UAV swarm task allocation algorithm based on the alternating direction method of multipliers network potential game theory

Compared with a single unmanned aerial vehicle (UAV), a large-scale UAV swarm can accomplish the unavailable, complex, and “1 + 1 > 2” tasks of traditional UAVs. To prevent the UAV swarm from falling into the dilemma of disorganized derailment and mission failure, higher requirements for the robustness and organizational scheduling capability of the UAV swarm were proposed. As one of the important components of the autonomous cooperative control technology of UAV swarms, task allocation refers to certain environmental situation information and UAV swarm status to maximize the overall efficiency of the swarm. To solve the task allocation problem of the UAV swarm, a UAV swarm task allocation algorithm based on the alternating direction method of multipliers (ADMM) network potential game theory was proposed. The ADMM is a typical algorithm that uses the idea of “divide and conquer.” The ADMM adopts the decomposition–coordination process, which coordinates the solutions of each subproblem step by step to determine the global optimum. In terms of problem modeling and algorithm design, the network potential game theory can solve the conflict and cooperation between multiple agents effectively. By combining the advantages of the ADMM and network potential game theory, UAV swarm task allocation can be divided into two parts: local and global benefits optimization. Firstly, considering the different resource constraints and execution capability factors of the UAV swarm, the task allocation problem was formulated as the problem of finding a minimum under inequality constraints, and the game model of the UAV swarm task allocation problem was constructed based on the network potential game theory. Based on the game model of UAV swarm task allocation, the equivalence of the optimum UAV swarm task allocation strategy and the Nash equilibrium solution of the evolutionary network was analyzed. Secondly, according to the UAV capability and task set characteristics, the local optimum execution efficiency of each UAV was determined using the ADMM. Moreover, each UAV was defined as a rational player, the local benefit maximization task combination of each UAV was used as the initial task allocation scheme, and the task allocation problem was transformed and solved by using the Nash equilibrium solution of the network potential game. Each UAV adjusts its strategy based on the information on the interaction between individuals in the neighborhood to maximize the global task benefits. Finally, the simulation experiments verified that the proposed UAV swarm task allocation algorithm can converge to the optimal solution stably within a limited step and assign all task target points without conflict. The feasibility and effectiveness of the method were also verified. The comprehensive verification platform for the 3D simulation process of UAV swarm task allocation and execution was given in the form of real-time deduction.