Division of. Control. and. Instrumentation

Transcription

1 Division of Control and Instrumentation

2 INTRODUCTION T he Division of Control and Instrumentation covers the broad areas of control, automation, robotics, computer vision and biomedical engineering in its teaching and research activities. It actively promotes the development and integration of these areas so as to bring about innovations and new applications. The Division has 32 full-time academic staff members and 11 laboratories to support its teaching and research activities. Website: The aims of the Division are to produce high quality graduates with specialized skills through good teaching and relevant programs, and to spearhead vibrant research activities that generate international recognition. The research activities of the Division are housed within two research centres: the Centre for Intelligent Machines Website: and the Centre for Modeling and Control of Complex Systems. Website: There are a total of 10 research fellows, 5 research associates, 7 project officers and 68 MEng/PhD students to support the research activities. The Division has 7 research areas: control of networks, control system technologies, robotics research, surveillance technology, machine learning, modelling and control of biological systems, modelling, optimization and control techniques. The major research achievements of the Division include running 35 research projects on research funding amounting to S$9.4 million from various sources, 18 associate editorships in international journals, the Best Paper award at the 6th International Conference on Intelligent Control and Automation (China, June, 2006), first prize in the Most Popular Research Project category of the College of Engineering Technology Competition at NTU, as well as producing 1 MEng and 18 PhD graduates in The Division continues to collaborate with overseas universities such as the University of Cambridge, the University of Newcastle, Tel-Aviv University, Harbin Institute of Technology, as well as with local research institutes and industry. Besides offering MEng and PhD programs, the Division further caters to the needs of working engineers in the industry by conducting a very successful MSc program in Computer Control and Automation. The current enrolment stands at 129, and 82 MSc students have graduated in The Division also organizes a biennial International Conference on Control, Automation, Robotics and Vision (ICARCV) that is well regarded by researchers in the relevant fields. The latest ICARCV conference (the 9th) was held on 5-8 December 2006 at the Grand Hyatt Singapore with more than 380 international participants. The research programs of the Division are described briefly in the following. Control of Networks Objectives In order to better understand the complexities of networked systems, and to develop techniques for improving the performance of these systems, there is a need to have good dynamic models of the networks to allow for real time and optimal control of these networks. One of the main objectives of this research programme is to develop practical dynamic models for networks that are suitable for real-time applications. The other objective is to develop and apply new system theories that will enhance the performance of networks. The areas of research involve dynamic resource allocation of internet traffic, modelling and control of optical communication systems and networked control systems. Highlights of Research Activities One of the key activities in this program has been the design and realization of a multi-port all-fibre optical interleaver. An interleaver/de-interleaver works like an optical router that allows existing DWDM filters designed for operations with wide channel spacing to be extended to system designs with narrow channel spacing in the range of 50GHz or even less. Basically, interleavers/de-interleavers can be scaled to match channel combinations or routings that might be needed. School of Electrical & Electronic Engineering 053

3 The multi-port interleaver developed in this programme can efficiently increase channel counts with fewer stages of interleavers and provide more flexibility for optical signal multiplexing. Its all-fibre structure has an extremely low insertion loss. Through a novel system theoretic approach realized by cascading multi-port directional fibre couplers, this multi-port all-fibre interleaver is designed to provide guaranteed gain and phase characteristics. Whereas past efforts were on the spectral gain characteristics, this design guarantees a linear phase response through its novel structural design considerations. In addition to the novel structure, a systematic quantitative filter design approach was developed and used to enable the interleaver to have desirable specifications such as linear phase, flattop passband/stopband, passband isolation, and minimal chromatic dispersion. In other optical instrumentation and communication system research activities, the main focus has been on exploring the use of optical technologies for all-optical communication networks and precision measurement. In terms of network hierarchy, optical CDMA technology is investigated at both the system and device levels. This includes the design of tuneable filters, optical interleavers, EDFA gain equalizers, dispersion compensators, optical switches, and optical add/drop multiplexers for optimal network performance. In addition to the novel interleaver design techniques, the other achievements of work in this area are the realization of multi-wavelength tunable lasers, and an all-fiber optical spectrum equalizer. Furthermore, by making use of optical feedback, the research team is hopeful of achieving dynamic all-optical EDFA gain clamping of a chain of EDFAs. Through proper controller design, the power transient and oscillations of the signal channels when subjected to add/drop operation can be greatly minimized, thus improving the performance of the EDFA chain. For precision measurement, a recent project has been on optical system design. The research group concerned worked closely with SIMTech on the development of enabling technologies for nanoscale measurements by using near field fiber. Another interesting project on networked control is the investigation of quantized feedback control of discretetime linear systems using a finite-level quantizer. The work is motivated by the fact that most feedback communication channels allow a moderate bit rate, and are not particularly concerned with the problem of finding the minimum bit rate of feedback for a given control objective. Instead, it is assumed that a moderate bit rate is available and the task is to find a quantization strategy which is easily implementable and can achieve a given control objective. The research group involved has developed a practically applicable dynamic scaling scheme for quantized feedback with finite-level quantizer and shown that the scheme can achieve asymptotic stability and certain robustness. Multiplexer EDFA Demultiplexer OADM Coupler Coupler 5% tap Optical tunable bandpass filter 5% tap Schematic diagram of dynamic all-optical EDFA gain clamping Photodetector computer Photodetector 054 Nanyang Technological University

