Unmanned aerial vehicle group task intention prediction and path recommendation method

Abstract

The invention discloses an unmanned aerial vehicle group task intention prediction and path recommendation method. The method comprises the following steps: step 1, constructing a scene model; step 2, scene formulation; step 3, fuzzy data fusion and scene task intention prediction; and step 4, designing a delay approximation algorithm based on a task area to optimize the total delay of all user tasks in the area completed by the unmanned aerial vehicle. According to the method, data of multiple sensors are fused by using a fuzzy theory, historical scene experiences such as a membership function set and a fuzzy rule are learned and utilized, and a key data capturing trend and a hot spot region are predicted. And performing unmanned aerial vehicle group path recommendation and intelligent decision making according to the generated task prediction model. The predictive path planning enables the system to optimize the path in advance, so that the system has certain universality and has autonomous decision-making and strain capabilities.

Description

Technical Field

[0001] The present invention relates to the field of drone swarm path planning, and specifically to a drone swarm task intention prediction and path recommendation system based on multi-source data and predicted scene task generation intention. The present invention also provides a method for using the system. Background Art

[0002] In recent years, drone technology has experienced rapid development, playing an increasingly important role not only in the military field, but also in civilian fields such as agriculture, disaster response, environmental monitoring, transportation hubs and other densely populated areas. However, in the face of an execution environment full of uncertainty and complexity, traditional static planning methods are difficult to adapt to this change, so more intelligent prediction and planning methods are needed. UAV systems can now carry more sensors, and it is important to predict mission requirements from these data and develop the ability to predict future mission requirements and make decisions based on them. In the case of autonomous systems, drones can make instant decisions intelligently without human intervention. Formation path planning can better solve the path planning problem of multiple drones. Specifically, formation path planning refers to each drone finding its own collision-free path while trying to maintain the formation structure. Therefore, it is necessary to design an intelligent drone swarm management and control system that can predict mission intentions and make path recommendations. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for efficiently executing tasks and recommending paths for drone swarms in new mission scenarios. Based on real drone swarm collection data, a drone swarm mission intention prediction and path recommendation method is developed through relevant algorithms.

[0004] The objective of the present invention is achieved through the following technical solution: A method for predicting the mission intention and recommending the path of a drone swarm comprises the following steps:

[0005] Step 1: Build a scenario model: Assume that the number of drones is M, and the drones need to perform tasks in K areas. Each area consists of N independent task sources, and the user i∈N;

[0006] Step 2: Scenario formulation, the steps are as follows:

[0007] Step 2-1: Construct the UAV maneuverability constraint: The maximum speed of the UAV is expressed as v max , e t represents the time slot duration, p t and q k denote the abscissa of the UAV in time slot t and the abscissa of user k respectively; the position constraint across consecutive time slots is written as:

[0008]

[0009] Step 2-2, define the completeness and atomicity of the task; since the task cannot be further compressed or partitioned, the drone can only decide whether to complete the entire task or reject the task request; I i,k,t and O i,k,t They represent the input and output data size of the task at time t, I i,k and O i,k They represent the total task input and output size of user i in region k respectively. In order to ensure the completeness and atomicity of the task, the numerical relationship is expressed as:

[0010]

[0011] Step 2-3, define the transmission rate limit: Due to the limitation of channel conditions, both uplink and downlink communications cannot exceed the maximum transmission rate. The transmission rate between user i and the drone is expressed as:

[0012]

[0013] where a i,k,t represents the task upload variable, B represents bandwidth, P i,k represents the transmission power of the ith user, g i,k,t represents the channel power gain of the i-th user in region k in time slot t, σ 2 represents additive white Gaussian noise;

[0014] Step 2-4: Define the effective time limit of the task; define the minimum delay constraint of user i in area k as τ i,k , use D i,k =e i,k +X i,k Represents the total task execution time of the user, including the transmission delay Use r i,k,t and M i,k They represent the computing resources allocated to user i in region k at time t and the computing resource requirements of the user, respectively. The task execution time is expressed as The user's total task execution time must meet the following requirements:

