CN111142553B

CN111142553B - Unmanned aerial vehicle cluster autonomous task allocation method based on biological predation energy model

Info

Publication number: CN111142553B
Application number: CN201911264700.0A
Authority: CN
Inventors: 段海滨; 张岱峰; 邓亦敏; 魏晨
Original assignee: Beihang University
Current assignee: Beihang University
Priority date: 2019-12-11
Filing date: 2019-12-11
Publication date: 2020-12-15
Anticipated expiration: 2039-12-11
Also published as: CN111142553A

Abstract

The invention discloses an unmanned aerial vehicle cluster autonomous task allocation method based on a biological predation energy model, which comprises the following implementation steps of: the method comprises the following steps: initializing cooperative task allocation; step two: leading-following global information interaction; step three: pre-allocating a hitting task based on the relation of total energy of biological predation; step four: collaborative fighting task allocation based on the biological predation energy balance model; step five: attack task conflict resolution based on the opportunistic hunting negotiation; step six: and outputting and executing the cooperative task distribution result. The method provides a distributed task allocation model supporting task load balance and ensuring task efficiency robustness, and can effectively support the unmanned aerial vehicle cluster to execute long-term and persistent search, attack and other resource consumption tasks in an unknown dynamic environment.

Description

Unmanned aerial vehicle cluster autonomous task allocation method based on biological predation energy model

Technical Field

The invention discloses an unmanned aerial vehicle cluster autonomous task allocation method based on a biological predation energy model, which is used for solving the task planning problem of unmanned aerial vehicle cluster search and multi-target attack under an unknown dynamic environment and belongs to the field of multi-unmanned system decision and control.

Background

Under the trend of the current unmanned aerial vehicle design pursuing miniaturization, light load and concealment, the possibility of realizing complex tasks by a single unmanned aerial vehicle is gradually reduced. Compared with the prior art, the unmanned aerial vehicle cluster utilizes the complementary performance advantage to complete the task in a cooperative manner, so that low-cost development can be maintained, the execution robustness of the system in the overall multi-task state is improved, and the unmanned aerial vehicle cluster has a good application prospect in the military and civil fields. However, how to allocate the subtasks of each unmanned aerial vehicle in the cluster under the drive of a specific task enables the subtasks to maximally improve the overall task performance of the cluster system on the premise of ensuring the constraints of individual movement, resource consumption and the like, and is a problem to be solved urgently in the field of intelligent decision of multiple unmanned systems at present. The balance of task load is used as a key index for maintaining task efficiency robustness of the system in an unknown environment, and is a problem that collaborative task planning and intelligent decision of the multi-unmanned system should be considered in a key mode.

Currently, the autonomous collaborative task allocation method for unmanned aerial vehicle cluster mainly focuses on centralized planning algorithms, including Mixed Integer Linear Programming (MILP), Markov Decision Process (MDP), Pigeon-Swarm Optimization (PIO), Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and other heuristic intelligent Optimization algorithms. However, the centralized planning method has a significant drawback, namely the dependency on the central planning node. When the communication demand is too high or the central node fails, the centralized planning algorithm cannot work normally, i.e. the task planning robustness is lacked. On the other hand, the distributed planning algorithm (such as an auction algorithm, a Bundle algorithm and the like) can overcome the defects of the traditional centralized algorithm, dependence on a central node is avoided through the autonomous decision of each unmanned aerial vehicle, and the consistency of the planning result and the task efficiency persistence and robustness in an unknown dynamic environment are difficult problems to be overcome.

Beast groups in nature, such as wolves, wild dogs, etc., will employ short-range, cooperative predation strategies in the process of predating prey. The energy of the population before and after predation can be balanced in absorption and consumption, namely the robustness of the energy of the predation population is ensured. Distributed decision making is adopted in the preying process of prey groups, and the balance characteristic embodied by the energy model of the prey groups is consistent with the task efficiency robustness requirement of unmanned aerial vehicle cluster autonomous task planning. Therefore, the unmanned aerial vehicle cluster distributed cooperative task allocation method based on the biological predation group energy model has important reference significance for effectively solving the autonomous cooperative task planning and intelligent decision problems of the unmanned aerial vehicle cluster system.

