I started my academic career on September 2009, when I was firstly introduced in the laboratory of Prof. Daniele Nardi. My very first assignment was working on the autonomous robots’ behavior. In particular, I was assigned to work in collaboration with the SPQR RoboCup team (site) to integrate the framework PNP (Petri Net Plans) into their code. The robotic platform used was the Aldebaran Nao (Standard Platform League) and my work was mainly focused on providing high level behaviours.
During this period, I participated to the RoboCup 2010 (site), held in Singapore, where our team has reached a middle position in the rank. A remarkable achievement was scoring the first backward kick goal in the RoboCup since 1998. Moreover, I participated to the 1st RoboCup Tournament (site) in Greece where our team has reached the second place; the Mediterranean Opens, held in Rome, where in the 2011 edition our team reached the third place; the RoboCup 2011 (site) held in Instanbul; the RomeCup 2012 held in Rome; the RoboCup Dutch Open 2012 (site) held in Eindhoven; the RoboCup 2012 (site) held in Mexico City; the Iran Open 2013 (site), held in Tehran, where our team won the tournament; the German Open 2013 (site), held in Magdeburg, where our team got the third place; the RoboCup 2013 (site) held in Eindhoven. Since September 2011, I am the Team Leader of the SPQR RoboCup team. One of the major aspects during these years was to acquire the competencies and the skills to work autonomously within the laboratory life.
On September 2010, I started to study Bayesian filtering and in particular Particle Filters. In fact, my Master thesis was about Distributed Data Fusion for Multi-Agent Multi-Object Tracking. The Multi-Agent Multi-Object Tracking (MAMOT) task consists in estimating the targets trajectory within the environment using a team of robots. Each one is capable of perceiving its environment and detecting the targets by using a sensor with a limited field-of-view. One of the main aspects in this work is the information sharing about the environment itself among robots. This is achieved by the Distributed Data Fusion: a robot estimates the posterior of target positions using its own perceptions and shared one. This lead to an improved global estimation. I developed a novel approach based on clustering for the MAMOT task, obtaining promising results. Currently, a real robot evaluation is ongoing to evaluate the impact of the algorithm on real-world environment.
I got the Ph.D in Computer Engineering at the Department of Computer, Control and Management Engineering “Antonio Ruberti”, Sapienza University of Rome.
I am currently a Postdoctoral fellow in the bioinformatics lab of the Italian National Research Council (IASI-CNR) under the supervision of Prof. Paola Bertolazzi and Prof. Giovanni Felici.
The main topics of the research project are Multi-Object Detection (MOD) and Multi-Object Tracking (MOT). Detection involves real-time extraction of moving objects from sensors while tracking involves a continuous observation of objects over time to form persistent objects’ trajectories.
The research project has the objective to create a system that is able to track objects (like people, robots, etc.) in an environment by using multiple sensors (like cameras, kinects, etc.) and that is able to learn tracked objects’ behaviours fusing the information gathered from the all sensors. Applying learning techniques in MOD and/or MOT is still an open field unexplored because it is very challenging from different points of view. Finally, systems without learning have constant performance over time whereas for systems that use learning, the better is the experience the better are the performance over time.
Curriculum Vitae (download here)
Ph.D Thesis (download here)
This thesis investigates two important problems for intelligent robotic interaction with other agents: (1) object tracking from multiple – and potentially heterogeneous – distributed sensors and (2) predicting future agent motions for interactive robotic navigation. These problems are motivated by the deficiencies of existing mobile robots to navigate amongst humans (or other agents) in an intelligent manner similar to how humans are able to co-navigate: by recognising other agents in the environment, inferring their intentions and planning complementary movement trajectories that lead to efficient joint optimisation for all agents. Many existing mobile robots do not reason about the goal-directed movements of others in the environment, leading to substantial sub-optimality in reaching target locations.
In order to address the first problem, we develop PTracking, an algorithm for tracking multiple objects from multiple sensors in a distributed manner using Bayesian filtering (and particle filtering specifically to approximate the generally intractable inference task). The main novelty of the proposed approach is the combination of clustering and mixture models to enable more computationally efficient asynchronous inference. We demonstrate the algorithm’s versatility in a number of realistic applications: robotic soccer, multiple object tracking with mobile sensors, multi-robot surveillance, networked camera tracking of people and maritime surveillance.
The second problem has been tackled by employing an Inverse Reinforcement Learning (IRL) approach in combination with PTracking to estimate the reward functions that motivate observed behaviour sequences. A key innovation is that unlike previous IRL methods, which typically assume a fixed state-space representation, the state-space representation is dynamically adapted in the proposed method, so that more modelling emphasis is placed on portions of the space that are frequently visited and less emphasis can be placed on rarely visited portions. This allows significant computational savings versus employing a uniformly detailed state-space representation. We show the benefits of the method for activity forecasting applications, intention prediction and for constructing interactive costmaps to guide robot navigation.
Master Thesis (download here)
The Multi-Agent Multi-Object Tracking (MAMOT) task consists in estimating the targets trajectory within the environment using a team of robots. Each one is capable of perceiving its environment and detecting the targets by using a sensor with a limited field-of-view. One of the main aspects in this work is the information sharing about the environment itself among robots. This is achieved by the Distributed Data Fusion: a robot estimates the posterior of target positions using its own perceptions and shared one. This lead to an improved global estimation. I developed a novel approach based on clustering for the MAMOT task, obtaining promising results.
Bachelor Thesis (download here)
The goal of my thesis was an implementation of a web application based on a famous Italian game named Fantacalcio. I developed both the server side and client side. In my country, this game is played by about 3 millions of people and the innovation of my work was an application completely free whereby everybody can use it without spending any kind of money. Moreover, I strictly followed a spiral model as Software Development Process that allowed me to understand very well, even in the early development, what and how I should have to implement in order to reach the goal.