Optimal Control of Switching/Hybrid Systems with Applications to Hybrid Electric Vehicles, DC-DC Converters, and Autonomous Mobile Robots

Organizers

Ray DeCarlo, Steve Pekarek - Purdue University, West Lafayette, IN (USA)

Miloš Žefran, University of Illinois-Chicago, Chicago, IL (USA)

This workshop will present recently developed results on the solution of the hybrid/switched optimal control problem using the embedding method developed by Bengea and DeCarlo (Automatica, January 2005).  Using a variation of the collocation method, a numerical solution of the problem via sequential quadratic programming is outlined.  Using these tools and a model predictive control approach, application of the techniques to the switching control of a boost converter using a sliding mode observer is then presented followed by the model predictive control of mobile robots and groups of autonomous aerial vehicles (AUVs).  Finally, a solution to the power management problem in a hybrid electric vehicle is presented with simulation studies for a variety of driving profiles including the new EPA driving profile.  The examples will not only describe appropriate models, MPC control methodologies, and simulation studies, but also highlight the broader appeal of these newly developed techniques for modeling, analysis, and design of hybrid/switched systems.

Description

Intelligent Systems for Modeling and Control: Advances in Design and Validation

Danil Prokhorov, Toyota Tech. Center, Ann Arbor, MI (USA)

Johann Schumann, Robust Software Engineering, RIACS/NASA Ames (USA)

Intelligent systems, or systems which include neural, fuzzy or evolutionary components, have to be designed or trained carefully, taking into account uncertainties, and verified/validated well before they are accepted for deployment.  This workshop intends to present an overview of the state of the art and recent advances in intelligent systems for modeling and control, with examples from automotive, aerospace and chemical industries.  With respect to automotive processes and in-vehicle systems, steps undertaken to design and validate intelligent control, diagnostics and prognostics will be discussed.  Among several automotive examples, fever-like symptoms in an engine suffering from an unknown fault will be demonstrated, recognized and mitigated through a novel approach of artificial immune system.  With respect to aerospace systems, advanced methods for verification, validation, and certification of intelligent control systems will be discussed and illustrated via examples including UAV and NASA Intelligent Flight Control System project.  An overview of application of computational intelligence solutions in the chemical industry will then be presented, with emphasis on the key technical, organizational, and political issues to be resolved for successful application of computational intelligence in industry in general.  Presentations by researchers from both industrial and non-profit organizations will ensure effective sharing of knowledge and cross-disciplinary relevance.

RACT - Randomized Algorithms Control Toolbox: a Tutorial Introduction

Fabrizio Dabbene, IEIIT-CNR, Politecnico di Torino, Torino (Italy)

Constantino Lagoa, The Pennsylvania State University, PA (USA)

Andrey Tremba, Pavel Shcherbakov, RAS Institute for Control Sciences, Moscow (Russia)

Probabilistic and randomized techniques for analysis of uncertain systems and design for robustly performing control systems have attracted considerable interest in recent years, and a significant amount of theoretical and algorithmic results have appeared in the literature. The starting idea in the probabilistic approach to the analysis of uncertain systems is to characterize the uncertain parameters as random variables, and then to evaluate the system performance in terms of probabilities. In an analogous sense, probabilistic synthesis is aimed at determining the design parameters so that certain desired levels of performance are attained with high probability. This probabilistic approach is complementary to the mainstream methods in robust control, which seek worst-case performance guarantees and consider the uncertainties as deterministic unknown-but-bounded quantities.

Specific randomized algorithms (RA) have been developed for solving a large class of probabilistic analysis and synthesis problems arising in control. These algorithms may help in overcoming the conservatism and computational complexity limitations of worst-case methods, especially in real-world situations where a large number of uncertain parameters enter the system description in a possibly nonlinear way. The goal of this workshop is to introduce the recently released Matlab Randomized Algorithms Control Toolbox (RACT). This package offers a convenient way for defining various types of structured uncertainties as well as formulating and analyzing the ensuing robustness analysis tasks from a probabilistic point of view. It also provides a full-featured framework for LMI-formulated probabilistic synthesis problems, which includes sequential probabilistic methods as well as scenario methods for robust design. The package can be freely downloaded from http://ract.sourceforge.net.

Details for registration can be found at the Registration Page.

All the workshops will be held on September 2.

One free workshop registration is available to all graduate students.

Workshop Chair