone computing task, and the transmission rate between the user and the cluster drone, the completion delay of the user unloading cluster drone task and the energy consumption of the user unloading cluster drone task are determined;

[0197] Determine the target task completion delay based on drone status data, user drone unloading data, user unloading calculation drone task completion delay, and fixed user unloading cluster drone task completion delay;

[0198] Based on drone status data, user drone unloading data, user unloading calculation drone task completion energy consumption and user unloading cluster drone task completion energy consumption, the target task completion energy consumption is determined.

[0199] Further, the construction module 304 is specifically used to:

[0200] Determine the task completion cost based on the target task completion delay and target task completion energy consumption;

[0201] The preset unloading decision constraints, task completion cost, preset UAV constraints, user UAV unloading data and UAV coordinate data are used to construct a joint optimization problem.

[0202] Furthermore, the user drone unloading data includes first unloading data, second unloading data, third unloading data and fourth unloading data; the preset unloading decision constraints include first unloading decision constraints, second unloading decision constraints, third unloading decision constraints, fourth unloading decision constraints and fifth unloading decision constraints; the preset drone constraints include first drone constraints, second drone constraints, third drone constraints, fourth drone constraints, fifth drone constraints, sixth drone constraints, seventh drone constraints and eighth drone constraints; the drone coordinate data also includes two-dimensional cluster drone coordinates; the joint optimization problem is specifically:

[0203]

[0204] Among them, P is a joint optimization problem; is the two-dimensional cluster UAV coordinate, representing the optimization variable of the cluster UAV trajectory, , is the horizontal coordinate of the mth swarm UAV at time slot t, is the ordinate of the mth swarm drone at time slot t; B is the user drone unloading data, which represents the unloading decision variables of the user and swarm drones, B= , is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the uth user offloads the task to the nth computing drone for calculation at time slot t. The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t. The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t. The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local calculation on the task at time slot t, where M is the set of cluster drones; is the task completion cost at time slot t; T is the time slot set; C1 is the first offloading decision constraint, indicating that the task of each ground user (user) can only be offloaded to a cluster drone or a computing drone, and U is the set of users; C2 is the second offloading decision constraint, indicating the constraint of the offloading decision variables calculated locally by the cluster drone; C3 is the third offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to the computing drone, and N is the set of computing drones; C4 is the fourth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to another cluster drone; C5 is the fifth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone; C6 is the first drone constraint, indicating the mobility constraint of the service range of each cluster drone, The maximum length of the service area divided by the mth cluster drone, is the maximum width of the service area divided by the mth swarm drone; C7 is the second drone constraint, indicating the coverage constraint that ground users can offload tasks to swarm drones. is the two-dimensional user coordinate of the u-th user at time slot t, is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal coverage of the mth cluster UAV; C8 is the constraint of the third UAV, which means the constraint to avoid collision between cluster UAVs. is the three-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional coordinates of a cluster of drones at time slot t, is the minimum safety distance; C9 is the fourth drone constraint, which indicates the constraint to avoid overlapping coverage between cluster drones. is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional cluster UAV coordinates of the cluster UAV at time slot t; C10 is the fifth UAV constraint, which represents the constraint of the horizontal flight angle of the cluster UAV. is the horizontal flight angle of the mth swarm UAV at time slot t; C11 is the sixth UAV constraint, which indicates the constraint on the horizontal flight distance of the swarm UAV. is the flight distance of the mth swarm drone at time slot t, is the maximum flight distance allowed by the cluster drone; C12 is the seventh drone constraint, indicating the state of the cluster drone. is the drone status data of the mth cluster drone at time slot t. When , it means that the mth cluster drone is in a fault state at time slot t. , indicating that the mth cluster drone is in a normal state at time slot t; C13 is the eighth drone constraint, indicating that the computing task in the tth time slot needs to be completed before the maximum tolerable delay. is the target task completion delay, is the maximum tolerable delay time.

[0205] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, modules and sub-modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0206] An embodiment of the present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory; when the computer program is executed by the processor, the processor executes the steps of the method for optimizing the flight strategy for calculating a drone mission as in any of the above embodiments.

[0207] An embodiment of the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon. When the computer program / instruction is executed by a processor, the steps of the method for optimizing the flight strategy for calculating a UAV mission as in any of the above embodiments are implemented.

[0208] An embodiment of the present invention also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for optimizing the flight strategy for calculating a drone mission as in any of the above embodiments.

