Pantelis Elinas, PhD

Science, technology, gadgets and academic life

Projects

By Pantelis • Jan 25th, 2009 • Category: Research

2006/2007

  • In 2006 and 2007, we continued to work on visual SLAM and occupancy grid mapping using stereo vision applied to the development of an intelligent wheelchair in collaboration with colleagues at the University of Toronto. We also explored planning under uncertainty using continuous actions which we demonstrated working on our interactive robot HOMER.
  • In November of 2007, I successfully defended my PhD thesis titled “On the design and implementation of decision-theoretic, interactive, and vision-driven mobile robots.”

2005

  • In 2005, we designed and implemented SLAM for robots using stereo vision. Our approach is designed to handle 6 degrees of freedom motion and it only requires as input a sequence of stereo images. It is based on the RBPF using the Scale Invariant Feature Transform for robust data association and efficient data structures for map management. We also show that given our well localized camera, we can trivially construct 2D occupancy grid maps from correlation-based stereo; these maps are useful for path planning and exploration. For more info see our RSS 2005 and ICRA 2006 papers.

2004

  • Currently, we are exploring a new method based on distributed MDPs for task specification and decision making on socially interactive mobile robots. We call these, Multiply Sectioned MDPs. We present our newest upgrade to our robot HOMER using this method. Go here for videos of HOMER delivering messages using the MS-MDP framework.

2003

  • Recently, I have been doing some work on building 3D occupancy grids from stereo data. The grids are centered around the robot i.e. the goal is not to build large volumetric maps of whole buildings. At first, I plan to use these grids for obstacle avoidance during robot navigation.

2002

  • In May of 2002, I finished my Master’s of Science degree in robotics. My thesis title was “Interactive Directed Exploration for Mobile Robots.” It involved creating a human-robot interface using speech and gestures that allowed a person to give commands to a robot. The task was one of directing the robot to different areas to explore. You can get a copy of my thesis in pdf format here.
  • In the summer of 2002, we started a new project the Human Oriented MEssenger Robot, or just HOMER, that is tailored for studies in human-robot interaction. HOMER is a powerful compute engine on wheels as it has 4 on board computers. He uses a Bumblebee Stereo Camera system by PointGrey Research for perceiving his surrounding environment.

2001

  • In August 2001 we were the winners of the AAAI “hors d’oeuvres anyone?” competition. For more information on the event including pictures and video follow this link.

2000

  • Until the summer of 2000, I worked on the real-time rendering of 3D clouds. Our efforts were focused on using off-the-selve 3D accelerator boards to render Gardner style clouds in real-time. The background image of this page is one generated using this method.

Tagged as: ,

Pantelis is
Email this author | All posts by Pantelis

Leave a Reply