Obstacle avoidance under relative localization uncertainty
Lyu Y, Pan Q, Hu J W, et al
Collision avoidance is one of the classical topics in robotics research. Collision avoidance control is guided by implementing typical collision avoidance methods, such as potential field, dynamic programming, and geometrical calculation, to various platforms. Since these methods usually assume the perfect localization of an obstacle, they may not guarantee collision-free operations during the stochastic operation of robots wherein the obstacle is localized using noisy sensors. In particular, for micro aerial vehicles (MAVs), the potential obstacles are detected using low-resolution sensors, such as cameras and low-cost radars, leading to obstacle localization with high uncertainty. This study considers collision avoidance tasks with relative localization uncertainty to ensure an optimum operation safety of MAVs using low-cost sensors, which have imperfect sensing mechanism.