SYSTeMS projects

Ongoing projects

  • Development and implementation of soft computing algorithms for classification, clustering and optimization of complex biological systems. (BOF project 011/051/04 - 4 years)
    Supervisors: Prof. L. Boullart, dr. M. Kuiper
    Contact info: Nicky Van Thuyne
    Start and end date: jan 2004 - dec 2007
    Short description: The aim of this research is to develop a new DSOM algorithm that will succeed in building a model for cell cycle regulation and prediction of hybrid vigour. Also, a new method based on genetic algorithms will be constructed capable of deterring the strength of each input parameter (both individual as in combination) as a function of its corresponding output.

  • Ordinal Classification with Support Vector Machines for automatic assessment of carpet wear. (IWT project - 4 years)
    Supervisors: Prof. B. De Baets, Prof. L. Boullart
    Contact info: Willem Waegeman.
    Start and end date: dec 2004 - nov 2008
    Short description: In many situations, humans are used to compare items or objects in order to choose the appropriate item for a specific goal. Think for example of buying clothes, listening to music, the dish one orders in a restaurant, etc.. Continually, we evaluate objects on criteria such as appropriateness, beauty, correctness, etc. In research areas like decision making, preference modeling, fuzzy modeling, statistics and machine learning, scientists have proposed various ways to characterize this human behavior with mathematical models. My research can be located on the edge between those domains with strong emphasis on machine learning.
    In the first part of my Ph.D the more specific machine learning setting of ordinal regression or ranking is considered, where categories typically correspond to quotations or linguistic terms --- varying from "very bad" to "brilliant" for example --- that express a difference in correctness, quality, beauty or any other characteristic of the analyzed objects. We proposed extensions of ROC analysis for this setting and we showed that such extensions can be useful for different applications.
    In the second part, the very related setting of preference modeling is studied. We show that under certain condition preference models coincide with ranking models and multi-class classification models. We also investigate the relationship between machine learning and decision making for preference modeling.

  • Algoritmic and computational models for studying EEG and local field potentials from deep brain core: correlation between signals from the cerebral cortex and the basal ganglia. (IWT project - 4 years)
    Supervisors(s): Prof. Dr. Ir. Luc Boullart en Prof. Dr. Patric Santens
    Contact info: Bruno Bauwens.
    Start and end date: jan 2006 - dec 2009
    Short description: The first part is a fundamental part. In the border between statistics and computability theory, tests for independence and directed influences between time series are studied by extending the theory of martin-lof tests, sumtests, martingales and time bound Kolmogorov complexity.
    The second part is devoted to the development of new independence and influence tests using general inference techniques and the theory above. These tests are evaluated by analysing interactions in dynamical systems with rapidly changing interaction-structures.
    The third part consists of applying the tests to detect interactions between the subthalamic nucleus and the motoric cortex measured during the treatment of Parkinson patient that are treated with deep brain stimulation. The absence of these interactions provides information about the development of the disease and possible future treatments.

  • Source localization of EEG signals for Brain Computer Interfaces. (BOF project - 4 years)
    Supervisors: Prof. Dr. P. Santens, Dr. B. Wyns, Prof. Dr. Ir. L. Boullart
    Contact info: Dieter Devlaminck.
    Start and end date: jan 2008 - dec 2011
    Short description: Brain Computer Interfaces can help people with severe motor disabilities to restore communication and control to a certain extent. Current systems have trouble with the high inter-subject variability. By using ICA, extended with spatial information of the electrodes, it is possible to implement a robust BCI. Aside of the technical realization we will look at the influence of BCI usage on changing brain patterns due to plasticity.