Disclosure of Invention

The invention discloses an unmanned aerial vehicle cluster autonomous task allocation method based on a biological predation energy model, and aims to provide a distributed task allocation model which supports task load balance and ensures task efficiency robustness in an unknown dynamic environment so as to solve key technical problems of dependence of a traditional centralized planning method on a central node, consistency of a traditional distributed planning method and the like.

Aiming at the problem of intelligent autonomous decision making of unmanned aerial vehicle clusters, the invention designs an unmanned aerial vehicle distributed cooperative task allocation method based on a biological predation energy model by combining an energy balance model in the predation process of beast groups, and the unmanned aerial vehicle distributed cooperative task allocation method has the advantages of no central dependence, consistency, task load balance and the like. The method comprises the following specific implementation steps:

the method comprises the following steps: collaborative task allocation initialization

The unmanned aerial vehicle cluster dynamic tasks in the unknown environment can be divided into reconnaissance (searching for a proposed target) and attack (releasing task load). Wherein, the reconnaissance task does not consume airborne resources, but instead strikes against task consumption resources.

S11, scout task initialization

For the scout mission, the following parameters and variables need to be initialized:

S_r(i)＝{V_r(i),θ_r(i),R_r(i),T(i),WP(i),E(i)}i∈N_u (1)

wherein i is the unmanned aerial vehicle number; n is a radical of_uThe unmanned aerial vehicle cluster scale; s_rIs a scout mission state set;

cruising airspeed for the unmanned aerial vehicle;

detecting the course of the unmanned plane;

reconnaissance of radii for the drone;

the target set searched by the unmanned aerial vehicle is obtained;

the coordinate of the three-dimensional position of the unmanned aerial vehicle in the reconnaissance task at the last time;

is the end point of the scout area.

S12, initializing the striking task

For the percussive task, the following parameters and variables need to be initialized:

wherein S is_aIs a set of percussion task states;

attacking a target airspeed for the drone;

attacking the course for the unmanned aerial vehicle;

striking the resource amount for the unmanned aerial vehicle;

an amount of resources expected to be consumed for the drone to strike the target;

attacking the radius for the drone;

waiting for an attack target set for the unmanned aerial vehicle;

expected simultaneous arrival times for the drone attack formation. The attack speed and direction of the unmanned aerial vehicle satisfy the following relations:

wherein p is_u(i)＝[p_ux(i),p_uy(i)]^TFor the ith unmanned plane horizontal position, p_ux(i) As its horizontal axis position coordinate, p_uy(i) Is the position coordinate of the longitudinal axis; p is a radical of_t(j)＝[p_tx(j),p_ty(j)]^TIs the jth target horizontal position, N_tFor the number of searched targets, p_tx(j) As its horizontal axis position coordinate, p_ty(j) Is the position coordinate of the longitudinal axis; and t is the current time.

S13, starting unknown environment target search

And initializing a task switching mode P e {1,2,3 }. Wherein, P1 represents in scout mission; p ═ 2 indicates that the percussion task is in progress; p-3 indicates that the task allocation phase is in progress. At the starting moment, each unmanned aerial vehicle starts target search, namely, the scout task P is entered to be 1. Meanwhile, initializing the following global communication planning state matrix as a zero matrix: leading unmanned aerial vehicle number matrix for sending team formation request

Global response pairing matrix

Wherein G is_res(i₁,i₂) 1 denotes the ith₁Frame following unmanned aerial vehicle response ith₂A team formation request for the guided drone is established; collaborative hit decision matrix

Wherein G is_not(i₁,i₂) 1 denotes the ith₁Frame leading unmanned aerial vehicle determination and ith₂The frame follows the unmanned aerial vehicle to carry out cooperative strike.

Step two: leader-follower global information interaction

S21, leading person broadcasting cooperation request

In the reconnaissance task state, if the ith unmanned aerial vehicle finds the jth target, i.e., | | p_u(i)-p_t(j)||＜＝R_r(i) Then the ith unmanned plane as the group leader to strike the jth target sends a cooperative strike request to the cluster, i.e. G_req(i) 1. At the same time, j is added to the sets T (i) and T, respectively_a(i) In that respect If N is found at the same time_m≤N_tThe targets are combined into a cooperative targetSet of targets

No. i unmanned aerial vehicle as set C_m(i) The group leader of (1). In addition, the task switching mode p (i) ═ 3 enters a task allocation stage; when multiple unmanned aerial vehicles find a target at the same time, the unmanned aerial vehicle closest to the target serves as a group leader of the unmanned aerial vehicles.

