Route planning for teams of unmanned aerial vehicles using Dubins vehicle model with budget constraint David Zahr´adka, Robert Pˇ eniˇ cka, and Martin Saska Czech Technical University, Faculty of Electrical Engineering, Technicka 2, 166 27, Prague, Tel.: +420-224357284 {zahrada2,penicrob,saskam1}@fel.cvut.cz Abstract. In this paper, we propose Greedy Randomized Adaptive Search Procedure (GRASP) with Path Relinking extension for a solution of a novel problem formulation, the Dubins Team Orienteering Problem with Neighborhoods (DTOPN). The DTOPN is a variant of the Orien- teering Problem (OP). The goal is to maximize collected reward from a close vicinity of given target locations, each with predefined reward, us- ing multiple curvature-constrained vehicles, such as fixed-wing aircraft or VTOL UAVs with constant forward speed, each limited by route length. This makes it a very useful routing problem for scenarios using multi- ple UAVs for data collection, mapping, surveillance, and reconnaissance. The proposed method is verified on existing benchmark instances and by real experiments with a group of three fully-autonomous hexarotor UAVs that were used to compare the DTOPN with similar problem for- mulations and show the benefit of the introduced DTOPN. Keywords: route planning, dubins team orienteering problem with neigh- bourhoods, dtopn, unmanned aerial vehicles, mapping, data collection, inspection, reconnaissance, surveillance 1 INTRODUCTION Due to the recent increase in usage of autonomous systems, such as Unmanned Aerial Vehicles (UAVs) for search and rescue, information gathering and even military applications [22–24], the demand for effective control rises proportion- ally. This includes route planning algorithms, which can increase the productivity of autonomous vehicle deployment when effective. Since optimization problems such as the studied multi-goal planning are computationally complex, various algorithms and heuristics surface and compete in which is closer to an optimal solution and in the required computational time. One of the problem formula- tions used for route planning optimization over multiple target location is the Orienteering Problem (OP) [19]. The OP features a set of target locations each with an assigned reward for visiting its particular position. Both the starting and the ending locations are predefined. The travel budget used to visit the locations is limited by a given