A multi-UAV multi-task allocation method combining improved particle swarm optimization and auction algorithm

Abstract

The present invention relates to a multi-unmanned aerial vehicle (UAV) multi-task allocation method that combines an improved particle swarm algorithm with an auction algorithm. The method first initializes the environment, tasks, and UAVs. Then, the improved particle swarm algorithm is used to pre-plan the UAVs' trajectories. Tasks are then assigned based on the specific tasks and pre-planned trajectory information, combined with the auction algorithm. Finally, a task allocation plan is determined, and the corresponding UAVs are dispatched to perform the corresponding tasks. This method combines the task allocation and trajectory planning problems for research, taking into account the characteristics of UAVs to better reflect actual conditions. By combining the improved particle swarm algorithm with the auction algorithm, the present invention achieves smoother UAV trajectory paths with good convergence.

Description

Technical Field

[0001] The present invention relates to the field of optimal allocation scheme and safe flight technology for unmanned aerial vehicles (UAVs) facing multiple tasks, and in particular to a multi-UAV multi-task allocation method combining an improved particle swarm algorithm and an auction algorithm. Background Art

[0002] Unmanned aerial vehicles (UAVs) are gaining popularity due to their low cost and compact size, rapidly gaining widespread application in applications such as traffic control, power inspections, and disaster relief. However, the capabilities of individual UAVs are limited, and application scenarios can be complex, leading to a growing trend in the field of multi-UAV systems. To ensure both reasonable task allocation and flight safety, a method for multi-UAV multi-task allocation is crucial.

[0003] Xu Ke, Gong Hua, Qin Xinli, Zhang Boyuan. Multi-UAV group task allocation based on distributed auction algorithm. Information and Control [J], 2018, 47(3): 341-346. An auction algorithm with a shared storage center was designed. The total revenue was maximized as the objective function. Tasks were grouped according to type, and a multi-UAV task allocation model was established; Jiang Shuo, Yuan Xiaoping. Application of improved PSO algorithm in multi-UAV collaborative task allocation [J]. Computer Application Research, 2019, 36(11): 3344-3347+3360. The redundant bidding scheme was used to reduce illegal inferior solutions. The improved PSO algorithm with hierarchical classification strategy combined with independent weights was used to solve the task allocation problem and obtain the optimal allocation sequence; X.Wu, Y.Yin, L.Xu, X.Wu, F.Meng and R.Zhen,"MULTI-UAVTask Allocation Based on Improved Genetic Algorithm," in IEEE Journal of Computer Applications, 2019, 36(11): 3344-3347+3360. Access, vol. 9, pp. 100369-100379, 2021, doi: 10.1109 / ACCESS.2021.3097094. This paper integrates an improved simulated annealing algorithm with a genetic algorithm. By raising the acceptance criteria for new solutions in the simulated annealing algorithm, the algorithm changes the way it escapes from local optimal solutions. Furthermore, the UAV task sequence is represented by encoding, and the genetic algorithm's selection operation is improved to a double selection operation, increasing individual diversity and improving algorithm efficiency. Most of the aforementioned studies treat the task allocation problem and the trajectory planning problem as two separate entities. Summary of the Invention

[0004] The present invention organically combines the improved particle swarm optimization algorithm with the auction algorithm, aiming to solve the optimal allocation scheme and safe flight problems of multiple UAVs facing multiple tasks in a three-dimensional environment.

[0005] The technical solution of the present invention:

[0006] A multi-UAV multi-task assignment method is proposed that combines an improved particle swarm optimization algorithm with an auction algorithm. First, the environment, tasks, and UAVs are initialized. Second, the improved particle swarm optimization algorithm is used to pre-plan the trajectory of the UAVs. Then, the auction algorithm is combined with the specific tasks and trajectory pre-planning information to assign tasks. Finally, the task assignment plan is determined and the corresponding UAVs are dispatched to perform the corresponding tasks.

[0007] Preferably, the initialization setting simulates the actual environment to establish a three-dimensional terrain; the threat existing in the environment is a mountain threat, and the mountain is modeled and defined as Threat = {x, y, h, r}, (x, y, h) represents the coordinates of the threat, and r represents the radius of the mountain threat; the drone set is defined as U = {Ux, Uy, Uz}, Ux, Uy, Uz respectively represent the sets of drone position coordinates x, y, z, and the load information ZH of each drone is specified, and Nt represents the number of drones; the task set is defined as R = {Rx, Ry, Rz}, Rx, Ry, Rz respectively represent the sets of position coordinates x, y, z of the task in the three-dimensional environment, and the task attributes include the value Value of executing the task, the threat Risk, and the drone load information HF required to execute the task, and N represents the number of tasks.