4 Control Systems Technologies Objective The objective of this research programme is twofold: (1) fundamental research in system, control and signal processing; and (2) industrial applications in related areas. Highlights of Research Activities The activities of the research programme mainly focus on adaptive and fuzzy-neuro control, iterative learning control, robust control, estimation theory, networked control systems, time-delay systems, and process control and optimization. In addition to theoretical development in the above areas, the research programme has also intensive activities in industrial control applications in the areas of flight control, control and optimisation of HVAC systems, sensor networks, intelligent sensing for chemical waste treatment, high precision motion control, and control of hard disk drives. The programme has generated new results of a theoretical nature during the year in areas such as quantized feedback, optimal and H-infinity control of systems with multiple input delays, adaptive control of systems with hysteresis, process control, and reliable control. For example, a duality between the linear quadratic regulation of systems with multiple input delays and smoothing estimation, and obtained simple explicit control laws for the complicated optimal control and H-infinity control problems for systems with multiple input delays via standard projection has been established. The results have been published or accepted for publication in Automatica and IEEE Transactions as regular papers. The programme s research in adaptive control has also led to some solutions to adaptive control problems for complex multi-input multi-output (MIMO) systems with non-smooth nonlinearities which are still largely open. One paper was accepted as a regular paper in Automatica. Further, the programme has also developed a new adaptive intelligent fault-tolerant controller which can handle unanticipated failures and uses an on-line learning neural scheme. In terms of industrial applications, a research team under the programme, in collaboration with Purechem Onyx Pte Ltd, has recently developed a portable Intelligent and Multi- Functional Machine for chemical waste treatment. The machine is capable of on-line monitoring of ph value, ion concentration, conductivity, dissolved oxygen, heavy metal concentration, density, temperature and heating value. The system can also be easily adapted to other applications including waste water treatment. The team also developed a total waste management system with functions such as sensor management; waste profile building; waste status tracking; waste database management; knowledge-based inference; guideline for treatment and disposal methods; guideline for safety measures and environmental protection; information enquiry for further testing; and on-line data sharing. The system is currently under testing in Purechem Onyx. The programme has had a very successful year in attracting funding from national funding agencies and industry. It has received a number of major grants from A*STAR, including $0.52m for a project on collaborative information processing in sensor networks. In collaboration with SIMTech, the programme secured a $1m grant for integrative manufacturing and service systems, and with DSI and the Centre for Advanced Power Electronics, it obtained $0.52m for a project for developing new servo control techniques for hard disk drives. Robotics Objective The objective of this research programme is to investigate, design and develop the mathematical, science and engineering foundations and enabling technologies that will pave the way for development and deployment of taskoriented intelligent autonomous systems. These systems include indoor, outdoor, land and underwater smart mobile robots and vehicles. Highlights of Research Activities This year, two new projects are underway, namely, one on the integration and demonstration of simultaneous localization and mapping (SLAM) for underwater vehicles, and another on the development of an integrated navigation system for ground vehicles. Both projects are externally funded. The first project aims to synthesise and implement navigation, localization and mapping algorithms exploiting naturally School of Electrical & Electronic Engineering 055

5 occurring features. This is for use in Smart Autonomous Underwater Vehicles (AUV). AUVs provide a new and synergistic means to observe the maritime environment and execute varied maritime missions which are difficult to achieve at present. These smart vehicles would be particularly well suited for mapping, inspection, monitoring, and exploration, in previously unmapped, unchartered, uncertain, dynamic and hostile underwater environments because of their ability to swim freely and stably without physical attachment to a surface ship or human intervention. A major obstacle in the realization of a new generation of AUVs is the availability of reliable and accurate navigation and mapping algorithms sustainable over long hours, and in large spaces without reliance on external artificial localization beacons or human intervention. This is because the current technology of navigation and mapping which is based on inertial sensing supplemented by frequent updates using artificial beacons, is unsatisfactory due to significant infrastructure costs and setup times. The project aims to overcome this obstacle. In addition to the two new projects, an IDeAS pilot project grant was also received for Autonomous Sensing Capabilities for Security Research. This pilot project is aimed at producing a proposal suitable of attracting funding from the European Union s (EU) Seventh Framework on Security Research. The A*STAR SERC funded project on Collaborative Autonomous Systems for Built Environments (CARSys) is continuing into its third year. An important outcome of this research effort is the development of mapping algorithms incorporating signal detection probability of the range sensors. Signal detection is rarely considered in mapping algorithms, as it is usually an internal function of the sensor and inaccessible to the user. (Above) Map constructed from 3D Laser data. (Below) Occupancy map estimated from RADAR rangeintensity data using signal detection probabilities At a more fundamental level several important results have been established in the popular localization framework known as Simultaneous Localization and Mapping (SLAM). The programme s research has shown that the typical approach to world centric robot localization and mapping is unobservable. Various practical alternative formulations to the robotic localization and mapping problem have hence been proposed. These results have prompted the SLAM research community to view and approach localization and mapping using robots differently. Further, closed form solutions to estimation theoretic exteroceptive and proprioceptive sensor bias correction problems in robot localization and mapping were established. These results provide greater insights into the selection and use of different types of sensors in robot localization and mapping. Surveillance Technology Objective This research programme studies problems in surveillance technology and aims to develop applications for surveillance related activities, such as those related to homeland security. In particular, the programme focuses on improving face and gesture recognition techniques, video surveillance capabilities, and visual information retrieval and analysis. Highlights of Research Activities The current specific focus of the programme is a project on multi-agent activity recognition. 056 Nanyang Technological University

