Multi-target unmanned swarm task coordination method, device and equipment

Abstract

The application discloses a multi-target unmanned bee swarm task coordination method, device and equipment, and relates to the technical field of unmanned aerial vehicles. The method comprises the following steps: acquiring distribution information of a target area and capability expression of each unmanned aerial vehicle in a current unmanned aerial vehicle bee swarm queue, wherein the capability expression comprises standard operation unit execution cost, a current state, capability reservation, cost constraint and basic parameters; acquiring available unmanned aerial vehicles from the current unmanned aerial vehicle bee swarm queue according to the capability expression for the distribution information; constructing a task cost matrix of the available unmanned aerial vehicles and target area operation based on the distribution information and the capability expression, and allocating task resources of the available unmanned aerial vehicles according to the task cost matrix; and planning a task positioning path and a task recovery path for each unmanned aerial vehicle with allocated task resources. The application realizes optimal solution of task planning through multi-machine overall planning.

Description

Technical Field

[0001] This invention belongs to the field of big data and artificial intelligence, and specifically relates to a task coordination method, device and equipment for multi-target unmanned bee swarms. Background Technology

[0002] Cities consist of numerous tall buildings, which suffer varying degrees of damage during disasters such as earthquakes, floods, and mudslides, trapping or burying large numbers of people inside. Because buildings provide some degree of concealment, it is impossible to quickly identify those buried by visual means. Therefore, rapidly determining the location of burial sites and the condition of those buried during urban disasters has become one of the major challenges in the field of emergency rescue.

[0003] Compared to other detection equipment, drones have significant advantages in terms of mobility and flexibility, making them an effective means of quickly identifying burial sites in disaster areas. Currently, drones are widely used in emergency response; however, due to limitations such as timeliness and management systems, their application is mostly limited to single-point patrol flights. Aerial imagery data is then analyzed to assess the disaster area. This method is less efficient when the disaster area is large, mainly due to the following reasons:

[0004] 1. Existing drones are mainly used for large-scale image data collection, fly at high altitudes, and lack the application capability to accurately ascertain the current situation on the ground;

[0005] 2. Existing drones primarily employ ground control methods, with flight operations initiated by the mission system and flight routes manually controlled by operators. In disaster areas, numerous targets are scattered across the region, resulting in low work efficiency due to factors such as human-machine collaboration and communication limitations during mission execution.

[0006] 3. Lack of coordination between different drones prevents effective coverage of the area during mission organization.

[0007] The aforementioned issues have a certain impact on the in-depth application of drones. From a developmental perspective, there is an urgent need for multi-drone collaborative and autonomous intelligent methods to organize aerial search operations in order to meet the requirements of future urban disaster relief. Summary of the Invention

[0008] To address the aforementioned problems, this invention discloses a multi-target unmanned swarm task coordination method, apparatus, and equipment. The method provides a hierarchical planning architecture. In this architecture, a UAV resource capability model is established, using UAV operational (time / energy consumption) resources as the basis for planning and allocation; task planning, execution command, and scheduling require a reasonable balance between cost constraints and resource reserves. Based on this, the optimal solution for task planning is achieved through multi-UAV coordination. This technology also has positive application value in defense, smart logistics, and smart cities.

[0009] The technical content of this invention includes:

[0010] A multi-target unmanned bee swarm task coordination method, the method comprising:

[0011] Obtain the distribution information of the target area and the capability representation of each UAV in the current UAV swarm queue, wherein the capability representation includes: standard operation unit execution cost, current status, capability reservation, cost constraints and basic parameters;

[0012] Based on the distribution information, available drones are obtained from the current drone swarm queue according to the capability representation;

[0013] Based on the distribution information and the capability representation, a task cost matrix for the available UAVs and the target area operations is constructed, and task resources for the available UAVs are allocated according to the task cost matrix.

[0014] For each drone allocated task resources, plan the task placement path and task recovery path.

[0015] Optionally, the step of obtaining available drones from the current drone swarm queue based on the distribution information and the capability representation includes:

[0016] Select drones that are currently available in the drone swarm queue;

[0017] For a drone that is in an available state, the positional relationship between the drone and the target is calculated based on the current state and the distribution information;

[0018] Based on the location relationship and the cost constraint, calculate the cost of different tasks for the current UAV;

[0019] If the cost of any task to the current drone is less than the capacity reservation, then the drone is considered a usable drone.

[0020] Optionally, constructing the task cost matrix for the available UAVs and the target area operations based on the distribution information and the capability representation includes:

[0021] Based on the distribution information and the capability representation, the in-situ cost, mission execution cost assessment, and recovery cost between each available UAV and each target area are calculated to obtain the corresponding cost T. t ;

[0022] Set a cost tolerance value for the path consumption between each available drone and each target area operation;

[0023] For any available drone and its corresponding target area, compare the capability reservation with the cost tolerance value, and

[0024] When the difference between the capacity reserve and the cost tolerance value is greater than the cost T t At that time, the cost between the available drone and the corresponding target area operation is set as the cost T. t ;

[0025] When the difference between the capacity reserve and the cost tolerance value is not greater than the cost T t At that time, the cost between the available drone and the corresponding target area is set to infinite;

[0026] The cost between the available drone and the corresponding task is used as an element in the task cost matrix.

[0027] Optionally, allocating the available UAV mission resources according to the mission cost matrix includes:

[0028] Based on the premise that drones cannot complete all operations in the target area and that no drones can complete the operations in the target area, the task cost matrix is ​​trimmed.

[0029] The pruned task cost matrix is ​​processed using either exhaustive traversal or ant colony algorithm to obtain several solution queue plans. p Where p represents the row number of the trimmed task cost matrix, and H is the row vector in the trimmed task cost matrix. p This represents the total cost vector for the p-th target area operation;

[0030] Statistical scheme queue plan p The sum of the target area range S p And when S p When the value exceeds the threshold setting, the plan queue will be... p This is considered a valid result;

[0031] Based on the cost of each valid result, obtain the optimal solution queue plan;

[0032] Based on the optimal queue plan, the available UAV mission resources are allocated.

