Toward Lifelong Visual Localization and Mapping.

摘要

In this thesis we address the problem of the temporal scalability of pose graphs models for long-term simultaneous localization and mapping (SLAM). We present a SLAM system using two different sensor modalities: imaging sonars for underwater navigation and vision based SLAM for terrestrial applications. In the underwater domain we consider two different applications. First, we describe our implementation of real-time imaging sonar aided navigation applied to in-situ autonomous ship hull inspection using the hovering autonomous underwater vehicle (HAUV). Second, we develop a feature-based navigation system supporting multi-session mapping, and provide an algorithm for re-localizing the vehicle between missions. We use a pose graph representation for the mapping. One of the problems with the pose graph formulation is that the state space continuously grows as more information is acquired. To address this we propose the reduced pose graph (RPG) model which partitions the space to be mapped and uses the partitions to reduce the number of poses used for estimation. To evaluate our approach, we present results using an online binocular visual SLAM system. We demonstrate long-term mapping using approximately nine hours of data collected in the MIT Stata Center, demonstrating mapping of a large environment an extended period of time.