[0015]

[0016] Indicates that any task completed within the system time period meets the corresponding deadline, where τ i,k represents the deadline of the i-th user in the k-th region;

[0017] Step 2-5: Define the optimization objective function:

[0018]

[0019] Step 2-6, fuzzy data: define the universe set U, a fuzzy set A on U to assign a membership degree A(x) between 0 and 1 to each element x, the degree to which element x belongs to the fuzzy set A is expressed as:

[0020] A={(x,μ A (x))|x∈U};

[0021] The universe set U represents the set of UAV airborne missions; x represents the value of the specific mission; the fuzzy set A represents the fuzzy classification of the mission; μ A (x) represents the degree to which the specific task quantity x belongs to the fuzzy set A;

[0022] Step 2-7, multi-source data fusion: define a data set MsD, which contains n m-dimensional vectors; define FD as a set, the elements of which are the results of mapping data from different data sources to each vector in MsD through f;

[0023] MsD={d i |d i =(x i1 ,x i2 ,...,x im ),i=1,2,...,n}

[0024] FD={f(x 11 ,x 12 ,...,x 1m ),f(x 21 ,x 22 ,...,x 2m ),...,f(x n1 ,x n2 ,...,x nm )};

[0025] where d i represents the task vector of user i, x i,m represents the characteristics of the mth task of user i;

[0026] Step 3: Fuzzy data fusion and scene task intention prediction. The steps are as follows:

[0027] Step 3-1, construct the membership function: extract the feature vector x of the task from the actual data set and use the prototype vector r to represent the center of the fuzzy set, calculate the Euclidean distance dst between x and the prototype vector r; use the formula μ x =1 / (1+dst) to calculate the fuzzy membership μ x , indicating the degree to which x belongs to a fuzzy set; the eigenvector x is the sum of x i,mThe vector formed;

[0028] Step 3-2, formulate fuzzy rules;

[0029] Step 3-3: Historical scene matching: using the task decisions in historical experience to quantify the similarity of the newly input tasks, and obtain the similarity of the task to different categories;

[0030] Step 3-4: Task area prediction based on fuzzy logic, using fuzzy membership tool to process the above fuzzy data fusion data, distinguish the spatial set and type of tasks according to membership and similarity, and obtain the scene task distribution intention;

[0031] Step 4: Based on the fuzzy data fusion and scenario mission intention prediction in step 3, a delay approximation algorithm based on the mission area is designed to optimize the total delay of the drone to complete all user tasks in the area.

[0032] The beneficial effects of the present invention are as follows: the present invention is based on the problem of efficient execution of tasks and path recommendation of drone swarms in new mission scenarios, based on real drone swarm collection data, and develops a drone swarm intelligent management and control method that can predict task intentions and perform path recommendation through relevant algorithms. It provides efficient, flexible and intelligent service capabilities in terms of intelligent decision-making, real-time data collection, multi-source fuzzy data processing and autonomous learning prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A flowchart of the method for predicting the mission intention and recommending the path of a drone swarm according to the present invention;

[0034] Figure 2 It is a schematic diagram of the system architecture of the present invention;

[0035] Figure 3 It is a schematic diagram of the UAV swarm simulation module of the present invention;

[0036] Figure 4 This is a schematic diagram of the drone control data transmission service module of the present invention;

[0037] Figure 5 This is a schematic diagram of a perception data visualization module of the present invention;

[0038] Figure 6 This is a schematic diagram of the platform comprehensive function service module of the present invention;

[0039] Figure 7 It is a simulated visual perception map of the UAV of the present invention;

[0040] Figure 8 It is a simulation diagram of unloading the UAV mission of the present invention;

[0041] Fig. 9 Comparison of the predicted task heat map generated by the algorithm designed for the present invention with the real task heat map;

[0042] Fig.10 Schematic diagram of recommended paths generated by the algorithm TDTS designed by the present invention and other three baseline algorithms;

[0043] Fig.11 A performance comparison diagram of the algorithm TDTS designed by the present invention and other three baseline algorithms in terms of system throughput under different numbers of households;

[0044] Fig.12 The performance comparison diagram of the algorithm TDTS designed by the present invention and other three baseline algorithms in terms of service percentage under different numbers of households. DETAILED DESCRIPTION

[0045] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0046] like Figure 1 As shown, a method for predicting the mission intention and recommending the path of a drone swarm of the present invention comprises the following steps:

[0047] Step 1, build a scenario model: Assume that the number of drones is M, and all drones need to perform tasks in K areas. Each area consists of N independent task sources, and user i∈N. The present invention is a mobile edge computing service system assisted by drones. In this embodiment, a commercial district scenario is constructed. There are three drones and a group of office building areas in the field. Each area consists of N independent task generation sources. The total number of areas is represented by K, that is, K office building areas, and user i∈N. When the scenario generates a large amount of data service demand, drones will provide edge computing services. The total service time is represented by T.

[0048] Step 2: Scenario formulation, the steps are as follows:

[0049] Step 2-1: Construct the UAV maneuverability constraint: Due to the limitation of physical equipment, the maximum speed of the UAV is expressed as v max , e t represents the time slot duration, p t and q k denote the abscissa of the UAV in time slot t and the abscissa of user k respectively; the position constraint across consecutive time slots is written as:

[0050]

[0051] Step 2-2, define the completeness and atomicity of the task; since the task cannot be further compressed or partitioned, the drone can only decide whether to complete the entire task or reject the task request; Ii,k,t and O i,k,t They represent the input and output data size of the task at time t (in bits), I i,k and O i,k They represent the total task input and output size of user i in region k respectively. In order to ensure the completeness and atomicity of the task, the numerical relationship is expressed as:

[0052]

[0053] Step 2-3, define the transmission rate limit: Due to the limitation of channel conditions, both uplink and downlink communications cannot exceed the maximum transmission rate. The transmission rate between user i and the drone is expressed as:

[0054]

[0055] where a i,k,t represents the task upload variable, which means that user i in time slot t area k cannot obtain channel resource a i,k,t =0, B represents bandwidth, P i,k represents the transmission power of the ith user, g i,k,t represents the channel power gain of the i-th user in region k in time slot t, σ 2 represents additive white Gaussian noise;

[0056] Step 2-4: Define the effective time limit of the task; define the minimum delay constraint of user i in area k as τ i,k , use D i,k =e i,k +X i,k Represents the total task execution time of the user, including the transmission delay Use r i,k,t and M i,k They represent the computing resources allocated to user i in region k at time t and the computing resource requirements of the user, respectively. The task execution time is expressed as The user's total task execution time must meet the following requirements:

[0057]

[0058] Indicates that any task completed within the system time period meets the corresponding deadline, where τ i,k represents the deadline of the i-th user in the k-th region;

[0059] Step 2-5: Define the optimization objective function:

[0060]

[0061] Step 2-6, fuzzy data: The formulation of fuzzy data relies on fuzzy set theory; define a universe set U, a fuzzy set A on which each element x is assigned a membership degree A(x) between 0 and 1. The degree to which element x belongs to fuzzy set A is expressed as:

[0062] A={(x,μ A (x))x∈U};

[0063] Corresponding to the UAV airborne mission scenario, the universe set U represents the set of possible values ​​of the UAV airborne mission volume; x represents the value of the specific mission volume; the fuzzy set A represents the fuzzy classification of the mission volume; μ A (x) represents the degree to which the specific task volume x belongs to the fuzzy set A.

[0064] Step 2-7, multi-source data fusion: define a data set MsD, which contains n m-dimensional vectors; define FD as a set, the elements of which are the results of mapping data from different data sources to each vector in MsD through f;

[0065] MsD={d i |d i =(x i1 ,x i2 ,...,x im ),i=1,2,...,n}

[0066] FD={f(x 11 ,x 12 ,...,x 1m ),f(x 21 ,x 22 ,...,x 2m ),...,f(x n1 ,x n2 ,...,x nm )};

[0067] where d i represents the task vector of user i, x i,m Represents the characteristics of the mth task of user i.