S22, responding the cooperation request by the follower

For any unmanned aerial vehicle i 'in the reconnaissance mission state, after receiving the cooperative attack request sent by the leader i, if the resource quantity of the unmanned aerial vehicle i' meets r_u(i′)＞_r(_rThe minimum resource amount reserved for each unmanned aerial vehicle is more than or equal to 0), response information G is fed back to the leader i_res(i', i) ═ 1, which becomes the follower of the population. The task switching mode P (i') is 3, the task distribution stage is entered, and the distribution result G is waited_not(i)。

S23, the leader enters the pre-allocation stage of the hit task

T after leader i sends cooperative hit request_wAlways receiving follower response information in one communication period, t_wAnd (5) after each communication period, turning to step three. The communication period is affected by the communication topology structure, and the communication topology needs to ensure that the broadcast information can be received by all unmanned aerial vehicles in a certain communication period.

Step three: hit task pre-allocation based on biological predation total energy relation

S31, generating a leading-following percussion formation total set

Is provided with

Set of followers for leading unmanned aerial vehicle i receives, where N_r＝|C_res(i) If the number of followers is |, leading the unmanned aerial vehicle i to generate a set { i, C according to the combination relation_res(i) Power set of } power set

Wherein

Represents a power set C_T(i) Original set of { i, C contained_res(i) And the number of subsets thereof, then C_T(i) Constituting a total set of percussion formations.

S32, pre-distributing the formation subset according to the total energy relation of the biological predation

For the total set of percussion formation C_T(i) The k-th subset of

Calculating the total energy of the mapped biological predation of the hit resources

If it satisfies

The subset corresponds to a total energy relationship for predation, where r_t(j) The amount of percussion resources that need to be consumed for the jth target. After the total energy of the biological predation of each subset is polled in sequence, the subsets meeting the total energy relation form a task pre-allocation set C of the ith unmanned aerial vehicle_U(i) In that respect If it is

The ith drone returns to the scout mission, i.e., p (i) ═ 1.

Step four: collaborative fighting task allocation based on biological predation energy balance model

S41, calculating expected consumption resource amount

Pre-allocating collections C for tasks_U(i) For each subset of drones it contains, the expected amount of resources consumed for each drone is calculated as:

wherein the content of the first and second substances,

to representC_U(i) The nth subset of (1); p is epsilon to C_nDenotes the pth drone, which is subset C_nAnd (5) medium element.

Represents a subset C_nThe unmanned aerial vehicle set meeting the condition of balanced energy for predation of living beings is provided

S42, calculating the pre-distribution subset biological predation energy profit value

Depending on the desired amount of resources consumed, C can be obtained by_U(i) The bio-predation energy yield value of each subset. For the unmanned aerial vehicles which participate in resource consumption, the residual resource amount is the same, and the unmanned aerial vehicles which do not participate in the resource consumption due to the lack of the resources reserve the attack resources, so that the balance of task load is ensured.

Wherein q (p) is the expected remaining resource amount of the pth drone; s_in(C_n) Represents a subset C_nA bio-predation energy yield value of; q_inThe value range of the energy gain balance coefficient is satisfied

S43, calculating a pre-distribution subset biological predation energy balance model

The subset C is calculated by_nThe biological predation energy cost value of (1):

wherein S is_out(C_n) Represents a subset C_nThe biological predation energy generation value of; l (p) is the p-thUnmanned aerial vehicle arrival target set C_m(i) 1 st required voyage for attacking the target; j is a function of₁Is C_m(i) 1 st attack target number in (1). All unmanned aerial vehicles adopt a continuous attack mode for C pair due to reconnaissance target attack_m(i) The ranges required for hitting the subsequent targets are the same, and the influence on the energy cost is neglected. h is_u(p) is the current height of the pth unmanned aerial vehicle; h is_aAttack altitude for the drone; q_outThe value range of the energy cost balance coefficient is influenced by the detection distance and the area of the hitting area.

