2016/12/05 by Michael Heindlmaier, Heindlmaier, Michael, Shirin Saeedi Bidokhti +1
Computer Science · Engineering · #Advanced MIMO Systems Optimization #Advanced Wireless Network Optimization #Cooperative Communication and Network Coding #FOS: Computer and information sciences #Information Theory (cs.IT)
paper · pdf · doi:10.48550/arxiv.1612.01487
openalex publication_date 2016/12/05 · openalex created_date 2022/10/05 · openalex updated_date 2026/07/28
The two-receiver broadcast packet erasure channel with feedback and memory is\nstudied. Memory is modeled using a finite-state Markov chain representing a\nchannel state. Two scenarios are considered: (i) when the transmitter has\ncausal knowledge of the channel state (i.e., the state is visible), and (ii)\nwhen the channel state is unknown at the transmitter, but observations of it\nare available at the transmitter through feedback (i.e., the state is hidden).\nIn both scenarios, matching outer and inner bounds on the rates of\ncommunication are derived and the capacity region is determined. It is shown\nthat similar results carry over to channels with memory and delayed feedback\nand memoryless compound channels with feedback. When the state is visible, the\ncapacity region has a single-letter characterization and is in terms of a\nlinear program. Two optimal coding schemes are devised that use feedback to\nkeep track of the sent/received packets via a network of queues: a\nprobabilistic scheme and a deterministic backpressure-like algorithm. The\nformer bases its decisions solely on the past channel state information and the\nlatter follows a max-weight queue-based policy. The performance of the\nalgorithms are analyzed using the frameworks of rate stability in networks of\nqueues, max-flow min-cut duality in networks, and finite-horizon Lyapunov drift\nanalysis. When the state is hidden, the capacity region does not have a\nsingle-letter characterization and is, in this sense, uncomputable.\nApproximations of the capacity region are provided and two optimal coding\nalgorithms are outlined. The first algorithm is a probabilistic coding scheme\nthat bases its decisions on the past L acknowledgments and its achievable rate\nregion approaches the capacity region exponentially fast in L. The second\nalgorithm is a backpressure-like algorithm that performs optimally in the long\nrun.\n