[0033] Optionally, the trimmed task cost matrix is ​​processed using an exhaustive traversal method or an ant colony algorithm to obtain several solution queues (plans). p ,include:

[0034] Constructing a queue plan p And it is initially empty;

[0035] From row vector H p Extract the element C with the lowest cost. pk , the element C pk Placed in the plan queue p And delete row vector H p With the corresponding column vector V k We obtain matrix MC0, where k represents the column number of the trimmed task cost matrix;

[0036] From matrix MC t row vector H e Extract the element C with the lowest cost. ek , the element C ek Placed in the plan queue p And delete row vector H e With the corresponding column vector V k , thus obtaining matrix MC e+1 Where t represents the extracted element C ek The number of times, e represents the matrix MC t Line numbers in the text;

[0037] After processing the trimmed task cost matrix, the solution queue plan is obtained. p .

[0038] Optionally, allocating the available UAV mission resources according to the optimal scheme queue plan includes:

[0039] Compare the number of available drones with the dimension of the optimal solution queue plan;

[0040] If the number of available drones is not greater than the dimension of the optimal solution queue plan, then the task resources of the available drones are allocated according to the optimal solution queue plan;

[0041] If the number of available drones is not greater than the dimension of the optimal solution queue plan, then the optimal solution queue plan is optimized, wherein the optimization includes:

[0042] Set a segmentation threshold limit;

[0043] Based on the job cost, the unassigned target area jobs are processed sequentially to obtain the queue reJob;

[0044] Determine whether the job cost of each target region job in the reJob queue is less than the segmentation threshold limit:

[0045] If so, the optimal solution queue plan will be used as the optimization result for task resource allocation;

[0046] If not, then:

[0047] By segmenting unassigned target area operations whose operation costs exceed the segmentation threshold, and returning to the distribution information, available drones are obtained from the current drone swarm queue based on the capability expression to obtain the optimal solution queue plan′;

[0048] By comparing the task coverage and cost of the optimal solution queue plan with that of the optimal solution queue plan', an optimized solution queue plan is obtained. 1 ;

[0049] By optimizing the queue plan t The corresponding queue reJob t If the cost of each target area operation is less than the segmentation threshold, determine whether to continue optimization, where t is the number of optimization attempts.

[0050] The optimal solution queue plan T The optimization results are used for task resource allocation, where T is the total number of optimizations.

[0051] Optionally, the planning of the task placement path for each UAV allocated task resource includes:

[0052] For the entire region, a situation map in bitmap form is constructed, wherein the bitmap is composed of several basic units;

[0053] Set the mission positioning path cost constraint for the UAV and the grayscale value threshold for the basic unit;

[0054] Based on the basic units corresponding to the UAV's starting position and target position, a vector is constructed.

[0055] Extract the vector All the basic units traversed are arranged into a sequence based on vector relationships and placed in the in-situ route result R;

[0056] Compare the cost of the positioning path result R with the cost constraint of the task positioning path:

[0057] If the cost of the positioning route result R is not less than the task positioning path cost constraint, then jump to output the UAV planning task positioning path;

[0058] If the cost of the in-place route result R is not less than the task in-place path cost constraint, then:

[0059] Extract the basic unit with the largest gray value from the in-place line result R;

[0060] The gray value of the basic unit with the largest gray value is compared with the gray value threshold:

[0061] If the gray value of the basic unit with the largest gray value is less than the gray value threshold, then proceed to determine whether all basic units in the positioning line result R have been processed.

[0062] If the gray value of the basic unit with the largest gray value is greater than the gray value threshold, and there is at least one basic unit with a nearby boundary, then a basic unit with a nearby boundary is selected as the basic unit with the largest gray value, and the process returns to comparing the gray value of the basic unit with the largest gray value with the gray value threshold.

[0063] If the gray value of the basic unit with the largest gray value is greater than the gray value threshold, and there are no basic units adjacent to the boundary, then the basic unit with the smallest gray value is selected to replace it and placed in the in-place line result R, and the process jumps to determine whether all basic units in the in-place line result R have been processed.

[0064] Determine whether all basic units in the in-place line result R have been processed:

[0065] If not, return to the comparison between the cost of the positioning route result R and the cost constraint of the task positioning path;

[0066] If so, then proceed to output the drone's planned mission positioning path;

[0067] Output the planned landing path for the UAV mission.

[0068] Optionally, the planning of the task recovery path for each allocated task resource UAV includes:

[0069] Optionally, the planning of the task recovery path for each allocated task resource UAV includes:

[0070] Obtain the current status and location of the drone;

[0071] Obtain the task positioning cost of the UAV executing the task positioning path, and compare the task positioning cost with a set threshold S;

[0072] If the task placement cost is greater than the set threshold S, select and determine whether the nearest retrieval point meets the requirements; if the requirements are met, plan the task retrieval path between the current position and the nearest retrieval point; if the requirements are not met, plan the task retrieval path between the current position and the original retrieval point.

[0073] If the task placement cost is not greater than the set threshold S, a task retrieval path is planned between the current location and the original retrieval point.

[0074] A multi-target unmanned swarm mission coordination device, the device comprising:

[0075] The information acquisition module is used to acquire the distribution information of the target area and the capability expression of each UAV in the current UAV swarm queue. The capability expression includes: standard operation unit execution cost, current status, capability reservation, cost constraints and basic parameters.

[0076] The drone filtering module is used to select available drones from the current drone swarm queue based on the distribution information and the capability expression.

[0077] The resource allocation module is used to construct a task cost matrix for the available UAVs and the target area operations based on the distribution information and the capability expression, and to allocate task resources for the available UAVs according to the task cost matrix.

[0078] The path planning module is used to plan the mission placement path and mission recovery path for each drone with allocated mission resources.

[0079] A computer device includes a memory and a processor, wherein the memory stores a computer program, which is loaded and executed by the processor to implement the above-described multi-target unmanned bee swarm task coordination method.

[0080] Compared with the prior art, the present invention has the following advantages and effects.

[0081] This invention constructs a closed-loop mission command and resource scheduling framework based on UAV swarms, achieving full-area, full-cycle coverage for urban disaster areas; it enables automatic mission adaptation of UAVs (swarms) based on target capability adaptability; and it forms a complete mission guidance scheme combined with the actual environment of the disaster area. It improves search efficiency through multi-level, multi-batch, and multi-entity collaborative methods. Attached Figure Description

[0082] Figure 1 Overall organizational framework diagram of the task.

[0083] Figure 2 Task phase division diagram.

[0084] Figure 3 Unmanned swarm mission objective allocation diagram.

