Bruno João Nogueira Guerreiro, with the workplan entitled “Sensor
based Control and Localization of Autonomous Vehicles in Unknown
Environments”
Supervisor: Paulo
Oliveira Co-supervisor: Profª. Rita Cunha
Abstract:
The proposed work program builds on the candidate's PhD. thesis
contributions to further address several problems related to the
control, navigation, and guidance of autonomous vehicles in unknown GPS
denied environments. Ultimately, these results will contribute to
develop innovative inspection tools for comprehensive inspection
procedures, reducing risks and operation costs. There are three
fundamental research directions: A. Perception, navigation, and
sensor-based simultaneous localization and mapping (SLAM) algorithms
with formal observability, convergence, and performance guarantees,
considering sensors and strategies that allow for consistent estimation
of increasingly unstructured/dynamic environments. B. Sensor-based
trajectory tracking control algorithms, considering new sensors, control
techniques, and error space formulations, with strong stability and
performance results for a wide set of flight regimes and the transitions
between them. C. Multi-vehicle and combined sensor-based control and
SLAM algorithms, guiding the vehicles to maintain observability of the
environment while guaranteeing the convergence to the desired paths,
possibly, within a formation of multiple vehicles.
FCT Pos-doc scholarship from 1/1/2016 until 31/12/2018.
Current position: Assistant Professor at NOVA University of Lisbon.
Supervision of PhD Students with the Thesis Completed
Doctoral Program in Electrical and Computer Engineering, Faculty of Science and Technology, University of Macao
Joel Oliveira, with the workplan entitled “Nonlinear Estimation Techniques for Underwater Navigation and Tracking”
Supervisor:
Paulo OliveiraCo-supervisor: Prof. Carlos SilvestreCo-supervisor:
Prof. Pedro Batista
Abstract:
This thesis reports, in the first part, the steps for
developing a Portable Navigation Tool for Underwater Scenarios (PONTUS) to be
used as a localization device for subsea targets. PONTUS consists of an
integrated ultra-short baseline acoustic positioning system aided by an
inertial navigation system. The tool’s architecture is fully disclosed,
followed by rigorous technical descriptions of the hardware ensemble and
software development. A localization technique is then developed to estimate
the position of a moving target based on discrete-time direction and biased
velocity measurements. A nonlinear system is first designed, followed by a
state augmentation that yields an equivalent linear time-varying system. The
final estimation solution resorts to a Kalman filter with globally exponentially
stable error dynamics. Its performance is assessed via realistic numerical
simulations and via a set of experimental results using PONTUS.
In the second part, four estimators are proposed to
tackle the problem of attitude estimation taking into account the rotational
motion of the planet. The solutions presented herein put emphasis on the fact
that only one vector measurement is explicitly employed in the estimators,
which are aided by angular velocity readings collected from a set of triaxial
high-grade gyroscopes sensitive to the Earth’s spin. Different strategies, such
as explicitly estimating the Earth’s angular velocity or considering
time-varying observer gains, are studied aiming at speeding up convergence
rates while maintaining high levels of accuracy and low computational
complexity. It is further examined the case when both the gyroscope readings
and the measurements of the reference vector are corrupted by biases. Realistic
simulation tests are detailed that illustrate the performance of all four
attitude estimators. In particular, the speed and efficiency of the fastest observers
are demonstrated through an extensive set of experimental results.
Current
situation: Post-doc Researcher at FCT, University of Macau.