[0068] Step 3: Fuzzy data fusion and scene task intention prediction. The steps are as follows:

[0069] Step 3-1, construct the membership function, newly generate the membership function or obtain it based on historical similar scene experience, the specific method is: extract the feature vector x of the task from the actual data set and use the prototype vector r to represent the center of the fuzzy set, calculate the Euclidean distance dst between the feature vector x and the prototype vector r; then, use the formula μ x =1 / (1+dst) to calculate the fuzzy membership μx , indicating the degree to which x belongs to a fuzzy set; the eigenvector x is the sum of x i,m The vector formed;

[0070] Step 3-2, formulate fuzzy rules. The formulation of fuzzy rules is divided into two categories: deterministic rules and gradual rules;

[0071] The general form of fuzzy rules is a set of "if-then" rules. The rules are expressed as:

[0072] R i :if x is A i and / or y is B i then z is C i

[0073] Where R i represents the i-th rule, x and y are input variables, A i and B i is the fuzzy set corresponding to the membership function defined on x and y, z is the output variable, C i is the fuzzy set corresponding to the membership function defined on z; the optimization experience of the entire rule set can be measured by defining a performance index P to measure the effect of the rule set, that is, P = MF (rule set), MF is the evaluation function. i To maximize or minimize the performance index P, that is, to minimize the total delay of the drone to complete all user tasks in the area, etc. The forms of the deterministic rule and the gradual rule are "the more x is in A, the more certain y is in B" and "the more x equals A, the more y equals B".

[0074] Step 3-3: Historical scenario matching, generating a decision template DT containing historical decision information by recording historical membership functions and fuzzy rules e , use the fuzzy membership function to quantify the new input task feature vector x to obtain the different feature memberships of the task, that is, the decision process DP(x). Then, use the similarity measurement function Measuring the decision process DP(x) and decision template DT e The similarity between represents the Hamming distance, and c represents the total task category. Through this process, the task decisions in historical experience can be reused to quantify the similarity of the categories of the newly input tasks, and the similarity of the task to different categories can be obtained;

[0075] Step 3-4: Task area prediction based on fuzzy logic, use the fuzzy membership tool to process the above fuzzy data fusion data, judge the spatial set and type of tasks according to membership and similarity, and obtain the scene task distribution intention.

[0076] The specific steps of fuzzy data processing and intention prediction algorithm are as follows:

[0077] Step 3-4-1, calculate the Euclidean distance dst between the feature vector x and the prototype vector r, and then use the formula μ x =1 / (1+dst) to calculate the fuzzy membership μ x , indicating the degree to which x belongs to a fuzzy set;

[0078] Step 3-4-2, formulate deterministic rules: ifμ x >tsd,return otpH orμ x ≤tsd,return otpL, select the output value according to the deterministic rule. If the input membership μ x If it is greater than the threshold value tsd, a high output otpH is returned, otherwise a low output otpL is returned and the output result is recorded as dtOp.

[0079] Step 3-4-3: Establish progressive rules:

[0080] ifμ x ≤ipM,return gdOp=otMi+(μ x / ipM)(otpMx-otMi). Define otpMx and otMi as the maximum and minimum values ​​of the progressive output, ipM as the maximum input membership, then according to the input membership μ x And the maximum input membership ipM can calculate a proportional factor scale = μ x / ipM, and then use this scaling factor to perform linear interpolation, i.e. gdOp = otMi + (μ x / ipM)(otpMx-otMi), returns a value that gradually changes between otMi and otpMx.

[0081] Step 3-4-4: First, calculate the membership degree μ through step 3-4-1 x , then use steps 3-4-2 and 3-4-3 to obtain the deterministic output dtOp and the gradual output gdOp respectively, and return these two results. dtOp represents the size of the hot spot, and gdOp represents the gradual change of the hot spot.