[0085] Figure 4 Target allocation process flowchart.

[0086] Figure 5 Diagram illustrating the positioning guidance for a drone (bee) mission.

[0087] Figure 6 Drone (Bee) mission positioning route planning map.

[0088] Figure 7 Unmanned aerial vehicle (UAV) mission recovery path planning map. Detailed Implementation

[0089] The technical solutions in the embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely specific embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0090] The multi-target unmanned swarm task coordination method of the present invention is applicable to various application scenarios such as urban disaster relief, national defense and military industry, smart logistics, and smart cities. The technical content of the present invention will be described below using urban disaster relief as an example.

[0091] This invention focuses on key technologies for unmanned swarm mission planning and collaboration, specifically addressing the characteristics of buildings in urban disaster relief. It integrates emergency rescue expertise to construct a mission scheduling framework, enabling unified command and dispatch of multi-robot collaborative missions. A unified swarm capability expression model is built to achieve unified measurement of swarm capability status, supporting rapid synthesis of search forces. Combining artificial intelligence and other methods, a mission requirement analysis and decision evaluation model is established. Through discretization algorithms for multi-target reconnaissance missions in complex spatial environments and resource matching optimization algorithms, optimal mission decision-making solutions are achieved based on multi-point collaboration. Finally, a path planning algorithm based on complex indoor and outdoor environments plans the mission paths of individual drones (swarms) under cost constraints, considering environmental and risk factors, thus meeting the requirements of multi-robot collaboration.

[0092] Following destructive disasters such as earthquakes, floods, and mudslides, on-site reconnaissance in disaster areas is a continuous operational process. It mainly comprises stages including task mobilization, task planning, task instruction communication, task deployment, operation implementation, and task recovery. From the perspective of the entire task system, this process includes three basic levels, such as... Figure 1 As shown, the entire technical system consists of three basic layers: the planning layer, the task layer, and the operation layer.

[0093] Planning Layer: Based on basic equipment information, and leveraging the disaster area's on-site communication network, the system gathers unmanned aerial vehicle (UAV) swarm mission status information to create a status and capability profile. Simultaneously, based on the disaster area's UAV search and detection system, it aggregates various summary search data and third-party intelligence data to form a general understanding of the disaster area's situation. On this basis, according to command instructions, targets in the disaster area environment are identified, and mission planning is conducted. During the planning process, target allocation and swarm resource matching are performed based on target distribution, swarm capabilities, and status, forming a combined UAV search force plan. Combining artificial intelligence algorithms, and under cost constraints, the optimal solution for force combination is achieved through overall planning, meeting the effectiveness and efficiency requirements of disaster area search missions.

[0094] Task Layer: During the search task planning process, path planning is performed for the participating UAVs (drones). In this process, with task time cost as the overall constraint, paths in the task positioning phase are enumerated, and the optimal path is selected based on task requirements, providing a basis for UAV task execution scheduling. Simultaneously, during the task recovery phase, the UAV withdrawal route is rapidly planned based on constraints such as communication and aircraft status, meeting the requirements of UAV (drone) task command and dispatch.

[0095] Operational Layer: After arriving at the designated operational area according to the mission plan, the UAV conducts reconnaissance operations. During this process, the UAV, based on the operational environment, uses a combination of autonomous, human-machine collaborative, and multi-UAV cooperation methods to complete data collection of the building space and execute operational recovery according to the mission plan.

[0096] The following points should be followed in the planning:

[0097] 1. Based on the given task objectives, satisfy the constraint of the number of tasks to be completed in one wave;

[0098] 2. Based on satisfying condition 1, the resulting solution has the lowest overall cost;

[0099] 3. Based on satisfying 1 and 2, the resulting drone (bee) swarm has the lowest overall risk.

[0100] Collaborative reconnaissance missions within built-up spaces involve multiple layers. Due to the complexity of the internal and surrounding environments of buildings, unmanned aerial vehicles (UAVs) need to possess strong scene adaptability during mission execution. Mission planning, execution command, and scheduling require a reasonable balance between cost constraints and resource reserves. Based on this, optimal solution for mission planning is achieved through multi-UAV coordination. Addressing this technical requirement, focusing on the planning and mission layers, and based on multi-target resource optimization and matching, this invention proposes a key technology for multi-target UAV swarm mission collaboration in urban disaster relief. The main contents are as follows:

[0101] I. Framework and Task Model Construction for Unmanned Bee Swarm Task Execution

[0102] As mentioned earlier, reconnaissance of architectural space is a continuous and progressive process, consisting of multiple waves. For example... Figure 2 As shown.

[0103] The following constraints apply to the mission waves:

[0104] 1. The scenario includes m target areas and n drones;

[0105] 2. In each wave, a single drone can only operate on one target area;

[0106] 3. In one wave, only one drone can operate in real time on a target area;

[0107] 4. If the target area is large, it can be divided into several partitions, each of which forms an independent target area with the same constraints as before.

[0108] like Figure 2 As shown, from a process perspective, a task wave can consist of the following stages:

[0109] 1. Mission Positioning: After the mission is determined, the UAV (drone) receives mission instructions and moves from its starting point to the operational entry point. During this process, the UAV completes its route movement through aerial maneuvers. In this mission command phase, the main considerations are the UAV's maneuvering time cost and environmental risk cost. Based on the overall mission cost constraint, a reasonable path is planned. Two main reference variables are established in this phase: T1 and D1, where: T1 is the total time consumed during the UAV's mission positioning (cruise); D1 is the total distance traveled by the UAV during the mission positioning process.

[0110] 2. On-site Operations: After aerial maneuvering to the designated operation location, the drone (or drone) uses its payload to conduct on-site operations, including surveying and data collection within the building space. During this process, the basic time cost of the task is calculated based on the area of ​​the operation zone and the operational capabilities of the drone unit.

[0111]

[0112] in:

[0113] A: This represents the set of all targets / partitions within the current work area;

[0114] ai: for independent target / partition;

[0115] dt: The time cost for a drone (bee) to complete a unit objective / area operation;

[0116] T A The total time cost for a drone (or bee) to complete a designated work area;

[0117] d: The core capabilities of unmanned aerial vehicles (UAVs);

[0118] S: Operation name, used to calculate the time it takes for a single drone to complete an area operation, which can be simply regarded as S = area / d.