[0209] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0210] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0211] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing flight strategy for unmanned aerial vehicle mission calculation, characterized in that: include: Obtain user computing task data, user coordinate data, drone status data, and drone coordinate data; Determine user-drone distance data based on the user coordinate data and the drone coordinate data, and determine a task transmission offloading rate according to the user-drone distance data and drone height data in the drone coordinate data; Based on the preset unloading decision constraints, determine the user drone unloading data, and determine the target task completion delay and the target task completion energy consumption according to the user drone unloading data, the drone status data, the task transmission unloading rate, and the computing task complexity and computing task size in the user computing task data; A joint optimization problem is constructed by using the preset unloading decision constraints, the target task completion delay, the target task completion energy consumption, the preset drone constraints, the user drone unloading data and the drone coordinate data; Transforming the joint optimization problem to determine a multi-agent Markov decision process model; A multi-agent deep reinforcement learning algorithm is used to optimize and solve the multi-agent Markov decision process model to generate a target UAV mission calculation flight strategy.

2. The method for optimizing the flight strategy of unmanned aerial vehicle mission calculation according to claim 1, characterized in that: The drone coordinate data includes three-dimensional cluster drone coordinates and three-dimensional calculation drone coordinates; the user drone distance data includes user-cluster drone distance data, user-calculation drone distance data, and drone-to-drone distance data; The determining the user-drone distance data based on the user coordinate data and the drone coordinate data includes: The three-dimensional user coordinates in the user coordinate data are used to perform Euclidean distance calculations with the three-dimensional cluster drone coordinates and the three-dimensional calculation drone coordinates, and the distance data between the user and the cluster drone and the distance data between the user and the calculation drone are output; The distance data between the drones is determined according to the three-dimensional cluster drone coordinates and the three-dimensional calculated drone coordinates.

3. The method for optimizing the flight strategy of unmanned aerial vehicle mission calculation according to claim 2, characterized in that: The task transmission offloading rate includes the transmission rate between the user and the UAV and the task offloading rate between the UAVs; The step of determining the task transmission offloading rate according to the user drone distance data and the drone height data in the drone coordinate data includes: Determine the connection probability between the user and the drone based on the distance data between the user and the cluster drone, the distance data between the user and the calculation drone, and the drone height data in the drone coordinate data; Determine the link channel power gain between the user and the drone according to the connection probability between the user and the drone, the distance data between the user and the cluster drone, and the distance data between the user and the computing drone; The preset Shannon formula is used to determine the transmission rate between the user and the UAV according to the link channel power gain between the user and the UAV; Based on the distance data between the drones, a task offloading rate between the drones is determined.

4. The method for optimizing the flight strategy of unmanned aerial vehicle mission calculation according to claim 3, characterized in that: The computing task complexity includes the computing task complexity of the computing drone and the computing task complexity of the clustered drones; the computing task size includes the computing task size of the computing drone and the computing task size of the clustered drones; The user-drone transmission rate includes the user-computing drone transmission rate and the user-swarm drone transmission rate; the target task completion delay and the target task completion energy consumption are determined according to the user drone unloading data, the drone status data, the task transmission unloading rate and the computing task complexity and computing task size in the user computing task data, including: According to the computing task complexity of the computing drone, the computing task size of the computing drone, and the transmission rate between the user and the computing drone, the completion delay of the user unloading the computing drone task and the energy consumption of the user unloading the computing drone task are determined; Based on the user drone unloading data, the complexity of the cluster drone computing task, the size of the cluster drone computing task, and the transmission rate between the user and the cluster drone, determine the completion delay of the user unloading cluster drone task and the energy consumption of the user unloading cluster drone task; Determine the target task completion delay according to the drone status data, the user drone uninstallation data, the user uninstallation calculation drone task completion delay and the fixed user uninstallation cluster drone task completion delay; The target task completion energy consumption is determined based on the drone status data, the user drone unloading data, the user unloading calculation drone task completion energy consumption and the user unloading cluster drone task completion energy consumption.

5. The method for optimizing the flight strategy of unmanned aerial vehicle mission calculation according to claim 1, characterized in that: The method of constructing a joint optimization problem by using the preset unloading decision constraints, the target task completion delay, the target task completion energy consumption, the preset drone constraints, the user drone unloading data and the drone coordinate data includes: Determining a task completion cost based on the target task completion delay and the target task completion energy consumption; A joint optimization problem is constructed by using the preset unloading decision constraints, the task completion cost, the preset drone constraints, the user drone unloading data and the drone coordinate data.

