Infrastructure Free City


What are the protocols and algorithms required for next generation opportunistic wireless networks?


Imperial AESE, ETH

Fixed infrastructures have limitations regarding sensor maintenance, placement and connectivity. Employing the ubiquity of mobile phones is one approach to overcoming some of these problems whereby the phone carries the data. This concept, “Data Mules,” explores how mobility and social patterns of phone owners can be exploited to optimise data forwarding efficiency.

This is innovative in this research space as prior work in Delay Tolerant Networks (DTNs) typically focus on individual packet routing rather than routing and control for streams of packets. Further no prior work has fully exploited underlying social networks pertaining to human relays, nor have they used opportunistic multi-hop-multi-radio human contacts.

Incentivisation is an important feature of this research in order to stimulate phone owners to serve as data relays.  We proposed a scheme which mimics a free market in which those who mule data get rewarded. By combining network science principles and Lyapunov optimisation techniques, we have shown that global social profit across a hybrid sensor and mobile phone network can be maximised. In order to minimise cheating the system we implemented Mechanism Design to weight the pay-back and show that cheating the system never results in reward.

The approach we have taken aligns with our principles to embrace distribution and agility in systems architecture. Therefore our algorithm is fully distributed and makes no probabilistic / stochastic assumptions regarding mobility, topology, and channel conditions, nor does it require prediction.

Phase 1:
Ad hoc experimentation was carried out in order to reason about this work and was compared with theoretical predictions and simulation in Hyde-Park using Mica-Z motes. This work demonstrated that in theory we can outperform other similar approaches using less computational resources.

Phase 2:

Experiments were carried out on Android mobile phones with a study of 15 student users relaying data across Imperial College to test the ability of the algorithms to ensure a profit for all and at the same time efficiently relay data. This highlighted that indeed the profit-based economic model was viable but that connection times using WiFi-Direct for mobile phones was problematic.