6 In the analysis of video content, one major problem in the classification of human activities is the shortage of modelling methods for multi-agent interactions. Work on human activity modelling, has focused mainly on individual activity recognition. Recently, however, interactions between multiple persons have been recognized as a significant source of information. The challenge in modelling multi-agent activities is that when the number of persons involved increases, the interactions between agents are much more complicated to model. Hence, there is a need for new modelling methods that can give a concise and extendable representation of multi-agent human activities. This project addresses this need. The project proposes an observation decomposed hidden Markov model (ODHMM) for the modelling of multi-agent activities. Compared with the traditional topology of HMMs, ODHMMs have multiple layers of observations, where each agent has a particular observation layer. A new parameter is added to the model which represents the alignment between an agent in the data set and the activity model. The proposed ODHMM has three advantages over existing models. Firstly, the dimension of the feature space can be significantly reduced so that the problem of dimensionality can be avoided even where there is a large number of agents. Secondly, the ODHMM is robust against changes in the number of the agents during the interactions. Thirdly, the role of each agent in the activity can be automatically estimated rather than manually assigned. Machine Learning Objective This research programme has two major aims: (1) fundamental research in machine learning algorithms; and (2) their applications in real world problems. Highlights of Research Activities The research activities of the programme comprise a number of in-house projects and one A*STAR funded project with a funding of around $707,000 and have generated a number of publications in refereed international journals and conferences in The activities focus mainly on machine learning areas such as neural networks, support vector machines, genetic algorithms, swarm intelligence, and fuzzy systems. In addition to theoretical development in the above areas, the research group has also put a significant effort into application oriented projects. One of the programmes most significant achievements is the recent development of a new efficient, fully automatic selfcontained machine learning technology called the Extreme Learning Machine (ELM). The ELM is extremely fast and totally free from control-parameters. This year, several variants of the ELM such as Online Extreme Learning Machine, Fuzzy Extreme Learning Machine, Fully Complex Extreme Learning Machine, and Evolutionary Extreme Learning Machine have been developed. These variants regularly outperform other methods in different applications. For example, Fuzzy ELMs have been used in the fault tolerant control of aircraft autolanding with actuator failures and have been shown to provide superior performance. Complex ELMs have also been shown to provide higher performance in QAM equalization problems in the area of communications. In addition, the ELM has been shown to produce accurate and fast performance in bioinformatics classification problems. In the area of micro array gene expression based cancer diagnosis problems, the ELM has produced excellent results. A sequential version of ELM called OS-ELM has been developed in the group and its superior performance based on the number of benchmark problem has been demonstrated. This algorithm can handle both additive and RBF neurons together and also can handle data coming one by one or chunk by chunk. Further applications of the ELM and its variants in different areas are currently being pursued in the group s research work. Another area in the machine learning research programme is the investigation of ways of sequentially accelerating the learning of linear classifiers such as SVM and LDA on very large data bases. Applications of 3D vision technology for face and expression recognition are also being actively explored in the group. School of Electrical & Electronic Engineering 057

7 One of the primary research areas of the programme has been the field of evolutionary computation. In this context, various evolutionary algorithms have been developed and applied to solve real world problems. In particular, two improved versions of the particle swarm optimization algorithm, namely the comprehensive learning particle swarm optimizer and the dynamic multi-swarm particle swarm optimizer, have been developed for solving single objective optimization problems and have been shown to perform significantly better than several other recent variants of the particle swarm optimizer. A novel adaptive version of the differential evolution algorithm has also been developed. This algorithm is capable of identifying the best strategy to solve a particular problem and of learning appropriate numerical values for various strategy parameters. These single objective algorithms have also been extended to solve multi objective optimization problems. The resultant novel algorithms have been applied to solve various optimization problems in the fields of photonics and microwave antenna design and bioinformatics. The research group also organized a special session on constrained realparameter optimization at CEC (Congress on Evolutionary Computation)-06 in Vancouver, Canada, and presented a tutorial at SEAL [Simulated Evolution and Learning]-06 on real coded evolutionary algorithms. In addition to these efforts in the field of evolutionary computation, the programme has also investigated the role of diversity measures for constructing classifier ensembles and developed an exact expression for the leave-one-out error for the support vector machines. The most attractive feature of the exact expression for the leave-one-out error is that the support vector machine has to be trained only once on the whole dataset instead of n times where n is the number of samples in the dataset. The research group has also developed techniques to generate pseudo support vectors different from the given training samples. This approach can increase the training / classification speed and potentially reduce the number of support vectors. Modelling and control of biological systems Objective This research programme focuses on developing technological capabilities for the modelling and control of biological systems. Its work has significance in biomedical applications. Highlights of Research Activities The research programme has made an important contribution to the assessment of anti-cancer drugs by medical imaging. It has, together with the National Cancer Center and the National Neuroscience Institute, jointly developed capabilities for perfusion imaging using dynamic contrast-enhanced computed tomography (CT) and magnetic resonance imaging (MRI). The developed algorithms allow for the assessment of blood flow parameters within tumours, which can be used as biomarkers for assessing the effectiveness of anti-cancer drugs. The research group was recently awarded a research grant from the Singapore Cancer Syndicate (SCS-CS0072) to employ these developed algorithms in the processing of clinical trial results using cancer drugs provided by Abbott. Another two clinical trials with other cancer therapy approaches are being planned. Modelling, Optimization and Control Techniques Objectives This research programme aims to develop general methods for modelling and controlling complex systems, as well as for optimizing the operation of such systems. It focuses on exploring both the fundamental and applied aspects of such methods. Highlights of Research Activities The current main activities under this program are modelling and design of steam ejector based air-conditioning systems; modelling, monitoring, control and optimization of water distribution networks; synchronization of chaotic systems with applications secure communication and multidimensional systems; and practical applications of fundamental theory developed. 058 Nanyang Technological University

