These projects are challenging; some so much so that I hesitate to list them for a ten-week class. Many of these projects, if done well, could form the basis for an interesting paper at an international conference, or could be extended into a senior thesis. (Unfortunately probably not possible for current seniors, since two terms of work is required for a senior thesis, and work cannot be double-counted.)
Robotics education: summer camp design
Develop a summer camp in robotics for middle school students. Every summer, we offer a robotics camp for local middle-school and high school students. The camps have had various themes, but typically the main challenge was to build lego robots. It would be great fun to introduce more robot programming into that course. For example, last year, students used Java to program the robot arm to draw pictures, and to program mobile robots to complete a cardboard maze.
Here is the 2008 robotics camp web page.
The project is to develop two one-week camps; one suitable for younger students, and one for older. Development of the camp includes serious technical work: the mobile robots need to be made easy to use, and the camp challenges have to be designed and implemented by the team. I prefer that members of this team be available this summer for two weeks at the end of June/ beginning of July to actually teach the camp, but this is not a requirement. (Camp instructors were paid a small but non-zero stipend last year.)
Simulation of folding cloth
Simulate Japanese shirt-folding! There's a lot going on in this video that we don't understand. How would you simulate this shirt-folding process, or design a robotic system to duplicate it? There has been work in the graphics community on how to simulate cloth (including work on collisions), but very little of it has focused on the problem of manipulation.
Design and analysis of planar motion
Here's a fun video from the MIT media lab connecting a planar dynamics simulator with a nice GUI for drawing physical systems.
In robotics, we're often interested in planar simulations, but the physical system might be something more complicated. For example, consider writing a simulator for a mobile robot. We might be mainly interested in how the sensing will work, or in the control law. I've often wanted to experiment with simulators for time-optimal motion of mobile robots. How about friction and contact? There are many many graphical methods for analyzing grasps and fixtures, and some less-graphical methods. Could these be implemented into a simulation or design package like the one above?
Robot arm control
Use Microsoft Robotics Studio to develop a simulation and control environment for the Adept arm. One problem with programming robot arms is that they are dangerous and slow — it's very hard to try out programs, and only one person can be developing software per arm at a time. Another problem is that the arm uses a programming language, V+, that is just flat-out bad. It's like BASIC, but broken. There are many basic features that are missing, as well. For example, coordinate transforms exist, but are hard to use.
This project might have two components: building a simulator for our robot arm using Microsoft robotics studio, and developing a library for programming the arm that would use ideas from our course, including Denevit-Hartenberg parameters, homogeneous coordinate transforms, and programs for automatic calibration of the robot. Some work would be needed to find out what already exists. For example, there could very well be an existing simulator in MSRS for our Adept arm; we should check with Adept. If this were the case, using existing work as a basis (with careful attribution, of course!) would be the right place to start.
A robotic assistant
Develop a mobile robot with an arm that can be directed to perform simple household tasks with an easy-to-use interface. (For example, a laser pointer could be used to describe the task.)
Virtual reality flight
The hobby RC airplane community has recently been attaching accelerometers to VR goggles and using head motions to control a model airplane for virtual reality flight.
Here's another RC airplane video, and here's one showing the goggles. There's also a forum devoted to this sort of thing.
Although I love the idea of building something similar, there are a few problems. First, you'd need to be pretty good at flying the airplane. Second, we'd have to buy a fair amount of hardware. Third, we'd want to think of how to extend the work to make it more interesting from a robotics perspective.
Robotic carton folding with fixtures
Matt Elwin did some initial work on this project, based on a research paper by Lu and Akella.
L. Lu and S. Akella, "Folding Cartons with Fixtures: A Motion Planning Approach,"
IEEE Transactions on Robotics and Automation Vol. 16, No. 4, pp. 346-356, August 2000.
Paper
Akella also has some movies posted on his site, but my viewer wouldn't play them.
It would be a really great thing to extend Lu and Akella's work to more cartons. We could use the rapid prototyper to build fixtures. Matt Elwin (a 2009) has already ordered some cardboard boxes, and made some initial designs — we'd want to talk to him to get his input and ideas.
