GLIO2 completes a route with weak LiDAR geometry

When LiDAR scans provide little information about movement in one direction, a mapping system can lose its trajectory. A scan match can also be biased and then treated as more certain than it is.

Can a system keep revising local scan geometry while satellite navigation still anchors its global position?

GLIO2 jointly updates relative LiDAR, inertial, and raw GNSS measurements so scan errors can be revised while GNSS anchors the global trajectory.

GLIO2 is a localization and mapping system that combines LiDAR, inertial measurements, and GNSS. It keeps relative scan constraints revisable as the estimate changes, while GNSS provides a global reference. The question is whether this design can maintain a trajectory when LiDAR geometry is weak.

On the 5.66 km Bridge route, GLIO2’s online estimate had 1.625 m horizontal and 3.586 m three-dimensional ATE RMSE, while the evaluated LiDAR-based systems diverged. The authors attribute this behavior to raw GNSS constraining the direction that the degenerate scan registration leaves unobserved. This is evidence from one route and its tested comparisons.

The evidence comes from eleven evaluated sequences and three tested LiDAR types. The authors also report that poor satellite-navigation conditions at startup, or starting while moving, can degrade the estimate; their current handling of multipath and blocked satellite signals remains limited.