A multi-UAV disaster rescue planning method

Optimizing drone task allocation through multi-objective quantum northern goshawk mechanism and non-dominant solution sorting has solved the local convergence problem in the existing technology, and improved the task execution efficiency of drone disaster rescue and disaster victim rescue.

CN114995492BActive Publication Date: 2024-11-08HARBIN ENG UNIV
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202210594253.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-27
Publication Date
2024-11-08
Estimated Expiration
2042-05-27

AI Technical Summary

Technical Problem

The existing UAV disaster rescue planning methods are prone to local convergence and cannot effectively optimize the allocation of multi-target tasks, resulting in low task execution efficiency and increased casualties for disaster victims.

Method used

The multi-objective quantum northern goshawk mechanism is adopted to optimize the mission allocation of the drone through quantum position update and non-dominant desorption, and combine fuel constraints and load constraints to optimize the overall flight cost and casualty cost.

Benefits of technology

The optimization search rate of the evolution mechanism has been improved, local convergence problems have been overcome, more reasonable task allocation has been achieved, the overall flight distance and casualties of the drone have been reduced, and the efficiency of disaster victims has been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114995492B_ABST
    Figure CN114995492B_ABST
Patent Text Reader

Abstract

The present invention discloses a multi-UAV disaster rescue planning method, step 1, establish a multi-UAV rescue planning model; step 2, initialize the quantum northern goshawk quantum position and set parameters; step 3, calculate the quantum northern goshawk objective function value; step 4, sort the non-dominated solutions according to the objective function values ​​of all quantum northern goshawk positions; step 5, calculate the crowdedness of the quantum northern goshawk position in each non-dominated level; step 6, update the quantum northern goshawk quantum position in the prey identification and attack stage; step 7, update the quantum northern goshawk quantum position in the pursuit stage; step 8, determine whether the maximum number of iterations of the quantum northern goshawk is reached, if yes, terminate the iteration, correspond the quantum northern goshawk position with a non-dominated level of 1 to a task allocation matrix, and output it as the disaster rescue planning task allocation result; otherwise, let k=k+1, and execute step 4. The present invention overcomes the disadvantage of being easily trapped in local convergence and improves the optimization rate of the evolutionary mechanism.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of unmanned aerial vehicle (UAV) applications, and relates to a multi-UAV disaster relief planning method, in particular to a multi-UAV disaster relief planning method of a multi-target quantum northern goshawk mechanism. Background Art

[0002] Unmanned Aerial Vehicle (UAV) is a powered, controllable, and unmanned aerial vehicle that can perform a variety of tasks. It has the advantages of low cost, flexibility, easy deployment, and long endurance. It is now widely used in military, agricultural, search and rescue, environmental monitoring and other fields. In life, natural disasters such as earthquakes, tsunamis, and mudslides occur from time to time. They are highly destructive, have a great social impact, and are difficult to defend. If they occur in villages or cities, they will inevitably endanger everyone. These disasters are often accompanied by secondary disasters such as aftershocks, floods, and fires, which bring great inconvenience to manual rescue.

[0003] The core of drone swarm technology is swarm intelligence, which means that under the control of artificial intelligence, many tasks can be completed autonomously. The inspiration of "swarm intelligence" technology comes from ant colonies, bee colonies and other swarm organisms in nature. These types of organisms have the same characteristics, that is, the individual organisms are very weak, but these organisms have very strong group collaboration capabilities and can complete such as beehives and ant nests. Therefore, the current drone swarm intelligence technology simulates the communication and collaboration between bees and ants to enhance the performance of drone swarms in executing tasks.

[0004] According to existing literature, Li Minglong et al. proposed an air-ground collaborative search and rescue framework for earthquake scenarios in the article "Research on Air-Ground Collaborative Unmanned Group Task Planning for Disaster Search and Rescue Scenarios" published in the Journal of Mechanical Engineering (2019, 55(11): 1-9). High-altitude reconnaissance aircraft are used for search missions to determine the location information of people in the disaster area; unmanned vehicles are used for rescue missions to rescue the affected people. An adaptive feedback-regulated genetic algorithm is proposed to plan the paths of drones and unmanned vehicles. However, although this method is air-ground collaborative planning, the target problem is considered to be single, and the algorithm is prone to local convergence.

[0005] In summary, the above literature has achieved certain results in the search and rescue of unmanned groups, but the problem is too single. Therefore, the present invention designs a multi-objective quantum northern goshawk mechanism multi-UAV disaster rescue planning method, with the overall flight cost and casualty cost of the UAV as two objective functions. After determining the location of the affected people and the characteristics that the affected people tend to gather together, several gathering centers of the affected people are obtained, and multiple UAVs are dispatched to reasonably plan the path to transport supplies to maximize the number of survivors of the affected people. Summary of the invention

[0006] In view of the above-mentioned prior art, the technical problem to be solved by the present invention is to provide a multi-UAV disaster rescue planning method of a multi-objective quantum northern goshawk mechanism, which overcomes the problem that the existing method is prone to fall into local convergence and improves the optimization rate of the evolutionary mechanism.

