This page shows the most relevant activities and events associated to the LAR, such as invited talks,
demonstrations, etc. Each tab below contains one year with multiple entries.
2013
2014
2015
2016
2017
2018
2019
2020
Invited Talk - 20 Feb 2013 - Dr. Ángel Sappa - Auditório DEM - 15:00
This talk will present recent work that has been doing at the CVC in this domain. The first part will describe a new framework for extracting disparity maps from a multispectral stereo rig. The system is constructed with an infrared and a color camera. It is intended to explore novel multimodal stereo matching approaches, going from sparse to dense representations (both disparity and depth maps). In the second part will be summarized some contributions in the shape representation and registration field. Both of them are unified under the umbrella of implicit functions. This compact representation becomes a robust tool to handle noisy data and outliers, hence it can be exploited in different computer vision applications. During the presentation, the Computer Vision Center as well as its main research activities will also be presented.
Invited Talk - 21 Feb 2013 - Dr. Miguel Ángel Sotelo - Anfiteatro IEETA - 10:00
click here
Toward recovering 3D structure from a static image
Developing a solution for accurate 3D structure recovery of generic objects in static
2D images from uncalibrated cameras is a complex challenge in the field of
computer vision. The final goal is to unveil the intrinsic 3D information contained in
static images by using prior knowledge obtained from experience by learning, just
as humans do. No motion cues will be used neither needed. Thus, virtual 3D
reconstruction from a monocular still image will be targeted, with special emphasis
on 3D recovery of human pose and other articulated objects. For this purpose,
statistical generative models and parts-based pattern recognition techniques can be
combined. By doing so, we will bridge the current gap between 2D-based objects
recognition techniques and 3D-based techniques for structure recovery and human
pose detection using low dimensional embeddings, such as latent spaces, and
probabilistic learning. Current state-of-the-art solutions for 3D human pose
detection are mostly based on multi-camera approaches, which are expensive and
difficult to calibrate, or 2.5D sensors, such as expensive laser-based sensors or the
more recent kinect range sensor. Although kinect is a cheap, high performance
sensor, it is restricted to indoor environments and requires supporting mechanic
systems to get full 3D object reconstruction by revolution. Structure From Motion
(SFM) techniques have traditionally been used for recovering 3D structure from a
sequence of images acquired by a monocular camera. However, the approach proposed in this talk is not constrained by the availability of motion. No motion
cues are considered at all in order to enable the managing of images provided by a
static camera and, even the management of a single static image. A proposal is
made to apply geometric and anthropometric constraints for lifting the feasible
poses into 3D, in order to significantly enhance the performance of state-of-the-art
monocular-based object recognition systems. Geometric and anthropometric
constraints will be learnt from experience and coded in a probabilistic framework.
The result will be the categorization of all relevant, previously learnt objects in the
scene together with the recovery of their 3D structure and relative pose. By doing
so, any monocular (2D) camera can be transformed into a virtual three-dimensional
(3D) camera. This result would allow for the massive use of any type of monocular
camera in a great variety of fields, such as robotics, image understanding and
activity recognition, context-based 3D scene interpretation, human recognition
through gait analysis, road safety, security and surveillance, sports and
rehabilitation medicine, motion pictures, markerless motion capture, humancomputer
interaction, and gaming.
PhD discussion - 21 Feb 2013 - Miguel Oliveira - Sala de Actos, Reitoria - 14:00
click here
Automatic Information and Safety Systems for Driving Assistance
by Miguel Oliveira
The main object of this thesis is the study of algorithms for automatic information
processing and representation, in particular information provided
by onboard sensors (2D and 3D), to be used in the context of
driving assistance. The work focuses on some of the problems facing
todays Autonomous Driving (AD) systems and Advanced Drivers Assistance
Systems (ADAS). The document is composed of two parts.
The first part describes the design, construction and development of
three robotic prototypes, including remarks about onboard sensors, algorithms
and software architectures. These robots were used as test
beds for testing and validating the developed techniques; additionally,
they have participated in several autonomous driving competitions with
very good results. The second part of this document presents several
algorithms for generating intermediate representations of the raw
sensor data. They can be used to enhance existing pattern recognition,
detection or navigation techniques, and may thus benefit future
AD or ADAS applications. Since vehicles often contain a large amount
of sensors of different natures, intermediate representations are particularly
advantageous; they can be used for tackling problems related
with the diverse nature of the data (2D, 3D, photometric, etc.), with the
asynchrony of the data (multiple sensors streaming data at different
frequencies), or with the alignment of the data (calibration issues, different
sensors providing different measurements of the same object).