  • Characterization of dry and boundary-lubricated friction for stabilization and control purposes. (FWO project)
    Supervisors: Patrick De Baets, Mia Loccufier, Al-Bender Farid (PMA, KUleuven)
    Contact person: Patrick De Baets (IR04, 3304), Mia Loccufier
    Short description: In dry and marginally lubricated surfaces the variation of the coefficient of friction as a function of sliding velocity can lead to instabilities occurring as quasi-harmonic vibrations and stick-slip. These phenomena negatively influence the behaviour of mechanical devices: unacceptable vibrations or noise occur or the lifetime of the construction is drastically reduced due to fatigue or wear. In the present project the friction of marginally lubricated surfaces will be characterised in dynamical conditions taking into account the mechanical parameters such as inertia, stiffness and damping. A mathematical model for the friction as a function of instantaneous velocity and acceleration as well as sticking and/or sliding history will be postulated. The model will be used as a basis for stability analysis. A research position for the execution of this project is available.

  • Stimuli for the long term viability of the Center of Excellence in Industrial Automation at Coruniversitaria in Colombia. (VLIR-UOS project)
    Promotors: Joos Vandewalle, Department (ESAT, KUleuven), Mia Loccufier
    Contact: Joos Vandewalle, Department (ESAT, KUleuven), Mia Loccufier
    Short description: A research project (VLIR project AEIN 1999 PR221) was successfully carried out in 2000-2005. A follow-up activity is required to ensure the quality and continuity of research in the Center of Excellence built with this project. The project aims at providing stimuli for the long term viability of the Center of Excellence in Industrial Automation at Coruniversitaria in Colombia by appropriate coordination missions and exchanges.

  • Verkeersmanagement (B/00848/02 - fonds IV2), Sustainability effects of traffic management systems. (DWTC project - 01.12.2001 -> 31.05.2005)
    Sponsor: DWTC PODO CP50
    Promotors for UGent: Rene Boel
    Short description: The goal of this project was to illustrate the potential of feedback control in improving the performance of freeway and mixed freeway-urban traffic. The contribution of the SYSTeMS group in Ghent was mainly focused on the use of modularity in making the large scale problem computationally feasible, on stochastic modelling of freeway traffic, and on the state estimation problem. A novel random cell transmission model for freeway traffic was developped. This model enabled the development of an efficient particle filter implementation, estimating the state of freeway traffic using data from a limited number of magnetic loop sensors. Model predictive control algorithms for optimal on-ramp metering were implemented using these estimators.

  • Model based distributed fault detection. (BOF project - 01.10.2003 -> 23.09.2005)
    Sponsor: BOF
    Short description: Large discrete event plants are modelled via Petri net models and timed Petri net models for each module, and via interaction between modules via exchange of tokens. This class of models is used for developing distributed algorithms for efficient fault detection. The minimal requirements for message exchange between agents connected to different modules has been investigated. An algorithm has been developed for local fault detection, and for exchange of asynchronous exchange of information between agents, that guarantees that immediately after each exchange of information the distributed fault detection algorithm achieves the same quality of fault detection (detects the same faults) as a (computationally much more complicated) centralised fault detector would do.

  • Control training site-project (Marie Curie Fellowships-NCN (2001 -> 2007) B/05271/01 - funds IV2)
    Short description: Any topic in the general area of modern non-linear control design can be included in these collaborative research projects with visiting doctoral students. Visiting doctoral students worked on modelling, estimation, and control of road traffic.

  • Optimal supervisory control of hybrid systems. (BOF project: B/00123/01 (01.10.1999 -> 30.09.2003) - funds IV1; formerly G06050, project number 01106299)
    Sponsor: BOF
    Short description: Petri net models were used for modelling timed and untimed discrete event systems, as abstractions of hybrid system models. Minimally restrictive supervisory controllers guaranteeing avoidance of dangerous states were designed. Modular design methods allowed application of the controllers to large plants.

