Utility-based map reduction for ground and flight vehicle navigation

Abstract

Maps used for navigation often include a database of location descriptions for place-recognition (to enable localization or loop-closing), which permits bounded-error navigation performance. A standard localization system must describe the entire operational environment in its place-recognition database. A standard pose-graph-based simultaneous localization and mapping (SLAM) system adds a new place-recognition database entry for every new vehicle pose, which grows linearly and unbounded in time and thus becomes unsustainable. To address these issues, this thesis proposes a new map-reduction approach that pre-constructs a fixed-size place-recognition database amenable to the limited storage and processing resources of the vehicle by exploiting the high-level structure of the environment and vehicle motion. In particular, the thesis introduces the concept of location utility - which encapsulates the visitation probability of a location and its spatial distribution relative to nearby locations in the database - as a measure of the value of potential localization or loop-closure events to occur at that location. While finding the optimal reduced location database is NP-hard, an efficient greedy algorithm is developed to sort all the locations in a map based on their relative utility without access to sensor measurements or the vehicle trajectory. This enables predetermination of a generic, limited-size place-recognition database containing the N best locations in the environment. A street-map simulator using city-map data and a terrain relative navigation simulator using terrestrial rocket flight data are used to validate the approach and show that an accurate map and trajectory reconstruction (pose-graph) can be attained even when using a place-recognition database with only 1% of the entries of the corresponding full database.