Log Barrier Method for SLAM Applications
Author(s)
Jayakumar, Nithya
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Simultaneous Localization and Mapping (SLAM) is an important robotics problem that entails mapping the environment of a robot at the same time as tracking the robot’s location in said environment as it moves. Representing SLAM (and other robotics) problems with Factor Graphs, a type of graphical model, is very useful as it allows for mathematical optimization techniques to be applied successfully. Factor graphs are also very helpful for understanding, representing, and visualizing robotics problems in order to further research in these fields. Representing robotics problems with factor graphs essentially cre- ates a mathematical optimization problem, which can be solved using various algorithms, and the implementation of these algorithms to a library that enables robots to apply these techniques to successfully navigate new environments au- tonomously is what my research is about.
The applications of solutions to the SLAM problem are countless. As it is an important factor in autonomous vehicle motion, applications to the implementations of these optimization techniques have impact in areas such as self-driving cars, walking and climbing robots, search-and-rescue robots, and more.
My research intends to further the current solutions for robot path-planning and mapping problems by looking at several algorithms and applications used to optimize factor graph-based representations of robotics problems. Especially the log-barrier method, which simply extends unconstrained or linearly constrained optimization problems by solving a sequence of unconstrained problems. The log-barrier method uses the last point found as the starting point for the next sub-problem, hence giving this category of optimization techniques the name “interior point” methods. I will be implementing the log-barrier method and integrating it within GTSAM (GT Smoothing and Mapping, Georgia Tech’s sensor fusion and computer vision applications library) and researching several other optimization strategies to do so with as well.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis