基于KnCMPSO算法的异构无人机协同多任务分配

随着无人机(Unmanned aerial vehicle, UAV)技术的广泛应用和执行任务的日益复杂,无人机多机协同控制面临着新的挑战.以无人机总飞行距离和任务完成时间为优化目标,同时考虑异构无人机类型、任务执行时序等多种实际约束,构建基于多种约束条件的异构无人机协同多任务分配模型.该模型不仅包含混合变量,同时还存在多个复杂的约束条件,因此,传统的多目标优化算法并不能有效地处理混合变量及对问题空间进行搜索并生成满足多种约束条件的可行解.为高效求解上述模型,提出一种基于拐点的协同多目标粒子群优化算法(Knee point based coevolution multi-objective particle swarm optimization,Kn CMPSO),该算法引入基于拐点的学习策略来更新外部档案集,在保证收敛性的同时增加种群的多样性,使算法能搜索到更多可行的任务分配结果;并基于二进制交叉方法,引入基于学习的粒子更新策略来提升算法的收敛性及基于区间扰动的局部搜索策略以提升算法的多样性.最后通过在四组实例上的仿真实验验证了所提算法在求解异构无人机协同多任务分配问题上的有效性.

A Knee Point Based Coevolution Multi-objective Particle Swarm Optimization Algorithm for Heterogeneous UAV Cooperative Multi-task Allocation

With the wide application of unmanned aerial vehicle (UAV) technology and the increasing complexity of UAV tasks, the multi-aircraft cooperative control on UAVs faces new challenges. In this paper, a heterogeneous UAV cooperative multi-task allocation model which takes the UAV total flight distance and task completion time as the optimization objectives is set up. The model includes mixed variables and multiple complex constraints, such as UAV type and task execution time. As a result, traditional multi-objective optimization algorithms cannot effectively search the problem space and generate feasible solutions. In this paper, a knee point based cooperative multi-objective particle swarm optimization algorithm, namely knee point based coevolution multi-objective particle swarm optimization (KnCMPSO), is proposed to solve the above model. In KnCMPSO, a knee point based learning strategy is employed to update the external archive set, which can help get more better solutions. The learning-based particle update strategy is proposed to improve the convergence and the interval disturbance based local search strategy is used to enhance the diversity. The experimental results on four sets of examples show that, the proposed KnCMPSO algorithm can solve the heterogeneous UAVs collaborative multi-task allocation problem more effectively than other existing methods.