Persistent autonomous exploration, mapping and localization

Abstract

In this thesis, we investigate methods for exploration, persistent autonomy, and simultaneous localization and mapping tasks for an autonomous mobile robot with battery constraints. First, we present modifications to baseline frontier exploration on an occupancy grid that makes the robot's frontier exploration more efficient. Second, we describe the new software structure and recovery behavior for an autonomous robot to navigate to its dock despite errors of uncertainty in its map. Third, we implemented a landmark-based topological mapping method using a state-of-the-art toolbox that maps the environment using visually unique tags to compare with metric mapping methods. Our analysis shows that the robot explores its environment more efficiently using our method than with previous frontier exploration methods, and that graph based mapping outperforms metric mapping against ground-truth accuracy tests.