一种基于风险感知策略的多节点任务调度方法

为解决无人机(UAV)集群任务调度时面临各节点动态、不稳定的情况，该文提出一种面向多计算节点的可尽量避免任务中断且具有容错性的任务调度方法。该方法首先为基于多计算节点构建了一个以最小化任务平均完成时间为优化目标的任务分配策略；然后基于任务的完成时间和边缘计算节点的存留时间两者的概率分布，将任务计算节点上的执行风险量化成额外开销时间；最后以任务的完成时间与额外开销时间之和替换原本的完成时间，设计了风险感知的任务分配策略。在仿真环境下将该文提出的任务调度方法与3种基准调度方法进行了对比实验，实验结果表明该方法能够有效地降低任务平均响应时间、任务平均执行次数以及任务截止时间错失率。证明该文提出的方法降低了任务重调度和重新执行带来的额外开销，可实现分布式协同计算任务的调度工作，为复杂场景下的无人机集群网络提供新的技术支持。

A Multi-node Task Scheduling Method via Risk Perception Strategy

In order to solve the dynamic and unstable situation of each node in Unmanned Aerial Vehicle(UAV)cluster task scheduling, a multi computing node oriented task scheduling method is proposed, which can avoid task interruption as much as possible and has fault tolerance. Firstly, a task allocation strategy based on multi computing nodes is constructed to minimize the average completion time of tasks. Secondly, based on the probability distribution of the completion time of tasks and the retention time of edge computing nodes, the execution risk on task computing nodes is quantified as the extra overhead time. Finally, the sum of the completion time and the extra overhead time of tasks is calculated instead of the original completion time, a risk aware task allocation strategy is designed. In the simulation environment, the proposed task scheduling method is compared with three benchmark scheduling methods. The experimental results show that the proposed method can effectively reduce the average response time, the average execution times and the miss rate of task deadline. It is proved that the proposed method can reduce the additional cost of task rescheduling and re-execution, realize the task scheduling of distributed collaborative computing, and provide new technical support for UAV cluster network in complex scenarios.