Distributed cooperative task allocation method for unmanned aerial vehicles in unmanned bee colony

Abstract

The invention discloses a distributed cooperative task allocation method for unmanned aerial vehicles in an unmanned bee colony. The method comprises the following steps: initializing unmanned aerial vehicle parameters, task parameters and GCAA parameters; creating an unmanned aerial vehicle and a task structural body, and uniformly packaging each parameter into the unmanned aerial vehicle and the task structural body for analog simulation; information beam communication between the unmanned aerial vehicles is constructed, and timely updating between information beams is realized; creating a graph window substrate, an unmanned aerial vehicle and task elements, and constructing a GCAA solving function; judging whether the communication of the unmanned aerial vehicle is limited or not, if the communication of the unmanned aerial vehicle is limited, firstly updating the communication matrix of the unmanned aerial vehicle and then calling the GCAA solving function to calculate the GCAA solution allocated by the task of the unmanned aerial vehicle, and if the communication of the unmanned aerial vehicle is not limited, directly calling the GCAA solving function to calculate the GCAA solution, recording the motion path of the unmanned aerial vehicle and the GCAA operation time, and calculating the total utility of the current simulation round; and dynamic task allocation and utility calculation are completed. According to the method, the overall effectiveness is optimal on the basis of effectively ensuring that the benefit of each unmanned aerial vehicle is as large as possible.

Description

Technical Field

[0001] The present application relates to a method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony, and belongs to the technical field of unmanned aerial vehicle task allocation. Background Art

[0002] In the current stage of UAV technology development, when unmanned swarms perform tasks in complex dynamic environments, there are still problems such as poor autonomous organization effectiveness, chaotic task execution, and insufficient reliability. In order to comprehensively consider the influence of multiple factors such as battlefield situation changes, problem solving scale, combat effectiveness objective function, weapon resource constraints, etc., in order to maximize the overall combat effectiveness value of all combat aircraft in performing tasks, multi-machine collaborative task allocation can be divided into two types: centralized allocation and distributed allocation according to different task allocation planning structures.

[0003] The centralized task allocation mechanism first gathers the situational awareness information of the battlefield to the central information processing unit, which develops a detailed task allocation plan for the entire combat fleet based on complex strategy generation algorithms, such as the principles of many classical mathematical models such as linear programming, covariance control, dynamic programming and game theory. Subsequently, these task allocation instructions are accurately issued to each fighter, and they each perform the assigned specific combat tasks. For simple task allocation scenarios, the industry often adopts optimal solution strategies, such as exhaustive search and mixed integer linear programming, to ensure the optimization of the allocation scheme. However, due to their inherent high time complexity, such methods are only applicable to scenarios with smaller task scales. When the battlefield environment becomes large and complex, the rapid expansion of the solution space leads to an exponential increase in the computational burden, making the optimal solution method infeasible in practical applications. In addition, the inherent defect of centralized task allocation is its strong dependence on the communication network. All fighters need to maintain close communication with the central processing center, which not only limits the scale of tasks that the fleet can handle, but also introduces the risk of single point failure - once the central system is damaged or fails, the entire task allocation system will be paralyzed, seriously affecting combat effectiveness. This limitation is particularly prominent in the context of today's increasingly expanding battlefield environment and increasingly complex and changeable task types, making it difficult to meet the urgent needs of modern warfare for efficiency, flexibility and robustness.