Within this scope, novel techniques are proposed for computing a multicamera
multi-modal inverse perspective mapping representation, executing
color correction between images for obtaining quality mosaics, or
to produce a scene representation based on polygonal primitives that
can cope with very large amounts of 3D and 2D data, including the
ability of refining the representation as new information is continuously
received.
Academia de Verão - 8 Julho 2013 - Sessão de robótica - 14:00-17:00 - LAR
click here
Activity inserted in the "Academia de Verão" event. This session will be coordinated by Dr. Ricardo Pascoal.
Masters discussion - 15 Jul 2013 - José Viana Bettencourt - Auditório DEM - 10:00
click here
E-Stop e consola de controlo e monitorização para o AtlasCar The AtlasCar vehicle is a robot developed at the Automation and Robotics laboratory in the Department of Mechanical Engeneering at the University of Aveiro. The main objective of this project is to develop both active and passive advanced security systems, for implementation in driver assistance system and autonomous vehicles. The main goal of this work is the development of a remote emergency E-stop, with a control console, and also the creation of a vehicle monitoring system. These twosystem complient each other in the question of safety. The E-Stop and the control console are part of a remote atuation system that manages three diferent states of atuation by user order: The three states are RUN, PAUSE and DISABLE. This system is mandatory to be eligle to participate in the international ELROB competion. Both software and hardware were developed with this goal in mind, acording to the competition directives. The monitoring system consists of a SQL database whose fields are published in a webpage. The refresh of the database is in charge of a ROS software module running on the atlascar using either a wifi or a 3g internet connection.
Both systems were tested and validated in real operation.
Deteção do Espaço Navegável para o
ATLASCAR usando informação 3D This work presents a method for detecting the navigability of the environment
using 3D data of the vehicle’s surroundings. In this work the 3D data is
achieved by a 2D laser on continuous rotation.
Initially the hardware of the 3D acquisition system was refined in order to
achieve better data quality due to the rotational movement of the planar
laser unit; afterwards, a software module to store the different 2D scans and
to create a point cloud of the environment was developed.
In the second stage, a software module for detecting the navigability of the
environment, using the elevation of the terrain as well as the normal average
was developed. The space was divided into a two-dimensional grid with cells
that contain several properties in order to calculate the accessibility average,
the result is a sort of extended elevation maps that may be used to calculate
navigability of any trajectory, that were designated accessibility maps. The
results indicate a clear improvement in the quality of the 3D data obtained
with the planar laser unit and were also made studies and analyzes of the
influence algorithms’ parameters’ for the calculation of accessibility maps.
Masters Discussion - 17 Jul 2013 - Pedro Silva- Auditório DEM - 14:30
click here
Visual Pedestrian Detection using Integral
Channels for ADAS This work exploits a vision based pedestrian recognition technique to build a framework capable of performing detection in images of urban setting. This method takes advantage of the richness of information present in multiple channels of an image to assemble different detection techniques in a simple and generic infrastructure. The general idea behind this method is that pedestrians present specific visual properties that can be used to differentiate them apart from a scene. So, to exploit such properties, features are extracted from the channels in an optimized manner through the use of integral images and a machine learning engine based on AdaBoost is trained through positive and negative examples the relationship of those features with the presence of pedestrians on visual data. This framework then integrates a multi-scale, sliding window approach to perform visual pedestrian detection. To evaluate the algorithm, tests were carried out in two distinct datasets and results confirm its validity.
