Research

With the advent of hand held devices wireless networks are struggling to deliver all the traffic that is demanded by mobile users. Future wireless networks are likely to be ultra-dense networks formed by a large number small-cell base station with small coverage areas. Due to practical and cost-related limitations small-cells are connected to the core network through a low capacity backhaul link, which can jeopardize the capacity of the whole wireless network.
We investigate techniques and algorithms for pushing popular content close to the mobile users, that is, caching popular content at the network edges (i.e, in small-cell base stations) in order to reduce delay and alleviate congestion in the backhaul link of these small-cell base stations.
Our research focuses into two areas: 1) fundamental limits of content caching by using tools from information theory and coding, and 2) practical algorithms for content caching: by using techniques from multi-armed bandit problems we design algorithms that learn the popularity profile while optimising the cache content.

Information and communication technologies are extremely integrated into our everyday environment. Embedded computers and networks monitor and control physical processes, actuators and sensors, and there are generally feed-back loops through which the physical world infuences the digital one and vice versa. Such interconnected systems are known as cyber-physical systems (CPSs). They bring many advantages, like energy savings, reduced operational and consumption costs, improved reliability. However, they collect and spread a massive amount of users' information. Thus, CPSs raise the issue of data security and especially privacy, since they may lead to the exposure of private details of our daily lives.
In our research group we study the theoretical limits of privacy-preserving techniques and we try to improve them. We focus on smart grids and we consider the presence of an energy harvester in the system in order to mask the real users' energy consumptions.

We investigate communications protocols for energy harvesting wireless networks. Our research focuses on the design of communication strategies that make an intelligent use of the available energy at the wireless nodes in order to guarantee exceptional communication performance.
Our research covers three major areas, which depend on the amount of information about the energy availability at the node, namely 1) offline optimization (predictable energy availability), 2) online optimization (energy availability is random and its distribution is known, 3) learning optimization (energy availability is random and its distribution has to be learn). We study communication for a number of communication system topologies, such as relay communications, multi-access channel communication, as well as point-to-point with energy harvesting.
We use techniques coming from Machine Learning and Operations research such as Reinforcement Learning and Multi-armed bandit optimization for the communication strategy design and node operation.

Please check the IEEE Special Interest Group.

The optimality of the source channel separation theorem does not generalize to multi user communication systems, but for a few special cases. So, for general problems like lossy transmission of the sources over multiuser networks, we have to optimize the source and channel coding jointly. Similarly, even in some point to point wireless communication scenarios, in which the exact value of the signal-to-noise ratio (SNR) at the receiver is not known at the transmitter, source channel separation theorem fails. In these scenarios, the transmitter might only know a range of SNRs at the receiver and tries to send the information such that the receiver reconstructs the source signal with the minimum possible end-to-end distortion. Generally, there are two ways of designing such a system: i) using analog transmission methods, and ii) using digital transmission methods. Analog transmission methods provide gradual quality degradation of the received signal with the decreasing SNR. However, these systems are in general suboptimal, in the sense that, they rarely achieve the optimal theoretical bounds. On the other hand, digital system are not only easy to implement, but they can be designed to asymptotically achieve the theoretically optimal performance. Unfortunately, digital systems suffer from threshold effect, meaning that when the SNR of the channel output is lower than the designed SNR, receiver fails to decode the information, and when the channel output SNR is higher than the designed SNR, the quality of the decoded information does not improve. An important practical application area for JSCC, is wireless sensor networks (WSN).