[0007] In order to solve the above technical problems, a multi-UAV disaster rescue planning method of the present invention comprises the following steps:

[0008] Step 1: Establish a multi-UAV rescue planning model;

[0009] Step 2: Initialize the quantum position of the quantum northern goshawk and set the parameters;

[0010] Step 3, calculate the objective function values ​​of all quantum northern goshawks;

[0011] Step 4: sort the non-dominated solutions according to the objective function values ​​of all quantum northern goshawk positions;

[0012] Step 5, calculate the crowding degree of the quantum northern goshawk position in each non-dominated level;

[0013] Step 6: Update the quantum position of the quantum northern goshawk during the prey identification and attack phase;

[0014] Step 7: Update the quantum position of the quantum northern goshawk during the pursuit phase;

[0015] Step 8: Determine whether the maximum number of iterations K2 of the quantum northern goshawk has been reached. If so, terminate the iteration, correspond the positions of the quantum northern goshawks with a non-dominated level of 1 to the task allocation matrix, and output it as the task allocation result of the disaster relief planning; otherwise, set k=k+1 and continue to execute step 4.

[0016] Furthermore, the multi-UAV rescue planning model established in step 1 includes:

[0017] Each person in the disaster area has a central location. M rescue drones will provide relief supplies to the people in the disaster area based on the central location. Assuming that after the disaster area information is determined, N central locations of the people are obtained, and M rescue drones perform N rescue tasks, an M×N-dimensional task allocation matrix A will be obtained, A={a m,n |a m,n ∈{0,1}} M×N ,in, Indicates whether the m-th UAV executes the n-th mission, m=1,2,…,M,n=1,2,…,N; different task allocation matrices A represent different task allocation schemes, optimizing the two objective functions of overall flight cost and casualty cost under the conditions of fuel constraints and load constraints;

[0018] The minimum joint optimization model is as follows: The total flight cost of the rescue drone and The casualties among the disaster victims, represents the total flight distance of all rescue drones, It represents the total cost of casualties caused by the disaster victims missing the rescue time; the overall flight cost of the drone: C1 and C2 represent two constraints, ξ1 and ξ2 are the influencing factors of the constraints, and D m represents the flight distance of the mth drone from departure to return to base; the casualties of the affected people: ξ3 and ξ4 are the influencing factors of the constraints, L n represents the total number of people in the nth disaster area, where γ and μ are the influencing factors of the two objective functions, indicating the importance of the corresponding objective functions, and μ>γ;

[0019] Assume that the mth drone is scheduled to execute rescue mission, the flight distance of the mth drone is: in, is the distance that the mth UAV flies from the base to the n1 mission location, After the mth UAV completes the n1 mission, it directly flies from n1 to n2, η m is the number of rescue missions of the mth UAV;

[0020] The victims are divided into four states: seriously injured, slightly injured, thirsty, and hungry. The rescue period for the four states is as follows: seriously injured has the shortest rescue period, followed by slightly injured, thirsty, and hungry. If the lth victim in the nth disaster area has not received the assistance of the drone when the rescue period arrives, he is considered dead. At this time, U n,l = 1, if drone assistance is obtained before the rescue deadline, then U n,l =0;

[0021] When M drones perform rescue missions, they must meet two constraints: one is the fuel constraint C1, which is: in, Indicates that two variables are compared. If the first value is greater than the second, the return value is 1, otherwise the return value is 0. m is the fuel carried by the mth drone, v m is the fuel consumption rate; the second is the rescue mission constraint. Multiple drones cannot perform the same mission repeatedly. The rescue mission constraint is: in, It is a function that determines whether two functions are equal. If they are equal, the return value is 0, and if they are not equal, the return value is 1. It is a non-zero function. When the variable value in the function is 0, it returns 0, otherwise it returns 1.

[0022] The task load balancing factor is designed to evaluate the scheme. When there is an unreasonable allocation scheme, the task load factor distinguishes whether the allocation scheme is reasonable through the judgment conditions, and the unreasonable scheme will be eliminated. The judgment conditions are as follows: in Represents the ceiling function.

[0023] Furthermore, in step 2, the quantum position of the quantum northern goshawk is initialized and the parameters are set, including:

[0024] Set the population size to K1, the maximum number of iterations to K2, and in the initial population, the quantum position of the random initial quantum northern goshawk, the first generation initial quantum position of the i-th quantum northern goshawk is Where S is the maximum dimension of the quantum position vector, and any dimension of all quantum positions is a random number between [0,1]. The position of the quantum northern goshawk is obtained by quantum position measurement; the quantum position of the i-th quantum northern goshawk in the k-th iteration is Then the measured position of the i-th quantum northern goshawk in the k-th iteration is in The h-th dimension variable representing the position of the i-th quantum northern goshawk, is a random number between [0,1], h=1,2,…,S.

