Using Negative Information in simultaneous localization and mapping

Abstract

The problem of autonomous navigation is one of efficiently utilizing available information from sensors and intelligently processing that information to determine the state of the robot and its environment. This thesis explores a topic often ignored in the Simultaneous Localization And Mapping (SLAM) literature: the utility of including Negative Information as a means of aiding state-estimation decisions and successfully re-localizing the autonomous agent. The work is motivated by a low-cost underwater mine neutralization project, which requires that an Autonomous Underwater Vehicle (AUV) successfully localizes itself in a difficult SLAM environment. This thesis presents a new Negative And Positive Scoring (NAPS) algorithm for comparing multiple localization hypotheses and then uses a large number of simulations to quantify the effect of including the often ignored Negative Information (NI). The ultimate conclusion of this thesis, that careful inclusion of Negative Information increases the chances of successful localization across a wide variety of difficult SLAM situations, extends beyond the intended target reacquisition application and is generally applicable to robotic navigation problems.