EURON  PhD theses with short discussion



 
Nonlinear Motion Control of Nonholonomic and Underactuated Systems
Instituto Superior Técnico, Instituto de Sistemas e Robótica, Lisboa, Portugal, April 2002  (www.isr.ist.utl.pt/~apra)
ABSTRACT

This thesis addresses the problem of stabilization of nonholonomic and underactuated systems. The key motivation for this research topic stems from the fact that nonholonomic systems pose considerable challenges to control system designers since those systems cannot be stabilized by smooth, time-invariant, state-feedback control laws. Furthermore, in spite of the number of methods available for the control of nonholonomic and underactuated mechanical systems, few address important practical topics such as the explicit inclusion of dynamics in the control problem formulation and the need to cope with model parameter uncertainty and external disturbances. This thesis tackles some of theses issues, formulates and solves the related control problems, and discusses the application of the new control methodologies derived to robotic land and underwater vehicles.

The first part of the thesis focuses on the control of the so-called extended nonholonomic double integrator (ENDI), which captures the kinematics and dynamics of a class of mobile robots. A solution to the problem of global stabilization of the ENDI is given that builds on logic-based hybrid control. The methodology derived is applied to the problem of stabilizing the dynamic model of an underactuated autonomous underwater vehicle (AUV) to a desired pose.

The second part of the thesis is devoted to the problem of steering a wheeled mobile robot and an underactuated AUV to a target point with a desired orientation, in the presence of parametric modeling uncertainty. Controller design relies on non-smooth coordinate transformations in the original state space, followed by the derivation of adaptive, smooth, time invariant feedback control laws in the new coordinates. The problem of positioning and way-point tracking of an underactuated AUV in the presence of an unknown ocean current disturbance is also
investigated and solved using a similar approach.

The last part of the thesis is devoted to the general problem of nonlinear system stabilization. A new controller design methodology is presented for a large class of nonlinear systems that builds on recently developed switching hybrid control techniques and classical Lyapunov based design tools.

Throughout the thesis, formal proofs of convergence of the algorithms derived are presented. Simulation results obtained with full nonlinear models of representative land and underwater robots are discussed.

Department of Electronic, Computer Science and Systems of the University of Bologna, Italy, March, 2002
( http://www-lar.deis.unibo.it/people/parcara )

Advisor: Prof. Claudio Melchiorri ( http://www-lar.deis.unibo.it/people/cmelchiorri/ )

PRESENTATION OF THE WORK

Haptic systems and robotic telemanipulators have been spreading out during the last years and many applications have been developed all over the world with the aid of telerobotics.
This thesis is inserted in this telerobotic field presenting both a theoretical part and a practical experimental part. Haptic and telemanipulation systems have been presented and detailed focusing in particular on their control algorithms; different control methods taken from the already existing literature have been presented and, after the definition of a set of indexes related to stability and performances, a comparison has been carried out leading to the underlining of the pros and cons of each presented scheme.

The experimental part is also presented for the validation of the theoretical results and it has been carried out in the Laboratory of Automation and Robotics of the University of Bologna.

DEFINITIONS

The word ``haptic'' does not have a rigorous definition and it means something related with the kinestetic and tactile human perception with the use of the hand. Haptic devices have to be in contact with the human operator giving him sensations as similar as possible to those he could perceive from the interaction, due to handling or grasping, with an object. Sometime, a visual or acoustic feedback, through appropriate displays, is added to the device to give a more complete and detailed perception to the user.

The good rendering that a person can perceive from a haptic system are mainly due to three concomitant factors: the simulator of the virtual reality, the control of the haptic device and the hardware interface of the same device. These factors contribute in an equal and important way to the good rendering of the task being simulated.

Telemanipulation schemes are bilateral systems which include both a local primarily side, or master, and a remote one said slave. The human operator interacts with the master manipulator with a double possibility, first to drive the remote side in the completing of the task and second to receive from the slave the correct force feedback of what exactly happening. Master and slave manipulators are governed by two local controllers and exchange their respective information with the opposite side through the communication channel. The communication channel has its proper bandwidth and delay, depending on the kind of transmission implemented.

