Toward an exploration-based probabilistic reasoning for a quadrotor

This work outlines a practically realizable (i.e., deployable and scalable) yet novel autonomous exploration strategy for unmanned aerial vehicles (UAV), which in our case, corresponds to multi-rotor configurations. Concretely, based on a probabilistic map, UAVs are able to modify their trajectory to localize the required target in unknown areas. This is thanks to the fact that the proposed exploration strategy uses the past and the actual perceived data in order to deduce the location of the target, and a dedicated control law allows the multi-rotor to reach the desired position. To realize the strategy, we developed a hierarchical control architecture that can be embedded in multi-rotors. We show its effectiveness by computer simulations and tests using real drones, against a forest-fire localization scenario for an unknown area.