Ph.D. Dissertation Defense Announcement

Title: The roles of the allocentric representation in local high-speed navigation

Duy-Nguyen Ta
School of Interactive Computing
College of Computing
Georgia Institute of Technology

Date: Thursday, January 08, 2015
Time: 11:00am-1:30pm EST
Location: TBD

Committee:
Prof. Frank Dellaert (Advisor, School of Interactive Computing, College of Computing, Georgia Tech)
Prof. Ronald Arkin (School of Interactive Computing, College of Computing, Georgia Tech)
Prof. Panagiotis Tsiotras (School of Aerospace Engineering, College of Engineering, Georgia Tech)
Prof. Tucker Balch (School of Interactive Computing, College of Computing, Georgia Tech)
Prof. Gabe Sibley (Department of Computer Science, University of Colorado, Boulder)

Abstract:
In this thesis, I study the computational advantages of the allocentric representation, as compared to the egocentric one, for autonomous high-speed local navigation. Whereas in the allocentric framework, all variables of interests are represented with respect to a coordinate frame attached to an object in the scene, in the egocentric one, they are always represented with respect to the robot frame at each time step. Geometrically, these two representations are related via only a simple coordinate frame transformation; hence, their advantages are often neglected in robotics research. However, they have been studied extensively in experimental psychology and cognitive science, although their computational advantages are largely unexplored. Inspired by many evidences for the dominant role of the allocentric representation in human navigation from these fields, I explore the benefits of the allocentric representation in state-of-the-art perception and control methods to improve the performance of autonomous navigation systems.

In contrast with well-known results in the Simultaneous Localization and Mapping literature, using Lie-group representations of poses, I show that the amount of nonlinearity of these two representations do not affect the accuracy of Gaussian-based filtering methods for perception at both the feature level and the object level. Furthermore, although these two representations are equivalent at the object level, the allocentric filtering framework is better than the egocentric one at the feature level due to its advantages in the marginalization process. Moreover, I show that the object-centric perspective, inspired from the allocentric representation, enables novel linear-time filtering algorithms, which significantly outperform state-of-the-art feature-based filtering methods with a small trade-off in accuracy due to a low-rank approximation. Finally, I show that the allocentric representation is also better than the egocentric one in Model Predictive Control for local trajectory planning and obstacle avoidance tasks.