An unmanned aerial vehicle group task allocation method based on a quantum annealing algorithm model

Abstract

The present application relates to a kind of unmanned aerial vehicle group task allocation method based on quantum annealing algorithm model, comprising the following steps: firstly confirm the unmanned aerial vehicle quantity required for unmanned aerial vehicle group task allocation, the position point of flight task;Then by increasing virtual position point, it is guaranteed that each unmanned aerial vehicle has flight task, and unmanned aerial vehicle group task allocation problem is converted into single salesman problem, the problem is quantized, so that the problem is converted into the system that can be characterized using quantum bit;Finally, using the target Hamiltonian set, the simulation quantum annealing evolution process is completed, and the unmanned aerial vehicle group task allocation result is obtained.The slow solving time of the intelligent optimization algorithm of traditional unmanned aerial vehicle group task allocation, the problem that global optimal solution cannot be found is solved, by using the combination of quantum annealing algorithm and unmanned aerial vehicle group task allocation problem, after the problem is quantized, the corresponding problem target Hamiltonian is constructed, and the optimal solution corresponding system quantum state is solved by adiabatic evolution.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence and quantum computing, in particular to a quantum annealing algorithm model applied to unmanned aerial vehicle group task allocation. The traditional unmanned aerial vehicle group task allocation is realized by intelligent optimization algorithm in the field of artificial intelligence, and the quantum annealing algorithm belongs to the field of quantum computing. Here, artificial intelligence and quantum computing are ingeniously combined to construct a quantum algorithm model that can be applied to unmanned aerial vehicle group task allocation. BACKGROUND

[0002] With the continuous development of information technology, the application of unmanned aerial vehicles is expanding. In civil use, it can realize forest monitoring, environmental protection, scientific irrigation, etc. Especially in military use, it can complete information reconnaissance, electromagnetic interference, precision strike, etc. However, in the face of more complex combat environment and diversified tasks, a single unmanned aerial vehicle will expose the problems of insufficient endurance, load capacity and computing capacity, and cannot complete complex task requirements. At this time, it is necessary to form an unmanned aerial vehicle group to work together to complete the combat task.

[0003] The traditional algorithm for unmanned aerial vehicle group task allocation is mainly intelligent combinatorial optimization algorithm, including genetic algorithm, particle swarm optimization algorithm, ant colony optimization algorithm, etc. In the problem of heterogeneous unmanned aerial vehicle group task allocation, Wang Ting designed a cyclic mutation operator and a chromosome gene position crossover operator according to the actual application scene, which accelerated the search efficiency of the algorithm for the optimal solution. Zhang Ruipeng designed an improved particle swarm optimization algorithm, which can consider the flight distance, flight time and task benefit of unmanned aerial vehicles, and improved the performance of the traditional particle swarm optimization algorithm. Su Meimei used the ant colony optimization algorithm, and added time, position, priority, etc. in the application scene, which further expanded the practicality of the ant colony optimization algorithm. However, these traditional algorithms only optimize the accuracy of the algorithm and expand the functional diversity, but when facing the expansion of the scale of the unmanned aerial vehicle group and the diversity of tasks, the computational complexity will increase sharply, the calculation speed will decrease rapidly, and the optimization result will frequently appear in the local optimal solution, which cannot effectively find the global optimal solution.

[0004] The traditional unmanned aerial vehicle group task allocation is solved by using intelligent optimization algorithm. For example, genetic algorithm, particle swarm optimization algorithm and ant colony optimization algorithm. However, these algorithms can only solve small-scale problems. When encountering large-scale problems, the solution time will increase rapidly. At the same time, these algorithms have a certain probability of falling into the local optimal solution trap in the solving process, and cannot find the global optimal solution. SUMMARY

[0005] Aiming at the problem that the traditional intelligent optimization algorithm for solving the task allocation of UAV group is slow in solving time and cannot find the global optimal solution, a solution using quantum algorithm is proposed. By combining quantum annealing algorithm with the task allocation problem of UAV group, a quantumized UAV group task allocation model is constructed. The super strong parallel computing capability of quantum algorithm is fully utilized to improve the calculation speed of large-scale problems. The global optimal solution is searched by using quantum tunneling effect to penetrate local barriers. After the problem is quantized, the corresponding problem Hamiltonian is constructed, and the optimal solution is solved by adiabatic evolution.