Subset C_nThe predation energy balance model of (a) can be expressed as E_S(C_n)＝S_in(C_n)-S_out(C_n)∈[-1,1]。

S44, selecting an optimal pre-distribution subset

Task pre-allocation set C for ith unmanned aerial vehicle_U(i) Calculating a biological predation energy balance model of each subset, and selecting the subset with the highest energy balance value as the ith unmanned aerial vehicle aiming at the target sequence C_m(i) And is noted as the optimal pre-allocated subset

Step five: attack task conflict resolution based on opportunistic hunting negotiation

S51, leading the unmanned aerial vehicle to share the optimal pre-distribution subset

Each lead drone will optimally pre-allocate subsets by broadcast

Number of task pre-allocation set subsets | C_UAnd the optimal energy balance value

And sending the information to other leading unmanned aerial vehicles, and receiving broadcast information of other leading unmanned aerial vehicles in the neighborhood. For the ith unmanned aerial vehicle (leading unmanned aerial vehicle), constructing a neighborhood leading unmanned aerial vehicle set

Where Ω is the set of unmanned aerial vehicles that are guided by the cluster at the current time. Therefore, for the ith unmanned aerial vehicle, the neighboring leading unmanned aerial vehicle is judged

To the optimal pre-allocation subset

Whether or not to satisfy

If yes, updating

And continues to perform step S51 until

Has already traversed; otherwise, go to step S52. If it is

Having traversed, go to step S53.

S52, determining cooperative strike decision according to the opportunity hunting negotiation mechanism

Judging to lead the unmanned aerial vehicle i and according to the opportunistic hunting negotiation mechanism

Whether or not to satisfy

If yes, updating

And returns to step S51; otherwise, it orders

Update C_res(i)＝C_res(i) - (i) and returns to step S31.

S53, determining the arrival time of the cooperative strike formation at the same time

For the

Calculating formation simultaneous arrival times

And step six, wherein eta is more than 1, namely the pursuit velocity gain. The higher η, the faster the pursuit speed, but the maximum airspeed constraint must be met.

Step six: outputting and executing cooperative task allocation results

S61, outputting a collaborative striking decision matrix

For the ith drone (leading drone),

let G_not(i, p) ═ 1, i.e., set of drones

A cooperative percussion formation is formed. Outputting a collaborative hit decision matrix G_notAnd broadcasts a notification to follow drone p. If WP (p) is not null, the three-dimensional position coordinates WP (p) of the reconnaissance mission of the unmanned aerial vehicle are respectively updated_u(p); the task switching mode is changed to a percussive task, i.e., p (p) ═ 2.

S62, executing the cooperative striking task

Cooperative strike formation according to V_a、θ_aAnd h_aFor target set C_m(i) In which each target performs a cooperative strike in turn. If the relative distances between the attack formation and the same target are smaller than the attack radius R_aThen the target is completed in coordination with the strike, T_a(i) Set the object removed and

when the target set C_m(i) After all targets are hit, the task switching mode is switched to the reconnaissance task, i.e. p (i) is equal to 1 and p (p) is equal to 1. Leading the unmanned aerial vehicle and the formation unmanned aerial vehicle to return respective waypoints WP; and when the reconnaissance area terminal E is reached, finishing the cluster task.

The invention provides an unmanned aerial vehicle cluster autonomous task allocation method based on a biological predation energy model, and provides a distributed task allocation model supporting task load balance and ensuring task efficiency robustness. The invention adopts a distributed planning architecture, and avoids the dependency of the traditional centralized planning method on the central planning node. Meanwhile, by using a biological predation energy balance model presented by beast groups in the predation process, the bottlenecks of consistency, task efficiency robustness and the like of the traditional distributed planning method are overcome, and the unmanned aerial vehicle cluster can be effectively supported to execute resource consumption tasks such as long-term and persistent search, attack and the like in an unknown dynamic environment.

