[hal-01137963] Image-based road network clearing without localization and without maps using a team of UAVs

We address the problem of clearing an arbitrary and unknown network of roads using an organized team of Unmanned Aerial Vehicles (UAVs) equipped with a monocular down-facing camera, an altimeter, plus high-bandwidth short-range and low-bandwidth long-range communication systems. We allow the UAVs to possibly split in several subgroups. In each subgroup a leader guides the motion employing a hierarchical coordination. A feature/image-based algorithm guides the subgroup toward the unexplored region without any use of global localization or environmental mapping. At the same time all the entry-points of the the explored region are kept under control, so that any moving object that enters or exits the previously cleared area. Simulative results on real aerial images demonstrate the functionalities and the effectiveness of the proposed algorithm. I. INTRODUCTION The use of coordinated teams of robots allows to require less from each hardware device and at the same time achieve higher performance and robustness. For such reasons the design of controllers for multiple robots gained great relevance in the literature, see, e.g., [1], [2] and references therein. Particular interest has been paid to the exploration of a given region [1], [3], [4], allowing the execution of several subtasks: guaranteed clearing for pursuit-evasion [4], [5], models for forest area coverage [6], autonomous or semi-autonomous control for rescue purposes [7], mine detection [8], exploration models for underwater regions [9], just to mention a few. Along this line, in this work we present an algorithm for the guaranteed clearing of a network of roads using a team of UAVs. The clearing problem is well established in the research on robotics: so far, however, there have been still few results suitable for real-world applications. In fact, a large fraction of such research work has developed models using assumptions far from those suitable for realistic applications, e.g., sensors with infinite range, a priori knowledge of the environment or target model, or perfect localization. Furthermore, From a theoretical point of view, important results have been achieved in this field using three main approaches: graph-based, visibility-based and sweeping research. In graph-based approaches the team is typically supllied with a



from HAL : Dernières publications http://ift.tt/1EIM9od