CV
Bio
My research experience is focused on how to model the vehicle surroundings, using computer vision and LIDAR based approaches. In particular, my activity is connected to the vehicle localization problem, which is fundamental to safely navigate Intelligent Vehicles in areas shared with human drivers. The final purpose is to provide a reliable vehicle awareness with respect to the surrounding environment.
Towards this goal, in the last few years I developed a probabilistic framework for ego-vehicle localization called Road Layout Estimation framework. Its main contribution is the synergistic exploitation of heterogeneous sensing pipelines in order to produce accurate lane-level vehicle localization in both difficult urban contexts and highway scenarios. The framework has been tightly coupled with the OpenStreetMap service, allowing the integration of priors into the localization process.
I have participated and supervised the creation of the whole system as well as the development of the detection components. The system leverages the interaction with the road cartography, building facades and road intersections, proving also its localization versatility by detecting the vehicle ego-lane in highways contexts. Moreover, the research has revealed the possibility to detect from images the number of arms and configuration entering the junction. The achievements were verified using well-known datasets such as KITTI or, if feasible, using the university’s autonomous vehicle.
The aforementioned research activities allowed me to deeply investigate robotics and computer vision techniques such as Kalman and particle filtering, HMM models, Visual Odometry systems and several convolutional neural network applications.
Education
- B.S. in Computer Science, Università degli Studi di Milano-Bicocca, 2008
- Thesis Title: A hierarchic Trajectory Planner for Indoor Robots download
- M.S. in Computer Science, Università degli Studi di Milano-Bicocca, 2011
- Thesis Title: 6DoF Monte Carlo Localization in a 3D World with Laser Range Finders
- Ph.D in Computer Science, Università degli Studi di Milano-Bicocca, 2017
- Dissertation Title: Matching heterogeneous sensing pipelines to digital maps for ego-vehicle localization
During the PhD program I attended the following courses and summer schools:
- Parallel Computing Using MPI and OpenMP - Cineca 17-25 June 2013.
- 1st Summer School on Critical Embedded Systems, 2-11 July 2013, Toulouse – France.
- 4th PAVIS School on Computer Vision, Pattern Recognition, and Image Processing, September 18-20, 2013 - Sestri Levante (GE), Italy.
- V-Charge Summer School on Perception and Planning for Autonomous Driving - July 7-10, 2014, ETH. Zurich, Switzerland.
- Computational approaches to Physical and Virtual Crowd Phenomena - UNIMIB DISCO PhD Course.
- Clustering Analysis - DISCO PhD Course.
- Advanced Techniques for Combinatorial Algorithms - UNIMIB DISCO PhD Course.
- Paradigms and Approaches to Computer Security - UNIMIB DISCO PhD Course.
Postdoctoral Experience
After the 4-years PhD period I worked as Research Associate in the Informatics and Robotics for Automation group (IRALAB) of the “Università degli Studi di Milano – Bicocca” in the following projects:
- 2017 – Autonomously driven road vehicles (Veicoli stradali a guida autonoma, university project code 16A129).
- 2018 – Perception for autonomously driven vehicles (Percezione per veicoli a guida autonoma, university project code 17A110).
- 2019 (ongoing) – “Reliable localization algorithms for autonomous driving car”, here I was awarded with Marie Skłodowska-Curie Actions research grant within the GOT ENERGY TALENT (GET) fellowship programme link
- 2022-2024 (ongoing) - “Effective and cooperative localization algorithms for self-driving cars”, research project awarded by the spanish ministry of Science Innovation and Universities under Maria Zambrano/NextGenerationEU grants. These grants were awarded following a competitive and objective selection process to attract international talent.
Knowledge Areas:
Localization, ITS, autonomous driving, point cloud registration, 3D reconstruction with stereo cameras, visual odometry, conditional random fields, hidden markov models, computer vision, image geometry and transformations, template matching, particle swarm optimization, Robotic Operating System (ROS), montecarlo localization within full 6DoF maps, path planning.
Skills
- 3D pose estimation and Visual Odometry algorithms
- Computer Vision, Image Formation, Geometry and Projection
- Kalman filtering, particle filtering
- Nonlinear optimization, using the g2o framework
- Proficiency in programming using C++ and the ROS framework (used since Boxturtle version)
- Comfortable with academic writing
- Semantic Classification of Road Features (road, sidewalks, traffic markers etc.) leveraging image classification algorithms such as Conditional Random Fields. In 2019 and I collaborated with Dr. Daniele Cattaneo in vehicle localization systems based on DNN approaches.
- I worked in several dataset acquisition and annotation projects in cooperation with the IRALAB team. The most important ones are related to the “A4 Highway dataset” collected in 2017 and the “Intersection Ground Truth and Training Set” sequences based on the KITTI Dataset.
Complementary skills & competencies
I worked and organized the following research and experimental/on-the-field activities, including public demos in different environments (closed areas, public areas, rural areas).
- Sept 2016 Autonomous car driving live demo at “MeetMeTonight 2016”.
- Sept 2016 Autonomous car driving live demo at “Parco Tecnologico Padano”. For this demo we had our autonomous car driving in outdoor/off-road scenarios.
- May 2013 Autonomous car driving live demo at “Wired Next Fest” (Innovation festival promoted by the Italian edition of Wired magazine). For this demo we had our autonomous car driving people around for 3 days in an off-road, heavily crowded environment at the city park of Porta Venezia in Milano.
- March 2013 Autonomous car driving demo, interview and photo session for “Quattro Ruote” (automobile magazine). In this demo we showed the ability of our autonomous car to drive inside an indoor parking lot.
- January 2013 Indoor autonomous robot navigation demo with “Geo Scienza”, RAI 3 (national television).
- January 2011 Outdoor autonomous car driving demo with RAI 3 (national television). In this demo we recoded for RAI 3 some sequences of autonomous driving within the University of Milano -Bicocca campus.
- November 2010 Indoor autonomous car driving demo at the Electrical Intelligent Vehicles Fair 2010 (EIV2010). In this demo we had our autonomous car driving people between stands of the fair for 4 days, moving through small spaces in a highly crowded environment.
Participation in scientific projects and role in them
- 2017 – Collaboration with the National Institute for Nuclear Physics (INFN) group on the MOSCAB experiment. I was responsible for the development of the computer vision toolkit. The tool allowed the MOSCAB research group to identify the 3DoF position of bubbles inside a double chamber system, filled with Freon and water, optimizing the refraction indices of the materials in an uncalibrated stereo camera configuration. During this project I had the opportunity to work with a completely different team, composed mainly of physicist and material science experts.
- 2017 – Collaboration with Magneti Marelli group for the development of a monocular visual odometry system. I was the main developer and responsible for the development, testing and acquisition phases.
- 2016 – Robotic Perception for autonomous driving (Percezione robotica per la guida autonoma, university project code 16A019). During this year I start working on the identification and classification of the upcoming intersections areas, given the images acquired using on-board cameras.
- 2012 – OMMAVE Project - Multi Sensor Odometry for autonomous vehicles. Università degli studi di Milano – Bicocca. I was responsible for the development of a 3D world simulator, necessary for the testing of a single-plane LIDAR-based localization algorithm.
Teaching
Publications
6DoF Monte Carlo Localization in a 3D world with Laser Range Finders
Your Name, You. (2009). "Paper Title Number 1." Journal 1. 1(1).
ira_laser_tools: a ROS LaserScan manipulation toolbox
ballardini2014iralasertools
An effective 6DoF motion model for 3D-6DoF Monte Carlo Localization
ballardini2012effective
A Framework for Outdoor Urban Environment Estimation
ballardini2015
An online probabilistic road intersection detector
ballardini2017
Ego-lane estimation by modeling lanes and sensor failures
ballardini2017b
Visual Localization at Intersections with Digital Maps
ballardini2019
CMRNet: Camera to LiDAR-Map Registration
cattaneo2019
Vehicle ego-lane estimation with sensor failure modeling
ballardini2020vehicle
A benchmark for point clouds registration algorithms
FONTANA2021103734
Model Guided Road Intersection Classification
ballardini2021model
WiFiNet: WiFi-based indoor localisation using CNNs
HERNANDEZ2021114906