[0004] Distributed task allocation planning disperses information processing and decision-making power to each fighter, realizes independent calculation and collaboration among fighters, effectively reduces the pressure of communication bandwidth and the computing load of a single processing center, makes up for the limitations of the centralized allocation model, and is more in line with the dynamic, complex and demanding characteristics of modern battlefields. Under this framework, traditional distributed task allocation algorithms often rely on auction algorithms and contract network protocols under market mechanisms. Auction algorithms focus on bidding for resources through the consistency of situational awareness among each allocation unit, while contract network protocols emphasize consistency in the task space to achieve cooperation. Compared with centralized solutions, these methods significantly shorten the task planning cycle. In the collaborative combat system, drones play the role of bidders and participate in resource competition with mission requirements as the goal. Auction algorithms can be divided into central control and decentralized types according to different management modes. The former relies on a central system to manage the auction process, evaluate and select the bidder with the highest bid; the latter innovatively selects temporary managers from the group of agents participating in the auction and executes the auction in a decentralized manner. Although auction algorithms have shown advantages in solving consistency problems, their solution process is relatively time-consuming. To this end, two new robust distributed consistency algorithms have emerged, namely, Consensus-Based Auction Algorithms (CBAA) and Consensus-Based Bundle Algorithm (CBBA), with the latter being an optimization and extension of the former in dealing with multi-task allocation problems. Both algorithms combine the wisdom of market decision-making with local communication consistency strategies, with the former being used for distributed task selection and the latter as a conflict resolution mechanism. Under reasonable assumptions and network connectivity guarantees, they have been proven to have excellent fast convergence characteristics, providing strong support for efficient task allocation in complex battlefield environments.

[0005] There are various methods for cooperative task allocation of UAVs, each with its own advantages and disadvantages. Among them, the method based on market mechanism competition adopts a competitive mechanism of bidding and tendering, which can coordinate the tasks between multiple UAVs and has the advantages of less communication volume and good robustness. However, this type of method also has the following problems, such as not considering the differences in tasks before the task auction, and may not guarantee the optimal overall utility. Summary of the invention

[0006] According to one aspect of the present application, a method for allocating distributed collaborative tasks among drones in an unmanned swarm is provided. The method effectively ensures that the overall utility is optimized on the basis of maximizing the benefit of each drone by combining a greedy method with an auction process.

[0007] A method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee swarm, characterized by comprising:

[0008] Step 1: Initialize UAV parameters, mission parameters and GCAA parameters;

[0009] Step 2: Create drone and mission structures, and encapsulate all parameters into the drone and mission structures for simulation;

[0010] Step 3: Build information beam communication between drones to achieve timely updates between information beams;

[0011] Step 4: Create a window base for displaying simulation, as well as drone and mission elements, and build a GCAA solver function;

[0012] Step 5: Determine whether the drone communication is restricted. If it is restricted, first update the drone communication matrix and then call the GCAA solver function to calculate the GCAA solution of the drone task assignment. If it is not restricted, directly call the GCAA solver function to calculate the GCAA solution, record the drone motion path and GCAA operation time, and calculate the total utility of the current simulation round.

[0013] Step 6: Repeat steps 4 to 5 until all dynamic task allocation and utility calculation are completed.

[0014] Furthermore, the drone parameters include drone state space, drone input space, drone type set, drone quantity and drone attribute values;

[0015] The task parameters include a task type set, a task quantity, and a task attribute value;

[0016] The GCAA parameters include the number of algorithm iterations M, the simulation time t s , simulation step length t s and the simulation start time t0.

[0017] Furthermore, the drone state space S i It is expressed as: represents the drone state space belonging to the m-dimensional vector space;

[0018] The drone input space U i Expressed as: u i ∈U i ;

[0019] Among them, x i is the state of the i-th drone;

[0020] u i is the input of the i-th UAV;

[0021] i∈[1,n] dIt represents the corresponding physical quantity of the i-th UAV in the multi-UAV system at time t ≥ 0;

[0022] but:

[0023] The joint state space S of multiple UAVs is expressed as: S = S1 × ··· × S n ;

[0024] The joint input space U: of multiple UAVs is expressed as: U: = U1×···×U n ;

[0025] The joint state x of the multi-UAV system is expressed as: x∈S, x:=(x1,...,x n );

[0026] The joint input u of the multi-UAV system is expressed as: u∈U, u:=(u1,...,u n );

[0027] The motion of the i-th UAV is expressed as:

[0028]

[0029] in is the initial state of the ith UAV, f i :S i ×U i →S i is the associated vector field of the ith UAV, It is the joint initial state of the multi-UAV system.

[0030] Furthermore, the drone type set includes attack drones D atc and reconnaissance drone D rec ;

[0031] The number of drones is represented by a drone set D = {D1, D2, D3, ..., D n};

[0032] The attribute values ​​of the drone include position (X i ,Y i ),speed The maximum number of tasks that can be executed L(D i ) and the flight drag coefficient k(D i ), where i represents the i-th drone, D i (i=1,2,3,...,n).