8 One major project of the programme is the development of steam ejector heat pump systems for automobile industry. The project involves the following: Conducting a fundamental study on the working principle of steam ejector heat pumps, providing a design guideline for equipment sizing, and predicting performance for available heat sources; Establishing both static and dynamic mathematical models of all elements in the combined ejector-vapour compression cycle (e.g. ejector, compressor, intercooler, condenser, evaporator, fans, coils and pumps) using the method of material and energy balances; Developing and integrating real-time optimization technologies to ensure that each steam ejector heat pump will always be running at the most efficient operating point, regardless of the change of available heat sources and environment; Developing and integrating dynamic control technologies for every set-point change determined by the optimization technique, so that each device can reach its new set-point smoothly and promptly; Constructing and testing the pilot steam ejector heat pump system to gain experiences on modifying/improving the system and to make the system industry-applicable and reliable. Another project concerns water distribution networks. The project proposes to solve the problems of reducing water leakage and improving efficiency through the following approach: modelling of the overall network and substations/building blocks; design of demand predictors; optimization of water distribution network operation based on established models; design of loop control modules; development of monitoring and management modules; and system integration. Work has also been carried out on synchronization in chaotic systems. Chaotic signals have been applied to secure communication systems, including chaotic encryption and chaotic spreading codes for multi-user access in spread spectrum systems. In a typical chaotic secure communication system a chaotic cryptosystem is often used. Each cryptosystem is composed of two main parts, the encrypter and the decrypter and each part includes a chaotic system. Synchronizing the two chaotic systems is an essential problem to be addressed in order to facilitate message recovery from the receiver. Current schemes to achieve synchronization have been obtained under the strong assumption that the parameters of chaotic systems are time invariant and that these parameters are identical for both chaotic systems. These assumptions, however, are impractical. Thus robust schemes against parametric uncertainties and mismatches are needed, and the work in this research programme aims to address this need. The research team involved has developed a practical impulsive synchronization scheme to address the existing problems. This was done by quantitatively establishing a relationship between a pre-specified bound of the synchronization error and the length of impulsive intervals in the presence of both parametric uncertainties and mismatch between the two systems, the first such finding in the area, and using the relationship as a guideline to choose impulsive intervals. The proposed scheme theoretically ensures that the magnitude of the synchronization error approaches and stays within the pre-specified bound, which can be arbitrarily small. One final area of research in the programme is in multidimensional systems whose input and output signals are functions of more than one variable. Multidimensional and 2-D systems exist in many practical applications, e.g. sound, seismic, underwater signal propagation, image data processing and transmission, visual recognition of School of Electrical & Electronic Engineering 059

9 robotic systems, and medical and biological information and image processing. In this area, the research group investigates methods and algorithms for analysis of fundamental properties and for processing, filtering, optimization and control of multidimensional signals, images and systems. Applications of its methods and algorithms include medical image processing, active noise cancellation and multirate filter banks. An Ejector Based Air-Conditioning System CENTRE FOR INTELLIGENT MACHINES T he Centre for Intelligent Machines (CIM) was setup in 2003 in the School of Electrical and Electronic Engineering, NTU. The director of the Center is Prof. Wang DW. The research activities of the Centre focus on areas of control, automation, robotics and computer vision. The common emphasis of these research activities is to develop practical and intelligent algorithms and techniques for industrial automation and manufacturing applications, as well as to integrate the different areas to lead to innovations and new applications. These activities are driven by Singapore s need to develop more efficient and competitive expertise and technologies relevant to the automation and control industries. The Centre s members include more than 25 academic staff, 2 adjunct staff, 12 research fellows and research associates, and 54 MEng and PhD students. Autonomous mobile robot technologies 060 Nanyang Technological University

