Visibility-based UAV path planning for surveillance in cluttered environments

Abstract

This paper focusses on the problem of determining near-optimal observation locations for an effective close-range UAV surveillance in terrains cluttered with buildings and trees. Use of Small-Unmanned Aerial Vehicles (S-UAVs) in civil defence applications has increased due to their portability and low operational costs. In close-range S-UAV surveillance in cluttered environments, there are two significant occlusions to visibility: complete (terrain) and partial (vegetation). However, in the existing literatures, the partial occlusions are generally neglected. In this paper, a probabilistic visibility model is proposed which considers both complete and partial occlusions to determine near-optimal surveillance path to enhance visibility of the desired regions on the ground using a two-step approach. In the first step, the waypoints are deployed in regions which provide near-uniform visibility of the desired target regions. This step involves finding the visibility space (region of space from which the desired target regions are visible) using the Fast Marching Method (FMM) and then deploying the waypoints in this region using Centroidal Voronoi tessellation (CVT). In the second step, flyable paths are constructed along the waypoints using an improved clustered spiral-alternating algorithm. Visibility with the proposed method is simulated for a synthetically generated terrain that resembles a residential area with buildings and trees. The results show the effectiveness of the proposed surveillance method in improving the visibility of the desired target regions.