MOBIX: System for managing mobility using information exchange

Abstract

It is evident that mobile devices of the future will have multiple wireless interfaces. For small, energy-constrained devices, determining network availability by keeping all radio interfaces turned on at all times will negatively impact battery lifetime even when these interfaces are idle. Predicting future network availability from user history requires a period of training and learning user habits. This method will fail when users deviate from their routines constantly or move to locations not visited before.We propose a different approach to determining network availability of mobile nodes which leverages on the fact that nodes on the move will meet other nodes who will be able to share conditions of networks they have recently encountered. This paper presents MOBIX, a system where nodes exchange information about network conditions using short-range communication such as Bluetooth. Our simulation results show that the required number of nodes needed for 100% success is not unrealistic of densely populated metropolitan areas. Even with relatively low population densities, we can expect a data store hit more than 50% of the time. Although our evaluation used WiFi, our scheme can easily be extended for other technologies such as GSM and WiMax.

Network availability prediction: Can it be done?

“…The information collected is shared with the other users who come within communications range. [16] shows how this can be effective in realtively densely populated environments. he information exchange methods are fully distributed and does not require updates and distribution.…”

“…In the following sections we summarise some of our previous work [23], [16] which show the viability of the proposed scheme. …”

“…With this scheme, uses the reports described in [16] to augment the local measurements for the determination of context variables described. Then use a Dynamic Baysean Network with hidden variables as described in [23] to predict the performance.…”

“…As described in [16] we used the information shown in Table III . It is assumed that there exists some function Y (repId R ) → k that returns the type of node k which generated report R with unique ID repId R .…”

“…[24] shows that it is possible to take "hints" from other layers to make better decisions. [16] shows that it is possible to use information from other nodes again to improve the decisions. [22] has shown that it is possible to use prediction to improve the decisions.…”