Automated negotiation for opportunistic direct cooperation between neighbouring wireless sensor networks

Abstract

As the Internet of Things grows, multiple wireless sensor networks (WSNs) are likely to coexist. From wearable health monitors to smart cities, WSNs will play an increasingly key role in most scenarios. In many of these applications, sensor nodes are likely to be battery-powered and hence limited in energy supply. Energy harvesting technologies have gained widespread attention to increase node lifetime. However, these exhibit spatio-temporal variations and expose a discontinuous power supply. Visioning a future with cooperative networks, this work proposes to extend network performance optimisation to an inter-domain approach by opportunistic cooperation among WSNs that share a common area. Since WSNs are highly heterogeneous and self-interested, cooperation is not guaranteed. The cooperation problem has been addressed using a game-theoretic approach. However, assumptions as full rationality or complete knowledge are not justified in this domain. Instead, this work utilises multi-agent design methods to provide a new methodology on negotiation-based cooperation that enables suitable agreements on energy sharing. With the aim to optimise a network’s power management using the suggested approach, a node’s own efficiency is computed. Thus, a self-organising algorithm capable of making optimal use of harvested energy is proposed. This power management technique is tested during every simulation presented. Such an algorithm enables self-organised nodes that can anticipate insufficient energy allocation schemes and identify the opportunity to start an energy negotiation (OEN). Experiments show the accomplishment of energy-neutrality when networks find energy flow agreements and adopt conciliatory behaviours. The effect on the power consumption and latency of establishing OEN is also quantified and proved to be insignificant (<0.01 J, <0.1 s). A novel partner selection method based on multi-armed bandits is also introduced to facilitate the estimation of successful negotiation agreements. The proposed model allows networks to maximise their energy allocation in the long run, while adapting to a highly dynamic and uncertain environment. The viability of the approach is measured through simulation, and results show that networks may improve their energy allocation by over 40% in the most challenging scenario considered in this thesis.

More information

Identifiers

ORCID for Geoffrey Merrett: orcid.org/0000-0003-4980-3894

Catalogue record

Export record