[0119] 3. Mission Recovery: After completing its mission, the UAV (drone) moves from its departure point to its recovery point according to the mission plan. During this process, the UAV completes its route movement through aerial maneuvers. In this mission command phase, a recovery path is pre-planned based on the mission plan. After completing its mission, the UAV matches the pre-planned path to its own status. If the current status meets the cost constraints of the pre-planned recovery path, the recovery is completed according to the pre-planned plan; if the current status does not meet the cost constraints, a new recovery plan and path are autonomously planned based on the conditions, and the UAV moves to the corresponding location. The main considerations are the UAV's maneuvering time cost and environmental risk cost, and the path is rationally planned based on the total mission cost constraint. In this phase, two main reference variables are formed: T2 and T2', where: T2 is the total time consumed during the UAV mission recovery process; T2' is the resource reservation for the UAV during the mission recovery process. Based on the above, the following processing procedure is formed:

[0120]

[0121] in:

[0122] Retrieve is the name of the recycling operation, which can be understood as calling a function;

[0123] Cap is the name of the capability acquisition operation. For example: After completing the task, acquire the remaining battery power of the drone d'.

[0124] d' represents the result after the drone d completes its task.

[0125] Based on the organization of the above task process, a basic task model is formed, as shown below:

[0126] Task = {job i |i=1,2,....p}

[0127] job={ID,location,jobScale,{risk j |j=1,2,....q|}}

[0128] risk = {ID, prop, risklevel}

[0129] in:

[0130] Task: The overall structure of a task, which consists of a set of jobs;

[0131] job: This is the basic definition of a task. Its detailed definition is as follows:

[0132] ID: A globally unique identifier for the job;

[0133] location: The location of the task;

[0134] jobScale: This represents the size of the work area. The total cost is estimated by comparing jobScale with the cost of drone (bee) unit operations.

[0135] Risk: This refers to the identified risks within the work area, consisting of an identifier, attribute set, and risk level. The overall risk of a work area is obtained by aggregating all identified risks.

[0136] risklevel: refers to the level of risk.

[0137] II. Unified Expression Model of Unmanned Bee Swarm Capabilities

[0138] Unmanned aerial vehicles (UAVs) are the main force in mission execution. During mission planning, UAVs are allocated based on the characteristics of the target area to achieve resource matching and force integration. During mission execution, UAVs maneuver and operate according to the planned action plan. Simultaneously, considering the actual impact of the scenario, their behavior is modified based on autonomous decision-making. From the above steps, it can be seen that the ability to express and measure the capabilities of UAVs becomes the foundation of the entire mission command and dispatch.

[0139] Based on this situation, a unified representation model for UAV capabilities is constructed. The basic definition of this model is as follows:

[0140] D = {ID, stdCap, curState, resKpt, capRuleSet, belongTo, prop}, where:

[0141] ID: A globally unique identifier for the drone;

[0142] stdCap: The cost corresponding to the execution process of a standard work unit;

[0143] curState: The current state of the drone; its definition is as follows:

[0144] curState = {preState, nowState, location}, where:

[0145] preState: Represents the previous state;

[0146] nowState: Represents the current state;

[0147] location: The current spatial location;

[0148] resKpt: The mission execution capability reserved for the drone at the current moment;

[0149] capRuleSet: A set of rules for the cost of drones performing tasks under different conditions;

[0150] capRuleSet={(cID,condDef,cost) i |i=1,2,.....n}, where:

[0151] cID: The identifier for the current rule;

[0152] condDef: The constraint condition corresponding to the current rule;

[0153] cost: The cost of a standard behavioral unit under the current constraints;

[0154] belongTo: Information about the bee colony to which the bee belongs;

[0155] prop: Basic parameters of the drone (bee), including cruising speed such as normalSpeed.

[0156] The above content achieves a unified expression of the basic state information and capabilities of the UAV. During mission execution, after the UAV (bee) is assigned a task, its state changes in certain ways, namely:

[0157] jobSet(job,D)={job,D'},D'=stateChange(D,s')

[0158] in,

[0159] jobSet: Drone task assignment operation, which binds a drone with a changed task status to the task.

[0160] stateChange: Operation to change the state of a drone mission.

[0161] III. Regional Multi-Objective Task Matching and Resource Optimization

[0162] During task planning, targets and drones (swarms) are first allocated based on the distribution of buildings and drones within the area. In this process, the cost of completing the task for drones (swarms) under different allocation schemes is calculated based on factors such as the size of the operational area, the reserved task execution capacity (resKpt) of the drones (swarms), and the influence of the surrounding environment. The ant colony algorithm is used to achieve the globally optimal solution, ultimately forming a target allocation and drone task orchestration scheme. For example... Figure 3 As shown:

[0163] Step 1: Construct the task set:

[0164] In this process, the first step is to fuse data from different sources to obtain all target objects within the region, forming a task set:

[0165] Task = {job i The task set (Task) consists of all targets to be observed (i = 1, 2, ..., n). Please refer to the preceding content for its detailed definition.

[0166] Step 2: Construct a candidate set of drone (bee) resources:

[0167] Based on the task set, all available drones (bees) are obtained according to status and cost constraints, forming a candidate resource set Ds:

[0168] Ds={D j |j=1,2,.....m},D j ∈AllD

[0169] curState∈D j ,nowState∈curState

[0170]

[0171] in:

[0172] AllD is the collection of all drone objects;

[0173] wp is the weight value;

[0174] c' represents the actual cost of the current task;

[0175] nowState is a component of curState;

[0176] Check is the function for calculating the value of the UAV during the construction of the cost matrix;

[0177] Easy is a specific value of nowState.

[0178] That is: first select different drones (bees) whose status is available (nowState = easy);

[0179] Then, based on the positional relationship and cost constraints between different drones (bees) and the target, the cost of different tasks for the current drone (bee) is calculated. If the cost of the current drone (bee) and the task is less than its own capacity reserve, it is regarded as a candidate and placed in the Ds set.