10 Centre for Intelligent Machines R&D Activites The R&D activities of the Centre are organized into five main research programs: Mobile Robotics Control Systems Technologies Surveillance Technology Machine Learning Real-time & Embedded Information Systems RESEARCH GROUPS RESEARCH GROUPS RESEARCH GROUPS RESEARCH GROUPS RESEARCH GROUPS Simultaneous localization & map building Robot control Underwater robotics Autonomous vehicles, evolutionary & bio robotics Robust control Adaptive and learning systems Intelligent systems & control Process control & automation Face & gesture Video surveillance Visual information retrieval Stochastic learning theory & applications Computational intelligence for machine learning applications Feature analysis & classification Model predictive control on a chip Framework for real-time monitoring & control Hardware architecture & software tools for mobile robots Service robots & e-toys Flight control systems Autonomous underwater robotics Optimization technology for building HVAC systems On-line laboratory F16 in a Herbst-like Maneuver School of Electrical & Electronic Engineering 061

11 Wireless Sensor Networks Research and Development Wireless Sensor Networks (WSNs) consists of inexpensive and low-power communication devices with inbuilt sensors called sensor nodes which are spatially distributed to cooperatively monitor physical and environmental conditions. In the Infocomm Technology Roadmap (ITR- 5, Mar 2005) of the Infocomm Development Authority of Singapore (IDA), WSN has been identified as a key technological area and new frontier which holds the potential to dramatically change the way we live. A prototype of target tracking scenario Simulated target tracking scenario A research team in the Centre for Intelligent Machines led by Prof Xie Lihua is exploring enabling technologies for WSNs. One aspect of its work involves fundamental theoretical studies on analysis and design to address the challenges of limited sensing, computing and networking resources. The main research issues that the team explores include uncertainty modelling in unreliable networked environments, energy-efficient quantization and compression for bandwidth assignment, distributed estimation with limited communications, QoS provisioning via cross-layer integration, target detection, localization, tracking and inception, and cooperation of multiple sensing agents. Another aspect of its work is prototype development. In this area, the team constantly seeks solutions for real world WSN applications by developing wireless sensor nodes and test beds for multitarget tracking and inception with limited energy, sensing, computing and wireless networking constraints. A sensor node Object detected by node Data Path Multi-target tracking of mobile robots Object detected by node Data Path 062 Nanyang Technological University

12 NTU-Purechem Onyx Intelligent Monitoring Systems for Chemical Waste Treatment In today s industrialized world, proper treatment of wastes, especially chemical wastes and waste water from industries such as semiconductor manufacturing and pharmaceutical industries, is of immense importance. The industrialization of countries such as China and India has thus created tremendous opportunities for the chemical waste and waste water treatment industry. A research team in the Centre for Intelligent Machines led by Prof Cai Wenjian and Prof Xie Lihua has been collaborating with Purechem Onyx Pte Ltd since The research consists of two phases, and is now in the second phase. In the first phase, an intelligent monitoring and control system for hazardous waste chemical treatment processing was developed and put into operation in The system consists of a smart laboratory with a multi-stream sampling apparatus, a micro-processor based integrated transducer system, and an intelligent control system. Its use has led to a sharp reduction of detection time and laboratory work and a significant increase in treatment capacity. novel density and heating value measurement devices, a flexible interface with PCs or laptops, and a powerful software system with automatic sensor calibrations, automatic ignition, novel signal processing and soft-sensing technologies, data recording and display, and data analysis. The parameters that are monitored include ph, ion concentration, conductivity, dissolved oxygen, heavy metal concentration, density, temperature and heating value. The modular design concept allows flexible addition of sensing subsystems for other environmental parameters. The WMS provides the following functions: sensor management; waste profile building; waste status tracking; waste database management; knowledgebased inference; treatment and disposal guide; safety measures and environmental protection guide; information enquiry for further testing; and on-line data sharing. The WMS and monitoring systems can be easily adapted or commercialized for other applications including waste water treatments. The system has been granted two patents by the World Intellectual Property Organization (WIPO). In addition, the Smart Laboratory component has also won two prestigious Occupational Health Best Practices Awards, namely the Innovation Award and the Excellence Award, from the Ministry of Manpower, Singapore. Veolia Environment, the world s largest waste treatment and management company has described the system as a unique fully automated Intelligent Monitoring and Control System (IMCS). PTK GUI interface of heavy metal measurement The second phase of the research focuses on the development of a Portable Test Machine (PTK) and a total Waste Management System (WMS). The PTK consists of a micro-processor controlled multi-probe testing chamber and School of Electrical & Electronic Engineering 063

13 C CENTRE FOR MODELING AND CONTROL OF COMPLEX SYSTEMS omplex systems is a field of science that cuts across all traditional disciplines of science, as well as engineering, management, and medicine. The complex systems concept provides a new unity of approaches to many different problems and applications. The study of complex systems may focus on certain questions about parts, wholes and relationships of these systems. The field of complex systems has provided a number of sophisticated tools, some of which are concepts that enable us to think about these systems, some of which are analytical for the purpose of understanding these systems in greater depth, and some of which are computer based for describing, modeling or simulating these systems. Under the directorship of Prof. Soh YC, the Center for Modeling and Control of Complex Systems coordinates research efforts in the areas of biological systems modeling, computational nanoelectronics, computational electromagnetics, modeling and control of networks, and modeling, optimization and control techniques. Modeling of Biological Systems This programme focuses on both fundamental and applied research into biological systems, with the view of extending research and development to practical implementations in health care, such as clinical diagnostics and treatment. The research projects initiated are multidisciplinary in nature, and currently include haemodynamic and vascular modeling of blood-tissue exchange for the study of cancer and stroke; development of drug delivery control systems; bio-electric signal analysis and modeling for the investigation of neuromuscular conditions; development of recognition and classification methods in cell informatics; and computational intelligent diagnostic systems for MR brain image scanning in mental health and illness. Diagnosis of MR brain image Functional imaging - modeling and imaging of cerebral haemodynamics 064 Nanyang Technological University

14 Computational Nano-Electronics The mission of this research programme is to establish a coordinated research effort in computational nanoelectronics among the local universities and research institutes and centers. The main objectives are to develop and deploy infrastructure in highperformance computational nanoelectronics for local and global researchers and industries, and to provide education and manpower training in the emerging field of nanoscale device modeling. Some recent projects include high current carbon nanotube based cathodes for miniature high power terahertz source applications; development and fabrication of a 1.55μm photodetector based on Ga(In)NAs/AlGaAs quantum well intersubband transition; compact modeling and characterization of sub-45nm active and passive CMOS devices; and quantum confined Zinc Oxide thin film for UV lasers applications. These efforts in nano-computation and atomic modeling will drive new concepts that help turn the promise of nanoscience into new nanotechnologies. (Left) Simulation of high current field emission of nanotubes FE current 3 nanotubes (Center) Nonbonding lone electron pairs zigzagged M-O-M monatomic chain derived from scanning tunneling microscopy/spectroscopy (STM/S) and low-energy electron diffraction (LEED) from the fcc(110) surface of Cu, Ni, and Ag (Right) Intense blue light (2.65 ev) emission of a synthetic metal oxide (bright spot) based on the model prediction Computational Electromagnetics The main objectives of this research programme are to develop high performance tools for simulation of the physical interactions and phenomena between electromagnetic waves and material structures, and to provide educational and professional services to local industries for the optimal design and innovative analysis of microwave, optoelectronic, and microelectronic devices. Major research carried out by the group includes the joint Programme for Electromagnetics (PEM) of the Institute of High Performance Computing (IHPC) and the University (NTU), and the CEM/Antenna Design project. Realistic simulation of electromagnetic response from a missile corvette. (Top: near field; Bottom from left: antenna impedance, geometrical model and current distribution, antenna radiation pattern) School of Electrical & Electronic Engineering 065

15 Computational Electromagnetics (continue) (Left) Electromagnetic field distribution inside a coaxial to microstrip line to coplanar waveguide transition (Right) A finite element mesh of photonic crystal fibre Modeling and Control of Networks The main focus of this research programme is to develop system theoretic approaches for networked control problems. This is to provide a systematic and efficient procedure for the design of systems and devices, and to allow for real time reconfiguration of network parameters in response to changing traffic loads and conditions. The areas of research involve dynamic resource allocation for multimedia applications, congestion control, modeling and control of optical communication systems and networked control systems. Lattice structure for implementing linear-phase two-port interleaves Multi-wavelength fiber ring laser Modeling, Optimization and Control Techniques Under this programme, the focus is on the areas of hybrid and switching systems, multidimensional systems, and modeling, simulation and optimization. In hybrid and switching systems, the main interest is in the modeling, analysis and control of inherently combined logical and continuous processes. The theory and applications involve several different fields in science and engineering, such as control and computer science. In multidimensional systems and modeling, problems involving signals and systems that depend on more than one variable are addressed. Theories related to modeling, simulation and optimization of general problems will also be developed. An adaptive Auto-Focus System 066 Nanyang Technological University

16 DIVISION OF CONTROL & INSTRUMENTATION Staff Members 1st Row (From left to right) 2nd Row (From left to right) 3rd Row (From left to right) Head of Division s s Wang Dan Wei, Professor Martin Adams Cai Wenjian Chan Chok You, John Cheah Chien Chern Chin Teck Chai Chu Yun Chung Chua Chin Seng Er Meng Joo Koh Tong San Lee Peng Hin Deputy Division Head Ling Keck Voon, Associate Professor Professors Soh Yeng Chai Narasimhan Sundararajan Xie Lihua School of Electrical & Electronic Engineering 067

17 DIVISION OF CONTROL & INSTRUMENTATION Staff Members 1st Row (From left to right) 2nd Row (From left to right) 3rd Row (From left to right) s s s Low Kay Soon Mao Kezhi Saratchandran P Sim Siong Leng Soh Cheong Boon Song Qing P N Suganthan Sung, Eric Teoh Eam Khwang Wang Han Wang Jianliang Wen Changyun Wijerupage Sardha Wijesoma Wong Jia Yiing, Patricia Zhang Cishen Zhu Kuanyi Assistant Professor Yau Wei Yun 068 Nanyang Technological University