[0025] Furthermore, the objective function values ​​of all quantum northern goshawks calculated in step 3 include:

[0026] The position of the i-th quantum northern goshawk of the k-th generation Each corresponds to a multi-UAV rescue mission allocation matrix, specifically: of Corresponding to the first row a in the rescue mission allocation matrix 1,1 ,a 1,2 ,…,a 1,N ; Corresponding to the second row a in the rescue mission allocation matrix 2,1 ,a 2,2 ,…,a 2,N ; and so on, Corresponding to the last row a in the rescue mission allocation matrix M,1 ,a M,2 ,…,a M,N , the task allocation matrix is ​​recorded as The maximum dimension S must satisfy S = N·M;

[0027] Will Mapping to task assignment matrix Substitute it into the model established in step 1 to obtain the objective function value of the i-th quantum northern goshawk in the k-th iteration: Among them, t=1 represents the total flight cost of the objective function, and t=2 represents the total casualty cost of the objective function.

[0028] Furthermore, in step 4, the non-dominated solution sorting is performed according to the objective function values ​​of all quantum northern goshawk positions, including:

[0029] There are two quantum northern goshawks labeled i1 and i2, whose positions are and If all objective functions satisfy And at least one objective function Then the quantum northern goshawk position Dominate the position of quantum northern goshawk is a non-dominated solution; if all objective functions satisfy And at least one objective function Quantum Northern Goshawk Domination Quantum Northern Goshawk is a non-dominated solution; if the above relationship does not exist, then these two quantum northern goshawks have no dominating relationship;

[0030] Traverse each quantum northern goshawk in the quantum northern goshawk population Get the position of the i-th quantum northern goshawk The number of quantum northern goshawks and quilt Dominant quantum northern goshawk position set Then count the number of dominated positions of all quantum northern goshawks, if This means that there is no quantum northern goshawk position dominance The non-dominated level of is 1; if but The non-dominated level is 2; by analogy, the non-dominated level of each quantum northern goshawk position is the number of dominated ones plus 1; finally, all quantum northern goshawks are sorted in ascending order according to their non-dominated levels.

[0031] Furthermore, the calculation of the crowdedness of the quantum northern goshawk position in each non-dominated level in step 5 includes:

[0032] Assume that there are b quantum northern goshawks in the quantum northern goshawk collection with a non-dominated level of w. They are arranged in ascending order according to the tth objective function value, t = 1, 2, After sorting The tth objective function value of the quantum northern goshawk, F tmax is the maximum value of the tth objective function in the quantum northern goshawk set, F t min is the minimum value of the tth objective function in the quantum northern goshawk set, then The crowding degree of the tth objective function in the northern goshawk when or hour, The first The total crowding value of the northern goshawk is Finally, the b quantum northern goshawks are arranged in descending order of crowding degree.

[0033] Furthermore, in step 6, updating the quantum position of the quantum northern goshawk in the prey identification attack phase includes:

[0034] In the prey identification and attack phase, the quantum position of the prey attacked by the i-th quantum northern goshawk is Where r is a random number between [1, N], and r≠i;

[0035] The quantum position of the quantum northern goshawk is obtained by updating the quantum rotation angle and the quantum revolving gate. The h-dimensional quantum rotation angle of the i-th quantum northern goshawk is in, is the h-dimensional quantum position of the i-th quantum northern goshawk in the k-th iteration, represents the h-th dimensional quantum position of the prey in the k-th iteration, which is the prey caught by the i-th quantum northern goshawk, τ is a random number between [0,1], and χ randomly takes the value 1 or 2;

[0036] Use the quantum revolving gate to update the h-dimensional quantum position of the i-th quantum northern goshawk individual: Through quantum position Measure the position Then calculate The objective function value of And discuss by category, if Dominate but Keep the result unchanged; otherwise, The details are as follows:

[0037] Furthermore, in step 7, updating the quantum position of the quantum northern goshawk during the pursuit phase includes:

[0038] In the pursuit phase, assuming that the hunting attack radius is R, the h-dimensional quantum rotation angle of the i-th quantum northern goshawk is obtained: The attack radius τ is a random number between [0,1];

[0039] Use the quantum revolving gate to update the h-dimensional quantum position of the i-th quantum northern goshawk individual: Then calculate The objective function value of And discuss by category, if Dominate but Keep the result unchanged, otherwise, let The details are as follows:

[0040] Beneficial effects of the present invention: Compared with the prior art, the present invention has the following beneficial effects:

[0041] In view of the complex environment of earthquake-stricken areas and the urgent need for supplies by the people in the disaster areas, the present invention designs a quantum northern goshawk mechanism to plan drones and rescue the people in the disaster areas, with minimizing the overall flight distance and the cost of casualties as two objective functions. The quantum northern goshawk mechanism is used to optimize the objective function and obtain a set of Pareto front solutions. Due to the conflict between the overall flight cost and the casualty cost objective functions, reducing the overall flight distance will increase the cost of casualties; reducing the cost of casualties will increase the flight distance of the drone. The fuel carried by each drone is limited. Too much emphasis on reducing the cost of casualties will affect the efficiency of the drone's mission execution and ultimately lead to great casualties. Therefore, the quantum northern goshawk mechanism is designed to overcome the problem that the past algorithm is prone to local convergence. The method of non-dominated solution sorting and crowding calculation is used to evaluate the quantum northern goshawk individuals, so that the entire quantum northern goshawk group evolves to quantum northern goshawk individuals with higher non-dominated levels and greater crowding.

