Optimal Transport for Visual Control and Tracking

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Tannenbaum, Allen ECS-0137412 The research program outlined in this proposal is driven by the need to develop novel means of acquiring, storing, manipulating and synthesizing signals and images for use in a feedback loop. This is the essential mandate of controlled active vision in which one uses a combination of techniques from systems and control, signal and image processing, as well as a computer vision for this purpose. The design of novel techniques for using visual information in control systems appears in a number of practical systems problems ranging from remote controlled weapons and vehicles to telesurgery, manufacturing systems, and ATR. Moreover, these efforts are leading to enhanced man-machine interfaces for interactions with computers. In this research, the PI will consider certain novel methods based on the notion of "optimal mass transport" for such problems in controlled active vision. This type of technique has appeared in econometrics, fluid dynamics, transportation, statistical physics, shape optimization, expert systems, meteorology, and nonlinear control analysis and synthesis.