18 Research Interests 1 Cai Wenjian Building automation, HVAC system control and optimization, Environmental process control. 2 Chan Chok You, John Power converter modelling and control, Nonlinear control. 3 Cheah Chien Chern Adaptive control, visual servoing, robotics. 4 Chin Teck Chai Time series, stochastic optimal control and statistical forecasting. 5 Chu Yun Chung Control, Artificial neural networks. 6 Chua Chin Seng 3D computer vision, Recognition and registration motion estimation and 3D deptu recovery. 7 Er Meng Joo Control Theory, Computational Intelligence, Robotics, Automation, Sensor Networks, Cognitive Systems. 8 Koh Tong San Medical functional imaging, Physiological modeling, Properties of semiconductor quantum structures Lee Peng Hin Ling Keck Voon Low Kay Soon H_infinity control, robust and reliable control. Model Predictive Control, Receding Horizon Estimation, Embedded System, GPS Receiver. Precision linear drive, UWB, Intelligent control system, wireless sensor network. 12 Mao Kezhi Predictive Modelling, Pattern Classification, Data Mining, Neural Networks. 13 Martin David Adams Land Based Vehicle, Sensing, Mobile Robot Navigation, Autonomous Underwater Vehicles (AUVs). 14 Narasimhan Sundararajan Professor Neural Networks, Fuzzy control, Evolutionary Algorithms, Aerospace Control, Bio-Medical/Informatics. 15 Paramasivan Saratchandran Artifical neural networks, Adaptive control, Computational Intellegence. 16 Ponnuthurai N Suganthan Evolutionary computation, Particle swarm, Pattern recognition, Bioinformatics, Computer games, Neural-nets. 17 Sim Siong Leng Array signal processing, Parallel processing Soh Cheong Boon Soh Yeng Chai Professor UWB Imaging, Distributed Diagnosis, Home Healthcare, E-medicine, Bioassay, Fluorescence Imaging Robust control, filtering and estimation; Information theory; Optical signal processing Song Qing Sung Eric Teoh Eam Kwang Wang Dan Wei Wang Han Professor Robotics, Control. Machine Learning, Computer vision, stereovision, structure-from-motion, face and expression recognition. Computer Vision and Pattern Recognition, Face Recognition and Classification, Biomedical Image Processing. Robotics, Control theory and applications. Robotics, Virtual reality, Multimedia, Machine vision, Computer vision. School of Electrical & Electronic Engineering 069

19 Research Interests 25 Wang Jianliang Reliable control, Fault detection and Identification, Flight control, and control of graphics rendering. 26 Wen Changyun Adaptive Control, Control of Nano-Systems, Control of Biomedical Systems. 27 Wijerupage Sardha Wijesoma Pattern recognition, Visualisation and image processing, Robotics and intelligent control systems, Control engineering. 28 Wong Jia Yiing, Patricia Differential, difference and integral equations. Dynamic equations on time scales. 29 Xie Lihua Professor Robust control, Networked control, Estimation theory and Applications. 30 Yau Wei Yun Assistant Professor Biometrics, Intelligent system, Biomedical engineering, Computer vision. 31 Zhang Cishen Control Signal Processing Medical imaging and instrumentation. 32 Zhu Kuanyi Process control, Predictive control, Biomedical engineering. PhD and MEng Degrees Awarded in 2006 PhD - Control & Instrumentation S/No. Project Title Student Supervisor/Co-supervisor 1 2 Two-point right focal boundary value problems on time scales Decentralized Multi-model Control for MIMO Processes with Applications in HVAC Systems Boey Kok Leong He Ming Wong Jia Yiing, Patricia Cai Wenjian 3 Vision-based target tracking for autonomous vehicles Jia Zhen Wang Dan Wei Arjuna Prabhath Balasuriya 4 Feature extraction and dimensionality reduction in patter recognition with applications in speech recognition Jiang Hai Er Meng Joo 5 Kernel machines and classifier ensemble learning for biomedical application Li Peng Krishnan, Shankar M Chan Kap Luk 6 Adaptive noise cancellation using soft computing approach Li Zhengrong Er Meng Joo 7 Adaptive control of robot manipulators with uncertain kinematics and actuator dynamics Liu Chao Cheah Chien Chern 8 Human activity recognition based on hidden markov models Liu Xiaohui Chua Chin Seng 9 10 Optical Filter Banks: Structure, Design and Implementation Robust coordinated multiple-axis motion control systems and applications Wang Qijie Xiao Yong Soh Yeng Chai Zhu Kuanyi 11 Face recognition using soft computing methods Chen Weilong Er Meng Joo 12 Vehicle following: Control Design, Simulations & Experiments Pham Minh Tuan Wang Dan Wei 070 Nanyang Technological University

