Augusto Luis Ballardini

Augusto Luis Ballardini

Profesor Ayudante Doctor and former Marie Skłodowska-Curie Fellow - INVETT Research Group

Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Intersection Dataset

less than 1 minute read

Published:

Together with the colleagues of the INVETT group of Universidad de Alcalá, I am preparing a new multi-season dataset focused on the intersection detection issue. More details will be provided in the official website of the group.

Paper Accepted! Model Guided Road Intersection Classification

1 minute read

Published:

It is our pleasure to inform you that the paper identified above, for which you are listed as one of the authors, has been accepted as a contributed paper to be presented at the 2021 32nd IEEE Intelligent Vehicles Symposium (IV) (IV 2021), July 11-15, 2021, in Nagoya University, Nagoya, Japan.

SSH Tunneling

2 minute read

Published:

From time to time, everyone has to deal with IT guys bizarre security habits. Among them, one of the most annoying are

Hello World (Blog Launch)

less than 1 minute read

Published:

This is the first post inside this blog. I will try to update this blog with some infos from general research activity, not strictly connected to a specific topic, just tech curiosities, news comments etc.

portfolio

publications

Paper Title Number 4

Published in GitHub Journal of Bugs, 2024

This paper is about fixing template issue #693.

Recommended citation: Your Name, You. (2024). "Paper Title Number 3." GitHub Journal of Bugs. 1(3).
Download Paper

talks

Know where you are, drive where you want

Published:

After approximately one century from the first attempts to create automated driving cars, the recent progress in the context of Intelligent Transportation Systems allows us to consider the autonomous driving cars just a matter of time. From automated driving systems like on-demand next-generation taxis to privately owned self-driving cars, an appropiate awareness of the surrounding environment is a common need to achieve safety, for both drivers and road users. This presentations seeks to provide a basic understanding of one of the key elements in the process of achieving this self-consciusness, presenting the current advances in the field of vehicle localization. We will see together the major challenges and the different approaches both industries and academies are investigating to enable robotize-cars to drive safely in a real world shared with human drivers.

teaching

Robotics and Computer Vision

Master course, Università degli Studi di Milano - Bicocca, 2014

From 2013 to 2018 I worked as teaching-assistant for the course Computer and Robot Vision, together with Prof. Domenico G. Sorrenti. For this course I prepared the material for the exam and the lectures for the robotics localization topics, including:

Vehiculos Inteligentes

Master course, Universidad de Alcala, 1900

During the academic year 2019-20 and 2020-21 I have been teaching in the Vehiculos Inteligentes course (Intelligent Vehicles) of the Máster Universitario en Ingeniería de Telecomunicación (Telecommunication Engineering Master Course).

Sistemas en Tiempo Real - Real Time Systems (590008)

Undergraduate, Universidad de Alcala, 1900

El incremento en fiabilidad y velocidad de proceso de los ordenadores, junto con una disminución en su tamaño y precio, ha hecho que los encontremos en todos los ámbitos de nuestra vida, desde un teléfono móvil hasta el control del tráfico aéreo. Una de las áreas de expansión más rápida de la explotación de ordenadores es la relacionada con los sistemas empotrados y de tiempo real. Estas áreas necesitan procesar información con el objetivo de controlar procesos. Se ha estimado que más del 90 por ciento de la producción mundial de microprocesadores se utiliza en este tipo de sistemas. Estas aplicaciones plantean requisitos específicos para lo lenguajes de programación necesarios para programarlos, ya que tienen características diferentes de las aplicaciones de procesamiento de información tradicionales.

theses

3D Mapping Ground Truth

Published:

Embedded Systems Development with ARM

The aim of this project is to develop a system that allows the creation of three-dimensional mapping models for autonomous navigation of IRALAB’s robots. To achieve this goal, we started our work with a device composed by a LIDAR (Laser Imaging Detection and Ranging) system, mounted on a rotating platform; this is capable of generating 3D points sets, also called point clouds, in a reference frame in agreement with the platform used.