[0082] Through steps 3-4, we can obtain the task heat map of the service scene to predict the service demand in each area. This step is to process the task data in the area first, and determine the area where service needs to be provided through previous task data. This process is scene task prediction, and the result is the task heat map. This process does not involve UAV, but paves the way for the subsequent UAV trajectory optimization, that is, area k is generated here. The purpose of doing this is that if there are hundreds or thousands of users in a scene, it is impossible to move to each user to provide services, so dividing the users into areas in advance can greatly improve the service efficiency of UAV.

[0083] Step 4: Based on the fuzzy data fusion and scene mission intention prediction in step 3, a delay approximation algorithm based on the mission area is designed to optimize the total delay of the drone completing all user tasks in the area. The delay approximation algorithm based on the mission area piecewise function has the following steps:

[0084] Step 4-1: The goal of the algorithm is to reduce the total delay of the drone completing all user tasks in the area The flight time and task processing time of the drone are also taken into account. This is to ensure that the maximum number of users can be served in the shortest time, thereby maximizing the overall efficiency of the system.

[0085] Step 4-2: Establish a delay calculation model that includes the UAV flight time, data transmission time, and data processing time. This model takes into account factors such as the distance from the UAV to the user, the flight speed of the UAV, the size of the data, the transmission rate, and the processing rate.

[0086] Step 4-3: Minimize the total delay of task execution in the area by determining the flight trajectory and regional positioning of the UAV. The optimization process takes into account the dynamic position of the UAV and the task requirements of users in the area;

[0087] Step 4-4: After determining the service area, optimize the execution time allocated by the drone to each user so as to minimize the flight time and waiting time while maintaining the highest efficiency.

[0088] Using the above algorithm, a drone swarm task intention prediction and path recommendation system can be generated based on multi-source data and predicted scene task generation intention, including system architecture, drone swarm simulation module, drone control data transmission service module, perception data visualization module, fuzzy data processing, intention prediction and path recommendation algorithm module, and platform comprehensive function service module.

[0089] The system architecture is designed based on the front-end and back-end separation architecture of B / S (Brower / Server). The service layer uses SpringBoot as the main back-end framework. The FastAPI framework is used to provide drone control data transmission services. The persistence layer uses the relational database MySQL, and Redis is used to solve the problem of multi-terminal user login and cache concurrent data. The front-end framework of the presentation layer uses Vue.js. The simulation module uses the Gazebo simulator and SITL to realize the simulation experiment capability. The hardware layer uses the MAVLink protocol to communicate with the drone. In summary, the system architecture is divided into the presentation layer, the back-end service layer, the algorithm module, the simulation module, the hardware module and the technical middle platform layer, such as Figure 2 shown.

[0090] The following is a summary of the functions of the drone swarm mission intention prediction and path recommendation system based on multi-source data and predicted scene mission intention:

[0091] 1. UAV swarm simulation module: Figure 3 As shown, the present invention uses the physical engine provided by the Gazebo simulation engine to implement an accurate flight dynamics model, taking into account factors such as the power, aerodynamics and inertia of the aircraft. The environment setting uses Gazebo to create a three-dimensional simulation environment, including terrain, buildings and weather conditions. The simulation runs by running the SITL version of ArduPilot on the computer to simulate the flight control system of the drone. The communication link establishes communication between the simulated drone and the drone control data transmission module through the MAVLink protocol. This module is used to support system testing, verification and performance evaluation.

[0092] 2. The drone control data transmission service module is based on the FastAPI framework, such as Figure 4 As shown in the figure, the asynchronous technology is fully applied to provide real-time control services for the UAVs in the simulation module and the real UAV swarm. The module can flexibly control the flight, status, unloading and uploading tasks of the UAVs, and is responsible for receiving environmental data of the simulation environment and the real world environment and passing it to the comprehensive function service module. The specific working process is:

[0093] (1) Define the running_tasks list to store the asynchronous tasks being executed;

[0094] (2) Create an empty list mission_items to store flight mission items;

[0095] (3) Traverse self.missionItemlist and add each mission item to the list;

[0096] (4) Create a MissionPlan object using these task items;

[0097] (5) Use the await keyword to asynchronously upload the mission plan to the drone and start the mission;

[0098] (6) The program asynchronously monitors the flight status of the drone and determines the flight status by detecting whether the drone is in the air. If the drone takes off, the variable was_in_air will be set to true. If the drone was in the air but has landed, all running tasks will be canceled;

[0099] (7) Close the asynchronous generator and clean up resources.

