home search a-z help
University of Cambridge Computer Laboratory
Thursday Feb 8th, 2007 - 4.30pm
Computer Laboratory > Research > Systems Research Group > NetOS > Seminars > Thursday Feb 8th, 2007 - 4.30pm

Understanding network requirements for mass socio technical sensor networks

Milena Radenkovic

There has been a proliferation of interest in ad hoc networking for providing communications among users and devices where essentially no pre-planned infrastructure is present. The large majority of the state of art research has focused on wireless sensor networks in remote areas, military and emergency response scenario, and interplanetary networking. In contrast, this talk is concerned with different ways in which many humans and sensors can collaborate as part of a mass 'socio technical sensor network' on the top of potentially disconnected environments. We desribe PARTICIPATE project that explores how pervasive computing can support mass participatory campaigns in which millions of people gather and share information about their local environments, contributing to a national picture of broad environmental issues, raising awareness, and encouraging debate. We aim to develop an underlying platform that supports such campaigns by integrating multiple devices, including mobile phones, handheld computers, broadband PCs and specialised sensors, with multiple communications technologies, including wireless and fixed networking and also broadcast television. The challenges facing the networking community for supporting distributed storage and processing of information in such mass-scale applications are fundamentally profound, especially where they involve mobile sensors, diverse sensing technologies, span multiple organisations and disconnected regions. We discuss the suitability of P2P overlays to help build applications on the top of mobile and unstable environments, and the applicability of delay tolerant network approaches to routing to help provide more reliable data communication. Building on the already published accounts of real field observations of large scale sensor networks and our own observations from Participate trails as they emerge, we adopt a methodological approach that is based on mathematical modelling, simulations, iterative development and evaluation of multiple prototypes. We describe how we plan to capture key workload parameters and reveal a gaming approach to help us build the workload model. The talk then identifies key requirements and objectives for the optimal routing protocol for mobile ad hod networks used to support such applications.