[0033] Furthermore, the mission type set includes strike missions and reconnaissance missions

[0034] The number of tasks is represented as a task set of m tasks

[0035] Task attribute values ​​include location Probability of completion Time required to be completed The hovering time required to be completed The required circling radius is completed Rewards for drone completion and the coefficient for converting the cost value unit required for the drone to complete the task into the reward received Among them, j represents the jth task,

[0036] Furthermore, constructing the GCAA solution function includes calculating the cost of the drone completing the task, calculating the reward of the drone completing the task, and calculating the utility of the drone completing the task.

[0037] Furthermore, the cost of calculating the UAV's mission is expressed as:

[0038] The i-th drone is Moment completion and status Related tasks j The cost:

[0039] Assume a i =T j ,

[0040] Among them, T j ∈T, i∈[1,n] d ;

[0041] make Indicates that T j The associated state;

[0042] Indicates T j The relevant completion time,

[0043] Obtaining the optimal control input

[0044] Minimize the function:

[0045]

[0046] Make

[0047] in, is a given C 1 Functions are respected;

[0048] Furthermore, the minimum cost is:

[0049] Furthermore, the reward for the drone to complete the task is calculated as:

[0050] Given an allocation profile a = (a1, ..., a n ), use T j -1 (a) represents the task T assigned under a specific profile j The index set corresponding to the drones of ∈T;

[0051] Let p ij ∈[0,1] is the task T j The probability of being successfully completed by the i-th drone;

[0052] At this time, the probability that at least one drone successfully completes the task increases with the number of drones assigned to the task. j The expected reward of is defined as:

[0053]

[0054] in, It is T j Nominal reward;

[0055] is the probability that no drone completes the task, where the probability that at least one drone completes the task is equal to the complement of the probability that no drone completes the task, that is,

[0056] Furthermore, the utility of the drone completing the task is expressed as:

[0057] Set relevant physical quantities to describe the motion state of the drone, that is

[0058] in, and p i ∈R 2 , They represent the acceleration, position and velocity of the i-th UAV at t0=0 in the two-dimensional vector space, i∈[1,n] d ;

[0059] represents the drone state space belonging to the 2-dimensional vector space;

[0060] Performance indicators are determined by control efforts given;

[0061] set up represents the drone state space belonging to the 4-dimensional vector space;

[0062] Then randomly select the terminal constraint function and take the difference between the two:

[0063] Indicates that the i-th drone attempts to Terminal velocity Arrives at its assigned task T j Related Locations

[0064] Indicates that the i-th drone attempts to Arrives at its assigned task T j The surrounding circle has a radius of Then wander around the target until

[0065] At this point, the best entry point into the circle is selected by discretizing the circle into 10 points and selecting the point with the smallest cost function;

[0066] Optimal control input:

[0067]

[0068] in

[0069] Task utility is characterized by completing task T j ∈T and the state-dependent cost reflecting the cost of completing the task, given an allocation profile a = (a1, ..., a n ), conversion constant Used to complete the mission of the drone j The unit of measurement of the required cost is converted to the same as the reward, j∈[1,p] d , Task T j Total cost to complete:

[0070]

[0071] Given x0 and task T j Related total task utility:

[0072]

[0073] in, is a constant used to convert the cost into the same units as the reward;

[0074] Denote the global utility as

[0075] Based on the task assignment a, the individual utility of drone i is set to its contribution to the global utility U(a; x 0 )'s marginal contribution:

[0076]

[0077] The beneficial effects of this application include:

[0078] The present application provides a method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned swarm, which can quickly allocate suitable drones for different types of tasks through a greedy alliance auction algorithm. In addition, by introducing a greedy method and combining it with the auction process, the final allocation result is more conducive to ensuring the optimal overall utility on the basis of maximizing the benefits of each drone, thereby optimizing the overall task execution efficiency. In an environment with the advantages of drone clusters, in a combat mode with a large number of task types and quantities, the drones in the swarm can be efficiently and quickly classified and allocated in an orderly manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of a method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned swarm in one embodiment of the present application;

[0080] Figure 2 A schematic diagram of a curve showing a task allocation simulation result and a measurement index change under a communication restriction in one embodiment of the present application;

[0081] Figure 3 This is a schematic diagram of the task allocation simulation results and measurement index change curves in an implementation manner of the present application without communication restrictions. DETAILED DESCRIPTION

[0082] The present application is described in detail below with reference to embodiments, but the present application is not limited to these embodiments.

