Home/Localization, Mapping, and Planning in 3D Environments

Localization, Mapping, and Planning in 3D Environments

Nathaniel Fairfield
PhD Thesis, Tech. Report, CMU-RI-TR-09-05, Robotics Institute, Carnegie Mellon University, January, 2009

View Publication

Copyright notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.


Building a map, localizing within the map, and planning using the map are fundamental prob-lems for mobile robotics. Every mobile robotic system must incorporate some type of solution to all three problems. While the interdependency between mapping and localization is well known as the Simultaneous Localization and Mapping (SLAM) problem, there is a growing understanding in the research community that planning how the robot goes about mapping and exploring an environment (and operating in the environment afterwards) can avoid degenerate conditions and significantly reduce complexity of SLAM. Thus the task of exploring a new environment combines all three problems, since the robot must plan to find actions that reduce uncertainty in both mapping and localization. This combined problem is known as Active SLAM. Independently, SLAM and planning have been solved in small, two dimensional, structured domains. Robots need to move beyond these simple environments. The challenge is to develop real-time Active SLAM methods that allow robots to explore large, three dimensional, unstructured environments, and allow subsequent operation in these environments over long periods of time. But scaling up to truly large environments requires a second key insight beyond Active SLAM: to circumvent the scale limitations inherent in SLAM, the world can be divided up into more manageable pieces, or submaps. The SLAM problem then becomes a segmented SLAM problem, which represents the world with a combined metric and topological map, building metric submaps as necessary and refining the topological relationships between submaps. The contribution of this thesis is a real-time Active SLAM approach that combines a novel evidence grid-based volumetric representation, a robust Rao-Blackwellized particle-filter, a topologically flexible submap segmentation framework, and an integrated stochastic planning method for reducing SLAM uncertainty and predicting possible loop closures based on local map structure. We demonstrate our methods on several robotic platforms in both structured and unstructured large, three dimensional environments.

author = {Nathaniel Fairfield},
title = {Localization, Mapping, and Planning in 3D Environments},
year = {2009},
month = {January},
school = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-09-05},
} 2017-09-13T10:41:20-04:00