[0042] (1) The multi-UAV rescue planning method designed takes the problem into consideration more comprehensively. Among the disaster victims, four states are set: seriously injured, slightly injured, thirsty and hungry, corresponding to different rescue deadlines. In real life, the victims always tend to gather together, so the central location of the victims is introduced, which is more in line with the actual situation of the victims in the rescue and shortens the simulation time. In the UAV, load constraints and fuel constraints are introduced to make the UAV closer to the real situation.

[0043] (2) The model established by the present invention belongs to a multi-objective optimization problem, which is less studied in the prior art. Compared with the previous single-objective multi-UAV rescue planning, it greatly increases the difficulty of simulation. The objective function established in the model is: the sum of casualty cost and flight cost. When optimizing one objective function, the other objective function often deteriorates. This is in line with the idea that problems cannot be considered one-sidedly in real life. Therefore, the present invention has practical research significance.

[0044] The present invention designs a multi-objective quantum northern goshawk mechanism method, designs a quantum-coded northern goshawk quantum position evolution mechanism, and obtains a new quantum northern goshawk mechanism method, which overcomes the drawbacks of the previous method that it is easy to fall into local convergence, and also improves the optimization rate of the evolution mechanism. And the non-dominated solution sorting method and the crowding degree calculation method are used to sort the population, and finally a uniform Pareto front solution can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Flowchart of the multi-UAV disaster rescue planning method based on the quantum Northern Goshawk mechanism designed for the present invention.

[0046] Figure 2 Comparison diagram of non-dominated solutions obtained by MOQNGO and MOQKH methods.

[0047] Figure 3 Comparison diagram of non-dominated solutions obtained by MOQNGO and MOQKH methods.

[0048] Figure 4 Comparison diagram of non-dominated solutions obtained by MOQNGO and MOQKH methods.

[0049] Figure 5 Comparison diagram of non-dominated solutions obtained by MOQNGO and MOQKH methods. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0051] Combination Figure 1 , the present invention comprises the following steps:

[0052] Step 1: Establish a multi-UAV rescue planning model.

[0053] According to the characteristics that people in disaster areas tend to gather together to wait for rescue, each person in the disaster area has a central location, and M rescue drones will provide relief supplies to the people in the disaster area based on these central locations. Assuming that the disaster area information is determined, N central locations of the people will be obtained. M rescue drones perform N rescue missions, and an M×N-dimensional task allocation matrix A will be obtained, A={a m,n |a m,n ∈{0,1}} M×N ,in, Indicates whether the mth UAV executes the nth mission, m = 1, 2, ..., M, n = 1, 2, ..., N. Different task allocation matrices A represent different task allocation schemes. Under the conditions of fuel constraints and load constraints, optimizing the two objective functions of overall flight cost and casualty cost is the focus of the present invention.

[0054] The minimum joint optimization model is as follows: The total flight cost of the rescue drone and The former represents the total flight distance of all rescue drones, and the latter represents the total casualties caused by the disaster victims after they missed the rescue time. The total flight cost of drones: C1 and C2 represent two constraints, ξ1 and ξ2 are the influencing factors of the constraints, and D m It represents the flight distance of the mth UAV from departure to return to the base. Since the center position of the crowd is determined, the flight distance can be predicted in advance. The casualty cost of the affected people is: ξ3 and ξ4 are the influencing factors of the constraints, L n Represents the total number of people in the nth disaster area. Among them, γ and μ are the influencing factors of the two objective functions, representing the importance of the corresponding objective functions, and satisfying μ>γ.

[0055] Assume that the mth drone is scheduled to execute rescue mission, the flight distance of the mth drone is: in, is the distance that the mth UAV flies from the base to the n1 mission location, After the mth UAV completes the n1 mission, it directly flies from n1 to n2, η m is the number of rescue missions of the mth UAV.

[0056] The disaster victims all have different health conditions and need different assistance. They can be roughly divided into four states: serious injury, minor injury, thirst, and hunger. The rescue period for the four states is different. Among them, serious injury is the most urgent and has the shortest rescue period, followed by minor injury, thirst, and hunger. If the lth victim in the nth disaster area has not received the assistance of the drone when the rescue period arrives, he is considered dead. At this time, U n,l = 1, if drone assistance is obtained before the rescue deadline, then U n,l =0.

[0057] When M drones perform rescue missions, they must meet two constraints. The first is the fuel constraint C1. The fuel carried by the drone is limited, and the fuel limits the total range of the drone. Therefore, the fuel constraint is: in, Indicates that two variables are compared in size. If the first value is greater than the second, the return value is 1, otherwise the return value is 0. m is the fuel carried by the mth drone, v mis the fuel consumption rate. The second is the rescue mission constraint. Each rescue mission can only be performed by one drone, so multiple drones cannot perform the mission repeatedly. The rescue mission constraint is: in, It is a function that determines whether two functions are equal. If they are equal, it returns a value of 0, and if they are not equal, it returns a value of 1. It is a non-zero function. When the variable value in the function is 0, it returns 0, otherwise it returns 1.