Road Layout Estimation - A probabilistic framework for autonomous driving Cars

Published:

Description of the work:

Use of car’s sensor measurements as well as car’s motion model for scene layout estimation; development of the software infrastructure; determination of appropriate detectors for the baseline road features, e.g., lane markings (horizontal road markings); customization of the framework to integrate these layout components, integration of the detectors for such components, training of these detectors if appropriate, extensive experimental activity.

3D Mapping with laser scanners

Published:

Embedded Systems Development with ARM

Performing automatic point cloud merging on the laser scanner SickLms 200 and performing odometry on Robosoft Robuter

A robotic mapping platform based on laser scanner

Published:

Aim

The work aims at integrating some existing software modules in order to build a single robotic mapping application based on a laser scanner, rotated by a mechanical turret about a vertical axis; the work includes the development of the graphical interface and the display of 3D results and this will be based on publicly available existing components. The application will be deployed on the iralab mapper robot. In a first version, the robot will be moved by hand to perform some preliminary mapping of some university areas. Afterwards, a joypad will be integrated so to make it possible to remotely control the robot; this task will also be performed exploing existing software modules. In addition, if the 4 wheels (2 traction and 2 castor) configuration of the robot will turn out to be unsuitable for the foreseen operations of the robot (mapping of urban outdoor environments), I’ll work out a mechanical modification to reduce the 2 castor wheel to 1, so to have a 3 wheeler (2 traction and 1 castor) platform. In the end, an extended experimental work will be performed with the remote version of the application and robot.

Leveraging the OSM building data in the Road Layout Estimation Framework

Published:

Description of the work:

In this project I will develop the detector systems for determining the presence of some static parts of the scene observed by a forward facing cameras mounted on a urban road vehicle. These developments will be integrated in the iralab Road Layout framework. In particular, I will focus on crossings and buildings.

A probabilistic intersection detector for the Road Layout Estimation Framework

Published:

Description of the work:

Detecting buildings in the surrounding of an urban vehicle and matching them to models available on map services is an emerging trend in robotics localization for urban vehicles. In this paper we present a novel technique that improves a previous work in this area by taking advantage of detected building fac¸ade positions and the correspondence with their 3D models available in OpenStreetMap (OSM). The proposed technique uses segmented point clouds produced using stereo images, processed by a Convolutional Neural Network. The point clouds of the fac¸ades are then matched against a reference point cloud, produced extruding the buildings’ outlines, as available on OSM. In order to produce a lane-level localization of the vehicle, the resulting information is then fed into our probabilistic framework, called Road Layout Estimation (RLE). We prove the effectiveness of this proposal testing it on sequences from the well-known KITTI dataset and comparing the results the respect to a basic RLE version without the proposed pipeline.

Robust Odometry for the USAD vehicles

Published:

Aim

The goal of this work is to carry out a USAD (Urban Shuttles Autonomously Driven) vehicle of a system that produced robust odometry. To achieve robust odometry, more heterogeneous sensors are used, as using only one technology would be unreliable. During the development of the work we added the covariance matrix to IMU messages using the information on the manual technician, the odometric messages presented the same problem that we solved by creating a matlab model that calculates the matrix covariance. The node communicating with the IMUs and posting messages had both issues in the frequency of message publishing and the time indicated in the message was incorrect. Both issues have solved them. We also made 3D and printed media for IMU.

Implementation of a motion planning algorithm for the Ackermann kinematics USAD vehicle

Published:

Descrizione della tesi

Lo scopo di questo lavoro è l’implementazione dell’algoritmo di pianificazione e controllo del moto per un veicolo a cinematica Ackerman. L’algoritmo è stato implementato all’interno dello stack di navigazione del framework ROS come un nuovo modulo di pianificazione locale, partendo da un prototipo sviluppato in MATLAB secondo un approccio noto in letteratura (U. Schwesinger, M. Rufli, P. Furgale, R. Siegwart - “A Sampling-Based Partial Motion Planning Framework for System-Compliant Navigation along a Reference Path”). La pianificazione del moto avviene basandosi su un percorso calcolato in precedenza da un altro modulo dello stack di navigazione, che si occupa della pianificazione globale.