[0008] Preferably, an objective function is established before improving the particle swarm algorithm. The objective function when the UAV performs task allocation is Cost = b1×f1+b2×f2+b3×f3+b4×f4, where b1, b2, b3, and b4 are weight coefficients, f1 is the track cost, f2 is the threat cost, f3 is the smoothing cost, and f4 is the task cost.

[0009] Preferably, the shortest track time is selected when the track is selected. Assuming that the speed of the drone remains unchanged, the shortest track is taken as the primary selection criterion, and the drone track X is specified. i It consists of n track points, and the track point P in the three-dimensional environment ij =(x ij ,y ij ,z ij ), the Euclidean distance between two track points is The track cost f1 is expressed as:

[0010] Preferably, the trajectory is required to be shortest and safe, so the trajectory planning requires avoiding obstacles. Assuming that the mountain threat set in the environment is T and it is assumed to be a cylinder, its projection center on the xy plane is C t , with a radius of R t , threat cost and the distance d between the drone and the threatt Proportional to the UAV diameter D and the dangerous distance S between the UAV and the threat, the threat cost f2 is expressed as: in

[0011] Preferably, the smoothing cost is mainly the cost of smoothing the turning angle and climbing angle in the track to make the track feasible; the turning angle is The climb angle is θ ij ,in, is the angle between the projections of two consecutive track segments on the horizontal plane, and Its projection on the XY plane; θ ij is the angle between the track segment and the horizontal plane projection, z ij 、z i,j+1 is the track point P ij 、P i,j+1 The Z-axis coordinate of

[0012] The smoothing cost f3 is expressed as: Where k1 and k2 represent the penalty coefficients of turning angle and climbing angle respectively.

[0013] Preferably, when the drone performs the corresponding task, it can obtain the benefits of the corresponding task, but at the same time it will also be threatened by the target task. According to the task attributes, the task cost f4 can be expressed as: f4 = k3 × Risk - k4 × Value, where k3 and k4 are the corresponding penalty coefficients; Risk is the threat value of the task, and Value is the value of the task.

[0014] Preferably, the improved particle swarm algorithm is different from the two-dimensional plane. In the three-dimensional environment, the flight angle of the drone is considered at the same time, so the spherical coordinate system is used, and the track points are Indicates, where L represents the distance between two track points, θ represents the climb angle, Represents the turning angle; spherical coordinate system The relationship with the rectangular coordinate system (x, y, z) is:

[0015]

[0016]

[0017] z ij =z i,j-1 +L ij cosθ ij .

[0018] The trajectory route P consists of N nodes i It can be expressed in spherical coordinate system as:

[0019] Position Represented as d ij ,speed Expressed as Δd ij , so the improved particle swarm formula can be expressed as: Where k represents the current iteration number; w k represents the inertia weight; c1 and c2 represent the individual self-cognition coefficient and social cognition coefficient, respectively, which determine the tendency of the particle to move toward the local optimum and the global optimum; r 1j and r 2j A random number in [0, 1]; Q i =(q i1 ,q i2 ,...,q iN ), Q g =(q g1 ,q g2 ,...,q gN ) represent the local optimum and global optimum of the particle, respectively.

[0020] Beneficial effects of the present invention:

[0021] Based on practical considerations, this invention combines task allocation and trajectory planning. A rational objective function is established based on the three-dimensional environment and the drone. The auction algorithm and particle swarm optimization algorithm are combined to ensure the rationality of task allocation and the feasibility of trajectory planning. The particle swarm optimization algorithm has a relatively small number of parameters, is simple and flexible, and can be integrated well with other algorithms. The auction algorithm, derived from real-life auction processes, is a classic algorithm for solving task allocation problems. This invention considers the flight characteristics of drones and improves the classic particle swarm optimization algorithm using a spherical coordinate system, eliminating the drone's treatment as a single point mass. After the drone completes its corresponding task, the invention promptly updates its position and payload information to facilitate the evaluation of subsequent tasks.

