WPR.10 - Network Theory

Objectives

The objectives of this WP are:

• Address fundamental theoretical aspects which are essential for wireless networks, including scaling laws and asymptotic behaviour of wireless systems when users and nodes dramatically increase; theoretical bounds to such metrics as capacity, network lifetime, minimal connectivity degree; trade-offs between in-network processing and communication complexity; and others. The objective is to study aspects of the Future Internet such as sensory networks based on derived results
• Develop a network information theory for understanding the theoretical foundations necessary for quantifying fundamental performance limits of wireless networks in capacity, throughput, and delay and devise techniques to closely approximate and even achieve them.
• Understand fundamental tradeoffs and dependences in wireless networks, including the interplay between capacity, energy consumption, stability and delay. To this end, information theory will be combined with non-classical constraints such as finite battery life, traffic characteristics, topology and mobility.
• Devise distributed optimization and game-theoretic methods for distributed self-regulating network paradigms, ranging from wireless sensory ones to wireless overlays and autonomous computing. Peer-to-peer networks will be used as an important class of networks.

Description of work

The work in this WP consists of the following tasks:

Task TR10.1: Exploration of state-of-the-art in network information theory

In this task we will extensively study and report existing research work and knowledge in network information theory with a view towards characterizing fundamental performance limits. This task will serve as a prelude for the other tasks of the WP.

Main players: IASA, Technion, all

Task TR10.2: Characterization of achievable throughput and delay performance limits

We will address and resolve the fundamental two-fold question: (i) how much capacity can we transfer through a network, (ii) how should we control a network in order to achieve this capacity. Asymptotic results will be provided and scaling laws will be identified, that characterize the dependence of throughput, delay and energy consumption on the number of nodes and resource allocation algorithms. We will also investigate the dependence on these metrics on statistical and deterministic network and topological parameters, such as node density, spatial distribution, node degree, connectivity. This task will dwell into characterizing performance limits both for sensory networks and hybrid networks. Different modes of transmission that are applicable to current and evolving standards (such as OFDMA, SDMA etc) will be studied and their impact towards shaping these limits will be investigated.

Main players: IASA, CNIT, all

Task TR10.3: Methods for achieving throughput and delay performance limits

In this task, we will design and evaluate methodologies to approach or achieve throughput and delay performance limits. We will address and quantify fundamental arising tradeoffs between required overhead and achievable performance. We will devise policies that satisfy maximum possible demand for data throughput under given resource constraints. The methods will operate in real time and will allow for optimal dynamic adjustments to changes in channel conditions and traffic patterns. We will dwell upon network aggregate utility as perceived by the network operator and will find the optimal tradeoff between maximizing utility and minimizing delay. We will control this tradeoff between wireless network capacity and wireless network delay at will. Specific subtasks include: (i) Light-weight backpressure-inspired resource allocation techniques for pushing the limits of information flow transfer over the network. (ii) Quantification of performance benefits of multi-layer-aware designs in network-wide throughput.  (iii) Use of tools from convex optimization theory to develop efficient and provably convergent methods for achieving various objectives. Besides throughput and delay objectives, we will consider optimization and control for other meaningful objectives such as that efficient detection of an event, action or motion. We will finally explore the interplay between detection performance and energy consumption. The objective is to study aspects of the Future Internet such as sensory networks based on derived results

Main players: IASA, CNIT, all

Task TR10.4: Theoretical foundations for  distributed network performance

The autonomous feature of envisioned wireless network architectures calls for establishing a theory that captures such an operation. We will model node interactions with game theoretic tools. Each node aims at maximizing its own utility, yet individual behavioral profiles of peers designate the collective behavior of the network. We will capture inherent competition of nodes for limited network resources such as bandwidth, energy, transmission power, service capacity, computational power, storage space. We will devise methods for network entities to act in an autonomous manner in optimizing their own utility, and capture how they affect utility of others. A first question concerns modeling strategies by which users independently take allocation decisions with no or partial knowledge. We will also investigate imposing mechanisms that effectively drive the network to equilibrium and we will use game theoretic tools and results to evaluate the performance of node strategies. Peer-to-peer networks will be used as an important class of real and operational networks whose performance will be evaluated.

Main players: IASA, CNIT, all