[0180] Step 3: Calculate the cost of the target task:

[0181] After the candidate set is selected, based on the drone (bee) (D) i D i ∈Ds) state, and different task objectives (job) j job j The relative positions of ∈Task) constitute the task cost matrix MC:

[0182]

[0183] Cost is the cost calculation function, and its calculation method is as follows:

[0184]

[0185] T t =T1'+T A +T2',

[0186] T A =jobCostAss(job,D)

[0187] T1'=T2'=shortcut(job,D) / normalSpeed,normalSpeed∈prop,prop∈D

[0188] Where T1' represents the resource reservation for the UAV during its positioning process, Shortcut is the cost calculation function for the straight-line distance between the UAV and the target, and jobCostAss is the task execution cost evaluation function, which comprehensively evaluates and calculates the cost (e.g., time and energy consumption) of the current UAV (bee) to complete the operation in the specified task area based on the UAV (bee) unit operation time, task area range, environmental impact, etc.

[0189] In the above calculation process, T offsetThis represents the cost tolerance value. When it is 0, it means that the cost tolerance for path consumption during the drone's (or bee's) deployment and recovery processes is zero. The drone can only maneuver in real time via direct flight from the endpoint. In urban environments, due to factors such as numerous tall buildings obstructing the view and the influence of microclimate weather, this value is obviously quite limited; when this value is increased, it indicates that the path cost tolerance is expanded, and an excessively increased tolerance will lead to a decrease in resource availability. In practical applications, an artificial intelligence-based training model is used to optimize the setting of this value.

[0190] Step 4: Task Resource Allocation

[0191] The target task cost matrix MC represents the time distribution of all available drones in the current area under the current conditions.

[0192] In this matrix, column vector V k , represents the total target area operation cost vector of the kth UAV;

[0193] V k =[cost(job1,D k ),.........,cost(job m D k )] T

[0194] cost(job j D k )∈MC,j≤m

[0195] Row vector H in this matrix p , represents the total cost vector of the p-th objective task;

[0196] H p =[cost(job) p ,D1),........,cost(job p D n )] T

[0197] cost(job p D j )∈MC,j≤n

[0198] In MovieCut (MC), some target areas cannot be completed due to excessively high operational costs, resulting in some drone (bee) target cost units having an infinite value. Therefore, MC needs to be trimmed for subsequent processing. This process follows these rules:

[0199] 1. When a drone (bee) D kWhen all objectives are impossible to complete, delete the column vector V corresponding to the drone (bee) from MC. k ;

[0200] 2. A target area job p No drone (swarm) can complete the operation targeting it in the current wave; therefore, the row vector H corresponding to the target is deleted from MC. P ;

[0201] ValidateMc(MC)=MC”,MC'∈MC,V k ∈MC',H p ∈MC'

[0202]

[0203]

[0204] ValidateMc is the operation for handling invalid vectors in the MC matrix.

[0205] The goal of task allocation is to achieve maximum target coverage at the lowest cost from an overall perspective. However, due to factors such as location, the costs of the participating drones (swarms) vary within a single task batch. Based on this, different task plans are generated by iterating through cost-based scenarios for the target. These task plans combine the task objective with specific drones (swarms). And in this process:

[0206] 1. Set i = 1, and let MC" = MC';

[0207] 2. Constructing a queue plan i And it is initially empty;

[0208] 3. Starting from the i-th line of MC, obtain H i Vector. From H i Extract the C with the minimum cost. ik ;

[0209] getMinCost(H i ) = C ik C ik =cost(job) i D k )

[0210] C ik <C jk i≠j,C ik ∈H i C jk ∈H i

[0211] 4. Place C ik Place it in the current plan queue, and remove the i-th row vector H from MC". i With the k-th column vector V k ;

[0212] 5. Since the current MC exceeds 1 dimension, let i = 2 and execute step 2;

[0213] 6. Place the remaining cost(job, D) of the current MC as a unit in the plan. i middle;

[0214] 7. Let i = i + 1. If the current i exceeds the number of row vectors of MC', then execute step 8; otherwise, execute step 2.

[0215] 8. Queue Formation Plan i Calculate the sum of the ranges of all target areas in the current plan, S. i If S i Greater than the threshold

[0216] If set, then the current plan will be... i If the result is true, it is considered a valid result; otherwise, it is considered an invalid result.

[0217] plan = {C ij，k |k=1,2,....q},C ik ∈MC',

[0218]

[0219] The above is the solution for finding the minimum cost by traversing MC' starting from a target region.

[0220] When the number of targets m and the number of drones (bees) n in the disaster area environment are small, the dimension of MC' is not high. All plans can be found by exhaustive traversal. Under the premise of prioritizing benefits, the plan with the lowest cost is selected as the result output.

[0221] When the number of targets m and the number of drones (swarms) n in the disaster area environment are large, the dimensionality of MC' is high, and the exhaustive traversal method has performance limitations. Antenna algorithm can be used to process MC', forming a fast optimal solution, and finally outputting the result.

[0222] The whole process is as follows Figure 4 As shown.

[0223] Step 5: Further optimization of the task plan:

[0224] During the aforementioned processing, some target areas may be abandoned due to excessive costs (exceeding the capacity limits of the drones / drones) or excessive distances. To address this, based on the aforementioned task resource allocation, larger or more distant areas can be divided into smaller target areas through regional subdivision, allowing for resource capacity adaptation and iterative development of standardized solutions. This process is as follows:

[0225] 1. Let plan T =plan;

[0226] 2. Determine whether the current solution needs further optimization.

[0227] • If the total number of available drones n in the current plan is less than the output plan T The vector dimension indicates that all drones have been assigned tasks, requiring no further adjustments. The program exits the current run and outputs the optimal result plan. T .

[0228] When n is greater than plan T When the vector dimension is [value], it indicates that there are available drones on standby. Further allocation optimization is then performed.

[0229] The process is as follows:

[0230] 3. Process the unassigned target objects jobk sequentially and form a queue reJob based on the job cost;

[0231]

[0232] jobScale k >jobScale p ,k<p<l,jobScale k ∈job k jobScale p ∈job p

[0233] 4. Set k=1 and let T=Task', where Task' represents the set of unassigned target objects;

[0234] 5. Extract the k-th job from reJob k Object, if jobScale k If the threshold for segmentation is exceeded, the current job will be terminated. k Process and output plan T ;

[0235] 6. Regarding the job k The process is performed to create a new target region object, and the job in T is then divided.k Replace it with a new target object to form a new task set T';

[0236]

[0237]

[0238] 7. Treat T' as a new task set, execute step 1, and obtain the corresponding output result plan';

[0239] 8. Compare the task coverage of plan and plan', and select the optimal one. If the benefits are the same, choose the plan with the lowest cost between plan' and plan' as the result plan. T ;

[0240] 9. Determine whether to stop optimization. If it is not necessary to stop optimization, execute step 1; otherwise, stop outputting the final result.