[0058] When allocating rescue tasks, it is possible that most tasks are assigned to one drone. However, the number of rescue tasks is limited, which results in some drones performing multiple tasks and some drones having no tasks to perform. This reduces the efficiency of task execution and causes most disaster victims to miss the rescue deadline. Based on this, a task load balancing factor is proposed. The task load balancing factor will evaluate the plan. When there is an unreasonable allocation plan, the task load factor will distinguish whether the allocation plan is reasonable through the judgment conditions, and the unreasonable plan will be eliminated. The judgment conditions are as follows: in Represents the ceiling function.

[0059] Step 2: Initialize the quantum position of the quantum northern goshawk and set the parameters.

[0060] Set the population size to K1 and the maximum number of iterations to K2. In the initial population, the quantum position of the random initial quantum northern goshawk is the first generation initial quantum position of the i-th quantum northern goshawk. Where S is the maximum dimension of the quantum position vector, and any dimension of all quantum positions is a random number between [0,1]. The position of the quantum northern goshawk can be obtained through quantum position measurement. The quantum position of the i-th quantum northern goshawk in the k-th iteration is Then the position of the i-th quantum northern goshawk in the k-th iteration can be measured as in The h-th dimension variable representing the position of the i-th quantum northern goshawk, is a random number between [0,1], h=1,2,…,S.

[0061] Step three, calculate the objective function values ​​of all quantum northern goshawks.

[0062] The position of the i-th quantum northern goshawk of the k-th generation Each corresponds to a multi-UAV rescue mission allocation matrix, specifically: of Corresponding to the first row a in the rescue mission allocation matrix 1,1 ,a 1,2 ,…,a 1,N ; Corresponding to the second row a in the rescue mission allocation matrix 2,1 ,a 2,2 ,…,a 2,N ; and so on, Corresponding to the last row a in the rescue mission allocation matrix M,1 ,a M,2 ,…,a M,N , the task allocation matrix is ​​recorded as Therefore, the maximum dimension S must satisfy S=N·M.

[0063] Will Mapping to task assignment matrix Bring it into the model established by the present invention to obtain the objective function value of the i-th quantum northern goshawk in the k-th iteration Among them, t=1 represents the total flight cost of the objective function, and t=2 represents the total casualty cost of the objective function.

[0064] Step 4: Sort the non-dominated solutions according to the objective function values ​​of all quantum northern goshawk positions.

[0065] The model established by the present invention belongs to the minimum optimization problem. Therefore, there are two quantum northern goshawks labeled i1 and i2, whose positions are respectively and If all objective functions satisfy And at least one objective function Then the quantum northern goshawk position Dominate the position of quantum northern goshawk is a non-dominated solution. On the contrary, if all objective functions satisfy And at least one objective function Quantum Northern Goshawk Domination Quantum Northern Goshawk is a non-dominated solution. If the above relationship does not exist, then these two quantum northern goshawks have no dominating relationship.

[0066] Traverse each quantum northern goshawk in the quantum northern goshawk population Get the position of the i-th quantum northern goshawk The number of quantum northern goshawks and quilt The set of quantum northern goshawk positions P dominated by i k Then count the number of dominated positions of all quantum northern goshawks, if This means that there is no quantum northern goshawk position dominance The non-dominated level of is 1; if but The non-dominated level is 2; and so on, the non-dominated level of each quantum northern goshawk position is the number of dominated plus 1. Finally, all quantum northern goshawks are sorted in ascending order according to their non-dominated levels.

[0067] Step 5: Calculate the crowding degree of the quantum northern goshawk position in each non-dominated level.

[0068] Assume that there are b quantum northern goshawks in the quantum northern goshawk collection with a non-dominated level of w. They are arranged in ascending order according to the tth objective function value, t = 1, 2, After sorting The tth objective function value of the quantum northern goshawk, F t max is the maximum value of the tth objective function in the quantum northern goshawk set, F t min is the minimum value of the tth objective function in the quantum northern goshawk set, then The crowding degree of the tth objective function in the northern goshawk when or hour, The first The total crowding value of the northern goshawk is Finally, the b quantum northern goshawks are arranged in descending order of crowding degree.

[0069] Step six, updating the quantum position of the quantum northern goshawk during the prey recognition and attack phase.

[0070] The quantum northern goshawk has two main behaviors during hunting: prey identification and attack phase and pursuit phase. This simulates the hunting habits of the northern goshawk in nature. The former is used for global search and the latter is used for local search. In the prey identification and attack phase: the quantum northern goshawk randomly selects a prey and quickly attacks it. Due to the random selection in the search space, this phase increases the global search of the search space in order to determine the optimal area. The quantum position of the prey attacked by the i-th quantum northern goshawk is Where r is a random number between [1,N] and r≠i.

[0071] In the quantum northern goshawk method, the quantum position of the quantum northern goshawk is obtained by updating the quantum rotation angle and the quantum rotation gate. Compared with the northern goshawk method, the optimization rate and the quality of the optimal solution are improved. The h-dimensional quantum rotation angle of the i-th quantum northern goshawk is in, is the h-th dimensional quantum position of the i-th quantum northern goshawk in the k-th iteration. represents the h-dimensional quantum position of the prey in the k-th iteration, which is the prey caught by the i-th quantum northern goshawk. τ is a random number between [0,1], and χ takes a random value of 1 or 2.