Drawings

FIG. 1 is a flow chart of the steps performed by the present invention

Figure 2 simulation test 5 unmanned aerial vehicle two-dimensional plane track chart

Figure 3 height variation curve of 5 unmanned aerial vehicles in simulation test

Fig. 4 shows a variation curve of the remaining resource amount of 5 unmanned aerial vehicles in the simulation test

The reference numbers and symbols in the figures are as follows:

C_Utask pre-allocation collections

C_resFollower set determined by piloting unmanned aerial vehicle according to response pairing matrix

Intersection of current leading unmanned aerial vehicle and neighborhood leading unmanned aerial vehicle optimal pre-allocation subset

Target order index of j-coordinated hit target set

Three-dimensional position coordinate of unmanned aerial vehicle when WP last time is in reconnaissance task

Detailed Description

The effectiveness of the cooperative task allocation method provided by the invention is verified through specific examples. In this example, N is given_uThe 5 drones participate in the target search and hit task for the 1km × 1km unknown area. The starting positions of 5 unmanned aerial vehicles are respectively set as S₁＝[0,100,50]^T，S₂＝[0,300,50]^T，S₃＝[0,500,50]^T，S₄＝[0,700,60]^T，S₅＝[0,900,60]^T(ii) a The reconnaissance end points are respectively set as E₁＝[1000,100,50]^T，E₂＝[1000,300,50]^T，E₃＝[1000,500,50]^T，E₄＝[1000,700,60]^T，E₅＝[1000,900,60]^TIn the unit m. The unmanned aerial vehicle motion model adopts the following second-order linear system:

wherein v is_uThe flying airspeed of the unmanned aerial vehicle is in the unit of m/s; theta_uIs the current course of the unmanned plane, unit rad; [ u ] of_v,u_θ,u_h]^TThe vector is controlled by the unmanned aerial vehicle, the forward acceleration, the course angular velocity and the vertical acceleration of the unmanned aerial vehicle are respectively controlled, and the value ranges of the forward acceleration, the course angular velocity and the vertical acceleration are respectively subjected to the maximum forward overload a_maxMaximum yaw angular velocity ω_maxAnd maximum vertical overload

And (4) limiting. Initial time v_u、θ_uAnd the control quantity is zero, and the maximum overload is respectively set as a_max＝5m/s²，

Common setting N in scout area_tThe 10 unknown targets meet the uniform distribution, and the striking resource amount of the unmanned aerial vehicle which needs to be consumed by each target in sequence is r_t＝[10,10,10,10,10,20,20,20,30,30]^THeight of attack h_a10 m. The simulation environment of the example is configured as an intel i7-4790 processor, a 3.60Ghz dominant frequency, a 4G memory, a software MATLAB 2010a version, and a simulation step size of 0.1 s.

The specific practical procedure of the examples is as follows:

S11, scout task initialization

For the ith unmanned aerial vehicle, i belongs to { 1.,. 5}, and the reconnaissance task changesQuantity S_r(i) Initialized to V_r(i)＝5m/s；

E(i)＝E_i；R_r(i)＝100m；

S12, initializing the striking task

For the ith unmanned aerial vehicle, i belongs to { 1., 5}, and a striking task variable S_a(i) Initialized to V_a(i)＝θ_a(i) When the cooperative strike formation and the simultaneous arrival time are determined, the amplitude is carried out according to the formula (3); r is_u(i)＝50；

R_a(i)＝20m；

S13, starting unknown environment target search

For the ith unmanned aerial vehicle, i belongs to { 1., 5}, initializing a task switching modality p (i) ═ 1; the global communication planning state matrix is a zero matrix: g_req＝0^5×1，G_res＝0^5×5，G_not＝0^5×5。

Step two: leader-follower global information interaction

S21, leading person broadcasting cooperation request

Under the reconnaissance task state, if the ith unmanned aerial vehicle discovers N_mEach object is then grouped into an object set C_m(i) As set C_m(i) Group leader of (1), set G_req(i)＝1，

P (i) ═ 3, and broadcasts G_req。

S22, responding the cooperation request by the follower

Given a_r0, in scout mission state, when receiving G_req(i) Then, any non-leading unmanned aerial vehicle i' judges whether r is satisfied_u(i′)＞_r. If yes, feeding back response information G_res(i', i) ═ 1, becoming the follower; setting P (i') to 3 at the same time, and waiting for an allocation result G_not(i)。

S23, the leader enters the pre-allocation stage of the hit task

Given t_wIf the communication period is 1 simulation step, the leader i is transmitting G_req(i) And the follower response information is always received in the last 0.5s, and then the step three is carried out.

