概要
To avoid collisions with distracted pedestrians, we propose a social navigation method that recognizes attention to adapt navigation costs. We use 3D skeleton data and a random forest classifier to categorize pedestrians as “normal walking” or “distracted.” It then dynamically adjusts the safety margin on the costmap. Real-world experiments verify the approach’s safety and efficiency.
収録
第26回計測自動制御学会システムインテグレーション部門講演会, 1E3-09