[0241] In the above process, step 6, the segmentation of the target region, has a certain impact on the calculation results. Specific segmentation methods can be implemented based on artificial intelligence to improve segmentation efficiency and the utilization efficiency of drones (bees). Specific methods will not be elaborated here.

[0242] IV. Unmanned Aerial Vehicle (UAV) Mission Path Planning

[0243] As mentioned earlier, after the drone (swarm) mission planning is completed, relevant information is disseminated to the command unit. Based on the command and control workflow, the mission plan is then issued to the corresponding drone (swarm) for implementation. The entire drone (swarm) operation is divided into three stages based on the objective: mission positioning, operation execution, and mission recovery. In light of the foregoing, this project primarily focuses on path planning and mission matching for the mission positioning and mission recovery stages.

[0244] 1. Task positioning path planning

[0245] Mission positioning refers to the process of a drone (or drone) moving from its starting point to its entry point. Typically, end-to-end linear movement is the most cost-effective and facilitates command and control of the drone. However, urban environments present complex building structures and spatial conditions. On one hand, tall buildings obstruct the flight path; on the other hand, microclimates can form between specific buildings, significantly impacting the flight safety of drones. These factors all pose risks to drone maneuvers. Therefore, appropriate avoidance measures are necessary during maneuvers to ensure the successful execution of the mission.

[0246] While drone (bee) evasive maneuvers improve their own safety, they also incur additional costs. In the aforementioned context, the cost T1 of the mission positioning process is relative to the total cost T. t This has a direct impact. Ignoring this issue during maneuvering will prevent the drone (or drone) from performing its mission and operations normally, leading to the failure of the plan. Therefore, during the deployment process, the drone (or drone) needs to rationally select its path under the constraints of the aforementioned overall mission planning cost coordination, such as... Figure 5 As shown.

[0247] As mentioned earlier, a drone (swarm) mission path planning consists of drone (swarm) behavioral cost constraints and objectives:

[0248] routePlanReq = {D i p1, p2, T 1i},D i ∈Ds,D i ∈plan T ,in:

[0249] D i : Drone (bee) to be planned for path;

[0250] p1: Starting position of the drone (bee);

[0251] p2: Target location (operation entry point) of the drone (bee);

[0252] T 1i Cost constraints are set for the positioning path planning behavior of drones (bees).

[0253] The disaster area environment is highly dynamic. Therefore, drones (or drones) gather information on various risks within the disaster area from multiple sources, forming a situational awareness based on unified labeling. Using this situational map, the drones (or drones) plan their mission paths. For ease of computation, this situational map is simplified to a bitmap format, with the following basic definition:

[0254] instEnvMap = {px ij |i=1,2,...n,j=1,2,....,m}

[0255] px ij ={center,riskGreyTag}

[0256] px ij A pixel is a basic unit of a bitmap. The entire bitmap is rasterized according to a fixed side length. ij A corresponding basic grid.

[0257] center: The center position of the grid;

[0258] `riskGrayTag`: This tag labels risks. The pixels (px) in the bitmap correspond to grid cells in the map. Within the map, risk objects within the corresponding grid area are mapped according to their risk level. Different risk levels are mapped to a corresponding grayscale value. These grayscale values ​​are then assigned to their corresponding pixels (px) in the bitmap. Here, risks refer to factors affecting the flight safety of drones (drones), such as tall buildings, complex spaces, and severe weather conditions. These risks are distributed according to their actual locations in the map and mapped in the grayscale image.

[0259] Based on this one-dimensional graph, rapid route planning for drones (swarms) is performed. The process is as follows:

[0260] 1. Set t=1, p=p1, route result R;

[0261] 2. Calculate the px corresponding to p1 and p2:

[0262] locate({p1,p2})={px1,px2},p1∈px1,p2∈px2

[0263] 3. Construct vector r, which represents the current p and p2:

[0264]

[0265] 4. Starting from p, Extract all pixels that pass through the sequence and store them in R.

[0266]

[0267] 5. Calculate the total cost C of R at present. R ,

[0268] If C R Not less than T 1i If so, then the current R will be output;

[0269] If C R Less than T 1i Then proceed to step 6;

[0270] 6. Extract the pixel with the largest grayscale value from R:

[0271] getMaxFromR(R) = px k ,px k ∈R,px j ∈R,

[0272] riskGrayTag k >riskGrayTag j riskGrayTag k∈px k riskGrayTag j ∈px j

[0273] Among them, getmaxfrom: the operation to get the cell with the largest gray value from R, and Riskgraytag: the gray value of the cell.

[0274] 7. Let p' = px k ;

[0275] 8. Determine the feasibility of p':

[0276] If the grayscale value of the current p' is less than the threshold setting, then the current px k To proceed, proceed to step 9;

[0277] If all the surrounding pixels of the current pixel p' have been calculated, then select the pixel with the smallest grayscale value to replace the pixel in the R queue and proceed to step 9.

[0278] If the current grayscale value of p' is greater than the threshold setting, then px k If the path is impassable, select the pixel closest to the boundary. pq ,make

[0279] p' = px pq Execute step 8;

[0280] 9. Whether to process all basic units stored in R. If not, proceed to step 5; otherwise, proceed to step 10.

[0281] 10. Complete the current route calculation, extract all px sequences from R, form the route Rline, and output it.

[0282] Its basic process is as follows Figure 6 As shown:

[0283] Once a route is planned, under certain circumstances, the cost of that route will exceed T. 1i In the aforementioned content, the cost tolerance value T is used. offset A certain amount of reservation has been achieved.

[0284] Simultaneously, based on the grayscale values ​​of all pixels in R, the risk distribution and severity along the route can be extracted and calculated, providing a basis for task evaluation.

[0285] riskMax(R) = riskGrayTag k riskGrayTag k ∈px k ,px k ∈R

[0286]

[0287] Based on these two indicators, an evaluation and analysis model is constructed, and the task risk is assessed through simulation training, providing support for task decision-making.

