Seminar by Dr. Francesco Maria Delle Fave
Title: Theory and Practice of Coordination Algorithms exploiting the Generalised Distributive Law
Abstract
The purpose of this seminar is to present some recent research achievements
on coordination technologies to allow interacting computer systems,
typically referred as agents, to coordinate their decisions whilst
operating in an environment with minimal human intervention. In more
detail, I will describe new problems and algorithms to address two key
research challenges.
The first challenge is practical in nature. All the current
state-of-the-art coordination algorithms have only been tested in
simulation. Thus, their practical performance still needs to be
demonstrated in the real world. To address this shortcoming, I will present
a case study on the deployment of the max-sum algorithm, a well known
coordination algorithm, on a system that is couched in terms of allowing
the first responders at the scene of a disaster to request imagery
collection tasks of some of the most relevant areas to a team of unmanned
aerial vehicles (UAVs). These agents then coordinate to complete the
largest number of tasks.
In contrast, the second challenge is a theoretical one. Indeed, none of the
existing algorithms are capable of solving problems where the agents need
to coordinate over complex decisions which typically require to trade off
several parameters such as multiple objectives or the parameters of a
sufficient statistic. To address this shortcoming, I will describe a new
framework, namely partially
ordered distributed constraint optimisation problems (PO-DCOPs) and a set
of solution techniques which employ the flexibility of the generalised
distributive law (GDL) to obtain either optimal or bounded approximate
algorithms to solve PO-DCOPs. Finally, I will present an empirical study of
some of fundamental properties of these algorithms on two instances of
PO-DCOPs to address problems involving multiple objectives and uncertainty.
Short Bio
I am a research associate (postdoc) at the University of Southern California (USC) (Los Angeles, USA). I work in the TEAMCORE group led by Professor Milind Tambe. My research focuses on deriving new solution techniques to solve large scale security games in a fast and effective manner. A key focus of my research is also to deploy such techniques into real systems capable of deriving patrol strategies to the security forces deployed to control transportation infrastructures such as ports, airports and trail stations.
Before attending USC, I was a PhD student at the University of Southampton (Southampton, UK). I was working in the Agents, Intercations and Complexity group (AIC) under the supervision of Professor Nick Jennings and Dr. Alex Rogers.
Before starting my PhD, I completed my studies at the "Sapienza" University of Rome, Italy, where I graduated and completed a masters in Computer Engineering.
During my studies, I became interested in the broad field of artificial intelligence. More specifically, I became interested in the field of multi-agent systems (MAS), a research area which studies problems where a potentially very large group of (robotic or software) agents interact in either a competitive or cooperative environment to solve a problem. Research in this area sits at the nexus between several disciplines such as computer science, machine learning, robotics and economics. For this reason, MAS have been studied from both a theoretical and a practical perspective.
The purpose of my work is then to produce both theoretical and practical studies capable of bridging the gap between research and real world systems.