Doctoral Program in Electrical and Computer Engineering, IST
Pedro António Duarte Marques Lourenço, with the workplan entitled ‘‘Globally Convergent Simultaneous Localization and Mapping: Design Techniques, Analysis, and Implementation”
Supervisor:
Paulo OliveiraCo-supervisor: Prof. Carlos SilvestreCo-supervisor:
Prof. Pedro Batista
Summary of the Thesis (partial):
This dissertation addresses the
problem of simultaneous localization and mapping (SLAM), with the main
aim of devising strategies to achieve global convergence and stability
guarantees, covering its main formulations in terms of the exteroception
sensors used. It is argued that, since the pose is unobservable as all
the information is relative to the vehicle, removing it from the filter
and exploring the relative nature of the measurements is advantageous
for that goal. Following that line of reasoning the SLAM problem is
divided in two: (i) design an observer for the relative map with global
convergence guarantees for its error dynamics; and (ii) find an
alternative strategy that allows the computation of the pose from the
sensor-based map and the initial pose. Regarding the first part of the
problem, this dissertation presents a series of algorithms deeply rooted
in a sensor-based approach to the SLAM problem that provide global
convergence guarantees. Finally, for the second part of the approach, it
is proposed an algorithm that builds on an optimization problem
equivalent to the orthogonal Procrustes problem, which has a closed-form
solution. Due to the uncertainty present in the sensor-based
estimation, the Procrustes problem is also studied thoroughly and a
novel uncertainty characterization of its results is presented and
validated through extensive Monte Carlo simulations. Simulation and
experimental results that complement those of the first part and are
included to illustrate the performance of the complete algorithm
comprising the Earth-fixed Trajectory and Map estimation in cascade with
each of the sensor-based SLAM filters under realistic conditions.
Thesis approved with Muito Bom com Distinção e Louvor (top rank), January 2019, PDF File.
Current
situation: GMV Aerospace and Defence.
Doctoral Program in Electrical and Computer Engineering, IST
Daniel Luís Laurens Viegas, with the thesis entitled “Distributed State Estimation for Multiple Autonomous Vehicles”
Supervisor:
Paulo Oliveira Co-supervisor: Prof.
Carlos SilvestreCo-supervisor: Prof. Pedro Batista
Abstract of the thesis (partial):
This thesis addresses problems on the subject
of distributed state estimation for multiple vehicles. In the scenarios
that are envisioned, each vehicle must rely on measurements provided by
sensors mounted on-board and limited communication with other vehicles
to estimate relevant variables such as its own inertial position and
linear velocity. For the case in which the vehicles have access to
relative position measurements, the proposed distributed state observer
features globally exponential stable estimation error dynamics with
performance guarantees in the presence of noise in the measurements, and
is able to cope with issues such as cycles in the measurement graph and
time-varying measurement topologies. Regarding the case in which the
vehicles have access to range measurements, conditions for global
observability of the nonlinear dynamics of the problem are derived, and
the proposed distributed state observer features error dynamics that
converge exponentially fast to zero for any initial condition for
acyclic measurement topologies. Simulation results for the practical
application of a formation of Autonomous Underwater Vehicles (AUVs)
working cooperatively are detailed for both cases to validate the
theoretical results.
Keywords: Distributed state estimation, Cooperative navigation, Autonomous vehicles, Kalman filtering, State observers.
Prémio da Sociedade Portuguesa de Robótica para a Melhor Tese de
Mestrado Nacional em Robótica de 2010, para o Eng. Daniel Viegas, de
que fui orientador no Mestrado em Engenharia Eletrotécnica e de Computadores
do IST, abril de 2011.
FCT scholarship.
Thesis approved with Muito Bom com Distinção e Louvor (top rank), June 2017, Ficheiro PDF.
PhD scholarship from FCT. Researcher at the Faculty of Science and Technology, Univ. Macao.
Portuguese Society of Robotics Award for the Best MSc Thesis in Robotics 2010, April 2011.
Current situation: Data Scientist at Feedzai, Lisbon.
Doctoral Program in Electrical and Computer Engineering, IST
Sérgio Daniel Gonçalves Gante Brás,with the thesis entitled“Deterministic Position and Attitude EstimationMethods”
Supervisor:
Prof. Carlos Silvestre Co-supervisor:Paulo Oliveira
Abstract:
This thesis proposes new deterministic attitude estimation methods for
autonomous platforms. The devised solutions are based on nonlinear
observers and on the set-valued observers (SVOs) methodology. Two
nonlinear observers are developed for indoor scenarios. The first
observer fuses inertial information with measurements from an active
vision system. Furthermore, a pan and tilt controller is proposed such
that the visual features are kept in the camera’s field of view. The
second observer is based on range measurements that are retrieved from
acoustic sensors fixed in the moving body reference frame to an array of
beacons installed in the inertial frame. Both solutions are validated
resorting to experimental prototypes and a high-accuracy motion rate
table, which provide ground truth data. Under the framework of SVOs, a
different solution for the problem of attitude estimation is proposed,
which provides a bounding set, which contains the true rotational state.
