Dynamics & Controls Seminars
Academic Year 2005-2006
* * * * * * * * * * * * *
Wednesday, January 11, 2006
10:00 a.m.
479 EBU-II
Eugenio Schuster
Laboratory for Control of Complex Physical Systems
Mechanical Engineering and
"Flow
Control in Magnetohydrodynamic (MHD) Channels"
A nonlinear Lyapunov-based
boundary feedback control law is proposed for mixing enhancement in a magnetohydrodynamic (MHD) channel flow, also known as Hartmann flow. This flow is
characterized by an electrically conducting fluid moving between parallel
plates in presence of an externally imposed transverse magnetic field. The
system is described by the MHD equations, a combination of the Navier-Stokes equation and the Maxwell equations under the
so-called MHD approximation. Pressure sensors, magnetic field sensors, and
micro-jets embedded into the walls of the flow domain are considered in this
work to find a feedback control law for mixing enhancement. The proposed
control law maximizes a measure related to mixing (that incorporates stretching and folding of material elements), while at the same time minimizes
the control and sensing efforts. Alternative control mechanisms will be
discussed.
* * * * * * * * * * * * *
Tuesday, December 6, 2005
12:00 p.m.
4307 EBU-I
Dr. Manfred Morari
Automatic Control Laboratory
Swiss Federal Institute of Technology
"Hybrid Systems: Theory, Computation, and
Applications"
Theory, computation and applications define the evolution of the field of control. This premise is first illustrated with some historical examples and then with the emerging area of hybrid systems, which can be viewed, loosely speaking, as dynamical systems with switches. Many practical problems can be formulated in the hybrid system framework. Power electronics are hybrid systems by their very nature, systems with hard bounds and/or friction can be described in this manner and problems from other domains, as diverse as driver assistance systems, anesthesia and active vibration control can be put in this form. We will highlight the theoretical developments and mention the extensive software that helps to bring the theory to bear on the practical examples. We will close with an outlook for hybrid systems and control more generally.
* * * * * * * * * * * * *
Tuesday, November 22, 2005
11:00 a.m.
479 EBU-II
Bram de Jager
Technische Universiteit
Eindhoven
“
It is shown that introducing additional freedom in selecting the geometry of the tetrahedral or pyramidal core, by
1) making the core units, or even all core members, disjoint
2) varying the distance between core units,
3) using multiple layers,
brings significant advantages.
For some conditions, the performance of the modified panel becomes better by a factor of almost two. A special case of the analysis is the mass optimal design of panels with 3D-Kagom’e truss cores.
* * * * * * * * * * * * *
Thursday,
November 17, 2005
3:00 p.m.
479 EBU-II
Joseph Bentsman
Department of Mechanical and
"Robust
Adaptive Control of Parabolic and Hyperbolic Spatially Varying Systems and its
Multi-resolution Finite-Dimensionalization"
Continuous steel casting admits approximation of boundary control of a two-dimensional parabolic PDE currently used to describe the process by a distributed control of a one-dimensional parabolic PDE. Parameters of the latter, such as heat transfer coefficients, are non-smooth slowly varying functions of spatial variable. Therefore, identification and adaptive control of this system and similar ones are of interest. With this motivation, a long-standing problem of the ill-posedness-induced instability in the model reference adaptive control (MRAC) of distributed parameter systems (DPS) with distributed sensing and actuation is solved for a class of parabolic and hyperbolic partial differential equations (PDEs) with spatially varying coefficients. A novel finite-dimensionalization multiresolution technique for the controller parameters adaptation law is then proposed. This technique permits efficient incorporation of the specific plant parameter characteristics such as non-smoothness into controller implementation, reducing computational demand, convergence time, and control signal magnitude. For this purpose, a new tool - the multiresolution Lyapunov functional is introduced. Using the latter, the stability of the closed loop system with the wavelet-based finite-dimensional parameter adaptation law and the infinite-dimensional plant is rigorously proven.
* * * * * * * * * * * * *
Thursday, October 27, 2005
9:30 a.m.
YC Fung Auditorium, Bioengineering
Dr. M. Vidyasagar
"The 4m (Mixed Memory Markov Model)
Algorithm for Finding Genes from Genome Sequences"
In this talk, we present a new algorithm, called the 4M (Mixed Memory Markov Model) algorithm, for classifying a stretch of genome (e.g., an ORF) as a coding or non-coding region (for prokaryotes). With minor modifications, the algorithm can also detect splice sites and be extended to eukaryotic genomes.
Popular algorithms such as Glimmer use multi-step Markov models, and then use combinations thereof. The basis of the 4M algorithm is the observation that in a multi-step Markov process, different past histories can have memories of different lengths; hence the name 4M. The effective memory of various past histories is computed using a simple rank condition.
The advantage of the 4M algorithm over other, existing methods is that its statistical properties (in terms of their significance) are easy to analyze.
This method has thus far been applied to several bacterial genomes, and yields predictions that are quite comparable to those of Glimmer. Moreover, some new genes have been “predicted” even in well-studied organisms. These predictions are now being verified experimentally by our partner laboratories.
* * * * * * * * * * * * *
Wednesday, September 28, 2005
10:00 a.m.
479 EBU-II
Dr. Aniruddha Datta
Dept. of Electrical
“Modeling
and Control in Cancer Genomics”
Genomics concerns the study of large sets of genes with the goal of understanding collective function, rather than that of individual genes. Such a study is important since cellular control and its failure in disease result from multivariate activity among cohorts of genes. Very recent research indicates that engineering approaches for prediction, signal processing and control are quite well suited for studying this kind of multivariate interaction. In this talk, we will present an overview of the research that has been accomplished thus far in this interdisciplinary field and point out some of the open research challenges that remain.
Among the recent paradigms
that have been proposed for modeling genetic regulatory networks are the so
called Probabilistic Boolean Networks (PBN’s).
Such rule-based networks provide a convenient tool for studying interactions
between different genes while allowing for uncertainty in the knowledge of
these relationships. This talk will first introduce PBN’s
as a modeling tool and then consider the issue of control in probabilistic Boolean
networks. First, we will consider the following control problem: given a
probabilistic Boolean network whose state transition probabilities depend on an
external (control) variable, choose the sequence of control actions to minimize
a given performance index over a finite number of steps. This is a standard
finite horizon optimal control problem for Markov Chains and can be solved
using the classical technique of Dynamic Programming. This choice
of the finite horizon performance index is motivated by cancer treatment
applications where one would ideally like to intervene only over a finite time
horizon, then suspend treatment and observe the effects over some additional
time before deciding if further intervention is necessary. Having established
the connection between optimal control theory and a problem in cancer therapy,
we will highlight several challenges that will have to be overcome before such
methods can be used in actual clinical practice. We will also report on
ongoing work and progress made in overcoming some of these challenges.
* * * * * * * * * * * *
For Information: (858) 822-1269