基于类脑决策模型的多无人机任务分配

Multi-UAV Task Assignment based on Fusion Improved Reinforcement Learning

摘要:

针对多无人机同时执行多个任务的执行时序不同和无人机任务完成效果以及无人机能耗效率低等问题，建立经验判断-任务预分配-任务重分配-在线决策的类脑决策模型，提出一种基于类脑决策模型的多无人机任务分配算法。融合调度改进组合聚类算法和混合强化学习等策略，将任务分配分为预分配和重分配两个阶段，生成决策网络和评价体系，并引入孤立点离散化策略进行在线决策，旨在解决孤立点的影响，减少状态空间和动作空间的复杂性，提高多无人机系统的分配效率。最后对算法进行了实验验证。

Abstract:

Aiming at the problems of different execution time sequences of multi-UAVs executing multiple tasks at the same time, the completion effect of UAVs tasks and the low energy efficiency of UAVs, a brain-like decision model of experience judgment - task pre-assignment - task reassignment - online decision is established, and a task assignment algorithm based on brain-like decision model is proposed. Fusion scheduling improves strategies such as combinational clustering algorithm and hybrid reinforcement learning, divides task assignment into two stages: pre-assignment and reassignment, generates decision network and evaluation system, and introduces isolated point discretization strategy for online decision-making, aiming to solve the influence of isolated points, reduces the complexity of state space and action space, and improves the allocation efficiency of multi-unmanned aircraft systems. Finally, the algorithm is verified by simulation.