Masters Discussion - 17 Jul 2013 - Daniel Coimbra- Auditório DEM - 16:00
click here
LIDAR Target Detection and Segmentation in
Road Environment In this project a comparative and exhaustive evaluation of several 2D laser
data segmentation algorithms in road scenarios is performed. In a rst stage, the segmentation algorithms are implemented using the ROS programming environment; the algorithms are applied to the raw laser scan data in order to extract groups of measurement points which share similar spatial properties and that probably will belong to one single object. Each algorithm has at least one threshold condition parameter that is con gurable, and one of the goals is to try to determine the best value of that parameter for road environments. The following stage was the de nition of the Groundtruth where multiple laser scans were hand-labelled. The next step was the comparison between the Ground-truth and the segmentation algorithms in order to test their validity. With the purpose of having a quantitative evaluation of the methods' performance, six evaluation metrics were created and compared.
Masters Discussion - 4 Dec 2013 - Telmo Rafeiro- Auditório DEM - 14:00
click here
Rede de Sensores Inerciais para Equilíbrio de um Robô Humanoide O objetivo principal desta dissertação é o desenvolvimento de uma rede de sensores inerciais composta por um conjunto de Inertial Measurement Units, constituídos por acelerómetros, giroscópios e magnetómetros (os últimos foram descartados no projeto). É pretendido que esta rede de sensores permita ao robô humanóide do PHUA - Projeto Humanoide da Universidade de Aveiro adquirir informação necessária ao seu equilíbrio. O primeiro passo deste trabalho passou pela caraterização de dois tipos de unidades inerciais, o RAZOR - SEN 10736 - 9 gdl e o POLOLU - MinIMUv2 - 9 gdl, com auxílio de um braço robótico antropomórfico. Os resultados desta avaliação são apresentados no trabalho. De seguida, foi desenvolvida a estrutura da rede inercial com recurso a uma unidade de processamento central, uma plataforma de desenvolvimento Arduino com controlo multiplexado de oito unidades assente no protocolo de comunicação I2C, acrescido de uma comunicação RS232 com a nona unidade. A rede desenvolvida suporta um máximo de nove unidades inerciais, com uma taxa de aquisição de 7 Hz. O funcionamento da rede não está dependente do número de unidades utilizadas, podendo operar com qualquer número de unidades compreendidas entre 1 e 9 unidades. A rede inercial foi instalada no robô humanóide, e testada durante a execução de movimentos simples, permitindo concluir da viabilidade da rede em diversas situações.
Invited Talk - 11 Dec 2013 - Dr. Luis Merino - Auditório DEM - 11:00
Service robots is one of the most promising applications of robotics. It is expected that technology will allow in the mid-term to deploy robots in everyday environments to offer services like guidance, transport of people and goods, etc. Several challenges lay ahead, though, before robots can act robustly in these scenarios. In the talk, different related aspects and techniques will be described, like localization and navigation in crowded environments. Furthermore, it will be shown how robots can be integrated within the infrastructure of cities, and collaborate with fixed sensors and actuators embedded in this infrastructure to perform their services. Finally, it will be described how human-awareness should be integrated into these robots to enhance their acceptance, becoming real social robots. The talk will describe real robot deployments in pedestrian scenarios in the frame of the EU projects URUS and FROG.
This talk is sponsored by Departamento de Engenharia Mecânica / Centro de Tecnologia Mecânica e Automação
PhD discussion - 11 Dec 2013 - Procópio Stein- Sala do Senado, Reitoria - 15:00
click here
Navigation in Dynamic Environments
taking Advantage of Moving Agents
by Procópio Stein
This thesis proposes a di erent form of robotic navigation in dynamic environments, where the robot takes advantage of the motion of pedestrians, in order to improve its own navigation capabilities. The main idea is that, instead of treating persons as dynamic obstacles that should be avoided, they should be treated as special agents with an expert knowledge of navigating in dynamic scenarios. To bene t from the motion of pedestrians, this work proposes that the robot select and follow them, so it can move along optimal paths, deviate from undetected obstacles, improve navigation in densely populated areas and increase its acceptance by other humans. To accomplish this proposition, novel approaches are developed in the area of leader selection, where two methods are explored. The rst uses motion prediction approaches while the second uses a machine learning method, to evaluate the leader quality of subjects, which is trained with real examples. Finally, the leader selection methods are integrated with motion planning algorithms and experiments are conducted in order to validate the proposed techniques.