The sensor readings are assumed corrupted by unknown bounded
measurement noise and constant gyro bias. A set containing the rate gyro
bias is also obtained and used to reduce the uncertainty of the
estimates. Sufficient conditions for boundedness of the estimated set
for the cases of multi- and single- vector observations are established.
Additionally, two fault detection and isolation (FDI) methods based on
bounding sets are proposed for navigation systems equipped with sensors
providing inertial measurements and vector observations.
Keywords: Inertial Navigation Systems, Position and Attitude Estimation,
Fault Detection and Identification, Nonlinear Observers, Set-valued
Observers.
Thesis approved with Muito Bom com Distinção e Louvor (top rank), November 2015, Ficheiro PDF.
Second Best Student in Electrical and Computer Engineering UTL/Santander Totta awards in 2009.
FCT scholarship.
Current occupation: Atitude and Orbirt Control System Performance Engineer at European Space Agency, Netherlands.
Doctoral Program in ECE, IST
Tiago Filipe Pires Gaspar with the thesis entitled ‘‘Monocular 3D Positioning and Tracking Systems”
Supervisor: Paulo
Oliveira
Abstract (partial):
This thesis addresses the 3D positioning and tracking problem, with
special emphasis on indoor applications. This problem consists in
estimating the state (composed of the 3D position, velocity, and
sometimes acceleration) of a target, typically a person or a moving
object, using remote measurements provided by one or more sensors, at
fixed locations or at moving platforms. In this case, a new positioning
and tracking architecture that solves this problem using measurements
provided by a single static pan and tilt camera is proposed. A dynamical
model for the target that depends on its angular speed is considered,
and the state of the target is computed using a suboptimal stochastic
multiple-model estimator. Depth from focus strategies are modified to
estimate the distance between a target with unknown dimensions and the
camera. This allows us to get rid of the first assumption. The
synthesis, design, and analysis, of a novel model based H2 adaptive
filter, with convergence guarantees, are also presented. The use of this
filter allows us to drop the second assumption. Two new filters that
estimate the state of a marine mammal that moves at the sea surface and
that is recorded with a camera installed on an Unmanned Aerial Vehicle
(UAV) are also proposed. Finally, a strategy that synchronizes video
sequences acquired by independently moving cameras that see (possibly)
different parts of a common rigidly moving object is also presented.
This strategy paves the way for a future extension of the proposed
monocular methods to multicamera configurations.
Keywords: Indoor Positioning and Tracking, Nonlinear Filters, Kalman
Filters, Complementary Filters, Linear Parameter Varying Observers,
Adaptive Estimation, Parameter Identification, Multiple Model Adaptive
Estimation.
Current occupation: Computer Vision Engineer at Reverse Engineering, Lisbon.
Doctoral Program in Electrical and Computer Engineering, IST
Pinar Oguz Ekim, with the thesis entitled ‘‘Simultaneous Localization and Tracking in Sensor Networks, in the presence of Mobile Robotic Platforms”
Supervisor: Prof. João Pedro Gomes Co-supervisor:
Paulo
Oliveira
Abstract (partial):
This thesis proposes several robust localization algorithms in WSN with a
special focus on Target Tracking. The first part of the thesis
considers the problem of locating a single source from noisy range
measurements of a set of nodes in a WSN. Several techniques are
presented for two noise distributions, namely Gaussian and Laplacian, in
different space dimensions. Moreover, algorithms based on l1-norm are
used to address the Laplacian noise case, which models the presence of
outliers in some practical ranging systems that adversely affect the
performance of localization algorithms designed for Gaussian noise. The
second part of the thesis presents Simultaneous Localization and
Tracking in WSN, which aims to determine the positions of sensor nodes
and a moving target in a network, given incomplete and inaccurate range
measurements between the target and each of the sensors. A modified
Euclidean Distance Matrix (EDM) completion problem is solved for a block
of target range measurements to approximately set up initial
sensor/target positions, and the likelihood function is then iteratively
refined through Majorization-Minimization. Simulation results and real
indoor experiments show that the proposed algorithms significantly
outperform state of the art methods in the presence of outliers as well
as Gaussian noise. Additionally, they attain the Cramer Rao Lower Bound
for small noise in the Gaussian case. The third part of the thesis
addresses sensor network localization problems with additional
difficulties.
