Fall 2012 - CATT Graduate Research Seminar VI


CATT Graduate Research Seminar VI
Wednesday, Nov. 7th, 2012, 12:00 pm – 1:00 pm
CATT Conference Room, Dibner Building LC-218

Speaker 1: Yu Jiang                                         [Advisor: Prof. Zhong-Ping Jiang]

“Robust adaptive dynamic programming and its applications in Biomedical and Power Systems” (Download Presentation)

Robust adaptive/approximate dynamic programming (robust-ADP) is a new methodology recently invented at NYU Poly, which aims to address emerging challenges arising from engineering and biology, such as biomedical and power systems. This method incorporates ADP in the past literature with tools from nonlinear control theory and is especially suitable for the optimal control design of uncertain interconnected systems. Specifically, novel tools for the computation of asymptotically stabilizing and suboptimal feedback control policies will be presented via online learning techniques. Finally, we will demonstrate the powerfulness of this methodology by means of a glucose regulation system and a multi-machine power system.

Speaker 2: Mustafa Riza Akdeniz                                 [Advisor: Prof. Sundeep Rangan]

“Wireless Scheduling with Dominant Interferers and Applications to Femtocellular Interference Cancellation” (Download Presentation)

We consider a general class of wireless scheduling and resource allocation problems where the received rate in each link is determined by the actions of the transmitter in that link along with a single dominant interferer. Such scenarios arise in a range of scenarios, particularly in emerging femto- and picocellular networks with strong, localized interference. For these networks, a utility maximizing scheduler based on loopy belief propagation is presented that enables computationally efficient local processing and low communication overhead. Our main theoretical result shows that the fixed points of the method are provably globally optimal for arbitrary (potentially non-convex) rate and utility functions. The methodology thus provides globally optimal solutions to a large class of inter-cellular interference coordination problems including subband scheduling, dynamic orthogonalization and beamforming whenever the dominant interferer assumption is valid. The paper focuses on applications for systems with interference cancellation (IC) and suggests a new scheme on optimal rate control, as opposed to traditional power control. Simulations are presented in industry standard femtocellular network models demonstrate significant improvements in rates over simple reuse 1 without IC, and near optimal performance of loopy belief propagation for rate selection in only one or two iterations.