作者: B. Peasley , S. Birchfield
DOI: 10.1109/WORV.2013.6521938
关键词: Mobile robot 、 Geography 、 Artificial intelligence 、 Obstacle avoidance 、 Robot 、 Ground plane 、 Robustness (computer science) 、 Specular reflection 、 Computer vision 、 Robot kinematics 、 Obstacle
摘要: This paper proposes a novel approach to obstacle detection and avoidance using 3D sensor. We depart from the of previous researchers who use depth images sensors projected onto UV-disparity detect obstacles. Instead, our relies on projecting points ground plane, which is estimated during calibration step. A 2D occupancy map then used determine presence obstacles, translation rotation velocities are computed avoid Two innovations introduced overcome limitations sensor: An infinite pole proposed hypothesize infinitely tall, thin obstacles when sensor yields invalid readings, control strategy adopted turn robot away scenes that yield high percentage readings. Together, these extensions enable system inherent Experiments in variety environments, including dynamic objects, varying heights, dimly-lit conditions, show ability perform robust real time under realistic indoor conditions.