Fast-LIO2 Tightly-Coupled Point Clouds, Loop Closure & Sub-Centimeter Relocalization

Tightly-coupled LiDAR-inertial odometry delivering drift-free localization in large, geometrically ambiguous industrial environments.

Iterated Extended Kalman Filter (IEKF) Tight Coupling

RoboticAMR fuses raw LiDAR points directly with high-frequency (200 Hz) IMU pre-integration terms inside an iterated Kalman filter. This direct formulation eliminates feature extraction delays, sustaining localization at speeds up to 3.5 m/s.

Incremental kd-Tree (ikd-Tree) Point Cloud Updating

By maintaining an ikd-Tree structure in RAM, new sensor points are inserted and dynamic points deleted dynamically with microsecond lookups, preserving crisp environmental geometry without map degradation.

Fast-LIO2 tightly-coupled LiDAR-inertial odometry running at 50 Hz
Sub-centimeter localization accuracy across 50,000 m² facilities
ikd-Tree spatial data structure supporting continuous dynamic updates
Automated loop closure detection via Scan Context descriptors

ドメインを見る