Step three: hit task pre-allocation based on total energy relationship

S31, generating a leading-following percussion formation total set

If lead unmanned aerial vehicle i to receive N_rThe follower is numbered to form a set C_res(i) In that respect Generating a set { i, C from the combinatorial relationship_res(i) Power set of } power set

In total

Seed percussion formation subset, then C_T(i) And forming a total set of percussion formation.

S32, pre-distributing the queuing subset according to the total energy relation

For C_T(i) Calculates its total hit resource energy SR. If it satisfies

Then the subset is included in task pre-allocation set C_U(i) In that respect If it is

Then p (i) ═ 1, i.e. the ith drone returns to the scout mission.

S41, calculating expected consumption resource amount

For C_U(i) Calculates the amount of hitting resources expected to be consumed by each drone it contains according to equation (4)

Given Q_inC was calculated according to equation (6) as 20_U(i) The bio-predation energy yield value S of each subset of_in。

Given Q_outC was calculated according to equation (7) 400_U(i) Biological predation energy cost value S of each subset of_outAnd further calculating a biological predation energy balance model E for each subset_S。

S44, selecting an optimal pre-distribution subset

Selecting an energy balance value E_SThe highest subset is the ith drone for target sequence C_m(i) To the optimal pre-allocation subset

Lead unmanned aerial vehicle broadcast

|C_UI and corresponding

Receiving broadcast information of other leading unmanned aerial vehicles in the neighborhood, and constructing a neighborhood leading unmanned aerial vehicle set

Go through

And judging whether the optimal pre-allocation subset of the neighborhood leading unmanned aerial vehicle contained in the self-adaptive cluster is empty with the local intersection. If yes, go to step S53; otherwise, go to step S52.

Judge the local i and neighborhood leader

Whether or not to satisfy

If yes, updating

And returns to step S51; otherwise, it orders

Update C_res(i)＝C_res(i) - (i) and returns to step S31.

Given η 2, for

Calculating the formation simultaneous arrival time of each unmanned aerial vehicle

And go to step six.

Step six: outputting and executing cooperative task allocation results

S61, outputting a collaborative striking decision matrix

For the

In any unmanned plane p, set G_not(i, p) ═ 1. Transfusion systemOutput and broadcast cooperative attack decision matrix G_not. If WP (p) is not empty, update WP (p) p_u(p); the task switching mode is changed to a percussive task, i.e., p (p) ═ 2.

S62, executing the cooperative striking task

Cooperative strike formation according to V_a、θ_aAnd h_aFor target set C_m(i) In which each target performs a cooperative strike in turn. If the relative distance between the percussion formation and the same target is less than R_aThen the target is completed in coordination with the strike, T_a(i) The set removes the object(s) and,

when the target set C_m(i) After all targets are hit, the task switching mode is switched to the reconnaissance task, i.e. p (i) is equal to 1 and p (p) is equal to 1. Leading the unmanned aerial vehicles and the formation unmanned aerial vehicles to return to a navigation point WP; and when the reconnaissance area terminal E is reached, finishing the cluster task.

Fig. 2 and 3 show two-dimensional flight paths and height variation curves of each unmanned aerial vehicle in the example, and as can be known from example initialization, each unmanned aerial vehicle executes reconnaissance tasks along the direction of the transverse axis. According to the reconnaissance radius, the unmanned aerial vehicles can mutually cover a search area, and targets are guaranteed not to be omitted. As can be seen from the target distribution in fig. 2, most of the targets are distributed in the upper half plane in this example. Thus, the drones 4, 5 will detect a plurality of targets. Due to individual resource amount restrictions, the drones 4, 5 need to perform cooperative strikes with drones whose remaining resource amounts are redundant. Example operation results show that the drone 1 detects the target 5, the drones 2,3 do not detect the target, and the drones 4, 5 detect the

targets

2, 4, 8, 9, 10 and the targets 1, 3, 6, 7, respectively. To achieve resource balance, drone 4 allocates percussion formation drones 1, 3 and 2 to