20 PhD and MEng Degrees Awarded in 2006 PhD - Control & Instrumentation S/No. Project Title Student Supervisor/Co-supervisor 13 Automated 3-D optic dusk reconstruction and feature extraction for glaucoma analysis Xu Juan Sung, Eric 14 Simultaneous Localization and Map Building for Outdoor Mobile Zhang Sen Xie Lihua Robot Navigation Lesion detection and classification in color fundus images Novel 3D statistical shape models for segmentation of medical images Adaptive backstepping control of practical uncertain systems Optical detection of optical code division multiple access Zhang Xiaohui Zhao Zheen Zhou Jing Kok Shaw Wei Song Qing Teoh Eam Khwang Wen Changyun Wen Changyun 19 All-fiber acousto-optical tunable filtering technology for optical communication Li Hao Wen Changyun 20 Multiple object tracking by stochastic method Li Jiang Chua Chin Seng MEng - Control & Instrumentation S/No. Project Title Student Supervisor/Co-supervisor 1 Design of control systems for regulation of mean arterial blood pressure of hypertension patients Zheng Hang Zhu Kuanyi Selected Publications in 2006 List of Selected Publications - Control & Instrumentation J. Mullane, E. Jose, M. D. Adams, W. Sardha Wijesoma, Including probabilistic target detection attributes into map representations, accepted, International Journal of Robotics and Autonomous Systems, M. J. He, W. J. Cai and B. Wu, Control structure selection based on relative interaction decomposition, International Journal of Control, Vol. 79, No. 10, pp , Oct Q. Xiong, W. J. Cai and M. J. He, Effective transfer function method for decentralized control system design of multi-input multi-output processes, Journal of Process Control, Vol.16, No. 8, pp , Sep G. Y. Jin, W. J. Cai, Y. W. Wang and Y. Yao, A simple dynamic model of cooling coil unit, Energy Conversion and Management, Vol. 47, No , pp , Sep C. C. Cheah, C. Liu and J. J. E. Slotine, Adaptive tracking control for robots with unknown kinematic and dynamic properties, International Journal of Robotic Research, Vol. 25, No. 3, pp , C. Liu, C. C. Cheah and J. J. E. Slotine, Adaptive Jacobian tracking control of rigid-link electrically driven robots based on visual task-space information, Automatica, Vol. 42, No. 9, pp , Sep C. C. Cheah, C. Liu and J. J. E. Slotine, Adaptive Jacobian tracking control of robots with uncertainties, dynamic and actuator models, IEEE Transactions on Automatic Control, Vol. 51, No. 6, pp , School of Electrical & Electronic Engineering 071

21 Selected Publications in 2006 List of Selected Publications - Control & Instrumentation H. Yu and C. S. Chua, GVF-based anisotropic diffusion models, IEEE Trans. Image Processing, Vol. 15(6), pp , Jun X. H. Liu and C. S. Chua, Multi-agent activity recognition using observation decomposed hidden Markov models, Image Vision Computing, Vol. 24(2), pp , Feb Y. J. Wang and C. S. Chua, Robust face recognition from 2D and 3D images using structural Hausdorff distance, Image Vision Computing, Vol. 24(2), pp , Feb C. Du, G. Guo, L. Xie and J. Zhang, Interaction rejection of multiple micro-actuators in dual-stage servos for hard disk drives, IEEE Transactions on Control Systems Technology, Vol. 14, no. 3, pp , C. Du, W. Wong and G. Guo, Experimental study of disk vibration reduction via stacked disks, accepted, Journal of Sound and Vibration, M. J. Er, R. Yan and J. Xu, Multi-period repetitive learning control for a class of unmatched systems with unknown control direction, accepted, Automatica, J. Zhou and M. J. Er, Adaptive output control of a class of uncertain chaotic systems, accepted, Systems and Control Letters, J. Zhou, M. J. Er and J. Zurada, Adaptive neural network control of uncertain nonlinear systems with non-smooth actuator nonlinearities, accepted, Neurocomputing, T. S. Koh, L. H. D. Cheong, C. K. M. Tan, and C. C. T. Lim, A distributed parameter model of cerebral blood-tissue exchange with account of capillary transit time distribution, NeuroImage, Vol. 30, pp , T. S. Koh, C. K. M. Tan, L. H. D. Cheong and C. C. T. Lim, Cerebral perfusion mapping using a robust and efficient method for deconvolution analysis of dynamic contrast-enhanced images, NeuroImage, Vol. 32, pp , T. S. Koh, X. Yang, S. Bisdas and C. C. T. Lim, independent component analysis of dynamic contrast-enhanced computed tomography images, Phys. Med Biol., Vol. 51, pp. N339-N348, C. Chen, Z. G. Li and Y. C. Soh MRF: A framework for source and destination based bandwidth differentiation service, accepted, IEEE/ACM Transactions on Networking, Y. Ji, Z. G. Li and C. Wen, Impulsive synchronization of chaotic systems via linear matrix inequalities, International Journal of Bifurcation and Chaos, Vol. 16, No.1, pp , H. Li, Z. G. Li and C. Wen, Fast mode decision algorithm for inter frame coding in fully scalable video coding, IEEE Transactions on Circuit and Systems for Video Technology, Vol. 16, No.7, pp , K. S. Low and T. S. Wong, A multiobjective genetic algorithm for optimizing hard disk drive performance, accepted, IEEE Transactions on Industrial Electronics, K. Zhou, K. S. Low and D. W. Wang, Zero-phase odd-harmonic repetitive controller for a single-phase PWM inverter, IEEE Transactions on Power Electronics, Vol. 21, No. 1, pp , Jan H. L. Zhuang, K. S. Low and W. Y. Yau, A pulse neural network with on-chip learning and its practical applications, accepted, IEEE Transactions on Industrial Electronics, Nanyang Technological University