Nonlinear Model Predictive Control for Repetitive Area Reconnaissance with a Multirotor Drone

This paper considers the problem of a reconnaissance mission in which a single multirotor drone must survey a given map by repetitively visiting different checkpoints. Several points of interest (POIs) are used to discretise the map, and each of them is associated with a time-varying heat value according to the specific application. In that way, each POI has a different visiting priority each time. The proposed solution considers a nonlinear model predictive control (NMPC) approach that minimises the map’s overall heat and considers several constraints related to the system dynamics and the environment (e.g., the presence of unknown obstacles). Possible applications are related to the research of gas leaks, area surveillance, patrolling, etc. The methodology is tested in a realistic simulation environment and through experiments.