targets

4, 10 and 8 at 49.6s, 188s and 349.1s, respectively; drone 5 assigns percussion formation drones 2, 1 and 4 to targets 1, 6 and 3 at 128.5s, 303.9s and 416.9s, respectively. Fig. 4 shows a variation curve of the remaining resource amount of 5 drones, which can be obtained by combining fig. 2, and the allocation result achieves the purposes of no task conflict, resource amount equalization and path consumption reduction. The results of FIG. 3 show that under the constraint of simultaneous arrival time, the attack formation can arrive at the attack site at a specified height in a limited time window, and the attack task can be completed smoothly. Because the residual resource amount of each machine is similar, when a subsequent target is found, if multi-machine cooperative attack constraint exists, the task allocation method provided by the invention is still applicable. Therefore, the method has the advantages of ensuring that the unmanned aerial vehicle cluster executes distributed, long-term and persistent tasks in an unknown dynamic environment and improving the robustness of task efficiency.

Claims

1. An unmanned aerial vehicle cluster autonomous task allocation method based on a biological predation energy model is characterized in that: the method comprises the following steps:

the method comprises the following steps: the collaborative task allocation initialization specifically includes:

initializing a scout task;

initializing a striking task;

starting unknown environment target search, namely initializing a task switching mode into a reconnaissance task;

step two: leading-following global information interaction specifically comprises:

the leader broadcasts a collaboration request;

the follower responds to the cooperative request;

leading the leader to enter a striking task pre-allocation stage;

step three: strike task pre-distribution based on biological predation total energy relation specifically comprises the following steps:

generating a leader-follower strike formation total set;

pre-distributing the formation subset according to the total energy relation of the biological predation;

step four: the cooperative attack task allocation based on the biological predation energy balance model specifically comprises the following steps:

s41, calculating expected consumption resource amount;

s42, calculating a pre-distribution subset biological predation energy profit value;

s43, calculating a pre-distribution subset biological predation energy balance model;

s44, selecting an optimal pre-distribution subset;

step five: attack task conflict resolution based on opportunistic hunting negotiation specifically comprises the following steps:

s51, leading the unmanned aerial vehicle to share the optimal pre-allocation subset;

s52, determining a cooperative attack decision according to the opportunistic hunting negotiation mechanism;

s53, determining the arrival time of the cooperative strike formation;

step six: outputting and executing a cooperative task allocation result, specifically comprising:

outputting a collaborative strike decision matrix;

executing a cooperative hit task;

the specific process of step S41 is as follows:

wherein the content of the first and second substances,

is represented by C_U(i) The nth subset of (1); p is epsilon to C_nDenotes the pth drone, which is subset C_nMiddle element;

represents a subset C_nUnmanned aerial vehicle set meeting the condition of balanced biological predation energy; i represents a leading drone number; r is_u(h)、r_u(p) represents the current hit resource amounts of drones h and p, respectively; r is_t(j) Representing the amount of percussion resources that the jth target needs to consume; c_m(i) Acquiring a cooperative target set for leading the unmanned aerial vehicle i; namely have

The specific process of step S42 is as follows:

depending on the desired amount of resources consumed, C can be obtained by_U(i) A biological predation energy yield value for each subset of; for the unmanned aerial vehicles which participate in resource consumption, the residual resource amount is the same, and for the unmanned aerial vehicles which do not participate in the resource consumption due to resource shortage, the attack resources are reserved, and the balance of task load is ensured;

The specific process of step S43 is as follows:

L(p)＝||p_u(p)-p_t(j₁)||+||h_u(p)-h_a||

wherein S is_out(C_n) Represents a subset C_nThe biological predation energy generation value of; l (p) is the target set C reached by the p-th unmanned aerial vehicle_m(i) 1 st required voyage for attacking the target; j is a function of₁Is C_m(i) 1 st attack target number in (1); since the reconnaissance target strikes in a continuous attack mode, all the targets do notMan-machine pair C_m(i) The ranges required by the striking of the middle and subsequent targets are the same, and the influence of the ranges on the energy cost is neglected; h is_u(p) is the current height of the pth unmanned aerial vehicle; h is_aAttack altitude for the drone; q_outThe value range of the energy cost balance coefficient is influenced by the reconnaissance distance and the area of the hitting area; then the subset C_nHas a prey energy balance value of E_S(C_n)＝S_in(C_n)-S_out(C_n)∈[-1,1]；p_u(p) represents the horizontal position of drone p; p is a radical of_t(j₁) Represents the target j₁(ii) a horizontal position of;

the specific process of step S52 is as follows: