This page contains a short list of topics for Laurea Theses
The list is not necessarily exaushive and may be also not frequently updated.
It is possible to contact me to evaluate other topics of reciprocal interest.
Digital and Wireless Communications
Topics are mostly focused on physical layer issues of wireless and wired communication systems (OFDM, LTE, 5G, WiFi, etc.).
In
the past we mostly focused on performance analysis of multicarrier
communication systems (OFDM, MC-CDMA, MC-DS-CDMA, etc.) both in linear
and non-linear (PA, A/D, etc) environments, subject to frequency
selective fading, frequency synchronization errors, impulsive noise,
etc..
Another widely investigated topic has been time-varying
channel estimation and equalization in wireless communication systems
subject to high Doppler spread, also in the presence of multiple
antennas (MIMO)., with an active scientific collaboration with Prof. G
Leus, from TU-Delft University, also within Erasmus + students
exchanges.
Below are highligted some of the physical layer
issues for future 5G cellular systems, which are expected to be
deployed worldwide by 2020, that may be objcet of a MS or BS
Thesis.
Signal Processing on Graphs and Distributed Signal Processing
Signal
processing on graphs (SPG) is a rather new theoretical subject, which
is rapidly emerging in the technical literature and where signal are
defined over a non-metric topological space such as a graph, rather
than a metric space such as the real/integer fields of classical signal
processing. This general tool, by capturing and exploiting the
structural relations among the nodes of the networks (graph), can
effectively model a wide class of phenomena such as the data or
parameters in a communication network, or the information associated to
nodes of a social network, or the measurements gathered by sensors in a
power-grid or a vehicular network, as well as the concentration of
proteins in a gene regulatory network.
Such a kind of data,
described as a signal on graph, are typically characterized by high
volumes (Big-data) in several application scenarios (communication
network, cloud, social networks, etc. SPG can be useful also in this
view because it is possible to extended to this new paradigm the
classical Shannon-Nyquist Sampling Theorem that, under mild conditions,
lets to reconstruct the overall data on all the network from samples
taken by a limited and opportunely selected number of nodes, with
controllable reconstruction error.
This fact actually suggests
that also the extraction of valuable, possibly statistical, information
from huge data-sets (big data analytics) may be possible by processing
only the data on a reduced number of nodes, by exploiting the
topological structure of the graph.
The SPG framework naturally
merges with adaptive and distributed signal processing, by modelling
each node (or part of them) not only as a source of information (data)
but also as a potential computational unit that can communicate with
neighborhoods nodes over a communication network. Thus, classical SP
algorithms, such as the LMS, RLS, Kalman filtering, and statistical
inference (e.g., estimation, detection, hypothesis testing, etc.)
can be extended to SPG to realize a decentralized computing and
communication network that is capable to execute complex tasks in a
distributed fashion, such as tracking the evolution of a dynamic field
or a moving target in a sensor network, infer the status and possibly
detect faults of a power-grid by a limited number of intelligent power
meters, as well as traffic analysis and control in a vehicular network,
or empower the functionality of an automated recommendation system,
such those used by Netflix for movies or by Amazon or Google for
electronic merchandising.
Our group is investigating this subject in collaboration wih Prof. S. Barbarossa, from University of Rome, "La Sapienza" .
Currently we are looking for motivated students that are willing to test and customize the general purpose SPG algorithms produced by the research community, on real applications such those highlighted before (enviromental sensor networks, power-grid networks, social networks, fault-detection in communication networks, fraud detection, vehicular networks, etc.).
The Thesis could be either signal and graph model characterization for a specific applications, data collection or access from public or private databases, developing of new algorithms for specific applications, as well as implementation of SPG algorithms on distributed, parallel, or cloud-based platforms.