Academia de Verão 2014 - LAR - 10 & 11 Jul, 9:30-12:30
Master Thesis Public Discussion- 16 Jul 2014, Auditório DEM - 10:00, Gonçalo Carpinteiro
Medição de Orientações Relativas de um Veículo em Movimento
Este trabalho descreve um método de obtenção das orientações de um veículo
em movimento. O facto de o veículo estar em movimento vai adicionar
variação das orientações ao sistema desenvolvido. Para obter as orientações
mediu-se a distância da estrutura do veículo à estrada por unidades locais de
medição: ULM. Foi implementada uma unidade central de integração, UCI,
que recebe a informação das ULM e envia para a unidade computacional do
veículo. A leitura da distância foi realizada através de sensores de medição
por infravermelhos da Sharp. Para que os dados provenientes desses sensores
fossem o mais preciso possível, foi levado a cabo um estudo e caracterização
destes mesmos sensores. Cada sensor foi calibrado individualmente, de
modo a se obter a relação da diferença de potencial lida com a distância à
estrada. Foi levado a cabo um estudo de qual o melhor método matemático
para obter as orientações com base nas distâncias à estrada. Foram ainda
conduzidos vários testes de modo a estimar o erro do sistema desenvolvido.
Por fim, testou-se o sistema final em ambiente real.
Master Thesis Public Discussion- 16 Jul 2014, Auditório DEM - 16:00, Pedro Mendes
Intelligent Monitoring of Human Actuations on
Automobile Vehicle Pedals
Nowadays, there is a huge socio-economic effort to reduce driving accidents.
To do so, there are many safety methods and emphasis has been given
to methods based on computer vision techniques. However, most of these
techniques require high computational power and economical cost. As an
alternative there are other techniques that exploit the vehicle-driver interfaces,
such as the vehicle pedals. Considering the pedals as a powerful tool
to monitor the driver actions, a system to monitor the driver actuations in
automobile vehicle pedals was developed. In doing so, several possible conflicting requirements were taking into account, such as: low cost, real-time
monitoring and accurate, robustness, non-invasiveness, flexible and easy-to use. To evaluate if the developed system fulfilled these requirements, tests
were carried out under both stationary and real driving conditions. Overall,
the system developed presented the necessary requirements and was able to
successful monitor the driver actuation on automobile vehicle pedals.
Master Thesis Public Discussion- 16 Jul 2014, Auditório DEM - 14:30, Sérgio Pinho
Caixa Automáatica e Manobras Especiais no ATLASCAR
O AtlasCar é um protótipo de um veículo desenvolvido pelo Laboratóorio de Automaçãoo e Robótica (LAR) do Departamento de Engenharia Mecânica
da Universidade de Aveiro, em que, o principal objetivo é o desenvolvimento
e estudo de sistemas de condução autónoma, de conduçãao assistida (ADAS)
e de segurança.
Entre todos os desafios deste projeto, a inexistência de um sistema de comutação de velocidades automático fiável limitou, em muito, a condução autóonoma. Este trabalho teve como objetivo a reconstrução de todo o
sistema de comutação, necessitando da construçãoo de um novo sistema
mecânico (baseado no modelo anterior), da calibração/reajustes do sistema
de comutaçãao (unidade de controlo), da implementação do sistema no veículo e do desenvolvimento de software no sistema computacional (para
realização de manobras especiais).
Para o efeito são utilizados protocolos de comunicação entre o firmware e
o software do AtlasCar, software de calibração e um modo ROS para comunicação com todo o sistema.
Esta dissertação aborda o desenvolvimento de um novo comutador de mudanças que teve como base o modelo existente. Após concluído o comutador,
implementou-se uma nova segurançaa no sistema eletrónico e tornou-se
a unidade de controlo mais compacta e robusta, juntando as várias unidades
de controlo num único bloco.
Por último, desenvolveu-se um software para a realizaçãoo de manobras especiais
como: avançar, recuar, estacionar, estacionar paralelamente, inverter
marchar e parar.
Apos realizados todos os trabalhos, concluiu-se que o comutador de mudanças é funcional e robusto, e as manobras além de serem realizadas não
contêm um grande grau de repetibilidade.
Master Thesis Public Discussion- 22 Jul 2014, Auditório DEM - 15:00, Luis Carrão
Sistema de visão para guiamento de AGV em
ambiente industrial
Este trabalho propõe um sistema de posicionamento utilizando visão artificial, baseada em marcadores, para guiamento em ambiente industrial.
