Prev-Talk | Next-Talk | All-Talks | Talks-Sorted

Talk-12: Data association towards persistent SLAM with underwater sideScan sonar implemented in MOOS-IvP

Mae L. Seto, Defence Research and Development Canada

Colin MacKenzie

Timothy Pohajdak

John J. Leonard, MIT

While critical to the simultaneous localization and mapping (SLAM) process data association can be unreliable and especially so in the noisy, dynamic underwater environment. This presentation discusses a data association algorithm to enhance underwater SLAM on autonomous underwater vehicles (AUV) with side-scan sonars through jointly associating the estimated relative position of a landmark to the AUV with the local seafloor elevation gradients surrounding the landmark. The local elevation gradients are extracted from the same side-scan sonar images that the landmarks are detected in. As environmental features, the gradients are less susceptible to gross changes due to the underwater environment compared to a small (~ few meters) landmark.

This scheme yields realistic correct associations when implemented in MOOS and validated in a hardware-in-the-loop AUV simulator using sidescan sonar data from earlier trials. The data association algorithm is now integrated into the iSAM (incremental smoothing and mapping) dynamic pose graph formulation and undergoing in-water validation as an on-line MOOS capability.


  • AUVs
  • Navigation/SLAM
  • Sensors/Sonar