Sponsors
Studying the dynamics of single and multiple molecules using confocal microscopy
The study of the trajectory of a single molecule can review many aspects of its motion, such as its diffusion coefficient, mechanisms of motion, and details on its interactions with the environment. Most single-particle tracking techniques rely on wide-field fluorescence microscopy. Even with ultra-fast CCD cameras, the temporal resolution for 3-D investigations is on the order of seconds. Methods involving confocal microscopes have temporal resolution on the order of microseconds or better. However, due to the small detection volume, information can only be obtained over a short period of time and about one molecule at a time.
We are developing techniques to track multiple fluorescing particles simultaneously in confocal laser-scanning microscopy. The goal is to obtain 3-D information on molecule motion and details about interactions between molecules over long periods of time. One focus of this work is a collaboration with Natalia Broude of the Boston University Center for Advanced Biotechnology. This collaboration is aimed at studying the kinetics of RNA transcription in living bacterial cells.
Non-raster approaches in AFM
The ability to study dynamics in systems with nanometer-scale features is a critical component for continued progress in a variety of disciplines, including molecular biology, medicine, materials science, and nanotechnology. Atomic force microscopy (AFM) has been an invaluable tool for investigating structure in such systems and is now being applied to explore motion as well. The standard approach is time-lapse imaging in which a sequence of images is acquired and post-processed to extract information about motion. Because images are acquired pixel-by-pixel, the time it takes to form a single image is typically on the order of seconds.This is far slower than the time scales of many important phenomena, severely restricting the applicability of the approach. As a result, there is great interest in improving the effective temporal resolution of the instrument.
Our approach focuses on the use of non-raster-scan methods. The standard raster-scan pattern can be viewed as an open-loop control scheme. However, by its very nature the microscope is acquiring information about the sample that can be used to drive the sampling process. We utilize this fact to develop fast-imaging algorithms and techniques for directly studying motion rather than structure.
Symbolic control of mobile robots
The design of feedback control laws that accomplish seemingly straightforward motion control tasks, such as navigation in environments of even moderate complexity, arguably remains a persistent challenge for automatic control. Efforts to overcome the complexity of specifying motion control tasks have included the development of symbolic approaches to control. Under this framework, one intentionally reduces the choice of feedback control laws to a small collection of specialized "primitives". The only allowable control inputs are compositions of symbols which are identified with primitives. Aside from reducing the complexity of a problem, this restriction carries intuitive appeal because tokenized descriptions are part of everyday experience. For example, a set of directions might include terms like "exit the room, turn right, walk down the hallway, and enter the third door on the right".
In many cases, multiple robots can perform a given task better than a single robot. Examples include data gathering, surveillance, and exploring unknown regions. We are interested in generating coordinate activity among multiple robots without explicit communication between them. By assuming the robots are limited to a common set of symbols, the problem is reduced to one of first identifying the current task, defined as a symbol sequence, being carried out by a robot, then predicting the future sequence, and designing a new control sequence for a second robot which will aid the first.
User manipulation of artificial potential fields
Artificial potential fields are a common tool for control and navigation of autonomous robots. A potential field is defined on the environment for which the goal region is given a low value while regions which should be avoided (such as obstacles) are assigned a high value. The robot is then steered according to a minimum-seeking control law. The utility of the approach is limited by the prevalence of local minima. In this project we are developing a user interface based on a haptics device to both feel and modify the potential field.