From the above descriptions, one can easily comprehend that the only difference between haptic and telerobotics systems is just due to the remote side to be rendered to the user; in the first case, the haptic case, the remote side is not physically existing but it is simulated via software; in the second case, the telemanipulation one, the remote side is composed by a manipulator with its controller and by the environment with which it is interacting. So the delays in the two systems are of different nature: they are due to the simulation computation time in the haptic case and to the transmission channel between local and remote side in the telemanipulation case.

Haptic systems have been mainly developed to give to the user a certain kind of perception, as realistic as possible, of what is going on in a simulated virtual scene. These kind of systems are mainly used for the training in the carrying out of specific complex tasks, for which it is too difficult or too expensive a direct training in the real operation; fields of application are for example virtual surgery, medical rehabilitation, flight simulators and so on; haptic devices are also used in the military and in the entertainment business to give to the persons a partial, or total, ``immersion'' in a virtual world being created artificially. Finally, haptic devices, once that they are built and tested, can be used, with great success, as
principal or master manipulators in a telemanipulation scheme, giving to the user a good control and feedback on the remote task being carried out.

As far as telerobotics is involved, possible telemanipulation systems include the following applications: dangerous environments for the human being, such as nuclear plants, underwater tasks, space operations and so on, in which it is not safe to operate directly for a person; differently scaled tasks, such as micro-surgery, high dimension or high load tasks, in which the user feels great difficulty due to the very different scale of the environment with respect to him; and finally, the ``simple'' possibility to operate remotely without being physically present in the real scene.

CONTENTS OF THE THESIS

A brief description of the chapters of this thesis is now presented, the work have been structured in the following way:

Chapter 1 presents the introduction of the work.

Chapter 2 contains the basic definitions for haptic systems with special focus on their control aspects, in particular this chapter deals with tendons/wires actuated devices and the aspects strictly connected to their peculiar structure.

Chapter 3 presents the definition of the elements of a complete telemanipulation scheme, the main properties and the ideal transparency aim and, finally, a detailed and organic presentation of all the principal control algorithms presented in the literature on this topic.

Chapter 4 defines stability and performances properties for a telemanipulation scheme and it proposes four different performance indexes in order to compare the algorithms previously presented in the benchmark example of two simple one degree-of-freedom (dof) linear manipulators.

Chapter 5 contains simulations and studies of a specific telemanipulation scheme, i.e. the intrinsically passive one; it holds with three dof manipulators, multiple input multiple output case, and enhances their performances using passivity definitions and concepts; this chapter contains also a passive approach for further performance enhancements in terms of position drift between manipulators.

Chapter 6 deals with two experiments carried out during the Ph.D. studies: the ViDet device, a particular haptic device actuated by tendons presented with its two different prototypes, and the telemanipulation device, built in order to test what above shortly described.

Chapter 7 contains some conclusions related to this work and it presents possible future work to be done.

Appendix A, finally, resumes the nomenclature of all the variables used in the mentioned chapters of this thesis.

ABSTRACT

A new methodology is proposed for the design of path following systems for surface and underwater autonomous ocean vehicles in the presence of constant but unknown ocean currents. Convergence to reference paths is achieved with a nonlinear control strategy that takes explicitly into account the dynamics of the vehicles as well as those of a current estimator. Formal convergence proofs are indicated. Simulation results with models of prototype ocean vehicles are presented to illustrate the performance of the path following system derived.

Additionally, the thesis presents a solution to the problem of combined trajectory tracking and path following for underactuated vehicles with non-negligible dynamics. The methods described borrow from and extend previous work by Hindman and Hauser on so-called maneuver modified trajectory tracking. Lyapunov based and backstepping techniques are used. Simulations with the models of a dynamic wheeled mobile robot, an
underactuated surface craft, and a fully actuated underwater vehicle are presented to illustrate the performance of the combined trajectory tracking and path following controllers. The same circle of ideas is applied to develop a control system for the coordinated operation of a surface craft and an underwater vehicle that are required to follow the same projected reference path at different depths while keeping a vertical channel for communications.