6. The method for optimizing the flight strategy of unmanned aerial vehicle mission calculation according to claim 5, characterized in that: The user drone unloading data includes first unloading data, second unloading data, third unloading data and fourth unloading data; the preset unloading decision constraints include first unloading decision constraints, second unloading decision constraints, third unloading decision constraints, fourth unloading decision constraints and fifth unloading decision constraints; the preset drone constraints include first drone constraints, second drone constraints, third drone constraints, fourth drone constraints, fifth drone constraints, sixth drone constraints, seventh drone constraints and eighth drone constraints; the drone coordinate data also includes two-dimensional cluster drone coordinates; the joint optimization problem is specifically: Among them, P is a joint optimization problem; is the two-dimensional cluster UAV coordinate, representing the optimization variable of the cluster UAV trajectory, , is the horizontal coordinate of the mth swarm UAV at time slot t, is the ordinate of the mth cluster UAV at time slot t; B is the user UAV unloading data, which represents the unloading decision variables of the user and cluster UAV, , is the first uninstall data, indicating that When , the uth user offloads the task to the mth cluster drone for calculation at time slot t. When , the uth user offloads the task to the nth computing drone for calculation at time slot t. The second uninstall data indicates that When , the mth cluster drone performs local calculations on the task at time slot t. When , the mth cluster drone unloads the task to other drones for auxiliary calculation at time slot t. The third uninstall data indicates that When , the mth cluster drone unloads the task to the nth computing drone for auxiliary computing at time slot t. When , the mth cluster drone performs local calculation on the task at time slot t. The fourth uninstall data indicates that At time t, the mth cluster drone unloads the task to the mth A cluster of drones performs auxiliary calculations. When , the mth cluster drone performs local calculation on the task at time slot t, where M is the set of cluster drones; is the task completion cost at time slot t; T is the time slot set; C1 is the first offloading decision constraint, indicating that the task of each ground user (user) can only be offloaded to a cluster drone or a computing drone, and U is the set of users; C2 is the second offloading decision constraint, indicating the constraint of the offloading decision variables calculated locally by the cluster drone; C3 is the third offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to the computing drone, and N is the set of computing drones; C4 is the fourth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone to another cluster drone; C5 is the fifth offloading decision constraint, indicating the constraint of the offloading decision variables of the cluster drone; C6 is the first drone constraint, indicating the mobility constraint of the service range of each cluster drone, The maximum length of the service area divided by the mth cluster drone, is the maximum width of the service area divided by the mth swarm drone; C7 is the second drone constraint, indicating the coverage constraint that ground users can offload tasks to swarm drones. is the two-dimensional user coordinate of the u-th user at time slot t, is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, is the horizontal coverage of the mth cluster UAV; C8 is the constraint of the third UAV, which means the constraint to avoid collision between cluster UAVs. is the three-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional coordinates of a cluster of drones at time slot t, is the minimum safety distance; C9 is the fourth drone constraint, which indicates the constraint to avoid overlapping coverage between cluster drones. is the two-dimensional cluster drone coordinate of the mth cluster drone at time slot t, For the The three-dimensional cluster UAV coordinates of the cluster UAV at time slot t; C10 is the fifth UAV constraint, which represents the constraint of the horizontal flight angle of the cluster UAV. is the horizontal flight angle of the mth swarm UAV at time slot t; C11 is the sixth UAV constraint, which indicates the constraint on the horizontal flight distance of the swarm UAV. is the flight distance of the mth swarm drone at time slot t, is the maximum flight distance allowed by the cluster drone; C12 is the seventh drone constraint, indicating the state of the cluster drone. is the drone status data of the mth cluster drone at time slot t. When , it means that the mth cluster drone is in a fault state at time slot t. , indicating that the mth cluster drone is in a normal state at time slot t; C13 is the eighth drone constraint, indicating that the computing task in the tth time slot needs to be completed before the maximum tolerable delay. is the target task completion delay, is the maximum tolerable delay time.

7. A UAV mission calculation flight strategy optimization system, characterized in that: include: The acquisition module is used to obtain user computing task data, user coordinate data, drone status data and drone coordinate data; A determination module, configured to determine user-drone distance data based on the user coordinate data and the drone coordinate data, and determine a task transmission offloading rate according to the user-drone distance data and drone height data in the drone coordinate data; The delay energy consumption determination module is used to determine the user drone unloading data based on the preset unloading decision constraints, and determine the target task completion delay and the target task completion energy consumption according to the user drone unloading data, the drone status data, the task transmission unloading rate and the computing task complexity and computing task size in the user computing task data; A construction module, for constructing a joint optimization problem by using the preset unloading decision constraints, the target task completion delay, the target task completion energy consumption, the preset drone constraints, the user drone unloading data and the drone coordinate data; A conversion module, used for converting the joint optimization problem and determining a multi-agent Markov decision process model; The solution module is used to optimize and solve the multi-agent Markov decision process model using a multi-agent deep reinforcement learning algorithm to generate a target UAV mission calculation flight strategy.

8. A computer device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method for optimizing the flight strategy for calculating a UAV mission as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for optimizing the flight strategy for calculating a UAV mission as described in any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the method for optimizing the flight strategy for calculating a UAV mission as described in any one of claims 1-6.

Images