[0288] 2. Task recovery path planning

[0289] After completing its mission, the drone (or drone) reaches its departure point according to the mission settings and begins the mission recovery process. However, during actual mission execution, due to factors such as exceeding the initial mission load, the drone (or drone) may not have the sustained flight capability to return to its initial starting point. Therefore, during this process, a recovery plan needs to be selected based on its own condition and the on-site situation, such as... Figure 7 As shown. The system detects and acquires the resources remaining on the drone (or drone / bee). Based on these resources, a recovery point is selected to ensure the drone (or drone / bee) has sufficient resources to support its return. On this basis, a flight path is planned according to the current location and the recovery point, forming a guidance scheme.

[0290] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Those skilled in the art can modify or make equivalent substitutions to the technical solutions of the present invention. The scope of protection of the present invention should be determined by the claims.

Claims

1. A task coordination method for multi-target unmanned bee swarms, characterized in that, The method includes: Obtain the distribution information of the target area and the capability representation of each UAV in the current UAV swarm queue, wherein the capability representation includes: standard operation unit execution cost, current status, capability reservation, cost constraints and basic parameters; Based on the distribution information, available drones are obtained from the current drone swarm queue according to the capability representation; Based on the distribution information and the capability representation, a task cost matrix for the available UAVs and the target area operations is constructed, and task resources for the available UAVs are allocated according to the task cost matrix. Plan the mission placement path and mission recovery path for each drone with allocated mission resources; The step of constructing the task cost matrix for the available UAVs and the target area operations based on the distribution information and the capability representation includes: Based on the distribution information and the capability representation, the positioning cost, mission execution cost assessment, and recovery cost between each available UAV and each target area are calculated to obtain the corresponding costs. ; Set a cost tolerance value for the path consumption between each available drone and each target area operation; For any available drone and its corresponding target area, compare the capability reserve with the cost tolerance value, and if the difference between the capability reserve and the cost tolerance value is greater than the cost... At that time, the cost between the available drone and the corresponding target area operation is set as the cost. When the difference between the capacity reserve and the cost tolerance value is not greater than the cost. At that time, the cost between the available drone and the corresponding target area is set to infinite; The cost between the available drone and the corresponding task is used as an element in the task cost matrix; The allocation of task resources for the available UAVs based on the task cost matrix includes: Based on the premise that drones cannot complete all operations in the target area and that no drones can complete the operations in the target area, the task cost matrix is ​​trimmed. The pruned task cost matrix is ​​processed using either exhaustive traversal or ant colony algorithm to obtain a queue of several solutions. ,in, This represents the row number of the trimmed task cost matrix, and the row vector in the trimmed task cost matrix... Indicates the first The total cost vector of each target area operation; the exhaustive traversal method is used to process the task cost matrix with a dimension smaller than the set value, and the ant colony algorithm is used to process the task cost matrix with a dimension larger than the set value. Statistical scheme queue The sum of the target area range , and when When the value exceeds the threshold setting, the solution queue will be... This is considered a valid result; Based on the cost of each valid result, obtain the optimal solution queue. ; According to the optimal solution queue Allocate task resources for the available UAVs; Specifically, the process of using exhaustive traversal or ant colony algorithm to process the pruned task cost matrix yields several solution queues. ,include: Construct scheme queue And it is initially empty; From row vectors Extract the element with the lowest cost. , the element Placed in the scheme queue And delete the row vectors With the corresponding column vector , to obtain the matrix ,in, Indicates the column number of the trimmed task cost matrix; From the matrix row vectors Extract the element with the lowest cost. , the element Placed in the scheme queue And delete the row vectors With the corresponding column vector , to obtain the matrix ,in, Indicates extracting elements Number of times, Representation matrix Line numbers in the text; After processing the trimmed task cost matrix, the solution queue is obtained. ; Wherein, the queue according to the optimal solution Allocating task resources for the available UAVs, including: Compare the number of available drones with the optimal solution queue. The dimension; If the number of available drones is not greater than the optimal solution queue The dimension is then determined according to the optimal solution queue. Allocate task resources for the available UAVs; If the number of available drones is not greater than the optimal solution queue The dimension of the optimal solution queue is then used. The optimization includes: setting a segmentation threshold limit; and performing sequential processing on unassigned target area jobs based on job cost to obtain a queue. ; Determine the queue Check if the job cost of each target region is less than the segmentation threshold: if so, queue the optimal solution. As an optimization result, task resource allocation is performed; otherwise: unallocated target area tasks with operation costs greater than the segmentation threshold are segmented, and the process is returned to the above. Based on the distribution information, available drones are obtained from the current drone swarm queue according to the capability expression to obtain the optimal solution queue. Compare the queue of the optimal solutions. With the optimal solution queue The task coverage and cost are used to obtain an optimization solution queue. ; By optimizing the queue of the scheme corresponding queues If the cost of each task in the target region is less than the segmentation threshold, determine whether to continue optimization. To optimize the number of times; Queue the optimal solution As an optimization result, task resource allocation is performed, where To optimize the total number of times; The planning of the drone mission recovery path for each allocated mission resource includes: Obtain the current status and location of the drone; Obtain the task positioning cost of the UAV executing the task positioning path, and compare the task positioning cost with a set threshold. Compare; When the cost of the task in place is greater than the set threshold In the case of [condition], select and determine whether the nearest recycling point meets the requirements; if the requirements are met, plan the task recycling path between the current location and the nearest recycling point; if the requirements are not met, plan the task recycling path between the current location and the original recycling point. The task placement cost is not greater than the set threshold. In this case, plan the task recovery path between the current location and the original recovery point.

2. The method as described in claim 1, characterized in that, The step of obtaining available drones from the current drone swarm queue based on the distribution information and the capability representation includes: Select drones that are currently available in the drone swarm queue; For a drone that is in an available state, the positional relationship between the drone and the target is calculated based on the current state and the distribution information; Based on the location relationship and the cost constraint, calculate the cost of different tasks for the current UAV; If the cost of any task to the current drone is less than the capacity reservation, then the drone is considered a usable drone.