ABSTRACT

The thesis presents a new control strategy that allows safe manipulation and transport of deformable object by nonholonomic mobile manipulators in an environment with obstacles. At first, the system is modeled kinematically
and dynamically. Then, a centralized nonholonomic kinematic controller that coordinates the motion of the complete system is designed. The desired motion is realized by decentralized dynamic controllers. Overall, the
control system is formed by two nested closed loops: the outer loop serves for feedback-based nonholonomic motion planning and the inner loop is used for decentralized dynamic control. The complete closed loop system is shown to be semiglobally asymptotically stable.
A nonholonomic mobile manipulator is modeled for the first time using a modern methodology known as "Kane's dynamical equations". This methodology is characterized by less complexity, both in derivation and in computation.
Furthermore, nonholonomic constraints can be directly incorporated in the dynamic equations without resorting to multipliers. Finally, Kane's method offers physical insight that reveals several subtle aspects of the dynamics
of a mobile manipulator, which are not obvious when alternative methodologies are used.
Motion planning and coordination is performed in a closed loop fashion, utilizing a new class of artificial potential fields. The workspace, the robots and the deformable object configurations are captured by an inverse
dipolar potential function. This artificial potential field guarantees obstacle avoidance, global kinematic convergence and local optimization of motion, subject to a number of secondary objectives such as singularity
avoidance and deformation minimization. This particular type of potential field, combined with a discontinuous kinematic controller is appropriate for nonholonomic navigation. By extending Lyapunov stability results, global
asymptotic convergence for the closed loop nonholonomic kinematic system is established.
The reference commands of the centralized kinematic controller are realized by several decentralized dynamic controllers. Dynamic control is designed based on new theoretic results that extend integrator backstepping to
nonsmooth and discontinuous systems. On the other hand, the deformable object is analyzed for the first time within the framework of underactuated mechanical systems. This treatment enables the application of advanced
nonlinear control techniques developed for this class of mechanical systems to deformable material handling.
Within this framework, sufficient conditions for output tracking with boundedness of the internal dynamics are developed for the deformable object.
Manipulation of the object is achieved through direct force control in a new control architecture of nested loops, in which the inner loop provides force feedback and the outer position feedback. Stability of the complete
system is established by application of singular perturbation theory, and lower bounds for the control gains are specified.
In this way a new complete control strategy for transporting and handling of deformable objects in cluttered environments by nonholonomic mobile manipulators is developed. The proposed methodology introduces
novelties, not only in motion planning and dynamic control of nonholonomic systems but in modeling and deformable object handling as well. The results can also be applicable to other problems such as coordination and control of formations of mobile robots, obstacle avoidance for multi-link mechanisms and shape control for deformable objects. Potential applications of autonomous robotic systems for deformable object manipulation, can range from robotic surgery and planetary exploration to construction and garment manufacturing

Dipartimento di Informatica e Sistemistica Universita' di Roma "La Sapienza" April 2001

http://labrob.ing.uniroma1.it/people/antoniali/antoniali.html

Abstract

The subject of this thesis is the development of a novel mathematical tool for  the localization of mobile robots equipped with range finders and heading sensing system, moving in two-dimensional indoor environments. The distinctive feature of this work is that the environment boundary is assumed to be only piecewise-linear; as a consequence, the measurement function modeling the range finders is discontinuous. It is well known that classical unimodal filters such as the Extended Kalman Filter and the Single Stage Iteration Filter do not provide reliable estimates in such a strongly nonlinear setting. Various attempts at performance improvement have been devised within the Bayesian framework, from the earlier Gaussian Sum Filter to the recent multimodal techniques such as Multi Hypothesis Tracking, Grid Based Markov Localization and Monte Carlo Localization. However, these techniques suffer from several problems and are not supported by a solid theoretical basis. Another common criticism is that many localization filters do not take advantage of the information conveyed by the structural nonlinearity that the measurement function inherits from the environment geometry. This thesis presents a solution to the above problems in the form of a multimodal filter called the Multi-Hypotheses Density Filter (MHDF), which applies to a class of discrete-time systems characterized by linear state equation and piecewise-linear measurement equation, and provides a sub-optimal approximation of the true state distribution. Moreover, the MHDF yields accurate and robust estimates even in presence of strong perturbations.
It is shown that the MHDF can be applied to the localization problem of nonholonomic mobile robots. The performance of the localization technique based on the MHDF is first assessed by means of an extensive simulation study consisting of several campaigns of Monte Carlo experiments in which different environments and noise conditions are used. Then, the thesis presents the results of the experiments performed with the ATRV-Jr robot equipped with a MHDF-based localization system. Both simulation and experimental results show that the MHDF provides superior performance with respect to the classic Extended Kalman Filter. Furthermore the proposed results confirm that this filtering technique allows to perform the re-localization task, thanks to the MHDF capability of dealing with strong o complete uncertainty on the initial state.

