Future communication networks will incorporate capabilities for routing encoded data to its destination along multiple simultaneous paths, in order to increase security, data rate, and reliability. A significant fraction of the data traversing these future networks will be associated with delay sensitive applications. Such applications include streaming video, video and teleconferencing, and remote control over a network, all of which will be required for other important future applications such as telepresence and telemedicine. There is a significant hurdle, however, in routing data for delay sensitive applications over several paths due to randomly varying delays, and packets of information arriving out of order at the destination. This research solves the important problem of matching delay intolerant applications to best effort multi-path routed and coded networks all the way from a fundamental perspective to implementation and testing on a network testbed to human interaction testing. The public awareness of live streaming technology and the work being performed will be enhanced through live streaming Philadelphia Orchestral concerts at Drexel University. The problem of matching delay sensitive applications to best effort multipath routed network architectures involves a clear fundamental tradeoff: the redundancy required to guarantee a decoding delay in spite of network packet reordering, delay, and loss comes at the cost of a reduced application data rate. This project investigates this fundamental tradeoff over all end-to-end coding schemes using information theory. The developed models and delay mitigating code transport protocol will be evaluated on a network tested, and the perceptual performance provided by applications running over the developed transport protocol are measured via an educational interactive web tool.
As communications networks and internetworks continue to grow and modernize, and an increasingly large portion of the traffic that they carry is video and multimedia signals, technologies like multipath routing, which routes packets between a source and a destination over multiple paths, and network coding, which enables routers to not only store and forward user information, but also to combine and code it, will grow in use. This project pointed out that time sensitive information broken up into packets being routed over multiple network paths could be encoded in ways that optimize the tradeoff between the average rate of information reception and a delay measuring deviation from a prescribed temporal information delivery profile consistent with this rate. It was shown that the study of the tradeoff, optimized over all codes, between this rate and delay, can be obtained by solving fundamental problems involving the region of entropic vectors and the capacity regions of graph networks. To attack these underlying fundamental problems, which also have applications in distributed information storage and in the limits of certain approaches to machine learning, novel computational methods and software were developed that can bound the region of entropic vectors, and can utilize these bounds to calculating the capacity regions of small graph networks. The results of the research were published in prestigious journals, and the software developed was made available to the public.
