Stability bounds on entropy rate for real-time tracking an unstable LTI process over a multi-access network
Ching-Ling Huang, Raja Sengupta, Xu Guan · 2009
We assume the separation of source/channel coding and propose a framework for real-time tracking an unstable, scalar, linear time-invariant (LTI) process via a G/G/1-¿ queueing network. We derive a stability bound on the entropy rate of an unstable LTI process such that tracking error can stay bounded for specified moment. This bound is a function of the moment to be stabilized, encoder efficiency, quantization accuracy, and network parameters. This bound is sufficient under all load conditions and tight when the network is heavily utilized. Proposed framework covers a broad class of unreliable networks, and it helps us tackle the tracking problem over a multi-access network. For three commonly used channel access protocols, we explicitly express their stability bounds on entropy rate in terms of channel data rate, erasure probability, node number, etc.