Wireless networks share a common channel that is spatially reused. This gives rise to interesting and complex interference phenomena that have substantial effects on performance and fairness. The shared channel makes wireless networks a strongly coupled system where decisions by one node can have a substantial and cascading effect on other nodes in the system. Developing effective distributed protocols in such an environment is challenging since local decisions can have significant effect making it difficult to converge to effective operating points.
SDN is establishing itself as a game changing technology in managing network infrastructure in a number of networking domains. SDNs makes it possible to carry out decisions that are globally effective and introduce the level of coordination not possible with distributed protocols. In addition, an SDN framework offers advantages with respect to service deployment, network monitoring and instrumentation as well as security.
In this research project, we review some of the recent developments in wireless mobility, investigate the possibility of predicting paths to pre-empt access point handoff and investigate the feasibility of using Raspberry PIs, a lowcost credit card-sized single-board computer, as an SDN enabled wireless routers for a Wireless Mesh Network.
Read the full proposal [PDF]
Read the full report [PDF]
The existence of accurate semantically rich indoor maps can lead to a significant growth in indoor, location based applications. In recent year the problem of indoor mapping has received a lot of attention but, while the existing systems can generate very accurate floorpans, the do not provide semantic tags for the spaces they map. Without knowledge of the environmental context of the user location based application would remain highly limited in their efficacy.
We propose using the acoustic response generated by the user's speech to infer the characteristics of his environment. We are working toward building a system that can detect when the user is talking, records a snippet of his speech and, using learning algorithms, infers the semantics of his environment.