Barrier Coverage using UAVs and Camera Sensors

Abstract

With advancements in camera sensor technology, camera sensor networks are increasingly utilized for border surveillance. UAVs equipped with downward-facing cameras also serve as effective sensors, and their features, such as rapid deployment and adjustable field of view, make them particularly suitable for border surveillance, often termed as barrier coverage in the literature. This thesis first addresses a barrier coverage problem using UAVs equipped with downward-facing cameras. It proposes both deterministic and optimization-based deployment strategies to ensure barrier coverage. In deterministic deployment, UAVs are initially aligned to a barrier line to maximize coverage based on the number of sensors and height constraints. For optimization-based deployment, a resolution cost of the belt is introduced to enhance the quality of existing barrier coverage. An optimization problem is also proposed to achieve barrier coverage with an overlapping constraint for UAVs placed arbitrarily within the belt. The approach is further demonstrated to be applicable to borders of any shape by considering a multi-belt problem. Additionally, a local fault-tolerance model is proposed to ensure continuous coverage if some UAVs become faulty. Also, the barrier coverage of a belt with varying resolution requirements is investigated. For a barrier covered network with certain regions within the belt that need higher resolutions, an optimization problem is formulated to determine the final placement of UAVs ensuring barrier coverage with the requirements. When handling vision-based sensors for tasks such as intruder detection, it is essential to account for occlusion caused by objects, as it can create a false sense of coverage. This thesis addresses the impact of occluders on barrier coverage networks involving UAVs. The study assesses whether a given UAV network, along with permeable and impermeable occluders in the belt, achieves barrier coverage. For permeable occluders, a dual graph approach, one along the length and one along the width of the belt is introduced to evaluate barrier coverage. Further, coverage metrics to distinguish any two barrier-covered networks are proposed. Overall, this thesis provides a comprehensive study of barrier coverage problems with UAVs, considering both the presence and absence of occluders. The thesis also tackles the barrier coverage problem for terrain-like borders using UAVs where achieving optimal UAV placement is challenging due to factors such as resolution, overlap constraints, and varying altitudes. We first simplify the 3D problem into an equivalent 2D model and introduce a resolution cost to assess terrain coverage quality. Additionally, we define the overlapping length and formulate an optimization problem to secure barrier coverage. In the second part of this thesis, the barrier coverage problem with camera sensor networks is investigated. For a network identified as barrier-uncovered, an optimization problem is formulated to determine the optimal positions and orientations of each sensor to ensure barrier coverage. An attraction force-based motion strategy is then used to relocate the sensors to the desired positions and orientations. Similar to UAVs, the impact of occluders on camera sensor networks is also considered. The limitations of conventional methods are highlighted, and a coverage model is proposed that introduces a novel sector division approach, utilizing the interaction between occluders and sensor regions. Additionally, the existing graph-based method is modified to assess the barrier coverage of a sensor network in the presence of occluders. Additionally, deployment strategies for ensuring barrier coverage when it is initially lacking are discussed. First, a deterministic approach using a barrier curve derived from a weighted graph is presented. Finally, an optimization-based deployment strategy, utilizing the sector division method for barrier coverage constraints, is proposed to ensure coverage in the presence of occluders.