Embedded Files
LiDAR Place Recognition & Odometry
LiDAR Place Recognition & Odometry
Place recognition is essential for merging multiple maps in different sessions.
In multi-robot PARK A, B (narrow scene) sequences, LoGG3D-NET and SOLiD are the most robust.
In multi-session (about one year) FOREST A, B sequences, M2DP using the principal component solves the place recognition problem well.
LiDAR odometry in legged robots is particularly vulnerable to z-axis errors.
CT-ICP (ICP in the continuous-time domain) is the most robust against vibration vulnerability, a characteristic of the legged robot.
Page updated
Google Sites
Report abuse