[0083] See also Figure 1-3 , a method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee swarm, characterized by comprising:

[0084] Step 1: Initialize UAV parameters, mission parameters and GCAA parameters;

[0085] Step 2: Create drone and mission structures, and encapsulate all parameters into the drone and mission structures for simulation;

[0086] Step 3: Build information beam communication between drones to achieve timely updates between information beams;

[0087] Step 4: Create a window base for displaying simulation, as well as drone and mission elements, and build a GCAA solver function;

[0088] Step 5: Determine whether the drone communication is restricted. If it is restricted, first update the drone connection matrix (construct an n×n-order matrix according to the number of drones n to reflect the on-off status of the information between drones. For example, when the kth drone and the lth drone can be connected and transmit information to each other because the distance requirement is met, the elements in the kth row and lth column of the matrix and the elements in the lth row and kth column are set to 1, otherwise they are set to 0) and then call the GCAA solver function to calculate the GCAA solution of the drone task allocation. If it is not restricted, directly call the GCAA solver function to calculate the GCAA solution, record the drone motion path and GCAA operation time, and calculate the total utility of the current simulation round;

[0089] Step 6: Complete dynamic task allocation and utility calculation.

[0090] Specifically, the relevant parameters of drones and tasks are encapsulated into structures for easy management and simulation. The structures can contain various attributes, such as status, capabilities, and task requirements. Information bundle communication is an efficient way of transferring information, especially suitable for distributed systems. It allows drones to exchange status, task progress, and decision information in real time or near real time. The update mechanism ensures the accuracy and timeliness of information, which is essential for dynamically changing environments and tasks. The figure base refers to the basic framework of the physical or logical environment used for visualization or simulation. Elements include drones and task nodes, as well as possible connections or relationships between them. The GCAA (Greedy Coalition Auction Algorithm) solver implements a specific task allocation algorithm and calculates the optimal or suboptimal task allocation scheme based on the current status of drones and tasks. Communication limitations caused by environmental factors (such as terrain occlusion, electromagnetic interference) or the drone's own status (such as insufficient energy, failure). When communication is limited, the connection status between drones is updated to ensure that these limitations are taken into account when allocating tasks. Whether communication is limited or not, the GCAA solver is used to calculate the task allocation scheme. The recorded data include the UAV’s motion path, GCAA operation time, and the total utility of the current simulation round, which are crucial for evaluating algorithm performance and optimizing strategies.

[0091] Among them, the basic parameters of the settings are imported and encapsulated together with the drone and task structures into a structure for simulation. Then, the two situations of whether there is communication restriction or not are traversed respectively, and a large loop is entered to create a window base and drone and task elements. Then, the drone and task are traversed to calculate the cost, reward, and utility of this round. Next, it is determined whether the communication is restricted. If it is restricted, the drone communication matrix is ​​updated first, and then the GCAA algorithm main function is called to calculate the GCAA solution of the drone task assignment. If it is not restricted, the function is directly called to calculate the GCAA solution, and the drone motion path and GCAA operation time are recorded. The total utility of the current simulation round is further calculated. Finally, it is determined whether the set number of iterations has been run. If not, the loop continues. If it has been run, the loop ends and the parameters and calculation results are stored in a table file.

[0092] The drone parameters include drone state space and input space, drone type set, drone quantity and drone attribute values;

[0093] The task parameters include a task type set, a task quantity, and a task attribute value;

[0094] The GCAA parameters include the number of algorithm iterations M, the simulation time t s , simulation step length t s and the simulation start time t0.

