Motion Planning and Visual-Inertial Target Tracking for UAV-based Radiation Detection

This paper addresses the problem of detecting radioactive material in transit using an unmanned aerial vehicle (UAV) of minimal sensing capability, where the objective is to classify the target's radioactivity as the vehicle plans its paths through the workspace while tracking the target for a short time interval. Assuming that the prior map of the workspace is available, this paper proposes a motion planning framework that integrates a navigation function based planner with a tightly-coupled visual-inertial localization and target tracking algorithm, and generates dynamically feasible trajectories that provably converge to a moving target while avoiding obstacles. The efficacy of the proposed approach is validated through Gazebo simulations.