Skip to main content

Non-equilibrium Information and Capacity Envelopes: Towards a Unified Information and Queueing Theory

Final Report Summary - UNIQUE (Non-equilibrium Information and Capacity Envelopes: Towards a Unified Information and Queueing Theory)

Recent trends like the Tactile Internet, cyber-physical systems, and the Internet of Things bring about various new applications in fields like robotics, road traffic, smart grid, telepresence, and virtual/augmented reality. A precondition is real-time human-to-machine and machine-to-machine interaction. This places major demands on the latency and reliability of future wired and wireless networks. The UnIQue project provides methods for network analysis that enable understanding the performance of these systems.

A main achievement of the UnIQue project are performance models of traffic sources and communication systems and methods for their composition. The findings extend the classical capacity results in information theory to include the notion of delay. In brief, information theory uses the concepts of entropy, i.e. the capacity that is required for transmission of a source, and channel capacity, i.e. the capacity offered by a system. If the entropy is smaller than the channel capacity, it is concluded that error-free communication is possible in principle. The UnIQue project extends this fundamental statement and introduces capacity-delay-error boundaries to characterize the capacity, that achieves defined statistical delay guarantees. The capacity-delay-error boundaries of sources (capacity required) and systems (capacity offered) are obtained individually and are composed by addition.

Models of a variety of sources and systems have been derived, including source and channel coders, fading channels, interference channels, different types of hybrid automatic repeat request protocols, MIMO channels, and cognitive radio systems. Specific attention is paid to multi-hop networks and to recent multi-path protocols that can take advantage of parallel connections to increase throughput, increase reliability, or decrease delay by performing load-balancing or using redundant transmissions.

The analysis is performed using either time-domain models in a min-plus algebra or space-domain models in a max-plus algebra and provides asymptotic and non-asymptotic statistical performance measures. In addition, the transient performance of systems is studied using a non-stationary service model that is applicable, e.g. in the case of discontinuous reception for energy saving in cellular networks.

The modelling approach is complemented by measurement methods that explore ways to excite a network with probe traffic such that the available network service can be estimated from observations of the traffic. A comprehensive measurement study of productive cellular LTE networks has been performed, including both stationary and mobile measurements.

Applications of the models and theory include the design and operation of scheduling algorithms and adaptive algorithms that adjust encoding and transmission parameters during operation. In the UnIQue project, we considered applications regarding the transmission rate strategies of cognitive radio networks and queue-aware uplink scheduling in LTE networks.