[0095] The drone state space S i It is expressed as: represents the drone state space belonging to the m-dimensional vector space;

[0096] The drone input space U i Expressed as: u i ∈U i ;

[0097] Among them, x i is the state of the i-th drone;

[0098] u i is the input of the i-th UAV;

[0099] i∈[1,n] d It represents the corresponding physical quantity of the i-th UAV in the multi-UAV system at time t ≥ 0;

[0100] but:

[0101] The joint state space S of multiple UAVs is expressed as: S = S1 × ··· × S n ;

[0102] The joint input space U: of multiple UAVs is expressed as: U: = U1×···×U n;

[0103] The joint state x of the multi-UAV system is expressed as: x∈S, x:=(x1,...,x n );

[0104] The joint input u of the multi-UAV system is expressed as: u∈U, u:=(u1,...,u n );

[0105] The motion of the i-th UAV is expressed as:

[0106]

[0107] in is the initial state of the ith UAV, f i :S i ×U i →S i is the associated vector field of the ith UAV, It is the joint initial state of the multi-UAV system.

[0108] The drone type set includes attack drones D atc and reconnaissance drone D rec ;

[0109] The number of drones is represented by a drone set D = {D1, D2, D3, ..., D n};

[0110] The attribute values ​​of the drone include position (X i ,Y i ),speed The maximum number of tasks that can be executed L(D i ), flight drag coefficient k(D i ), where i represents the i-th drone, D i (i=1,2,3,...,n).

[0111] The mission type set includes strike missions and reconnaissance missions

[0112] The number of tasks is represented as a task set of m tasks

[0113] Task attribute values ​​include location Probability of completion Time required to be completed The hovering time required to be completed The required circling radius is completed Rewards for drone completion and the coefficient for converting the cost value unit required for the drone to complete the task into the reward received Among them, j represents the jth task,

[0114] Specifically, in the present invention, u i ∈U i , where i∈[1,n] d They are the state space and input space (S) of the ith UAV in the multi-UAV system (MAS) at time t≥0. i For its state space, U i input space for it), On this basis, we define x∈S as the joint state of MAS, where x:=(x1,...,x n ), S: = S1 × ··· × S n (joint state space), and define u∈U as the joint input of MAS, where u:=(u1,...,u n ), U:=U1×···×U n (Joint input space). The motion of the i-th drone is

[0115]

[0116] To describe, is the initial state of the ith UAV, f i :S i ×U i →S i is its associated vector field. Therefore, is the joint initial state;

[0117] Set the mission type to strike mission Reconnaissance mission There are two types, the drone types are divided into attack drones D atc , reconnaissance drone D rec Two types;

[0118] Suppose there are n drones in the drone set D = {D1, D2, D3, ..., D n}, a task set containing m tasks In the case of the present invention, n=10, m=10 are selected. The number of reconnaissance tasks is half of the total number of tasks, that is, m1=5;

[0119] Drone D i (i=1,2,3,...,n) has seven attributes: position (X i ,Y i ),speed The maximum number of tasks that can be executed L(D i ), flight drag coefficient k(D i ), set the attribute values ​​corresponding to each drone. There are seven attributes: Position Probability of completion Time required to be completed The hovering time required to be completed and the required circling radius to be completed Nominal reward for drone completion The coefficient for converting the cost value unit required for the drone to complete the task into the reward obtained. Set the attribute values ​​corresponding to each task. According to the actual combat scenario, the horizontal and vertical coordinates of the initial positions of all drones and tasks in the present invention are set to random numbers generated between 0.1 times the map width and 0.9 times the map width. and The values ​​of are all set to -0.1 to 0.1. The maximum number of tasks that a single UAV can complete is L(D i )=1, the UAV flight drag coefficient k(D i )=0.3, the probability of each task being completed Set to a random number between 0 and 1, the required circling radius for the reconnaissance mission to be completed Set to 0.04 times the map width, the nominal reward for the drone to complete the task Set to a random number between 0 and 0.2, the coefficient for converting the cost value unit required for the drone to complete the task into the reward obtained Set to 1.