Implementation of a global planner operating with OSM for the USAD vehicle

Published:

Aim

Currently the global planner of the IRALAB’s autonomous driven vehicle operates on image format maps (.png file). In these maps is represented the area where you want the vehicle to carry out the global motion planning. The stage activity will be focused on changing the global planner in order to operate with “openstreetmap” (https://www.openstreetmap.org/) so that the output would be a map on which it can then complete the global planning movement. The framework used will be ROS, changes to the navigation system will be done using C ++, all in Linux environment. It will be necessary to test the work that will be done directly on the vehicle, as the goal is to put the vehicle in the street.

Evaluation / Experimentation of mapping and localization systems in the university campus

Published:

Aim

The work involves an initial part that consists of mapping roads between the U9 and the U14 buildings, collecting and analyzing the data provided by the LIDAR sensors mounted onboard the vehicle. The acquired data will then be processed using open source software available on the ROS.org platform to create a true two-dimensional map of the explored environment needed for autonomous navigation. This activity is aimed at analyzing the current state of the system boundaries, designed to be used on areas not overly extensive. The possibility of making a modification to the existing software architecture will be evaluated to allow real-time navigation on maps, always related to the current position, via dynamic loading of maps of manageable dimensions. There is therefore an extensive testing phase on the autonomous machine of the IRALAB laboratory. As regards the training objectives, the student is expected to acquire the ability to understand the functioning of the ROS mapping (gmapping) and localization (amcl) stack as well as the operation of the framework itself; Moreover, any programming activity will take place in C ++, relying on the same framework and therefore it is expected that the student will also reach a discreet familiarity with this language.

Curb detection in urban environments using LiDAR pointclouds

Published:

Aim

Integration and implementation of a simple curb detection algorithm within the ROS navigation stack. The aim was to implement and integrate an algorithm in the vehicle’s local planner which allow to perform distance checks with reference to the roadway limits to ensure that the vehicle, in addition to follow a path set by the global planner, could be able to move in relationship with its position within the surrounding environment. The work and the analysis of the data acquired by the vehicle were processed using the ROS framework within Linux environment and the algorithm implementation was done in C++. Tests were carried out both in simulation and on the vehicle to verify the effective “operation” of the implemented work.

Integration in the navigation stack of the USAD vehicle of a vision-based tool for lane detection

Published:

Aim

The project involves the analysis of a detection algorithm for horizontal signage, in particular the lane detection. The aim is to create the limit zones beyond which the transit of a USAD cinema machine is not allowed, thus guaranteeing security in movement. A local planner of the autonomously guided vehicle is already present which decides instant by instant all the maneuvers that the kart must perform to maintain one trajectory; My work consist of integrating in local planner a control for not invading the opposite lane on the street. The framework used is ROS, the algorithm implementation is written in C ++ language, all in a Linux environment.

An extension of the ROS Navigation Stack for the management of moving obstacles

Published:

Aim

Integration and implementation of a dynamic obstacle detector in the ROS Navigation Stack. The project involves the development of an algorithm for the detection of dynamic obstacles commonly found in the urban environment. The developed software is integrated in the ROS navigation stack as a plugin that interfaces with the local planer and it’s able to represent static and dynamic obstacle on a 2D map.

A pedestrian detection module for the USAD vehicle

Published:

Descrizione della tesi

Il progetto consiste in un primo momento nel testare alcune reti neurali al fine di scegliere la migliore per rilevare ostacoli dinamici che intersecano la traiettoria del cart. Le reti neurali lavoreranno sulle rilevazioni ottenute da 3 telecamere montate sul veicolo a guida autonoma, due camere in bianco e nero sono in modalità stereo, mentre la camera a colori funziona singolarmente ed è montata tra le due precedenti. Le camere andranno quindi montate e calibrate correttamente al fine di ottenere dei bag da usare per testare gli algoritmi di riconoscimento. Scelto l’algoritmo migliore per le nostre necessità si procederà analizzando l’output ottenuto da esso per ricavarne informazioni relative al numero, tipo, posizione e velocità degli ostacoli presenti nell’ambiente. Queste informazioni saranno poi inviate al sistema in grado di gestirle e prendere decisioni sui movimenti del cart.

