论文标题
部分可观测时空混沌系统的无模型预测
Mission planning for emergency rapid mapping with drones
论文作者
论文摘要
我们介绍了一个任务计划概念，用于通过一组采样位置在诸如火灾或化学事故之类的事件之后，通过一组抽样位置将无人驾驶汽车（UAV）路由。使用插值方法来说明被调查现象固有的空间相互依赖性，这些样本允许预测危险物质在整个受影响区域的分布。我们定义了广义相关的团队定向式问题（GCORTOP），用于选择{信息性}样品考虑观察到的位置和未观察到的位置之间的空间相关性以及被调查区域中的优先级。为了快速在时间敏感的情况下提供高质量的解决方案，我们提出了两阶段的多开始自适应大型邻里搜索（2ML）。我们使用基准实例来表明解决方案方法的竞争力，并根据新介绍的“任务计划问题”基于基准实例进行了广泛的研究，并研究了拟议模型和解决方案方法的性能。
We introduce a mission planning concept for routing unmanned aerial vehicles (UAVs) through a set of sampling locations in the immediate aftermath of an incident such as a fire or chemical accident. Using interpolation methods that account for the spatial interdependencies inherent in the surveyed phenomenon, these samples allow predicting the distribution of hazardous substances across the affected area. We define the generalized correlated team orienteering problem (GCorTOP) for selecting {informative} samples considering spatial correlations between observed and unobserved locations as well as priorities in the surveyed area. To quickly provide high-quality solutions in time-sensitive situations, we propose a two-phase multi-start adaptive large neighborhood search (2MLS). We show the competitiveness of the solution approach using benchmark instances for the team orienteering problem and investigate the performance of the proposed models and solution approach in an extensive study based on newly introduced benchmark instances for the mission planning problem.
