Cognitive Robotics Final Project Ideas for 2009
Here are some suggestions for final projects. You can pick something
from this list, or you can come up with your own idea and talk it over
with Dave and Leigh Ann. All projects must be approved in advance; we
want to make sure that what you propose is challenging but feasible in
the time available.
Experiments with an Infra-Red Range Finder and Sound Detector
Students: Charles Ruhland and Jin Su Kim
- The Dynamixel AX-S1 sensor module mounted directly below the
webcam contains a three-direction infra-red rangefinder and a sound
detector.
- Determine the range and acccuracy of the rangefinder.
- Can the rangefinder be used for obstacle detection (cliffs and/or walls)?
- Can the rangefinder be used to detect human motion?
- Can the rangefinder be used to measure the dimensions of an object?
- What are the characteristics of the sound detector?
- Write demo programs to show off the AX-S1 capabilities you've
investigated.
Using the Right Front Leg as a Pincer
Student: Richard Ha
- The right front leg contains an extra degree of freedom
thaht allows it to rotate upward and be used as a pincer.
- Write code to use the pincer to swat a large object such as a toy
basketball.
- Modify the walk parameter file to let the robot walk on
five legs instead of six, so the pincer is free for manipulation.
Denavit-Hartenberg Kinematics Wizard
Student: Leland Thorpe
- Interactive tool for creating and displaying kinematic descriptions of robots
using the Denavit-Hartenberg conventions and Tekkotsu .kin files.
- Displays and edits kinematic chains in a GUI window and draws the
robot as a 3D structure.
Depth From Stero
Students: Chandrasekha Bhagavatula, David Klionsky, and Ian Lenz
- Use multiple camera images, taken from different positions
as the robot sways back and forth, to derive a depth map and
do object segmentation.
- Use the depth map to determine precise object locations, or
do obstacle detection for navigation.
Sheep Herding: Boundary Navigation
Students: Daegon Won and Siyuan Feng
- Create a complex but convex boundary using pink tape.
- The robot must keep all the blue easter eggs (sheep) inside the
boundary and all the orange eggs (wolves) outside. Eggs on the wrong
side of the boundary must be pushed to the other side.
- Tricky issue: determining which is the "inside" of the boundary,
especially when you're standing on top of it. This is the heart
of the project.
- Robot patrols the boundary and surrounding area looking for stray
sheep that must be herded back to where they belong.
- Human can move eggs around; robot must notice and respond.
The Ultimate Ball-Chaser
Student: TBD
- Use the Lookout to track the ball (and modify the Lookout
so it works well on the Chiara and Create robots).
- If the ball goes off screen, use its direction of travel to
intelligently guess where to look for it.
- If ball isn't where we expect, execute a visual search for it.
- If visual search fails, go off and hunt for the ball, avoiding
obstacles along the way. Possible wall-following behavior. Should be
able to eventually find a ball hidden anywhere in the playpen.
- Use audio signals or speech to indicate when the ball is lost and reacquired.
- When the robot gets close to the ball, execute one of several
behaviors: kick the ball, circle the ball, try to push the ball.
Porting the OpenCV Face Detector to Tekkotsu
Student: Ilya Matiach
- Get the OpenCV face detector running.
- Modify the face detector to interface to Tekkotsu's dual-coding vision system.
Particle Filter for Playpen Navigation
Student: TBD
- Decorate the playpen with distinctive markers.
- Write code to allow the robot to determine its position by first
visually acquiring markers and then using Tekkotsu's particle filter
to localize itself based on the marker positions.
- Write code that allows the user to instruct the robot to travel
to any desired location in the environment. Calculate a rough
trajectory and use the particle filter to guide the robot's progress.
- The goal is to achieve fast, accurate playpen navigation. This
will require intelligent strategies for fast landmark search, and
tuning of the particle filter.
SIFT-Based Object Sorter
Student: TBD
- Given a collection of two types of objects, use the arm to sort
them into two piles,
- First problem is to isolate one item from the collection. Then
identify the item class. Then push it to the correct location.
- Can use SIFT to distinguish objects based on shape, e.g., spoons
vs. forks. Must train up spoon and fork models from examples.
Rewrite the Pilot
- The pilot was written using old-style state machine notation.
- The code should be updated to use newer programming conventions
and provide more and better functionality.
- Produce some demos showing off what the Pilot can do.
Dave Touretzky
Last modified: Sat Apr 11 07:54:05 EDT 2009