[0022] This paper combines the task allocation and trajectory planning problems, taking into account the characteristics of drones to better reflect the actual situation. By combining an improved particle swarm optimization algorithm with an auction algorithm, the proposed method achieves smoother and more convergent drone trajectories. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of threat cost;

[0024] Figure 2 Schematic diagram of the turning angle and climbing angle of the UAV;

[0025] Figure 3 It is the algorithm flow chart;

[0026] Figure 4 Assigning track charts to missions;

[0027] Figure 5 This is a bird's-eye view of the track;

[0028] Figure 6 This is the convergence curve diagram for Task 1;

[0029] Figure 7 This is the convergence curve diagram for Task 2;

[0030] Figure 8 This is the convergence curve diagram for Task 3;

[0031] Figure 9 This is the convergence curve diagram for Task 4;

[0032] Figure 10 This is the convergence curve diagram for Task 5. DETAILED DESCRIPTION

[0033] The algorithm flow chart is as follows Figure 3 As shown in the figure. First, the environment, tasks, and drones are initialized. Then, the improved particle swarm algorithm is used to pre-plan the drone's trajectory. Then, based on the specific tasks and pre-planned trajectory information, the auction algorithm is combined to allocate tasks. Finally, the task allocation plan is determined, and the corresponding drone is dispatched to perform the corresponding task. This is specifically reflected in the following points:

[0034] 1.1.1 Initialization

[0035] Simulate the actual environment to create a three-dimensional terrain. The primary threat in the environment is mountainous terrain, so the mountain is modeled and defined as Threat = {x, y, h, r}. Define the drone set U = {Ux, Uy, Uz}, and specify the payload information ZH for each drone. Nt represents the number of drones. Define the task set R = {Rx, Ry, Rz}. Task attributes include the value of executing the task, the threat risk, and the drone payload information HF required to execute the task. N represents the number of tasks.

[0036] 1.1.2 Objective Function

[0037] When selecting the trajectory, we will take the shortest trajectory time as the first selection criterion, assuming that the speed of the drone remains constant. i It consists of n track points. Track point P in the three-dimensional environment ij =(x ij ,y ij ,z ij ), the Euclidean distance between two track points is The track cost f1 is expressed as:

[0038] In addition to the shortest possible trajectory, the trajectory also requires a certain degree of safety, so the trajectory planning requires avoiding obstacles. Assume that the mountain threat set in the environment is T and is assumed to be a cylinder. Its projection on the xy plane is as follows: Figure 1 As shown, the center is C t , with a radius of R t Threat cost and the distance d between the drone and the threat t Proportional to the UAV diameter D and the dangerous distance S between the UAV and the threat, the threat cost f2 is expressed as: in

[0039] The smoothing cost is mainly the cost of smoothing the turning angle and climbing angle in the trajectory to make the trajectory feasible.

[0040] in

[0041] like Figure 2 shown is the angle between the projections of two consecutive track segments on the horizontal plane, and Its projection on the XY plane; θ ij is the angle between the track segment and the horizontal plane projection, z ij 、z i,j+1 is the track point P ij 、P i,j+1 The Z-axis coordinate of .

[0042] The smoothing cost f3 is expressed as: Where k1 and k2 represent the penalty coefficients of turning angle and climbing angle respectively.

[0043] When a drone performs a corresponding task, it can obtain the benefits of the corresponding task, but it will also be threatened by the target task. According to the task attributes, the task cost f4 can be expressed as: f4 = k3 × Risk - k4 × Value, where k3 and k4 are the corresponding penalty coefficients.

[0044] In summary, the objective function of each UAV when performing task allocation is:

[0045] Cost = b1×f1+b2×f2+b3×f3+b4×f4, where b1, b2, b3, and b4 are weight coefficients.

[0046] 1.1.3 Improvement of Particle Swarm Optimization

[0047] Different from the two-dimensional plane, in the three-dimensional environment, the flight angle of the drone is taken into consideration, so the spherical coordinate system is used, and the track points are represented by Indicates, where L represents the distance between two track points, θ represents the climb angle, Represents the turning angle. The relationship between the spherical coordinate system and the rectangular coordinate system is:

[0048] The trajectory route P consists of N nodes i It can be expressed as:

[0049] Position Indicated as d ij ,speed Expressed as Δd ij , so the particle swarm formula can be expressed as: in

[0050] Q i =(q i1 ,q i2 ,...,q iN ), Q g =(q g1 ,q g2 ,...,q gN ) represent the local optimum and global optimum of the particle, respectively.

[0051] 1.2 Experimental plan

[0052] Matlab was used for simulation verification. The population size was set to 500, and the maximum number of iterations was 200. The three-dimensional environment was set to 1000 km * 1000 km * 500 km. The terrain threat parameters in the environment are shown in Table 1, the basic parameters of the UAV are shown in Table 2, and the mission value and threat are shown in Table 3. Each mission is performed by a UAV, and a UAV can perform multiple missions.

