arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2008.01144eess.SPcs.DC
收藏

软件定义空地一体化车载网络中能量感知图任务调度

Energy-Aware Graph Task Scheduling in Software-Defined Air-Ground Integrated Vehicular Networks

Minghui LiWang, Zhibin Gao, Xianbin Wang

摘要:

软件定义的空地一体化无人机(SD-AGV)网络是一种很有前途的网络模式,它可以实现灵活的地面资源共享,以支持具有高计算开销的无人机创新应用。在本文中,我们研究了SD-AGV网络中的车辆云辅助任务调度问题,其中无人机承载的计算密集型任务和车辆云通过基于图的表示建模。为了将图任务的每个组件映射到可行的车辆,同时实现最小化无人机的任务完成时间,能量消耗和移动车辆之间的数据交换成本之间的权衡,我们将问题表示为混合整数非线性规划问题,这是NP-难的。此外,与保持任务结构相关联的约束提出了解决动态车辆拓扑上的子图同构问题,这进一步使算法设计复杂化。出于这一动机,我们提出了一种有效的解耦方法,通过分离的模板(组件和车辆之间的可行映射)搜索的传输功率分配。对于前者,我们提出了一个计算复杂度较低的搜索所有同构子图的有效算法。对于后者,我们引入了一个功率分配算法,通过应用$p$ -范数和凸优化技术。大量的仿真表明,该方法优于基准方法考虑各种问题的大小。

英文摘要:

The Software-Defined Air-Ground integrated Vehicular (SD-AGV) networks have emerged as a promising paradigm, which realize the flexible on-ground resource sharing to support innovative applications for UAVs with heavy computational overhead. In this paper, we investigate a vehicular cloud-assisted task scheduling problem in SD-AGV networks, where the computation-intensive tasks carried by UAVs, and the vehicular cloud are modeled via graph-based representation. To map each component of the graph tasks to a feasible vehicle, while achieving the trade-off among minimizing UAVs' task completion time, energy consumption, and the data exchange cost among moving vehicles, we formulate the problem as a mixed-integer non-linear programming problem, which is Np-hard. Moreover, the constraint associated with preserving task structures poses addressing the subgraph isomorphism problem over dynamic vehicular topology, that further complicates the algorithm design. Motivated by which, we propose an efficient decoupled approach by separating the template (feasible mappings between components and vehicles) searching from the transmission power allocation. For the former, we present an efficient algorithm of searching for all the isomorphic subgraphs with low computation complexity. For the latter, we introduce a power allocation algorithm by applying $p$-norm and convex optimization techniques. Extensive simulations demonstrate that the proposed approach outperforms the benchmark methods considering various problem sizes.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