Những Người Bạn

Descriptif:
Wireless networks are increasingly carrying video traffic.  A major difficulty of live video streaming in wireless networks is the variability in user data rates.  Data rates in wireless networks depend strongly on user location and environment.  As such, a constant feedback to the source is necessary in order to appropriately adapt the sending rate.  At the radio link, retransmissions for packet losses and acks increase congestion at a link which already suffers from limited capacity.  Further, in the case of multicast or broadcast transmissions such as TV, multiple transmissions of varying rates create redundancy and additional congestion at the radio link.

Digital fountain-like source coding with appropriate buffer management and bandwidth sharing has been proposed to address the issue of congestion control with feedback.  Such transmission mechanisms seem to be a natural fit for the wireless medium.  The source sends a stream of coded packets, whose order is not important.  Once the wireless destination has received a sufficient number of packets in order to decode, it signals to the source or the base station (in case of multicast or broadcast) an indication that it has received the data.

Considering the significant differences between wireless and wired networks, application of such source coding is not entirely straightforward.  The role of the base station as a simple relay or smart scheduler may need to be redefined.  Furthermore, latency requirements of video delivery may limit expected gains.  Evaluation of the performance gains of source coding in the wireless environment, and video applications in particular is crucial to a better understanding of such a system of delivery.

The objective of the proposed thesis will be the evaluation of performance gains of source coding techniques for live streaming delivery in wireless networks.  While performance analysis of source coding techniques for data has been initiated, many question remain concerning fairness policies, scheduling policies in intermediate nodes, and latency requirements for streaming applications, in addition to the applicability of such mechanisms in wireless networks.  A major portion of the thesis will be devoted to the performance analysis of the impact of digital fountain-like source coding techniques on network congestion, with the aim of characterizing gains, in both a live video streaming scenario and a wireless environment.    The results will potentially uncover system parameters that are crucial to performance gains.  The performance study will naturally lead to the proposals of mechanisms for better design.

The evaluation of performance gains will begin with an appropriate model of the system that includes the random nature of traffic.  The specificities of wireless networks will be included as seen fit.  The study will then proceed on two fronts:  analysis of the applicability of source coding without explicit congestion control to the wireless scenario; and the analysis of such a source coding mechanism for streaming applications with latency requirements.
The thesis will begin with a bibliographical study of the state of the art.  Once a proper understanding the concepts at play is established, depending on the abilities of the student, analysis or simulations will be carried out.  This phase is expected to last for a long duration, and will be followed by the characterization of design principles and new mechanisms for delivery of live streaming services over wireless networks.

It is important that the candidate be knowledgeable on networks in general, and wireless networks in particular, including congestion control schemes.  The candidate must  be very comfortable developing simulations and have a strong  interest and aptitude for applied probability.

The thesis work will be carried out at Orange Labs R & D, in Paris, France.  Note that this is a fully funded Ph.D position.

Contact:  Nidhi Hegde, Orange Labs
nidhi.hegde@orange-ftgroup.com