Keywords: Source localization, sensor network localization, simultaneous
localization and tracking, no convex and no smooth minimization,
semidefinite programming, maximum likelihood estimation, Laplacian
noise, central processing.
Thesis approved with Muito Bom com Distinção e Louvor (top rank), September 2013, PDF File. Presentation PDF File.
2012 IBM Scientific Award for the work entitled ‘‘Algoritmos robustos localização em redes de sensores com aplicações a seguimento de alvos,” (in Portuguese) May 2013.
Current occupation: Assistant Professor at İzmir Ekonomi Üniversitesi, Turkey.
Doctoral Program in Electrical and Computer Engineering, IST
Marco Martins Morgado, with the thesis entitled ‘‘Advanced Ultra-Short Baseline Inertial Navigation Systems”
Supervisor: Paulo OliveiraCo-supervisor: Prof.
Carlos Silvestre
Abstract:
This thesis addresses the synthesis, design, and analysis of modern
navigation systems with application to underwater vehicles, focusing on
small arrays of acoustic receivers as the main sensor suite installed
on-board the underwater vehicle. Novel ultra-short baseline (USBL)
tightly-coupled and integrated filtering solutions are proposed to
enhance error estimation in strap-down inertial navigation systems
(INSs). The improvements of a new tightly-coupled filter, that estimates
the INS navigation errors and sensors bias calibration uncertainties,
are evidenced when compared to a more conventional loosely- coupled
solution, while also being shown to operate closer to state-of-the-art
theoretical performance lower bounds. A new approach to the design of
globally asymptotically stable (GAS) position and velocity filters, is
also presented based directly on the nonlinear ranges readings of the
USBL acoustic array. A novel attitude filtering solution is also
presented for an intervention underwater vehicle working in tandem with
an autonomous surface craft. The design, development, and testing of an
integrated USBL and INS to be used as a low-cost navigation system for
underwater robotic vehicles is also presented. Experimental results
obtained at sea are analyzed and discussed to assess the performance and
feasibility of the navigation system.
Keywords: INS, underwater acoustic positioning, USBL, marine robotic
vehicles, autonomous underwater vehicle (AUV), position and attitude
estimation, Kalman filters, nonlinear observers, underwater acoustic
signal generation, matched filters.
Thesis approved with Muito Bom com Distinção e Louvor (top rank), December 2011, PDF File. Presentation PDF File.
Portuguese Society of Robotics Award for the Best PhD Thesis in Robotics 2011, April 2012. PDF File.
Best student paper award for the paper: Morgado M., Oliveira P., and
Silvestre C., ‘‘Posterior Cramér-Rao bounds analysis for INS/USBL
navigation systems,” 8th IFAC Conference on Manoeuvring and Control of
Marine Craft–MCMC 2009, Guarujá, Brasil, September 2009. Best student
paper award.
Atualmente: Diretor Geral do Gigajoule Group, Moçambique.
Doctoral Program in Electrical and Computer Engineering, IST
Pedro Tiago Martins Batista, with the thesis entitled ‘‘Sensor-based Navigation and Control of Autonomous Vehicles”
Supervisor: Prof. Carlos SilvestreCo-supervisor: Paulo Oliveira
Abstract:
This thesis addresses the problems of sensor-based
navigation and control of autonomous vehicles. Due to the broad range of
topics present in the thesis, the contributions are organized in three
main categories, which share the sensor-based concept as a common
ground. The first part of the thesis, entitled Sensor-based Linear
Motion Estimation, is devoted to estimation of linear motion quantities
(position, linear velocities, and linear accelerations), in 3-D, of
autonomous vehicles. Novel sensor-based navigation solutions are derived
based on different sensors, in particular, position sensors, both in
body-fixed and inertial coordinates, single range measurements, and
multiple range measurements. The second part of the thesis, entitled
Sensor-based Angular Motion Estimation, concerns the estimation of the
attitude of a vehicle. A novel sensor-based general framework is devised
and exploited in different mobile platforms. The third part of the
thesis, entitled Sensor-based Control of Autonomous Underwater Vehicles
(AUVs), presents novel sensor-based integrated guidance and control
strategies for homing of AUVs to a base station based on an Ultra-short
Baseline position sensor. The solutions proposed in the thesis are
strongly rooted in well-known control and filtering system theories, and
are deeply focused on asymptotic stability properties and achieved
performances.
