This paper addresses the sensor-placement problem (SPP) within the context of discretizing large, complex continuous 2D environments into graphs for efficient task-oriented route planning. The SPP aims to minimize the number of sensors required to achieve a user-defined coverage ratio while considering a general visibility model. We propose the hybrid filtering heuristic (HFH) framework, which enhances or combines outputs of existing sensor-placement methods, incorporating a filtering step. This step eliminates redundant sensors or those contributing marginally to the coverage, ensuring the coverage ratio remains within the desired interval. We implement two versions of HFH: the basic version and a variant, HFHB, incorporating a preprocessing technique known as bucketing to accelerate region clipping. We evaluate HFH and HFHB on a dataset of large, complex polygonal environments, comparing them to several baseline methods under both unlimited and limited-range omnidirectional visibility models. The results demonstrate that HFH and HFHB outperform baselines in terms of the number of sensors required to achieve the desired coverage ratio. Additionally, HFHB significantly reduces the runtime of more competitive baseline methods. We also adapt HFHB to a visibility model with localization uncertainty, demonstrating its effectiveness up to a certain level of uncertainty.

翻译：本文研究了传感器布局问题（SPP），其背景是将大型、复杂的连续二维环境离散化为图，以实现高效的任务导向路径规划。SPP旨在最小化达到用户定义覆盖率所需的传感器数量，同时考虑一个通用的可见性模型。我们提出了混合过滤启发式（HFH）框架，该框架通过引入过滤步骤来增强或结合现有传感器布局方法的输出。该步骤剔除冗余传感器或对覆盖率贡献微小的传感器，确保覆盖率保持在期望区间内。我们实现了两个版本的HFH：基础版本以及一个变体HFHB，后者采用称为分桶的预处理技术以加速区域裁剪。我们在一个包含大型复杂多边形环境的数据集上评估了HFH和HFHB，并在无限范围和有限范围的全向可见性模型下，将其与多种基线方法进行了比较。结果表明，在达到期望覆盖率所需的传感器数量方面，HFH和HFHB均优于基线方法。此外，HFHB显著降低了更具竞争力的基线方法的运行时间。我们还使HFHB适应于存在定位不确定性的可见性模型，证明了其在特定不确定性水平下的有效性。