[0100] 3. The perception data visualization module is used to process and present the perception data in the UAV system. The schematic diagram is as follows: Figure 5 As shown. It includes environmental information collected by sensors, data during task execution, etc. Through the visual interface, users can intuitively understand the system status and environmental conditions. It realizes data integration and display, and supports the simultaneous display of perception data from different sensors and drones; it includes the following modules:

[0101] (1) System comprehensive visualization screen: The thermal area displays the scenario mission intent predicted by the system;

[0102] (2) Drone group and equipment management page: displays the group, status, equipment parameters and usage time of the drone group, and displays the equipment type, parameters, status and usage time of sensors carried by different drones;

[0103] (3) Task management page: displays the task type, task assignment, start time and task progress of the drone group's pending tasks, ongoing tasks, and completed tasks.

[0104] 4. Fuzzy data processing, intention prediction and route recommendation algorithm modules face fuzzy and uncertain information data from different sources, such as Figure 6 As shown in the figure, through fuzzy logic modeling and reasoning, more accurate and robust data fusion results are provided; at the same time, in response to new scenario environments where drone swarms perform tasks, the task distribution intention model of the current scenario can be generated through historical task experience, membership function set experience, etc.; finally, a variety of paths with different focuses are generated through utility optimization paths to support decision-making under different needs.

[0105] 5. The platform comprehensive function service module serves as the core service of the UAV system, integrating and providing the functions of each module of the system, such as Figure 7 As shown. Spring Cloud is used to achieve communication and collaboration between microservices to ensure the elasticity and scalability of the system. Nacos is used as the service registration and discovery center to ensure dynamic management and discovery of each service node. The steps of the platform comprehensive function service module are as follows:

[0106] Step 1): The user management module implements basic operations such as user registration, login, and information modification. Use SpringSecurity to authenticate users and ensure the security of user information. Combined with Nacos, dynamic role and permission configuration is implemented to achieve flexible user permission management. Use Redis to cache user information to improve the access speed of user information and system performance.

[0107] Step 2): The system management module provides management of system configuration information, including database connection information, service ports, etc. Use Nacos as a distributed configuration center to achieve centralized management of system configuration.

[0108] Step 3): The drone simulation service provides drone simulation services by accessing the drone simulation module, including drone flight trajectory simulation, mission scenario environment simulation, and sensor data simulation.

[0109] Step 4): The drone control data transmission service controls the simulation module by accessing the drone control data transmission module, and provides control services for the flight and state changes of drones in the simulation and real drone swarms, including flight control, task loading, and perception data feedback.

[0110] Step 5): Algorithm service integrates algorithm modules into the system to provide algorithm processing services for drone data, including fusion services for perceived fuzzy data, mission intent prediction, and path recommendation capabilities.

[0111] Figure 7 It is a simulated visual perception map of the UAV of the present invention; Figure 8 It is a simulation diagram of unloading the UAV mission of the present invention; Fig. 9 Comparison of the predicted task heat map generated by the algorithm designed for the present invention with the real task heat map; Fig.10 Schematic diagram of recommended paths generated by the algorithm TDTS designed by the present invention and other three baseline algorithms; Fig.11 A performance comparison diagram of the algorithm TDTS designed by the present invention and other three baseline algorithms in terms of system throughput under different numbers of households; Fig.12 The performance comparison diagram of the algorithm TDTS designed by the present invention and other three baseline algorithms in terms of service percentage under different numbers of households. The schematic diagram shows that the recommended trajectory generated by the present invention has a higher system throughput while ensuring a shorter trajectory distance.

