Safe vehicle motion design with learning for moving in environment with uncertainties
Dénes Tompos, Balázs Németh, Tamás Hegedűs, Vũ Văn Tấn, Péter Gáspár · IFAC-PapersOnLine · 2024
In this paper a motion profile design for unmanned aerial vehicles is proposed which method is able to guarantee safe collision-free motion. The motivations of the work are provided by the uncertainties of covered areas by the vehicles, and also the need of high performance fast vehicle motion. The uncertain information on the environment for detecting Conflict areas is processed through clustering and Mahalanobis-distance-based filtering methods. The resulted Conflict areas are involved in the motion design method, which is facilitated through reinforcement learning. This paper shows the application of the method on a drone that moves together with a mobile robot in the same environment. The safe and high performance motion of the drone is illustrated through simulation example.