[0053] Table 1 Terrain threat parameters

[0054]

[0055] Table 2 Basic parameters of UAV

[0056]

[0057] Table 3 Task parameters

[0058]

[0059] 1.3 Simulation Results

[0060] The simulation results are as follows: the task allocation results are shown in Table 4, and the task allocation trajectory diagram is shown in Figure 4 As shown, Figure 5 is the track top view, and the relevant convergence curve is as follows Figures 6-10 shown.

[0061] Table 4 Task allocation results

[0062]

Claims

1. A multi-UAV multi-task allocation method combining an improved particle swarm optimization algorithm and an auction algorithm, characterized in that: First, the environment, tasks, and drones are initialized. Then, the improved particle swarm algorithm is used to pre-plan the drone's trajectory. Then, based on the specific tasks and pre-planned trajectory information, the auction algorithm is combined to allocate tasks. Finally, the task allocation plan is determined, and the corresponding drone is dispatched to perform the corresponding task. The initialization settings simulate the actual environment to establish a three-dimensional terrain; the threat in the environment is a mountain threat, and the mountain is modeled and defined as Threat = {x, y, h, r}; the drone set is defined as U = {Ux, Uy, Uz}, and the payload information ZH of each drone is specified, where Nt represents the number of drones; the task set is defined as R = {Rx, Ry, Rz}, and the task attributes include the value Value of executing the task, the threat Risk, and the drone payload information HF required to execute the task, where N represents the number of tasks; Before improving the particle swarm algorithm, the objective function is established. The objective function of the UAV for task allocation is Cost = b1×f1+b2×f2+b3×f3+b4×f4, where b1, b2, b3, and b4 are weight coefficients, f1 is the track cost, f2 is the threat cost, f3 is the smoothing cost, and f4 is the task cost. When selecting a track, the shortest track time is selected. Assuming that the speed of the drone remains unchanged, the shortest track is taken as the primary selection criterion. The drone track X is specified. i It consists of n track points, and the track point P in the three-dimensional environment ij =(x ij ,y ij ,z ij ), the Euclidean distance between two track points is The track cost f1 is expressed as: In addition to the shortest possible trajectory, the trajectory must also be safe, so the trajectory planning requires avoiding obstacles. Assume that the mountain threat set in the environment is T and it is assumed to be a cylinder, and its projection center on the xy plane is C. t , with a radius of R t , threat cost and the distance d between the drone and the threat t Proportional to the UAV diameter D and the dangerous distance S between the UAV and the threat, the threat cost f2 is expressed as: in The smoothing cost is mainly the cost of smoothing the turning angle and climbing angle in the track to make the track feasible; the turning angle is The climb angle is θ ij ,in, is the angle between the projections of two consecutive track segments on the horizontal plane, and Its projection on the XY plane; θ ij is the angle between the track segment and the horizontal plane projection, z ij 、z i,j+1 is the track point P ij 、P i,j+1 The Z-axis coordinate of The smoothing cost f3 is expressed as: Where k1 and k2 represent the penalty coefficients of turning angle and climbing angle respectively.

2. The multi-UAV multi-task allocation method combining an improved particle swarm optimization algorithm and an auction algorithm according to claim 1 is characterized in that: When a drone performs a corresponding task, it can obtain the benefits of the corresponding task, but it will also be threatened by the target task. According to the task attributes, the task cost f4 can be expressed as: f4 = k3 × Risk - k4 × Value, where k3 and k4 are the corresponding penalty coefficients; Risk is the threat value of the task, and Value is the value of the task.

3. The multi-UAV multi-task allocation method combining an improved particle swarm optimization algorithm and an auction algorithm according to claim 2 is characterized in that: Improved particle swarm algorithm, different from the two-dimensional plane, in the three-dimensional environment, taking into account the flight angle of the UAV, so the spherical coordinate system is used, and the track points are Indicates, where L represents the distance between two track points, θ represents the climb angle, Represents the turning angle; spherical coordinate system The relationship with the rectangular coordinate system (x, y, z) is: The trajectory route P consists of N nodes i It can be expressed in spherical coordinate system as: Position Indicated as d ij ,speed Expressed as Δd ij , so the improved particle swarm formula can be expressed as: Where k represents the current iteration number; w k represents the inertia weight; c1 and c2 represent the individual self-cognition coefficient and social cognition coefficient, respectively, which determine the tendency of the particle to move toward the local optimum and the global optimum; r 1j and r 2j A random number in [0, 1]; Q i =(q i1 ,q i2 ,...,q iN ), Q g =(q g1 ,q g2 ,...,q gN ) represent the local optimum and global optimum of the particle, respectively.

Images