Current occupation: Assistant Professor at SDC/DEEC/IST.
Doctoral Program in Electrical and Computer Engineering, IST
José Maria Fernandes Vasconcelos, with the thesis entitled ‘‘Nonlinear Navigation System Design with Application to Autonomous Vehicles”
Supervisor: Prof. Carlos SilvestreCo-supervisor: Paulo Oliveira
Abstract:
This thesis addresses the design of nonlinear
navigation systems for autonomous vehicles, following two main
approaches: Kalman filter based estimators, and Lyapunov theory based
nonlinear observers. The proposed Kalman filter architectures are
designed for accurate position and attitude estimation using low-cost
sensor suites. An extended Kalman filter is adopted to merge a high
accuracy inertial navigation system with advanced aiding information,
namely i) frequency contents of vector measurements, ii) vehicle model
dynamics, and iii) LASER range measurements. In alternative, simple yet
effective multirate complementary Kalman filters are proposed, endowed
with stability and performance properties. The navigation systems are
validated using realistic vehicle simulators, and experimental data
collected onboard the DELFIMx autonomous surface craft. The second
approach addresses the design of nonlinear observers in non-Euclidean
spaces. The observers are derived resorting to Lyapunov theory, bearing
stability and robustness properties in the presence of inertial sensor
non-idealities. The considered sensor readings are provided by an
inertial measurement unit and i) landmark measurements, ii) vector
observations, and iii) GPS receivers. The regions of attraction are
explicitly characterized, and an output feedback configuration is
proposed, allowing for the practical implementation of the algorithms.
Current occupation: Project Manager at Altran, Lisbon.
Doctoral Program in Electrical and Computer Engineering, IST
Alex Alcocer Penas, with the thesis entitled ‘‘Positioning and Navigation Systems for Robotic Underwater Vehicles’’
Supervisor: Paulo Oliveira Co-supervisor:Prof. António Pascoal
Abstract:
This thesis addresses the problem of underwater
navigation of robotic vehicles using acoustic positioning systems.
Several estimation problems are considered that are based on Range-Only
measurements obtained from the Times of Arrival of acoustic signals.
First, the Range-Only localization problem is addressed, which consists
of determining the position of a vehicle given ranges to a set of
landmarks with known locations. This problem arises in acoustic
positioning systems such as GIB (GPS Intelligent Buoys). Several
solutions based on Least Squares, Maximum Likelihood, and Extended
Kalman filtering are presented and applied to real experimental data
obtained during sea trials. Special attention is given to performance
issues and practical problems related to acoustic positioning systems
such as sound speed estimation and multipath mitigation. Second, the
problem of pose estimation with Range-Only measurements is addressed, in
which the vehicle is equipped with an array of beacons with known
relative position and uses range measurements to a set of Earth fixed
landmarks. A Maximum Likelihood estimator is derived that requires
solving a constrained minimization problem on the Special Euclidean
group SE(3). Borrowing tools from optimization on Riemannian manifolds,
generalized gradient and Newton methods are derived to solve this
problem. An alternative solution is derived in a system theoretic
setting by adopting a suitable Lyapunov function that is a function of
range measurements only, yielding convergence conditions. Finally, the
thesis addresses the post-processing of acoustic positioning data. An
extension of diffusion-based trajectory observers is derived that
incorporates measurement error information.
Current occupation: Associate Professor, Electrical Engineering and Robotics @ OsloMet - Oslo Metropolitan University, Norway.