[0006] The technical scheme of the application is:

[0007] A UAV group task allocation method based on quantum annealing algorithm model, comprising the following steps:

[0008] Firstly, the number of UAVs required for UAV group task allocation and the position points of flight tasks are confirmed;

[0009] Then, virtual position points are added to ensure that each UAV has a flight task, and the UAV group task allocation problem is converted into a single traveling salesman problem, and the problem is quantized to convert the problem into a system that can be represented by quantum bits;

[0010] Finally, the simulation quantum annealing evolution process is completed by using the set target Hamiltonian, and the UAV group task allocation result is obtained.

[0011] Further, the method comprises the following steps:

[0012] Suppose that there are n UAVs, n is a positive integer, m+1 position points, an origin point, m is a task point, P0 is used as the starting point, (P1,…,P m ) represents m position points, n UAVs perform tasks, n-1 virtual position points (P m+1 ,…,P m+n-1 ) are added, (P0,P m+1 ,…,P m+n-1 ) represent starting points, and the distance between the n points is set to be infinite when calculating the distance, so that each UAV has a flight task; the m+n points are traversed as P0…P i …P m+1 …P j …(P m+2 …P m+n-1 )…P k …P0, so that the n starting points (P0,P m+1 ,…,P m+n-1 ) divide the path into n segments, and each segment is a task path executed by a UAV.

[0013] Each quantum bit in the system will collapse to a specific |0> state or |1> state; since the quantum bit can only be in the |0> or |1> state after measurement, it is defined that:

[0014]

[0015] Where x t,i Indicates whether the UAV is at position i at time t, and takes the value 1 at position i and 0 at position i;

[0016] In order to make the problem valid and accurate mathematical expression, the following two constraints need to be met: the UAV can only be at one position point at each time; each position point can only be traversed once;

[0017] Where the UAV can only be at one position point at each time, and x t,i is defined as: Each position can only be traversed once, which is mathematically expressed as:

[0018] The total distance of the problem is:

[0019] Where d is the total distance, N=(m+n+1) is the total number of positions, and l i,j Indicates the distance between position i and position j;

[0020] Therefore, the target Hamiltonian H of the system is:

[0021]

[0022] Where H represents the target Hamiltonian of the system, and alpha and beta are adjustment parameters of the constraint condition, which must be positive;

[0023] The quantum annealing algorithm for the UAV group is developed by simulating the quantum annealing evolution process using the set target Hamiltonian, and the optimal path assignment result is obtained.

[0024] Preferably, the quantum annealing algorithm is simulated and calculated by using the PYQUBO software package to obtain the optimal path assignment result.

[0025] Preferably, the flight path is converted into a legend form according to the data.

[0026] The beneficial effects of the present application are:

[0027] Due to the inherent properties of quantum state evolution of the system, i.e. super strong parallel computing capability and tunneling effect, the improved model has the following two advantages compared with the existing model:

[0028] It boasts high computational speed and, through an adiabatic evolution process, its evolution time does not increase rapidly with the size of the problem, effectively handling large-scale task allocation problems.

[0029] The calculation results are more reliable. Since the model uses the principle of quantum annealing, the quantum tunneling effect can be effectively utilized by appropriately adjusting the parameters during the evolution process, so that the system evolution can penetrate the barrier, escape the local optimum trap, and find the global optimum solution. Attached Figure Description

[0030] Figure 1 This invention provides a flight path planning diagram for three unmanned aerial vehicles (UAVs) facing ten mission points.

[0031] Figure 2 The image shows the result obtained after running the program of this invention. Detailed Implementation

[0032] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0033] First, we need to introduce the issue of task allocation for drone swarms, such as... Figure 1 As shown, the solid black dots represent the starting point (0) of the drone swarm, and the black circles represent the task locations (1-10). Assuming there are 3 drones and 10 task locations, the goal is to find the flight paths of the three drones that minimize the total flight distance. This example uses... Figure 1 The document details the implementation of the three drones and ten mission points. For other drone swarm mission allocation issues, only the number of mission points and drone sorties need to be adjusted.

[0034] The drone swarm task allocation problem described above can be equated to the multi-person traveling salesman problem (TSM). The TSM problem involves a person starting from a point, visiting several cities, and returning to the starting point. The goal is to choose the path that minimizes the total distance traveled, visiting each city once. The multi-person TSM problem involves multiple people starting simultaneously, each visiting a portion of the cities, and returning to the starting point. The goal is to ensure that every city is visited by at least one person, minimizing the total distance traveled. Our drone swarm task allocation problem can be transformed into a multi-person TSM problem. Solving the multi-person TSM problem can be further simplified into a single-person TSM problem. For example... Figure 1 As shown, there are a total of 11 location points (one starting point and 10 task points), 3 drones, and (P0, P1, ..., P...) 10 This indicates 11 location points, with 3 drones performing the mission, adding 2 (3-1=2) virtual location points (P).11 ,P 12 ), (P0,P 11 ,P 12 The three points represent the starting positions. When calculating the distance, the distance between these three points is set to infinity (in programming, you only need to set the distance between the three points to be much greater than the distance between the other 10 task points; here we set it to |P0P). 10 |=100max(|P i P j This ensures that every drone has a flight mission. By adding virtual location points, the multi-person traveling salesman problem is transformed into a single-person traveling salesman problem. Specifically, traversing all 13 points once results in P0…P… i …P 11 …P j …P 12 …P k …P0, so (P0,P 11 ,P 12 The three starting points divide the path into three segments, each of which is a task path executed by a drone.

[0035] After transforming the drone swarm problem, the next step is to quantize it, transforming it into a problem that can be characterized using a qubit system. Since a quantum system composed of multiple qubits collapses after measurement, changing from a superposition state to a definite quantum state, each qubit in the system collapses into either a |0> or |1> state. Because a qubit can only be in either a |0> or |1> state after measurement, we define:

[0036]

[0037] Where, x t,i This indicates whether the drone is at position i at time t. If it is at position i, the value is 1; otherwise, the value is 0.

[0038] For a problem to be expressed mathematically effectively and accurately, the following two constraints must be met:

[0039] (1) At any given moment, the drone can only be at one location.

[0040] (2) Each location point can only be traversed once.

[0041] At any given moment, the drone can only be at one location point, denoted by the defined x. t,i Represented as: Each position can only be traversed once, mathematically represented as:

[0042] The total distance of this problem is:

[0043] Where d is the total distance, N is the total number of positions 13, l i,j represents the distance between position i and position j.

[0044] So the target Hamiltonian H of the system is:

[0045]

[0046] Where H represents the target Hamiltonian of the system, and alpha and beta are adjustment parameters of the constraint condition (usually between 0-100, and the appropriate value can be adjusted according to the program running result), which must be positive, because alpha and beta are multiplied by a constraint condition, so the larger the parameter value, the more the constraint condition needs to be satisfied in the solving process of the system.

[0047] Here we use the PyQUBO software package developed by DWave company to complete the quantum annealing evolution process, and develop a quantum annealing algorithm for UAV group using the target Hamiltonian set by the invention. The result after running the program is as shown in Figure 2

[0048] Where c[t][i] represents x t,i We extract all the data from 0 time to 13 time that is 1:

[0049] x 0,3 =1, x 1,0 =1, x 2,4 =1, x 3,5 =1, x 4,6 =1, x 5,11 =1, x 6,10 =1, x 7,9 =1, x 8,8 =1, x 9,7 =1, x 10,12 =1, x 11,1 =1, x 12,2 =1, and other values are 0. According to the data flight path is:

[0050] 3→0→4→5→6→11→10→9→8→7→12→1→2, so after sorting, 0, 11, and 12 are all starting points 0, we adjust the time from the starting point, then the flight path circular clockwise and counterclockwise is the same, so the flight paths of the three UAVs are 0→1→2→3→0, 0→4→5→6→0, 0→7→8→9→10→0, respectively. The result is as shown in Figure 1

[0051] ​​For more general cases, i.e. n drones (n is a positive integer), m+1 location points (one origin, m is the task point), the above case can be generalized. Using P0 as the starting point, (P1,..., P m ) represents m location points, n drones perform tasks, and n-1 virtual location points (P m+1 ,..., P m+n-1 ) are added. (P0, P m+1 ,..., P m+n-1 ) all represent the starting point position, and the distance between the n points is set to be infinite when calculating the distance (only need to set the distance between n points to be much larger than the distance between the other m+1 real points in programming, we set |P0P m+1 | = 100xmmax(|P i P j |)) in this way, it is guaranteed that each drone has a flight task. By adding virtual location points, the multiple traveling salesman problem is converted into a single traveling salesman problem. Traverse the m+n points as P0...P i ...P m+1 ...P j ... (P m+2 ...P m+n-1 )...P k ...P0, so that the n starting points (P0, P m+1 ,..., P m+n-1 ) divide the path into n segments, and each segment is a task path executed by a drone. For the general case of the system target Hamiltonian, the expression is the same as above. The only difference is that the total number of locations N = (m+n+1). Then, using the PYQUBO quantum annealing algorithm to calculate the target Hamiltonian, the optimal result of the path assignment under the general case can be obtained.

[0052] The above-described embodiments only express one embodiment of the present application, which is described in detail and in detail, but it cannot be understood as limiting the scope of the patent. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A method for task allocation in a drone swarm based on a quantum annealing algorithm model, characterized in that, Includes the following steps: First, confirm the number of drones required for drone swarm mission allocation and the location of the flight mission; Then, by adding virtual location points, it was ensured that each drone had a flight mission, transforming the drone swarm mission allocation problem into a single-person traveling salesman problem. This problem was then quantized, making it so that it could be characterized using a quantum bit system. Finally, using the set target Hamiltonian, the simulated quantum annealing evolution process is completed, and the task allocation results of the drone swarm are obtained. Specifically, the following steps are included: Assume you have n drones, where n is a positive integer, m+1 location points, an origin, and m is the mission point. As the starting point, ( () represents m location points, n drones performing a mission, and n-1 virtual location points added. , , ), ( , The points (m+n) represent the starting and ending points. When calculating distances, the distance between these n points is set to infinity to ensure that each drone has a flight mission. Traversing the m+n points once... Thus, by ( , The path is divided into n segments by n starting points, and each segment is the mission path executed by a drone. Each qubit in the system will collapse into a specific state or State; because a quantum bit can only be a state after measurement. or State, therefore defined as: in, This indicates whether the drone is at position i at time t. If it is at position i, the value is 1; otherwise, the value is 0. To achieve an effective and accurate mathematical expression of the problem, the following two constraints must be met: the drone can only be at one location point at any given time; and each location point can only be traversed once. At any given moment, the drone can only be at one location point, as defined... Represented as: Each position can only be traversed once, mathematically represented as: The total distance in this problem is: Where d is the total distance, and N = (m + n + 1) is the total number of positions. This represents the distance between position i and position j; Therefore, the target Hamiltonian of the system yes: Where H represents the target Hamiltonian of the system. and The adjustment parameter for the constraint condition must be a positive number; The quantum annealing evolution process was simulated, and a quantum annealing algorithm for UAV swarms was developed using a set target Hamiltonian to obtain the optimal path assignment result.

2. The UAV swarm task allocation method based on the quantum annealing algorithm model according to claim 1, characterized in that, The PYQUBO software package was used to simulate the quantum annealing algorithm to obtain the optimal path assignment result.

3. The UAV swarm task allocation method based on the quantum annealing algorithm model according to claim 1, characterized in that, The flight path is converted into a legend based on the data.

Images