Evaluation and experimentation of mapping and localization systems in the university campus

Published:

Aim (starting from the work by Nicolò Vecchio)

The involves an initial part that consists of mapping roads between the U9 and the U14 buildings, collecting and analyzing the data provided by the LIDAR sensors mounted onboard the vehicle. The acquired data will then be processed using open source software available on the ROS.org platform to create a true two-dimensional map of the explored environment needed for autonomous navigation. This activity is aimed at analyzing the current state of the system boundaries, designed to be used on areas not overly extensive. The possibility of making a modification to the existing software architecture will be evaluated to allow real-time navigation on maps, always related to the current position, via dynamic loading of maps of manageable dimensions. There is therefore an extensive testing phase on the autonomous machine of the IRALAB laboratory. As regards the training objectives, the student is expected to acquire the ability to understand the functioning of the ROS mapping (gmapping) and localization (amcl) stack as well as the operation of the framework itself; Moreover, any programming activity will take place in C ++, relying on the same framework and therefore it is expected that the student will also reach a discreet familiarity with this language.

Creación de un mapa digital 3D de zonas urbanas a gran escala empleando sensores LiDAR y GNSS

Published:

Description of the work:

El objetivo de este proyecto es la creación de una nube de puntos de la ciudad de Alcalá de Henares en la que se registrará la precisión del sistema GPS. Asimismo, se pondrán a prueba distintas técnicas de SLAM basadas en algoritmos “LOAM”, tales como LegoLOAM, capaces de generar mapas 3D. Para el desarrollo de este proyecto se empleará el framework ROS (Robot Operating System), se adquirirán los datos mediante un sensor LiDAR, una IMU y un GPS. Cabe destacar que el algoritmo para la creación del mapa no se ejecutará en tiempo real, sino que los datos obtenidos por los sensores se recopilarán en un bag para su posterior procesamiento. Para determinar cual de los algoritmos es mejor se ha calculado el error medio de la trayectoria respecto a los datos del GPS, previamente transformados a coordenadas UTM, y se ha representado mediante “plot”de matlab ambas trayectorias juntas. El que menor error cometa, es el mejor algoritmo. En conclusión, se busca crear un sistema que permita la localización y mapeo con capacidad de identificar objetos en la ciudad de Alcalá de Henares. Para lograrlo, se implementarán técnicas de SLAM y análisis de datos, utilizando ROS, sensores LiDAR y herramientas de visualización.

Estudio de ablación para un detector de intersecciones viales

Published:

Description of the work:

En este trabajo se ha llevado a cabo un estudio de ablación detallado para un detector de tipologías de intersecciones viales basado en redes neuronales, centrado en el análisis exhaustivo del procesamiento de los datos de entrada, incluyendo transformaciones como rotaciones, traslaciones y variaciones en la resolución de las imágenes. A través de diversos experimentos, se han evaluado cuidadosamente las ventajas y desventajas que ofrece cada una de estas transformaciones, así como el impacto en el rendimiento de los modelos previamente definidos. Además, se proponen y discuten posibles mejoras y futuras direcciones para optimizar el estudio.

Geoestadística y Sistema de Información Geográfica para realizar predicciones y análisis de datos espaciales

Published:

Description of the work:

El trabajo consiste en la implementación de un Sistema de Información Geográfica para la predicción de precipitación mensual (mm de lluvia) en ubicaciones específicas a partir de los valores de las ubicaciones cercanas aplicando técnicas de geoestadística y diferentes métodos de interpolación espacial. La visualización de las predicciones se hará en un software de Sistema de Información Geográfica para presentar los resultados en mapas y gráficos que ayuden a visualizar los hallazgos de manera clara y comprensible para poder tomar decisiones.

Probabilistic Camera-to-LiDAR Map Registration for Autonomous Vehicle Localization

Published:

<!– This project aims to enhance a deep-learning based system capable of localizing a vehicle inside a LiDAR map using images captured by onboard cameras. The latter is a well-established method first introduced by Daniele Cattaneo in the paper CMRNet: Camera to LiDAR-Map Registration and further expanded in the subsequent papers by Daniele. The systems hinges on the capability of deep-neural-netoworks to find a correspondences between the points of an existing map made with a LiDAR sensor and the correspondent pixels on the camera image. This matching is completely done by the neural-network acting like a black box, with no information about the reliability of the matching.