CML/SLAM studies and “Velocity-Over-Ground” SLAM

Simultaneous Localization and Map Building (SLAM), also known as Concurrent Mapping and Localization (CML), is a tracking technique with the following goals: 1) Construct maps of the locations of strong point scatterers in the sensors´ field-of-view 2) Continually improve these maps as more sensor data arrives 3) Simultaneously achieve improved platform self-localization with respect to scatterer locations CML/SLAM´s refinement of the platform location estimates is directly linked to its exploitation of the stationary nature of the scatterers of opportunity. However CML/SLAM is often computationally prohibitive as its architecture attempts to minimize the position error for each scatterer, i.e. “map-building.” This requires all pairwise correlation information to be maintained and updated leading to very large memory and computation requirements. Nearly every Unmanned Undersea Vehicle (UUV) mission involves a transiting stage. Here “map-building” is unnecessary but can improve platform estimates. Can we capture relevant platform information without explicitly tracking strong point scatterers? We derive a CML/SLAM-inspired “velocity-based” estimator that provides virtually optimum performance in transit scenarios. The method is of the order of complexity of the single contact Kalman Filter. A reformulation of the CML/SLAM problem leads to new insights into performance and earlier results. In addition, this reformulation stresses the fundamentally relative nature of the available measurements and helped lead to the new “velocity-based” method. Theoretical arguments and real-data results are provided to reveal performance. A UUV with a forward-looking sonar along with an INS suite is the platform for this work.