Visual-Acoustic SLAM for Underwater Caves

Rahman, Sharmin, Li, Alberto Quattrini, Rekleitis, Ioannis
In Underwater Robotics Perception Workshop - ICRA 2019 2019.


Abstract:Underwater caves are extremely challenging environment for perception, due to the absence of natural light and the highly unstructured nature of such environment, making it also dangerous and cognitively heavy even for highly skilled divers. This paper presents an overview of our previous works for underwater cave mapping which combines data from multiple sensors to assist the divers by reducing the cognitive loads. The challenges of the underwater environment augmented by the complete absence of natural light and the effects of sharp shadows are discussed together with the contributions of the different sensing modalities. A tightly-coupled keyframebased SLAM framework with loop-closing and relocalization capabilities combining visual, inertial, depth, and acoustic sensors has been described together with the design of a sensor suite for collecting data in the challenging environment of underwater caves. Experimental results illustrate the accuracy and robustness of the proposed methodology from a cavern at Ballroom, Ginnie Springs, FL, USA.



  title        = {Visual-Acoustic SLAM for Underwater Caves},
  author       = {Rahman, Sharmin and Li, Alberto Quattrini and Rekleitis,
  booktitle    = {Underwater Robotics Perception Workshop - ICRA 2019},
  year	       = {2019}

Rekleitis's home page
Rekleitis's publication list

Thu Jan 23 06:21:02 EST 2020