[0120] Set the algorithm iteration termination number M. When the number of iterations is reached, stop the auction and output the final task allocation result. In the case of the present invention, set the algorithm iteration termination number M = 200. Set the simulation time t s , then press Set the simulation step size And set the simulation start time t0. In the present invention, set the simulation time t s =10, simulation step length And set the simulation starting time t0=0.

[0121] The construction of the GCAA solution function includes calculating the cost of the drone completing the task, calculating the reward of the drone completing the task, and calculating the utility of the drone completing the task.

[0122] The cost of calculating the UAV's mission is expressed as:

[0123] The i-th drone is Moment completion and status Related tasks j The cost:

[0124] Assume a i =T j ,

[0125] Among them, T j ∈T, i∈[1,n] d ;

[0126] make Indicates that T j The associated state;

[0127] Indicates T j The relevant completion time,

[0128] Obtaining the optimal control input

[0129] Minimize the function:

[0130]

[0131] Make

[0132] in, is a given C 1 Functions are respected;

[0133] Furthermore, the minimum cost is:

[0134] The reward for the drone to complete the task is calculated as:

[0135] Given an allocation profile a = (a1, ..., a n ), use T j -1 (a) represents the task T assigned under a specific profile j The index set corresponding to the drones of ∈T;

[0136] Let p ij ∈[0,1] is the task T j The probability of being successfully completed by the i-th drone;

[0137] At this time, the probability that at least one drone successfully completes the task increases with the number of drones assigned to the task. j The expected reward of is defined as:

[0138]

[0139] in, It is T j Nominal reward;

[0140] is the probability that no drone completes the task, where the probability that at least one drone completes the task is equal to the complement of the probability that no drone completes the task, that is,

[0141] Specifically, given an allocation profile a=(a1,...,a n ), use T j -1 (a) represents the task T assigned under a specific profile j ∈T corresponding to the index set of drones. A task is not necessarily completed when it is assigned to a drone, so let p ij ∈[0,1] is the task T j The probability of being successfully completed by the i-th UAV. At this time, the probability of at least one UAV successfully completing the task increases with the number of UAVs assigned to the task. j The expected reward is defined as

[0142]

[0143] in It is T j The nominal reward is , and because the probability that at least one drone completes the task is equal to the complement of the probability that no drone completes the task, that is,

[0144] The utility of calculating the UAV's task completion is expressed as:

[0145] Set relevant physical quantities to describe the motion state of the drone, that is

[0146] in, and They represent the acceleration, position and velocity of the i-th UAV at t0=0 in the two-dimensional vector space, i∈[1,n] d ;

[0147] represents the drone state space belonging to the 2-dimensional vector space;

[0148] Performance indicators are determined by control efforts given;

[0149] set up represents the drone state space belonging to the 4-dimensional vector space;

[0150] Then randomly select the terminal constraint function and take the difference between the two:

[0151] Indicates that the i-th drone attempts to Terminal velocity Arrives at its assigned task T j Related Locations

[0152] Indicates that the i-th drone attempts to Arrives at its assigned task T j The surrounding circle has a radius of Then wander around the target until

[0153] At this point, the best entry point into the circle is selected by discretizing the circle into 10 points and selecting the point with the smallest cost function;

[0154] Optimal control input:

[0155]

[0156] in

[0157] Task utility is characterized by completing task T j ∈T and the state-dependent cost reflecting the cost of completing the task, given an allocation profile a = (a1, ..., a n ), conversion constant Used to complete the mission of the drone j The unit of measurement of the required cost is converted to the same as the reward, j∈[1,p] d , Task T j Total cost to complete:

[0158]

[0159] Given x0 and task T j Related total task utility:

[0160]

[0161] in, is a constant used to convert the cost into the same units as the reward;

[0162] Denote the global utility as

[0163] Based on the task assignment a, the individual utility of drone i is set to its contribution to the global utility U(a; x 0 )'s marginal contribution:

[0164]

[0165] Specifically, a set of drones with dual integrator dynamics is constructed, i.e. where p i ∈R 2 and Respectively represent the i-th unmanned

[0166] Then randomly select the terminal constraint function and take the difference between the two:

[0167] Indicates that the i-th drone attempts to Terminal speed

[0168] Arrives at its assigned task T j The surrounding circle (radius is ), then wander around the target until In this case, the best entry point into the circle is chosen by discretizing the circle into 10 points and selecting the point with the smallest cost function.