Integration of vision and force for robotic servoing
Department of Mechanical Engineering Katholieke Universiteit Leuven, Leuven, Belgium, 2001.
(http://www.khlim.be/~jbaeten/indexE.html)

Abstract:
This work shows how integrated vision/force robotic control improves the task quality, in the sense of increased accuracy and execution velocity, and widens the range of feasible tasks. It demonstrates the fitness of the task frame formalism and the hybrid control structure as a basis to easily model, implement and execute robotic tasks in an unknown workspace. To
this end, the high level task description divides the control space in vision, force, tracking and velocity controlled directions, possibly augmented with feedforward control.

Several tasks illustrate traded, hybrid and shared vision/force control with both sensors mounted on the end effector. This combined mounting and control is classified into four meaningful camera/tool configurations being parallel or non-parallel endpoint closed-loop and fixed or variable endpoint open-loop (EOL). The EOL configuration with variable task/camera frame relation is fully explored, since it is the most adequate one for `simple' image processing and the most challenging
in the sense of control.  For the variable EOL configuration, special control issues arise, due to the time shift between the moment the contour is measured and the moment these data are used.  Furthermore, keeping the image feature in the camera field of view, while maintaining a force controlled contact, imposes additional requirements on the controller. This is adequately solved using the redundancy for rotation in the plane, which exists for rotationally symmetric tools, in order
to control independently task and camera orientations. The presented variable EOL tasks demonstrate that adding vision based feedforward to the tracking direction reduces tracking errors, hereby enabling a faster and more accurate execution of the task.

Control of flexible manipulators with extensions to  non-minimum phase systems

Ecole Centrale de Nantes - France

This thesis concerns the control of the flexible manipulators. We have generalized some of the proposed control schemes to the case of non-minimum phase systems.
The results proposed in this work are:
1- An approach for stable inversion of linear SISO non-minimum phase systems, through output trajectory planning. This method has been applied to the one flexible link robots, and validated through experimental results.
2- The control of the one flexible link robots via an algebraic scheme, based on the parameterization of linear differential operators. With experimental results validation.
3- Output tracking via stable inversion for a particular class of nonlinear non minimum phase systems , through boundary value problem formulation. This method has been applied to the PVTOL (Planar Vertical Take off and Landing aircraft) system and to the multi-link flexible planar manipulators. The results have been validated trough experiments on a two-link flexible robot.
4- A solution for the rest to rest motion in fixed time for the multi-link flexible planar robot. This method based on optimal trajectories planning has been validated via experiments on the two-link flexible manipulator.
5- A solution for the joint trajectories tracking for multi-link flexible planar robot, based on non-causal controls. The results have been validated through simulation tests.

Key words: Robotics, control, manipulator robots, trajectory tracking, non minimum phase systems, stable inversion, optimal trajectories planning.
Members of the commission
H.  ABOU-KANDIL      (Professor  at  ENS de Cachan- France)
B. d'ANDREA NOVEL (Professor at  Ecole des Mines de Paris)
E. BAYO                       (Professor at the university "Navarra", Spain)
A. DE LUCA                 (Professor at the université "La Sapienza", Italy)
C. SAMSON                 (Research director at  INRIA Sophia-Antipolis- France)
Supervisors:
W. KHALIL                   (Professor at the Ecole Centrale de Nantes- France)
G. LE VEY                    (Assistant professor at the Ecole des Mines de Nantes)
State Estimation and Limited Communication Control for Nonlinear Robotic Systems
henrikr@math.kth.se, OPTSYST
In this thesis, two different control theoretic topics are studied, nonlinear observer design for orientation estimation and control design under limited communication. The specific problems studied originate from control of mobile robots in difficult terrain, especially walking robots.

In Paper A, the fusion of rate gyro data and inclinometer data with a high-gain observer and a time-varying observer is studied. The Euler angle representation of rigid body motion is used. It is shown that even in the process of rate gyro offsets and slow inclinometer dynamics it is possible to derive exponentially convergent attitude observers with arbitrarily small bounded errors. The high-gain observer is evaluated experimentally.