[0112] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for predicting mission intention and recommending paths for drone swarms, characterized in that: The following steps are involved: Step 1: Build a scenario model: Assume that the number of drones is M, and the drones need to perform tasks in K areas. Each area consists of N independent task sources, and the user i∈N; Step 2: Scenario formulation, the steps are as follows: Step 2-1: Construct the UAV maneuverability constraint: The maximum speed of the UAV is expressed as v max , e t represents the time slot duration, p t and q k denote the abscissa of the UAV in time slot t and the abscissa of user k respectively; the position constraint across consecutive time slots is written as: Step 2-2, define the completeness and atomicity of the task; since the task cannot be further compressed or partitioned, the drone can only decide whether to complete the entire task or reject the task request; I i,k,t and O i,k,t They represent the input and output data size of the task at time t, I i,k and O i,k They represent the total task input and output size of user i in region k respectively. In order to ensure the completeness and atomicity of the task, the numerical relationship is expressed as: Step 2-3, define the transmission rate limit: Due to the limitation of channel conditions, both uplink and downlink communications cannot exceed the maximum transmission rate. The transmission rate between user i and the drone is expressed as: where a i,k,t represents the task upload variable, B represents bandwidth, P i,k represents the transmission power of the ith user, g i,k,t represents the channel power gain of the i-th user in region k in time slot t, σ 2 represents additive white Gaussian noise; Step 2-4: Define the effective time limit of the task; define the minimum delay constraint of user i in area k as τ i,k , use D i,k =e i,k +X i,k Represents the total task execution time of the user, including the transmission delay Use r i,k,t and M i,k They represent the computing resources allocated to user i in region k at time t and the computing resource requirements of the user, respectively. The task execution time is expressed as The user's total task execution time must meet the following requirements: Indicates that any task completed within the system time period meets the corresponding deadline, where τ i,k represents the deadline of the i-th user in the k-th region; Step 2-5: Define the optimization objective function: Step 2-6, fuzzy data: define the universe set U, a fuzzy set A on U to assign a membership degree A(x) between 0 and 1 to each element x, the degree to which element x belongs to the fuzzy set A is expressed as: A={(x,μ A (x))|x∈U}; The universe set U represents the set of UAV airborne missions; x represents the value of the specific mission; the fuzzy set A represents the fuzzy classification of the mission; μ A (x) represents the degree to which the specific task quantity x belongs to the fuzzy set A; Step 2-7, multi-source data fusion: define a data set MsD, which contains n m-dimensional vectors; define FD as a set, the elements of which are the results of mapping data from different data sources to each vector in MsD through f; MsD={d i |d i =(x i1 ,x i2 ,...,x im ),i=1,2,...,n} FD={f(x 11 ,x 12 ,...,x 1m ),f(x 21 ,x 22 ,...,x 2m ),...,f(x n1 ,x n2 ,...,x nm )}; where d i represents the task vector of user i, x i,m represents the characteristics of the mth task of user i; Step 3: Fuzzy data fusion and scene task intention prediction. The steps are as follows: Step 3-1, construct the membership function: extract the feature vector x of the task from the actual data set and use the prototype vector r to represent the center of the fuzzy set, calculate the Euclidean distance dst between x and the prototype vector r; use the formula μ x =1 / (1+dst) to calculate the fuzzy membership μ x , indicating the degree to which x belongs to a fuzzy set; The eigenvector x is the i,m The vector formed; Step 3-2, formulate fuzzy rules; Step 3-3: Historical scene matching: using the task decisions in historical experience to quantify the similarity of the newly input tasks, and obtain the similarity of the task to different categories; Step 3-4: Task area prediction based on fuzzy logic, using fuzzy membership tool to process the above fuzzy data fusion data, distinguish the spatial set and type of tasks according to membership and similarity, and obtain the scene task distribution intention; Step 4: Based on the fuzzy data fusion and scene mission intention prediction in step 3, a delay approximation algorithm based on the mission area is designed to optimize the total delay of the drone to complete all user tasks in the area.

Images