  • Decentralised control and reliability of large systems, with applications to electrical energy grids. (1/2006-> 12/2009)
    Promotors: Rene Boel and Lieven Vandevelde
    Short description: Extremely large networks (Internet, electrical power systems, water transport networks or road traffic networks) are vital resources in modern society. The push generated by ever more stringent economical and ecological constraints, and the push from the growing IT capabilities, have led to the operation of these resources at working points that are ever closer to the capacity limits. This leads inevitably to reliability problems. Small errors may lead to cascading failures whereby large parts of the network become unavailable. This is very damaging both for the human users, and for the economy. Recent examples are the major blackouts in the U.S. and Italy in 2003, where minor local faults, followed by erroneous state estimation and faults in the communication systems led to major blackouts for millions of users over periods of several hours. Similar cascading failures occur in the internet and in road traffic. As a result of these cascading failures the research dealing with the design and the control of large reliable networks is very topical. The very nature of the large networked systems, and the robust reliability specification, make it inevitable that decentralised design and implementation of the controllers is considered in this research. This research proposal will emphasize the safety and reliability of the operation of an electrical power grid. This is currently a very lively research topic. The deregulation of the energy market and the increasing cost of infrastructure require new management tools for network operators. Recent developments in the fields of power electronics and data communication allow the implementation of novel concepts for network control and operation. The current research proposal tries to demonstrate that a combination of novel decentralised control design tools combined with new capabilities (both hardware and software) of the control devices used for electrical power grids can reduce the risk of cascading failures.

  • Direct torque control for permanent magnet synchronous machines. (10/2006 -> 9/2008)
    Promotor: Jan Melkebeek, co-promotor: Rene Boel
    Short description: Application of direct torque control to permanent magnet synchronous machines, in order to achieve reliable speed control at various speeds. Accurate estimation of the state of the machine (initial position, magnetic fluxes) is necessary in order to achieve this goal.

  • Development and Clinical Validation of a MIMO model-based Predictive Closed-Loop Systems to Guide Patient-Individualized ICU Sedation. (IWT project - 2 years (17.09.2007 -> 17.09.2009))
    Supervisor: Robin de Keyser
    Contact: S. Syafiie
    Short description: Standard drug dosing during anesthesia causes inappropriate under -or over- sedation leading to increase morbidity and mortality due to huge inter-patient pharmacological variability. Manually patient-individualization of dosing schemes are actually not accurate applied due to lack of care resources. Lighter dosage may make a patient more prone to wake up during surgery, contrarily, higher dosage puts patient in deep hypnosis. Therefore, an optimal and automated dosing is required for general anesthesia. The IWT-TBM anesthesia project is to optimize ICU sedation by applying automation technology in drug administration. At the patient lever, patient individualized and automated sedative drug administration will optimize ICU sedation by minimizing the accumulation of sedatives in critically ill patients with altered organ functions. When optimizing ICU sedation at a patient-individual level, pharmacodynamical and physiological stability and influence outcome are improved. At the ICU level, optimizing patients' sedation will result in increased turn-over, decrease long-term sedation complications and reduce cost. At the end, the aim is to improve the overall ICU outcome.

  • Learning Control for Production Machines (sept 2009 -> aug 2013)
    Supervisors: Prof. L. Boullart, prof. R. De Keyser
    Contact info: prof. L. Boullart
    Short description: Up till now, only a few implementations of learning controllers have been realized for mechatronic systems. Moreover, only model-based techniques were adapted and only implementations were done on small, relatively simple mechatronic subsystems. In a first track of the LeCoPro-project, the potential of different learning control algorithms for more complex mechatronic (sub)systems will be investigated. The advantages and drawbacks of model-based learning control methods (Iterative Learning Control ILC; Model Predictive Control - MPC), suitable for the control of systems of limited complexity, and non-model-based methods (Machine Learning), suitable for the control of more complex systems, will be compared. State-of-the-art techniques in these research domains will be extended and adapted to address the specific features of mechatronic systems (fast non-linear dynamics, uncertainty of system parameters, stringent stability and robustness requirements, etc.).

top

Finished projects

top
Please send comments to: webmaster
SYSTeMSSYSTeMS
News