In Paper B, rate gyro data and accelerometer data are fused in order to obtain attitude estimates. Two exponentially convergent observers are presented. The underlying kinematic model used for the observer design is global and the state evolves on the unit sphere. The observers are evaluated experimentally on a walking robot.

In Paper C, the use of inertial sensors and computer vision is investigated. Orientation estimation is studied and it is shown that orientation can be estimated in the presence of an unknown time-varying position. The problem is phrased as an observer problem for an implicit output system evolving on $SO(3)$. Observability is studied and a generalization of the unobservable subspace, the unobservable subgroup, is derived. A locally convergent exponential observer is presented and the domain of attraction is estimated numerically. It is shown that position estimation can be formulated as a linear implicit output problem.

In Paper D, the problem of control design under limited communication is studied. It is shown how a limited communication channel can be modeled and how the constraint due to this limitation is incorporated into the system model. An algorithm for computing optimal communication schedules and corresponding LQ-controllers is given.


patric@nada.kth.se  S3, KTH; Stockholm
Opponent: Raja Chatila, LAAS, F

The thesis deals with all aspects of mobile robot localization for indoor applications. The problems span from tracking the position given an initial estimate, over finding it without any prior position knowledge, to automatically building a representation of the environment while performing localization. The theme is the use of minimalistic models which capture the large scale structures of the environment, such as the dominant walls, to provide scalable and low-complexity solutions.

In many cases it is enough to only maintain an estimate of the robot position. For such situation an extensively tested low-complexity, robust and accurate pose tracking method is presented which utilizes the minimalistic model in combination with a laser sensor.

When the initial position is unknown the robot must perform global localization. Two different methods are investigated. The first one is a novel localization scheme, based on the ideas of Multiple Hypothesis Tracking. The second is an, experimentally verified, significant improvement of the standard Monte Carlo Localization technique.

To automatically generate an environmental representation an hierarchical approach to simultaneous localization and mapping (SLAM) is presented. The map scaling issue is here addressed by dividing the environment into submaps, each representing a small area.

Service robots are gradually extended to operation in everyday environments. To be truly useful, a mobile robot should include facilities for interaction with the environment, in particular methods for manipulation of objects. One of the most flexible sensory modalities to enable this is computational vision. In this thesis the issue of visual servoing and grasping to facilitate such interaction is investigated. A notorious problem for use of vision in natural environments is robustness with respect to variations in the environment. It is also well-known that no single technique is suitable for different tasks a robot is supposed to perform.

Robustness is here investigated using several different approaches. The issues of variability are formulated with respect to
visual features, the number of cameras used, and task constraints. It is argued that integration of methods facilitate construction of more robust visual servoing systems for realistic tasks.

Traditionally, fusion of visual information has been based on explicit models for uncertainty and integration. The most dominating technique has been use of Bayesian statistics, where strong models are employed. Where a large number of visual features are available it is suggested that it might be possible to perform tasks such as visual tracking using weak models for integration. In particular, integration using voting based methods is analyzed.

If the object to be manipulated is known or has been recognized, it is possible to use explicit geometric models to facilitate the estimation of its pose. Consequently, a methodology for tracking of objects using wire-frame models has been developed and evaluated in the context of grasping.

Visual servoing can be carried out in the image domain and/or using 3D information. In this context, the tradeoff between explicit models and use of multiple cameras strongly influences the performance of a visual servoing system. The relation between visual features, the number of cameras and their placement has been studied to provide guidelines for a design of such a system.

An integration of a multi-ocular vision system, suitable visual techniques and task constraints facilitate flexible manipulation of everyday objects. To demonstrate this the developed techniques have been evaluated in the context of manipulation for opening/closing of doors in an everyday setting. In addition, it is demonstrated how the techniques, together with model based information, may be used for grasping and grasp monitoring in the context of a well-known set of objects.

In summary, a toolkit for interaction with everyday objects has been investigated and evaluated for real-world tasks. The developed methods provide a rich basis for real-world manipulation of objects in everyday settings.

Last modified by Rosario Marino

on2002-07-15 @ 09:45