[0072] Use the quantum revolving gate to update the h-dimensional quantum position of the i-th quantum northern goshawk individual: Through quantum position Measure the position Then calculate The objective function value of And discuss by category, if Dominate but Keep the result unchanged; otherwise, The details are as follows:

[0073] Step seven, update the quantum position of the quantum northern goshawk during the pursuit phase.

[0074] In the second stage, after the quantum northern goshawk attacks the prey, the prey will try to escape, and the prey will continue to be chased. Due to the high-speed flight of the northern goshawk, it can catch almost any prey and hunt. This behavior simulation improves the local search ability of the algorithm. Assuming that the attack radius of this hunting is R, the h-th dimension quantum rotation angle of the i-th quantum northern goshawk is obtained: The attack radius τ is a random number between [0,1].

[0075] Use the quantum revolving gate to update the h-dimensional quantum position of the i-th quantum northern goshawk individual: Then calculate The objective function value of And discuss by category, if Dominate but Keep the result unchanged, otherwise, let The details are as follows:

[0076] Step 8, determine whether the maximum number of iterations K2 of the quantum northern goshawk has been reached. If so, terminate the iteration, correspond the positions of the quantum northern goshawks with a non-dominated level of 1 to the task allocation matrix, and output it as the task allocation result of the disaster relief planning; otherwise, set k=k+1 and continue to execute step 4.

[0077] The following examples are given in combination with specific parameters to illustrate the performance of the present invention. The multi-objective quantum northern goshawk optimization method is recorded as MOQNGO, and the multi-objective quantum krill method is recorded as MOQKH. In order to verify the performance of the multi-objective UAV rescue method based on the quantum northern goshawk mechanism, the present invention conducted four sets of simulation experiments. Set the population size to K1 = 100, the maximum number of iterations K2 = 300, the flight cost influence factor γ = 0.1, the casualty cost influence factor μ = 0.3, ξ1 = 150, ξ2 = 500, ξ3 = 1, ξ4 = 2, the dimension S = M × N, the UAV base position is (0, 0), the UAV range constraint D = 1000, the rescue period for seriously injured victims is 12h, the rescue period for slightly injured victims is 24h, the rescue period for thirsty victims is 96h, the rescue period for hungry victims is 120h, and the UAV flight speed is 20 units / h. In the simulation experiment, two assumptions need to be met: the relief supplies carried by each drone are sufficient for the disaster area, and the time the drone takes to provide supplies to the victims is negligible. Table 1 gives the central location of the people in the disaster area.

[0078] Table 1 Location of the mass centers in the disaster area

[0079]

[0080] The status of the people in the disaster area is as follows:

[0081] Table 2 Status of disaster victims

[0082]

[0083]

[0084] In the first set of experiments, N = 4 drones performed T = 9 rescue missions for a total of 378 disaster victims, most of whom were seriously injured in disaster areas T1 to 4. The rescue period for seriously injured victims is the shortest and their lives are in danger, so they should be rescued first. Figure 2 This is a comparison chart of non-dominated solutions obtained by using the MOQNGO and MOQKH methods when the total flight cost of the drone and the casualty cost of the disaster victims are used as the optimization targets. From the figure, it can be seen that the optimal solution generated by the MOQNGO method is slightly smaller than the optimal solution generated by the MOQNGO method in terms of total flight cost, and at the same time, the rescue mission is completed with zero casualty cost. The rescue mission allocation scheme is shown in Table 3:

[0085] Table 3 Task allocation schemes corresponding to some individuals in the Pareto front end

[0086]

[0087] In the second group of experiments, N = 6 drones performed T = 14 rescue missions, with a total of 586 victims. Among them, T1-T4, T13 and T14 disaster areas had the most seriously injured victims, while T11 had the most slightly injured victims. In the rescue mission planning, the areas with the most seriously injured victims were given priority, followed by the areas with the most slightly injured victims. The non-dominated solutions obtained by the MOQNGO and MOQKH methods are as follows: Figure 3 As shown in Table 4, since the number of seriously injured disaster areas is equal to the number of drones, and the number of rescue missions is small, only one optimal solution is generated. It can be seen that the optimal solution generated by the MOQNGO method also achieves the rescue mission with zero casualties, and can dominate the optimal solution generated by the MOQKH method. The rescue mission allocation scheme is shown in Table 4:

[0088] Table 4 Task allocation schemes corresponding to some individuals in the Pareto front end

[0089]

[0090] In the third group of experiments, N = 8 drones performed T = 18 rescue missions, with a total of 748 victims, of which T1 to T4, T13, T14 and T18 had the most seriously injured victims, while T11 and T15 had the most injured victims. The non-dominated solutions obtained by the MOQNGO and MOQKH methods are as follows: Figure 4 As shown. It can also be seen that the optimal solution generated by the MOQNGO method can dominate the optimal solution generated by the MOQKH method. The rescue task allocation scheme is shown in Table 5:

[0091] Table 5 Task allocation schemes corresponding to some individuals in the Pareto front end

[0092]