[0169] Both terminal constraints are associated with an optimal control problem with non-zero initial and terminal velocities, and the optimal control input is given by

[0170]

[0171] Given, it defines a second-order differential equation with time-varying coefficients, where The numerical solution was performed using the integration tool (ODE45) in Matlab.

[0172] While the problem with zero terminal velocity has an analytical solution, the problem with non-zero terminal velocity requires more computation time due to numerical integration, which is then performed and presented as a dynamic task allocation through a dynamic mapping of the assignments. The task utility is characterized by the number of tasks completed, T j ∈T and the state-dependent cost reflecting the cost of completing the task (e.g., the transfer cost due to the motion of the drone), given an allocation profile a = (a1, ..., a n ), Task T j The total cost of completion is given by

[0173]

[0174] Given x0, this leads to the task T j The definition of the relevant total task utility

[0175]

[0176] in is a constant used to convert the cost to the same units as the reward (e.g., from energy loss to money loss).

[0177] Denote the global utility as The goal is to set the utility of the team equal to the sum of the utilities of each individual in order to maximize the utility of each individual separately. To this end, based on the task assignment a, set the individual utility of drone i to its contribution to the global utility U(a; x 0 )'s marginal contribution:

[0178]

[0179] This application imports the basic parameters set, and encapsulates them together with the drone and task structures into a structure for simulation. Then, the two situations of whether there are communication restrictions or not are traversed respectively, and a large loop is entered to create a window base and drones and task elements. Then, the drones and tasks are traversed to calculate the cost, reward, and utility of this round. Next, it is determined whether the communication is restricted. If it is restricted, the drone communication matrix is ​​updated first, and then the GCAA algorithm main function is called to calculate the GCAA solution of the drone task assignment. If it is not restricted, the function is directly called to calculate the GCAA solution, and the drone motion path and GCAA operation time are recorded. The total utility of the current simulation round is further calculated. Finally, it is determined whether the set number of iterations has been run. If it has not been run, the loop continues. If it has been run, the loop ends and the parameters and calculation results are stored in a table file. In the case of the present invention, the communication distance radius with communication restrictions is set to 0.3 times the map width, and the communication distance radius without communication restrictions is set to 1 times the map width. The map width is uniformly set to 1. The simulation results of the final task allocation and the cost, reward, and utility curves of the completed tasks are shown as follows. Figure 2 and Figure 3 shown.

[0180] The above are only a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application is disclosed as above with preferred embodiments, it is not intended to limit the present application. Any technician familiar with the profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee swarm, characterized in that: include: Step 1: Initialize UAV parameters, mission parameters and GCAA parameters; Step 2: Create drone and mission structures, and encapsulate all parameters into the drone and mission structures for simulation; Step 3: Build information beam communication between drones to achieve timely updates between information beams; Step 4: Create a window base for displaying simulation, as well as drone and mission elements, and build a GCAA solver function; Step 5: Determine whether the drone communication is restricted. If it is restricted, first update the drone communication matrix and then call the GCAA solver function to calculate the GCAA solution of the drone task assignment. If it is not restricted, directly call the GCAA solver function to calculate the GCAA solution, record the drone motion path and GCAA operation time, and calculate the total utility of the current simulation round. Step 6: Repeat steps 4 to 5 until all dynamic task allocation and utility calculation are completed.

2. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 1, characterized in that: The drone parameters include drone state space, drone input space, drone type set, drone quantity and drone attribute values; The task parameters include a task type set, a task quantity, and a task attribute value; The GCAA parameters include the number of algorithm iterations M, the simulation time t s , simulation step and the simulation start time t0.

3. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 2, characterized in that: The drone state space S i It is expressed as: represents the drone state space belonging to the m-dimensional vector space; The drone input space U i Expressed as: i ∈U i ; Among them, x i is the state of the i-th drone; u i is the input of the i-th UAV; i∈[1,n] d It represents the corresponding physical quantity of the i-th UAV in the multi-UAV system at time t ≥ 0; but: The joint state space S of multiple UAVs is expressed as: S = S1 × ··· × S n ; The joint input space U: of multiple UAVs is expressed as: U: = U1×···×U n ; The joint state x of the multi-UAV system is expressed as: x∈S, x:=(x1,...,x n ); The joint input u of the multi-UAV system is expressed as: u∈U, u:=(u1,...,u n ); The motion of the i-th UAV is expressed as: i∈[1,n] d in is the initial state of the ith UAV, f i :S i ×U i →S i is the associated vector field of the ith UAV, It is the joint initial state of the multi-UAV system.

4. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 2, characterized in that: The drone type set includes attack drones D atc and reconnaissance drone D rec ; The number of drones is represented by a drone set D = {D1, D2, D3, ..., D n }; The attribute values ​​of the drone include position (X i ,Y i ),speed The maximum number of tasks that can be executed L(D i ) and the flight drag coefficient k(D i ), where i represents the i-th drone, D i (i=1,2,3,…,n).

5. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 2, characterized in that: The mission type set includes strike missions and reconnaissance missions The number of tasks is represented as a task set of m tasks Task attribute values ​​include location Probability of completion Time required to be completed The hovering time required to be completed The required circling radius is completed Rewards for drone completion and the coefficient for converting the cost value unit required for the drone to complete the task into the reward received Among them, j represents the jth task, 6. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 1, characterized in that: The construction of the GCAA solution function includes calculating the cost of the drone completing the task, calculating the reward of the drone completing the task, and calculating the utility of the drone completing the task.

7. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 6, characterized in that: The cost of calculating the UAV's mission is expressed as: The i-th drone is Moment completion and status Related tasks j The cost: Assume a i =T j , Among them, T j ∈T, i∈[1,n] d ; make Indicates that T j The associated state; Indicates T j The relevant completion time, Obtaining the optimal control input Minimize the function: Make in, is a given C 1 Functions are respected; Furthermore, the minimum cost is:

8. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 6, characterized in that: The reward for the drone to complete the task is calculated as: Given an allocation profile a = (a1, ..., a n ), use T j -1 (a) represents the task T assigned under a specific profile j The index set corresponding to the drones of ∈T; Let p ij ∈[0,1] is the task T j The probability of being successfully completed by the i-th drone; At this time, the probability that at least one drone successfully completes the task increases with the number of drones assigned to the task. j The expected reward of is defined as: in, It is T j Nominal reward; is the probability that no drone completes the task, where the probability that at least one drone completes the task is equal to the complement of the probability that no drone completes the task, that is, 9. The method for allocating distributed collaborative tasks of unmanned aerial vehicles in an unmanned bee colony according to claim 6, characterized in that: The utility of calculating the UAV's task completion is expressed as: Set relevant physical quantities to describe the motion state of the drone, that is in, and p i ∈R 2 , They represent the acceleration, position and velocity of the i-th UAV at t0=0 in the two-dimensional vector space, i∈[1,n] d ; represents the drone state space belonging to the 2-dimensional vector space; Performance indicators are determined by control efforts given; set up represents the drone state space belonging to the 4-dimensional vector space; Then randomly select the terminal constraint function and take the difference between the two: Indicates that the i-th drone attempts to Terminal velocity Arrives at its assigned task T j Related Locations Indicates that the i-th drone attempts to Arrives at its assigned task T j The surrounding circle has a radius of Then wander around the target until At this point, the best entry point into the circle is selected by discretizing the circle into 10 points and selecting the point with the smallest cost function; Optimal control input: in Task utility is characterized by completing task T j ∈T and the state-dependent cost reflecting the cost of completing the task, given an allocation profile a = (a1, ..., a n ), conversion constant Used to complete the mission of the drone j The unit of measurement of the required cost is converted to the same as the reward, j∈[1,p] d , Task T j Total cost to complete: Given x0 and task T j Related total task utility: in, is a constant used to convert the cost into the same units as the reward; Denote the global utility as Based on the task assignment a, the individual utility of drone i is set to its contribution to the global utility U(a; x 0 )'s marginal contribution:

Images