Brief description of projects that I have worked on. For further details please take a look at my publication list.
In the beginning of my PhD I spent a decent amount of time developing the Markovian Agent Spatial Stochastic Process Algebra (MASSPA), which simplifies the use of Davide Perotti’s and Marco Gribaudo’s Markovian-Agent formalism.
Mean field analysable models for Wireless Sensor Networks (WSNs)
As a spatial application for mean field analysable spatial performance models I looked at WSNs. These networks tend to have of a large number of individual nodes that communicate autonomously via radio. In particular I studied how mean field techniques can approximate flow of information in presence of interference as the number of messages produced by nodes becomes large.
Time inhomogeneous models for bicycle migration prediction
In this project I studied the use of white box CTMC based forecasting techniques in order to predict bike movements for bicycle hire schemes, such as the London Barclays Bike one. Moreover, I compare our modelling and analysis techniques with those for traditional regression and times series models that are predominantly used in the literature for bike sharing system analysis.
Hybrid models for infrastructure planning in large crowd sourcing systems
As part of the Intel PhD Fellowship program that I was awarded in 2013, I am looking at infrastructure analysis and planning models for large crowd sourcing systems. Examples of such systems would be a Smart City setup, where sensors attached to bicycles or cars, which could complement stationary sensors as a cheap, easy to maintain data source. In such models mobile agents sample data as they move in space, which is assumed to be uploaded whenever the agent reaches a wifi hotspot. Alternatively, agents can exchange data with other agents in a gossip like manner, so that the time to upload data depends on the passage time of the agent that has sampled the data, as well as on the time it takes agents that it has communicated with to reach an upload location. Analysing the performance of such systems naturally only makes sense, when assuming that network wide mobile internet coverage is infeasible, e.g. due to bandwidth or cost constraints. Having worked on the hybrid modelling approaches for representing and analysing performance of large systems, I am currently preparing a large case study. Ultimately, I envisage the resulting models and analysis techniques as a means to reduce expensive stochastic simulation, when having to try out different infrastructure networks in large, mobile, crowd sourcing networks.