Ypodimatopoulos, P. "Cerebro: Forming Parallel Internets and Enabling Ultra-Local Economies"

Abstract

Internet-based mobile communications have been increasing rapidly [5], yet there is little or no progress in platforms that enable applications for discovery, context-awareness and sharing of data and services in a peer-wise manner among collections of devices in the same physical area. This is important because proximate devices may need to communicate directly when no infrastructure is available, and because such local access may be an efficient alternative to connecting a large number of sensors, effectors, and people to a readily accessible, universal central system.

This thesis presents the design, implementation and evaluation of Cerebro, a system that allows suitably equipped humans and objects in the same physical area to discover each other and share data and services. Cerebro offers two basic services: a presence service that propagates information about local devices through an automatically generated mesh network and an innovative data transport service to transfer data via this network. On top of these services, Cerebro offers an extensible Application Programming Interface (API). In a mobile mesh network with N devices, Cerebro offers an upper bound of O(N) on traffic overhead to maintain presence information at any part of the network and responsiveness to device arrival/departure events that take at most O(N) time to propagate throughout the network. This makes Cerebro a scalable and useful addition to mobile service delivery.