A reinforcement learning approach for UAV target searching and tracking

Tian Wang (1) , Ruoxi Qin (1) , Yang Chen (2) , Hichem Snoussi (3) , Chang Choi (4)

1 School of Automation and Electrical Engineering [Beijing] (University of Science and Technology Beijing)
2 Nanyang Technological University [Singapour]
3 LM2S - Laboratoire Modélisation et Sûreté des Systèmes
4 GIST - Gwangju Institute of Science and Technology

Owing to the advantages of Unmanned Aerial Vehicle (UAV), such as the extendibility, maneuverability and stability, multiple UAVs are having more and more applications in security surveillance. The object searching and trajectory planning become the important issues of uninterrupted patrol. We propose an online distributed algorithm for tracking and searching, while considering the energy refueling at the same time. The quantum probability model which describes the partially observable target positions is proposed. Moreover, the upper confidence tree algorithm is derived to resolve the best route, with the assistance of teammate learning model which handles the nonstationary problems in distributed reinforcement learning. Experiments and the analysis of the different situations show that the proposed scheme performs favorably.

Domaines: Automatique / Robotique

Tian Wang, Ruoxi Qin, Yang Chen, Hichem Snoussi, Chang Choi. A reinforcement learning approach for UAV target searching and tracking. Multimedia Tools and Applications, 2019, 78 (4), pp.4347-4364. ⟨10.1007/s11042-018-5739-5⟩. ⟨hal-03320597⟩