Bats navigate dense forests using only sonar, and modern in-air 3D imaging sonars can now capture the entire frontal hemisphere with a single pulse. Sonar-based simultaneous localisation and mapping (SLAM), however, has so far been planar: existing systems estimate only a 2D pose, limiting their use in complex 3D environments.
The new approach, named RTISSLAM-3D, extends sonar SLAM to full six-degree-of-freedom (6-DoF) pose estimation. This allows a robot to track both position and orientation in three dimensions, rather than being confined to a flat plane. The algorithm leverages the rich hemispherical data provided by each sonar pulse to build a more complete map of the surroundings.
By moving beyond planar assumptions, RTISSLAM-3D brings autonomous systems closer to the echolocation abilities of bats, which navigate cluttered forests with ease. The work is described in a recent arXiv preprint.