[0093] In the fourth group of experiments, N = 10 drones performed T = 25 rescue missions, with a total of 994 victims, of which T1-T4, T13, T14, T18, T19 and T23 were the most seriously injured, followed by T11, T15, T20 and T25. The non-dominated solutions obtained by MOQNGO and MOQKH methods are as follows: Figure 5 As shown. Overall, in the four groups of experiments, the optimal solution generated by the MOQNGO method can dominate the optimal solution generated by the MOQKH method, and the rescue task allocation is more uniform and reasonable. The rescue task allocation scheme is shown in Table 6:

[0094] Table 6 Task allocation schemes corresponding to some individuals in the Pareto front end

[0095]

[0096]

Claims

1. A multi-UAV disaster rescue planning method, It is characterized in that The following steps are involved: Step 1: Establish a multi-UAV rescue planning model; Each person in the disaster area has a central location. M rescue drones will provide relief supplies to the people in the disaster area based on the central location. Assuming that after the disaster area information is determined, N central locations of the people are obtained, and M rescue drones perform N rescue tasks, an M×N-dimensional task allocation matrix A will be obtained, A={a m,n |a m,n ∈{0,1}} M×N ,in, Indicates whether the m-th UAV executes the n-th mission, m=1,2,…,M,n=1,2,…,N; different task allocation matrices A represent different task allocation schemes, optimizing the two objective functions of overall flight cost and casualty cost under the conditions of fuel constraints and load constraints; The minimum joint optimization model is as follows: The total flight cost of the rescue drone and The cost of casualties among the disaster victims, represents the total flight distance of all rescue drones, It represents the total cost of casualties caused by the disaster victims missing the rescue time; the overall flight cost of the drone: Represents two constraints, ξ 1 , 2 is the impact factor of the constraint condition, D m represents the flight distance of the mth drone from departure to return to base; the casualties of the affected people: ξ 3 , 4 is the impact factor of the constraint condition, L n represents the total number of people in the nth disaster area, where γ and μ are the influencing factors of the two objective functions, indicating the importance of the corresponding objective functions, and μ>γ; Assume that the mth drone is scheduled to execute rescue mission, the flight distance of the mth drone is: in, For the mth drone to go from the base to n 1 The flight distance of the mission location, Complete n operations for the mth drone 1 After the task, directly from n 1 Location Fly to n 2 The flight distance, η m is the number of rescue missions of the mth UAV; The victims are divided into four states: seriously injured, slightly injured, thirsty, and hungry. The rescue period for the four states is as follows: seriously injured has the shortest rescue period, followed by slightly injured, thirsty, and hungry. If the lth victim in the nth disaster area has not received the assistance of the drone when the rescue period arrives, he is considered dead. At this time, U n,l = 1, if drone assistance is obtained before the rescue deadline, then U n,l =0; When M drones perform rescue missions, they must meet two constraints: one is the fuel constraint C 1 , the fuel constraints are: in, Indicates that two variables are compared. If the first value is greater than the second, the return value is 1, otherwise the return value is 0. m is the fuel carried by the mth drone, v m is the fuel consumption rate; the second is the rescue mission constraint. Multiple drones cannot perform the same mission repeatedly. The rescue mission constraint is: in, It is a function that determines whether two functions are equal. If they are equal, the return value is 0, and if they are not equal, the return value is 1. 2 () is a non-zero function. When the variable value in the function is 0, it returns 0, otherwise it returns 1. The task load balancing factor is designed to evaluate the scheme. When there is an unreasonable allocation scheme, the task load factor distinguishes whether the allocation scheme is reasonable through the judgment conditions, and the unreasonable scheme will be eliminated. The judgment conditions are as follows: in represents the ceiling function; Step 2: Initialize the quantum position of the quantum northern goshawk and set the parameters; Step 3, calculate the objective function values ​​of all quantum northern goshawks; Step 4: sort the non-dominated solutions according to the objective function values ​​of all quantum northern goshawk positions; Step 5, calculate the crowding degree of the quantum northern goshawk position in each non-dominated level; Step 6: Update the quantum position of the quantum northern goshawk during the prey identification and attack phase; Step 7: Update the quantum position of the quantum northern goshawk during the pursuit phase; Step 8: Determine whether the maximum number of iterations K of the quantum northern goshawk has been reached 2 , if yes, the iteration is terminated, and the position of the quantum northern goshawk with a non-dominated level of 1 is corresponded to the task allocation matrix, and is output as the task allocation result of the disaster relief planning; otherwise, let k = k + 1 and continue to execute step 4.

2. A multi-UAV disaster rescue planning method according to claim 1, Features: Step 2 of initializing the quantum position of the quantum northern goshawk and setting parameters includes: Set the population size to K 1 , the maximum number of iterations is K 2 , in the initial population, the quantum position of the random initial quantum northern goshawk, the first generation initial quantum position of the i-th quantum northern goshawk is Where S is the maximum dimension of the quantum position vector, and any dimension of all quantum positions is a random number between [0,1]. The position of the quantum northern goshawk is obtained by quantum position measurement; the quantum position of the i-th quantum northern goshawk in the k-th iteration is Then the measured position of the i-th quantum northern goshawk in the k-th iteration is in The h-th dimension variable representing the position of the i-th quantum northern goshawk, is a random number between [0,1], h=1,2,…,S.

