TOPICS FOR MASTER AND/OR DOCTORAL THESIS

PROJECT SUGGESTIONS USING ROBOTS AT IOC

The Institute of Cybernetics has some robots that may be used for project work.

• Khepera: http://www.k-team.com/robots/khepera/
• Pioneer: http://www.activrobots.com/ROBOTS/p2dx.html
• Lego: http://mindstorms.lego.com/eng/products/ris/

Here are some project suggestions. I am also open to suggestions from students, especially projects that aim to reproduce results obtained with Khepera, Pioneer or Lego robots at other Universities. I would expect students to prepare and submit a conference paper describing their work. Successful project work may lead to PhD work, abroad or here in Estonia.

For further details or to arrange a meeting, please mail me, Adam Eppendahl, at ae@cs.ioc.ee.

1. Title: Focusing Occupancy Models
   Robots: Khepera, Pioneer.
   Background: Work by Hakim Miah in London.

   Aim: Determine the prospects for parameter recovery by model focusing.

   Objectives: Reproduce Miah's results with a 1d camera on the Khepera, and then with a 2d camera on the Pioneer. Automate the focusing.

   Demo: Robot navigates successfully without prior knowledge of camera and drive parameters.

2. Title: Combined Feature/Occupancy Models for Robot Vision.
   Robots: Khepera, Pioneer.
   Background: Work by Hakim Miah in London and the Vision Group in Oxford.

   Aim: Integrate the complementary strengths of feature and occupancy based models.

   Objectives: Reproduce feature based model construction with a 1d camera on the Khepera (or with a 2d camera on the Pioneer if necessary). Extend with the feature model with occupancy information. Produce a simple mathematical and programming framework for combined models.

   Demo: Robot performs feature based tasks using occupancy information, and vise versa.

3. Title: Seeing Empty Space
   Robots: Khepera, Pioneer
   Background: ?

   Aim: Use an occupancy model to detect empty space and avoid collisions in real-time.

   Objectives: Program the Khepera to build a map of empty space using an occupancy model. Establish the speed limits for a Khepera building and using such a map. Reproduce these results on the Pioneer.

   Demo: Robot drives about faster and faster without bumping into anything.

4. Title: Learning in the Presence of Irreversible Actions
   Robots: Pioneer, Lego.
   Background: Work by Maarja Kruusmaa in Sweden (now in Tartu).

   Aim: Demonstrate the significance of irreversible actions in the performance of learning algorithms.

   Objectives: Reproduce the results Kruusmaa obtained in simulation on the pioneer. Obtain similar results a variety of Lego robots.

   Demo: Robot preforms a series of tasks in a cautious way.

5. Title: Exploring Reversibility Space
   Robots: Khepera, Lego.
   Background: ?

   Aim: Establish the feasibility of mapping a robot's space of reversible actions.

   Objectives: Program the Khepera to make a simple map of its reversibility space. Test the same algorithm on a variety of Lego robots.

   Demo: Robot makes a series of larger and larger movements checking that each can be reversed.

6. Title: Abstract Motivation Principles
   Robots: Lego.
   Background: Work by Frederic Kaplan in Paris.

   Aim: Adapt the use of abstract motivational principles to a small processor.

   Objectives: Reproduce the results Kaplan obtained on the Aibo on a simple Lego robot. Obtain similar results with a motivational principle based on reversibility.

   Demo: Robot develops coherent movements without prior knowledge of its body or environment.