Projects of Burchan Bayazit

OBPRM: An Obstacle-Based Probabilistic Roadmap Method

Although OBPRM project started as the implementation of the motion planning algorithm with the same name, it evolved into a motion planning library with several probabilistic roadmap algorithms included. (More detail)

Haptic Interaction in Motion Planning for Accessibility of CAD models

Usually, the accessibility of particular parts of a CAD design is tested using physical mock-ups of the system. This is highly expansive process. Our work, which treats the parts to be removed as robots and applies motion planning techniques, could ultimately remove the need for physical mock-ups by enabling such tests to be performed virtually on three--dimensional CAD/CAM models. Indeed, with a human operator's guidance, our automated planner was able to find accessibility paths for the parts in very confined situations. To help the human operator's perception of the environment we used a haptic device (PHANToM) which let the user sense the virtual object as if the user were touching it. Almost all applications of haptic interaction concentrate on simple representations of virtual objects. One of our achievements in this research was to develop an algorithm to provide a realistic sense of touch for very complex 3D virtual object pairs. We also suggested some simple visualization techniques which proved to be very helpful to show the planner's progress. (More detail)

Ligand Binding with OBPRM and Haptic User Input
A successful treatment of a disease requires an efficient drug molecule. Efficiency of the drug molecule (ligand) can be described as its ability to bind the correct site within the disease protein where it would disable the disease protein. In this case, the drug molecule is a very flexible robot. The problem is very high dimensional with constraints imposed by energy related, geometrical and biochemical properties. In this research we successfully generated configurations within a few A of the original site (which is considered to be in the binding site). Our approach has an advantage over most of the other well known docking algorithms since our algorithm does not require a priori knowledge of the studied molecules. Similar to our work on previous research, we also used a haptic device in this problem. Using the haptic device, an operator felt the molecular forces applied on the drug molecule as it was moved around the protein and suggested some binding sites. The haptic device also helped the user to understand the molecular interactions and the binding process itself which is an important factor in drug design. There are very few haptic applications ever implemented for this challenging problem and even in its preliminary stage, our implementation is promising. We have been collaborating with researchers from the Department of Biochemistry to improve our algorithms. (More detail)

Motion Planning for Deformable Objects
Although, the assumption that a robot is made up of rigid parts is generally correct, there are several cases where a deformable robot is preferable. In this research, we investigate the problems where such robots are desirable (especially in animation) and find solution to such problems by using roadmap techniques. (More detail)

Flocking Behaviors in Complex Environments
Group behavior can be observed everywhere. For example, birds fly in flocks, fish swim in schools, sheep move as a herd steered by a dog, and ants explore until they find a food source and then all ants follow the same path to the food source. The ability to simulate such behaviors plays an important role in computer animation, robotics and artificial life applications. In this research, we want to capture global information of the environment to simulate such behaviors in more complex environment. We are interested in Homing, Exploring (goal searching or area covering) and Shepherding behaviors. (More detail)