Optimal control with limited communication: modeling, tools, and experiments with a dexterous robot

Prof. Dimitris Hristu, University of Maryland, Department of Mechanical Engineering

Control and communication issues have traditionally been decoupled in discussions of decision and control problems because this simplifies the analysis and generally works well for classical models. Recent work has begun to re-examine this fundamental assumption as control applications spread into modern engineering systems, including smart structures, communication networks, and formations of autonomous robots. Such systems are distributed, in the sense that their actuators, sensors and computing elements communicate via a shared medium, be it a radio frequency, a computer bus or pins on a MEMS device. Restrictions on access to this medium often have a profound effect on the performance of the overall system, underscoring the need for analytical tools that bridge communication and control.

Using real-time manipulation as a benchmark problem, I will discuss recent work involving a language-driven manipulator. This dexterous ``hand'' relies heavily on vision, not only for tracking objects in its workspace but also for tactile sensing. However, the availability of a ``rich'' set of sensors and actuators comes at a price: components must ``share the attention'' of the controller, thus limiting performance due to lack of time. I will present a framework for understanding this interplay of communication and control, using an experiment in trajectory tracking to highlight the key ideas.