Communications in Mobile Wireless Networks: A Finite Time-Horizon Viewpoint

Abstract

In mobile wireless networks (MWNs), short-term communications carry two key features: 1) Different from communications over a large time window where the performance is governed by the long-term average effect, the short-term communications in MWNs are sensitive to the instantaneous location and channel condition caused by node mobility. 2) The short-term communications in MWNs have the finite blocklength coding effect which means it is not amenable to the well-known Shannon's capacity formulation.

To deal with the short-term communications in MWNs, this thesis focuses on three main issues: how the node mobility affects the instantaneous interference, how to reduce the uncertainty in the locations of mobile users, and what is the maximal throughput of a multi-user network over a short time-horizon.

First, we study interference prediction in MWNs by proposing and using a general-order linear model for node mobility. The proposed mobility model can well approximate node dynamics of practical MWNs. Unlike previous studies on interference statistics, we are able through this model to give a best estimate of the time-varying interference at any time rather than long-term average effects. In particular, we propose a compound Gaussian point process functional (CGPPF) in a general framework to obtain analytical results on the mean value and moment-generating function of the interference prediction.

Second, to reduce the uncertainty in nodal locations, the cooperative localization problem for mobile nodes is studied. In contrast to previous works, which highly rely on the synchronized time-slotted systems, this cooperative localization framework we establish does not need any synchronization for the communication links and measurement processes in the entire wireless network. To solve the cooperative localization problem in a distributed manner, we first propose the centralized localization algorithm based on the global information, and use it as the benchmark. Then, we rigorously prove when a localization estimation with partial information has a small performance gap from the one with global information. Finally, by applying this result at each node, the distributed prior-cut algorithm is designed to solve this asynchronous localization problem.

Finally, we study the throughput region of any MWN consisting of multiple transmitter-receiver pairs where interference is treated as noise. Unlike the infinite-horizon throughput region, which is simply the convex hull of the throughput region of one time slot, the finite-horizon throughput region is generally non-convex. Instead of directly characterizing all achievable rate-tuples in the finite-horizon throughput region, we propose a metric termed the rate margin, which not only determines whether any given rate-tuple is within the throughput region (i.e., achievable or unachievable), but also tells the amount of scaling that can be done to the given achievable (unachievable) rate-tuple such that the resulting rate-tuple is still within (brought back into) the throughput region.

This thesis advances our understanding in communications in MWNs from a finite-time horizon viewpoint. It establishes new frameworks for tracking the instantaneous behaviors, such as interference and nodal location, of MWNs. It also reveals the fundamental limits on short-term communications of a multi-user mobile network, which sheds light on communications with low latency.