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

Talk-11: Implementing on-line AUV SLAM using side-scan sonar and automated target detection with MOOS- IvP

Mae L. Seto, Defence Research and Development Canada

Timothy Pohajdak, Dalhousie University

Colin MacKenzie, Dalhousie University

John J. Leonard, Department of Mechanical Engineering, Computer Science and AI Lab, MIT

Simultaneous localization and mapping (SLAM) using side-scan sonar on-board autonomous underwater vehicles (AUV) for mapping mine-like objects (or targets) was implemented and tested in-water. The work is motivated by persistent SLAM for change detection over longer periods.

The payload autonomy was implemented in MOOS-IvP on-board the backseat processor of the IVER AUV. iSAM (incremental smoothing and mapping) was the SLAM engine used for the core state estimation. The DRDC automated target detection (ATD) tools were used to perform the on-line target detection and identification as well as provide an estimate for the target location. The data association used was joint compatibility branch and bound. These tools and algorithms were integrated and/or developed with a hardware-in-the loop simulator for the Iver AUV. Newly developed MOOS modules are discussed.

Three sets of in-water trials, separated by weeks, were performed to collect the change detection data sets. A year later a newer data set was collected over the same area with the AUV fusing a high calibre INS into the dead-reckoned navigation solution. The presentation discusses the use of the MOOS framework for implementing in-water SLAM using the DRDC ATD tools and iSAM on an embedded processor. As a measure of localization accuracy, the internal dead-reckoned AUV position estimate (with and without INS) just before surfacing is compared with the GPS fix on the surface and with the SLAM position estimate.


  • AUVs
  • MCM
  • Navigation/SLAM
  • Sensors/Sonar