Numa primeira instância foi determinado o tipo de informação que seria
representada pelos marcadores, assim como de que forma esta seria interpretada.
Posteriormente foi escolhido que tipo de marcadores existentes
seriam os mais adequados para as condições da aplicação desejada.
De seguida foi utilizada uma biblioteca, de uso livre, para leitura de códigos
DataMatrix, com a qual é possível descodficar e obter a posição dos marcadores
na imagem.
Posteriormente, conhecendo a informação relativa à geometria dos marcadores
tanto real como na imagem foram medidas as suas distâncias.
Sabendo o seu posicionamento na imagem foram ainda calculados os ângulos
entre estes e o eixo ótico da câmara.
Finalmente, utilizando a informação relativa a estas medições e sabendo a
posição de apenas dois marcadores, foi possível calcular o posicionamento
(X, Y e ϕ) do AGV, com recurso a um método híbrido entre trilateração e
triangulação.
Os resultados foram aferidos com um sistema de medição baseado em laser.
Master Thesis Public Discussion- 22 Jul 2014, Auditório DEM - 16:30, Ricardo Silva
Removable Odometry Unit for Vehicles with
Ackerman Steering
The main objective of this work is to develop a solution of odometry for
vehicles with Ackermann steering. The solution had to be portable, flexible
and easy to mount. After the study of the state of the art and a survey of
solutions, the solution chosen was based on visual odometry. The following
steps of the work were to study the feasibility to use line scan image
sensors for visual odometry. The image sensor was used to compute the
longitudinal velocity; and the orientation of motion was computed using
two gyroscopes. To test the method, several experiments were made; the
experiments took place indoor, under controlled conditions. It was tested
the ability to measure velocity on straight line movements, diagonal movements,
circular movements and movements with a changing distance from
the ground. The data was processed with correlation algorithms and the
results were documented. Based on the results it is safe to conclude that
odometry with line scan sensors aided by inertial sensors has a potential for
a real world applicability.
Master Thesis Public Discussion- 20 Nov 2014, Auditório DEM - 15:00, Ricardo Morais
Parametrização de algoritmos para deteção de estrada a bordo do ATLASCAR
O veículo AtlasCar é um protótipo desenvolvido pelo Laboratório de Automação e Robótica do Departamento de Engenharia Mecânica da Universidade de Aveiro, para o estudo de sistemas de segurança ativos e passivos e soluções de condução
autónoma. O objetivo deste trabalho é dotar o AtlasCar de um sistema de visão artificial capaz de reconhecer os limites da estrada, sendo para isso necessária a procura de marcas na estrada e/ou características das mesmas.
Este trabalho encontra-se dividido em duas partes principais: o desenvolvimento de módulos de software para a implementaçãao de algoritmos com metodologias de deteçãao distintas e o estudo dos parâmetros que os compõem. O primeiro algoritmo baseia-se no método de RANSAC para efetuar a aproximação de uma spline aos limites da estrada, por sua vez o segundo algoritmo utiliza a transformada de Hough para efetuar esta aproximação. Os métodos apresentados foram testados em condições reais e foram analisadas as situações de falha, vantagens e desvantagens da aplicação de um método em relativamente ao outro.
Com a implementação deste sistema será possível a distinguir as zonas navegáveis
das não navegáveis em contexto de estrada, permitindo assim a outros processos
fazer um planeamento local da navegação.
Events for 2015 not listed
C++ Workshop, 22 Feb 2016
Workshop lectured by Prof. Miguel Oliveira to prepare the attendees for a ROS workshop. Check further details and materials here.
Workshop on ROS, 29 Feb 2016 & 11 Mar 2016
Workshop sessions lectured by Prof. Miguel Oliveira. Check for aditional details and materials here.
C++ & ROS Workshops, 2 and 3 Mar 2017
C++ Workshop lectured by Prof. Miguel Oliveira to prepare the attendees for a ROS workshop. Check further details and materials here.
Workshop sessions lectured by Prof. Miguel Oliveira. Check for aditional details and materials here.
C++ & ROS Workshops, 1, 2 and 8 Mar 2018
C++/ROS Workshop lectured by Prof. Miguel Oliveira. Check further details and materials here.