Supervision of PhD Students with the Theses Ongoing
Doctoral Program in Mechanical Engineering, IST
João Pedro Dias Madeiras, with the workprogram entitled “Inspection with Unmanned Aerial Vehicles”
Start date: January 2021
End date: December 2024
Supervisor:Paulo
Oliveira
Supervisor: Prof. Carlos Cardeira
Summary of the Workplan (in Portuguese):
A scientific research plan is proposed for
the development of Advanced Navigation Systems design methodologies with
applications to Inspection with Unmanned Aerial Vehicles. For this
work, it is foreseen the operation of unmanned, dynamic and
high-performance platforms in unstructured environments and with high
social impact objectives such as the inspection of infrastructures, the
location in relation to conspicuous points in the environment, the
surveillance of sensitive areas and tracking uncooperative targets. With
strong inspiration and motivation in energy criteria, we intend to use:
i) techniques for synthesis of dynamic systems based on recent
developments in the theory of Optimization for Linear and Non-Linear
Systems, ii) Sensory Fusion to obtain Non-Linear Navigation Systems and
iii) non-linear techniques (Lyapunov Theory) or total linearization
techniques (Lyapunov Transformations, Polynomial Systems, ...). The
obtained solutions will be validated in prototypes existing in IDMEC /
LAETA.
Scholarship from FCT from January 2021 until December 2024.
Doctoral Program in Mechanical Engineering, IST
Pedro Miguel Gomes Outeiro, with the workplan entitled “Guidance, Navigation, and Control Systems for Unknown Load Transportation by UAVs”
Start date: November 2019
End date: October 2023.
Supervisor:Paulo
Oliveira
Supervisor: Prof. Carlos Cardeira
Summary of the Workplan:
The development of methodologies for the
project of integrated Guidance, Navigation and Control systems with
application of Unmmaned Aerial Vehicles is proposed. In this work the
use of umanned platforms, dynamic and of high performance in
unstructured environments is foreseen. High societal objectives are
pursued namely on the transportation and distribution of unknown loads
in isolated regions, search and rescue, intervention on the environment,
inspection of critical infrastructures, and surveillance of sensitive
areas. Inspired on energetic criteria the work will resort to: i)
synthesis techniques for linear systems based on recent advances in the
Optimization Theory for Linear and Nonlinear Systems; ii) Linear
Parametric Varying Systems; iii) Nonlinear Lyapunov techniques or full
linearization (feedback linearization or flat systems). The solutions
obtained will be validated in prototypes available in the
CSI/IDMEC/LAETA.
Scholarship from FCT from November 2019 until October 2023.
Supervision of MSc Students
Master in Science in Electrical and Computer Engineering, IST
Bruno Miguel Simões Carvalho Cardeira, with the thesis entitled ‘‘Arquiteturas para Navegação Inercial/GPS com Aplicação a Veículos Autónomos” (in Portuguese)
Supervisor: Prof. Carlos
SilvestreCo-supervisor : Paulo Oliveira
Abstract:
This
thesis addresses the development and integration of a strapdown
navigation system to determine the position and attitude of unmanned
vehicles, using accelerometers, magnetometers and rate gyros triads
aided by Global Positioning System (GPS) measurements. The current work
resorts to Complementary Filtering techniques to implement the
navigation system developed on Earth frame coordinates for the position
estimation, with the velocity estimated in the body frame and with the
attitude described using Euler angles. Special features include bias
estimation and removal in inertial sensors. An attitude-aiding device,
referred to as Magneto-Pendular Sensor, is implemented and the synthesis
of the multirate complementary filters is outlined. Stability and
performance properties of the proposed filters are derived to solve the
position and attitude estimation problem. The proposed filters
parameters are synthesized based on optimality results, regarding the
available sensor suite characteristics, that can be stochastic or
frequency based. It is also presented and briefly discussed a nonlinear
transformation that finds application in the stability and performance
analysis of the position and attitude complementary filters. Formulated
in discrete-time, the position and attitude complementary filters allow
for practical implementation without requiring high performance signal
processing hardware and maintaining all their stability and performance
properties. The hardware architecture for the implementation of the
real-time navigation system is presented and the different hardware
modules are described. Integration issues, both electromagnetic and
mechanical, that arise from the interaction of different systems are
also addressed. The on-board systems and ground station design made from
Commercial-Off-The-Shelf (COTS) sub-systems and custom developed
hardware/ software modules are briefly presented. Finally, the overall
system performance is evaluated both in simulation and in sea trials
using the DELFIMx catamaran developed at the Institute for Systems and
Robotics - Lisbon.