3. A multi-UAV disaster rescue planning method according to claim 1, Features: The objective function values ​​of calculating all quantum northern goshawks in step 3 include: The position of the i-th quantum northern goshawk of the k-th generation Each corresponds to a multi-UAV rescue mission allocation matrix, specifically: of Corresponding to the first row a in the rescue mission allocation matrix 1,1 ,a 1,2 ,…,a 1,N ; Corresponding to the second row a in the rescue mission allocation matrix 2,1 ,a 2,2 ,…,a 2,N ; and so on, Corresponding to the last row a in the rescue mission allocation matrix M,1 ,a M,2 ,…,a M,N , the task allocation matrix is ​​recorded as The maximum dimension S must satisfy S = N·M; Will Mapping to task assignment matrix Substitute it into the model established in step 1 to obtain the objective function value of the i-th quantum northern goshawk in the k-th iteration: Among them, t=1 represents the total flight cost of the objective function, and t=2 represents the total casualty cost of the objective function.

4. A multi-UAV disaster rescue planning method according to claim 1, Features: The non-dominated solution sorting according to the objective function values ​​of all quantum northern goshawk positions described in step 4 includes: There are two labels i 1 and i 2 The positions of the quantum northern goshawk are and If all objective functions satisfy And at least one objective function Then the quantum northern goshawk position Dominate the position of quantum northern goshawk is a non-dominated solution; if all objective functions satisfy And at least one objective function Quantum Northern Goshawk Domination Quantum Northern Goshawk is a non-dominated solution; if the above relationship does not exist, then these two quantum northern goshawks have no dominating relationship; Traverse each quantum northern goshawk in the quantum northern goshawk population Get the position of the i-th quantum northern goshawk The number of quantum northern goshawks and quilt The set of quantum northern goshawk positions P dominated by i k , and then count the number of dominated positions of all quantum northern goshawks, if This means that there is no quantum northern goshawk position dominance The non-dominated level of is 1; if but The non-dominated level is 2; by analogy, the non-dominated level of each quantum northern goshawk position is the number of dominated ones plus 1; finally, all quantum northern goshawks are sorted in ascending order according to their non-dominated levels.

5. A multi-UAV disaster rescue planning method according to claim 1, Features: The calculation of the crowdedness of the quantum northern goshawk position in each non-dominated level described in step 5 includes: Assume that there are b quantum northern goshawks in the quantum northern goshawk collection with a non-dominated level of w. They are arranged in ascending order according to the tth objective function value, t = 1, 2, is the tth objective function value of the i-th quantum northern goshawk after sorting, F t max is the maximum value of the tth objective function in the quantum northern goshawk set, F t min is the minimum value of the tth objective function in the quantum northern goshawk set, then The crowding degree of the tth objective function in the northern goshawk ( and ),when or hour, The first The total crowding value of the northern goshawk is Finally, the b quantum northern goshawks are arranged in descending order of crowding degree.

6. A multi-UAV disaster rescue planning method according to claim 1, Features: Step 6 of updating the quantum position of the quantum northern goshawk during the prey identification and attack phase includes: In the prey identification and attack phase, the quantum position of the prey attacked by the i-th quantum northern goshawk is i=1,2,…,K 1 ,r∈{1,2,…,i-1,i+1,…,K 1 }, where r is a random number between [1, N] and r≠i; The quantum position of the quantum northern goshawk is obtained by updating the quantum rotation angle and the quantum revolving gate. The h-dimensional quantum rotation angle of the i-th quantum northern goshawk is in, is the h-dimensional quantum position of the i-th quantum northern goshawk in the k-th iteration, represents the h-th dimensional quantum position of the prey in the k-th iteration, which is the prey caught by the i-th quantum northern goshawk, τ is a random number between [0,1], and χ randomly takes the value 1 or 2; Use the quantum revolving gate to update the h-dimensional quantum position of the i-th quantum northern goshawk individual: Through quantum position Measure the position Then calculate The objective function value of And discuss by category, if Dominate but Keep the result unchanged; otherwise, The details are as follows:

7. A multi-UAV disaster rescue planning method according to claim 1, Features: Step 7 describes updating the quantum position of the quantum northern goshawk during the pursuit phase, including: In the pursuit phase, assuming that the hunting attack radius is R, the h-dimensional quantum rotation angle of the i-th quantum northern goshawk is obtained: The attack radius τ is a random number between [0,1]; Use the quantum revolving gate to update the h-dimensional quantum position of the i-th quantum northern goshawk individual: Then calculate The objective function value of And discuss by category, if Dominate but Keep the result unchanged, otherwise, let The details are as follows:

Citation Information

Patent Citations

  • UAV group task allocation method based on quantum crow group search mechanism

    CN108549402A

  • An unmanned aerial vehicle resource allocation method based on a quantum bird flock evolution mechanism

    CN109190978A

  • Multi-unmanned aerial vehicle task allocation method based on multi-target quantum krill group mechanism

    CN112926825A

  • Unmanned aerial vehicle group task allocation method based on quantum sea lion mechanism

    CN113608546A

  • Method and apparatus for joint optimization of multi-UAV task assignment and path planning

    US10140875B1