3. The method as described in claim 1, characterized in that, The planning of the drone's positioning path for each allocated task resource includes: For the entire region, a situation map in bitmap form is constructed, wherein the bitmap is composed of several basic units; Set the mission positioning path cost constraint for the UAV and the grayscale value threshold for the basic unit; Based on the basic units corresponding to the UAV's starting position and target position, a vector is constructed. ; Extract the vector All the basic units traversed are arranged into a sequence based on vector relationships and placed in the in-situ route result R; Compare the cost of the positioning path result R with the cost constraint of the task positioning path: If the cost of the positioning route result R is not less than the task positioning path cost constraint, then jump to output the UAV planning task positioning path; If the cost of the in-place route result R is not less than the task in-place path cost constraint, then: Extract the basic unit with the largest gray value from the in-place line result R; The gray value of the basic unit with the largest gray value is compared with the gray value threshold: If the gray value of the basic unit with the largest gray value is less than the gray value threshold, then proceed to determine whether all basic units in the positioning line result R have been processed. If the gray value of the basic unit with the largest gray value is greater than the gray value threshold, and there is at least one basic unit with a nearby boundary, then a basic unit with a nearby boundary is selected as the basic unit with the largest gray value, and the process returns to comparing the gray value of the basic unit with the largest gray value with the gray value threshold. If the gray value of the basic unit with the largest gray value is greater than the gray value threshold, and there are no basic units adjacent to the boundary, then the basic unit with the smallest gray value is selected to replace it and placed in the in-place line result R, and the process jumps to determine whether all basic units in the in-place line result R have been processed. Determine whether all basic units in the in-place line result R have been processed: If not, return to the comparison between the cost of the positioning route result R and the cost constraint of the task positioning path; If so, then proceed to output the drone's planned mission positioning path; Output the planned landing path for the UAV mission.

4. A multi-target unmanned swarm task coordination device, characterized in that, The device includes: The information acquisition module is used to acquire the distribution information of the target area and the capability expression of each UAV in the current UAV swarm queue. The capability expression includes: standard operation unit execution cost, current status, capability reservation, cost constraints and basic parameters. The drone filtering module is used to select available drones from the current drone swarm queue based on the distribution information and the capability expression. The resource allocation module is used to construct a task cost matrix for the available UAVs and the target area operations based on the distribution information and the capability expression, and to allocate task resources for the available UAVs according to the task cost matrix. The path planning module is used to plan the mission placement path and mission recovery path for each drone with allocated mission resources. The step of constructing the task cost matrix for the available UAVs and the target area operations based on the distribution information and the capability representation includes: Based on the distribution information and the capability representation, the positioning cost, mission execution cost assessment, and recovery cost between each available UAV and each target area are calculated to obtain the corresponding costs. ; Set a cost tolerance value for the path consumption between each available drone and each target area operation; For any available drone and its corresponding target area, compare the capability reserve with the cost tolerance value, and if the difference between the capability reserve and the cost tolerance value is greater than the cost... At that time, the cost between the available drone and the corresponding target area operation is set as the cost. When the difference between the capacity reserve and the cost tolerance value is not greater than the cost. At that time, the cost between the available drone and the corresponding target area is set to infinite; The cost between the available drone and the corresponding task is used as an element in the task cost matrix; The allocation of task resources for the available UAVs based on the task cost matrix includes: Based on the premise that drones cannot complete all operations in the target area and that no drones can complete the operations in the target area, the task cost matrix is ​​trimmed. The pruned task cost matrix is ​​processed using either exhaustive traversal or ant colony algorithm to obtain a queue of several solutions. ,in, This represents the row number of the trimmed task cost matrix, and the row vector in the trimmed task cost matrix... Indicates the first The total cost vector for operations in each target area; Statistical scheme queue The sum of the target area range , and when When the value exceeds the threshold setting, the solution queue will be... This is considered a valid result; Based on the cost of each valid result, obtain the optimal solution queue. ; According to the optimal solution queue Allocate task resources for the available UAVs; Specifically, the process of using exhaustive traversal or ant colony algorithm to process the pruned task cost matrix yields several solution queues. ,include: Construct scheme queue And it is initially empty; From row vectors Extract the element with the lowest cost. , the element Placed in the scheme queue And delete the row vectors With the corresponding column vector , to obtain the matrix ,in, Indicates the column number of the trimmed task cost matrix; From the matrix row vectors Extract the element with the lowest cost. , the element Placed in the scheme queue And delete the row vectors With the corresponding column vector , to obtain the matrix ,in, Indicates extracting elements Number of times, Representation matrix The row numbers in the table; the exhaustive traversal method is used to process the task cost matrix with a dimension smaller than the set value, and the ant colony algorithm is used to process the task cost matrix with a dimension larger than the set value; After processing the trimmed task cost matrix, the solution queue is obtained. ; Wherein, the queue according to the optimal solution Allocating task resources for the available UAVs, including: Compare the number of available drones with the optimal solution queue. The dimension; If the number of available drones is not greater than the optimal solution queue The dimension is then determined according to the optimal solution queue. Allocate task resources for the available UAVs; If the number of available drones is not greater than the optimal solution queue The dimension of the optimal solution queue is then used. The optimization includes: setting a segmentation threshold limit; and performing sequential processing on unassigned target area jobs based on job cost to obtain a queue. ; Determine the queue Check if the job cost of each target region is less than the segmentation threshold: if so, queue the optimal solution. As an optimization result, task resource allocation is performed; otherwise: unallocated target area tasks with operation costs greater than the segmentation threshold are segmented, and the process is returned to the above. Based on the distribution information, available drones are obtained from the current drone swarm queue according to the capability expression to obtain the optimal solution queue. Compare the queue of the optimal solutions. With the optimal solution queue The task coverage and cost are used to obtain an optimization solution queue. ; By optimizing the queue of the scheme corresponding queues If the cost of each task in the target region is less than the segmentation threshold, determine whether to continue optimization. To optimize the number of times; Queue the optimal solution As an optimization result, task resource allocation is performed, where To optimize the total number of times; The planning of the drone mission recovery path for each allocated mission resource includes: Obtain the current status and location of the drone; Obtain the task positioning cost of the UAV executing the task positioning path, and compare the task positioning cost with a set threshold. Compare; When the cost of the task in place is greater than the set threshold In the case of [condition], select and determine whether the nearest recycling point meets the requirements; if the requirements are met, plan the task recycling path between the current location and the nearest recycling point; if the requirements are not met, plan the task recycling path between the current location and the original recycling point. The task placement cost is not greater than the set threshold. In this case, plan the task recovery path between the current location and the original recovery point.

5. A computer device comprising a memory and a processor, the memory storing a computer program which is loaded and executed by the processor to implement the method of any one of claims 1-3.

Images