Projects

Research Interests
I am generally interested in developing intelligent systems that can move around in and manipulate their environments. I try to understand such systems through a combination of techniques from computer science, mathematics, and engineering. In order to plan a task, one must first be able to reliably predict the outcomes of the application of possible actions. For example, if a robot pushes against a box on the floor, we should be able to predict if the box will stick, slide, or tumble. I have worked to develop new multibody dynamics algorithms to answer this sort of question. Once we can make such predictions, we can then think about designing a series of actions to accomplish a specific goal such as, "Put new staples in the stapler." One common approach to this type of problem is to decompose the possible states of the stapler into equivalence classes, with each class represented by a node in a graph. The nodes can then be connected with directed arcs determined by predicting the outcome of the application of basic manipulation actions (such as "push left") to the states. Once the arcs are determined, the planning problem can be solved by searching the graph for a path connecting the nodes containing the starting and goal states of the stapler.

A new research interest of mine is human-robot interaction.  The motivation is still the same – build robots that can manipulate their environments – but do this with consideration for humans that share space with robots.  Robot controllers must be designed that are “human-friendly.”

The daVinci Project
In the DaVinci project (the acronym came from our NSF grant entitled, “Differential Algebraic and Variational Inequalities for Engineering Applications”), my colleagues I are studying a new class of problems, differential algebraic inequalities, which are useful in modeling dynamic systems with unilateral constraints, such as robotic systems that contact their environments.  Below are some of the results obtained on this project under the support of the National Science Foundation.

Human-Friendly Robot Controllers
This work is a piece of a larger effort on Robotic Sensor Networks conceived and led by Prof. Volkan Isler.

Here are some other problems I've worked on: