Brief description of my current project :

Efficient Video Multicast over WLANs - MAC layer optimizations

Problem Statement: The challenges for real time video streaming over Wireless LANs are limited bandwidth, high bit error rates as compared to wired links and fluctuations in channel quality. In addition, we also have to consider receiver heterogeneity as each user will have different channel conditions, power limitations and processing capabilities. The video quality at each of the received mobile terminals would depend on the channel condition between the AP and the respective mobile terminal (we assume the video server and the access point are co-located). Receiver heterogeneity can be dealt in one of the two ways: simulcasting the content at different rates or by layered coding [which forms the basis of Receiver Driven Layered Multicast scheme pioneered by McCanne]. The ways to improve the video quality at the mobile terminals would be to employ forward error correction (FEC) schemes so that the receivers can reconstruct lost packets from the additional redundant packets. Another option is to use ARQ. However in multicast scenarios there is a possibility of an implosion of repeat requests at the sender if ARQ is used. Our work is focused on MAC layer multicast and error control techniques that provide service quality and/or capacity improvements.

Our Approach: In our system model we have the video server co-located with the AP and N mobile terminals. We consider that multi-resolution video streams at four different rates 1.5 Mbps, 768 kbps, 384 kbps and 128 kbps are available from the sever i.e. the AP is multicasting video streams at four different rates. The mobile terminals subscribe to one of the multicast streams based on their PHY bit rate, bit error rate and SNR measurements. We have formulated the problem as a general feedback control system with the following observation and control variables: Bit Error Rate (long term and short term BER), SNR, MAC multicast groupings and Percentage/type of FEC and ARQ used. The algorithm under consideration aims to maximize the total system utility, Q_system=sum(Q_i) i=1,2...N where N is the total number of mobile receivers, subject to the condition Q_i >= Q_threshold by selecting the appropriate multicast group and also adjusting the FEC and/or ARQ) as a function of the observed PHY bit rate, BER and SNR for each wireless client.

Simulation Tool used: OPNET

Brief description of my previous project :

NETWORK TOMOGRAPHY

Network Tomography - 4 Node Problem

Network Tomography -How to choose a subset of paths from the source node to the  destination node in a network for estimating link success probabilities

PAPERS

Performance Analysis of Downlink Power Control in CDMA Systems (accepted for the IEEE Symposium Digest for Student Papers/Posters, 2004 )

Rapid deployment emergency infrastructure for a disaster site with multiple types of emergency personnel (police, fire, national guard, etc)

Image Histograming and Connected Components Labeling in Parallel

PRESENTATIONS

Talk at Thomson Multimedia Research Lab, Princeton on March 1, 2004

Talk at WINLAB on March 5, 2004 (Part of Special Problems Presentation)

Talk at the Networking Group Meeting, WINLAB on October 7, 2003

Talk at the Networking Group Meeting, WINLAB on February 3, 2004