The Organizing Committee intends to arrange workshops/tutorials to be held in conjunction with the IEEE CIRA 2009. Proposals are solicited for appropriate subjects. Proposals for half day or full day tutorials and workshops should be submitted to the Tutorial/Workshop Chair. The tutorials and workshops will be held on December 15, 2009. In order to evaluate a proposal, please send an email with the following information to the Workshops & Tutorials Chair, Prof. Tzuu-Hseng S. Li, at thsli@mail.ncku.edu.tw before July 31, 2009.
 
  1. Title
  2. Category: half day tutorial, full day tutorial, half day workshop, full day workshop
  3. Abstract 200-300 words
  4. Organizer (electronic and physical mail address)
  5. Confirmed presenters with affiliation
  6. Motivation and objectives
  7. List of topics to be covered
  8. Primary and secondary audience
  9. Relation to previous CIRA workshops
 
 
 

There will be two tutorials in the morning and the afternoon of Tuesday, December 15, 2009. The tutorials are free to attend for conference registrants.
 

Presenter: Jong-Hwan Kim (KAIST, Korea)
(09:30AM-12:30PM / Tuesday, December 15, 2009)
 

Human beings will be living in a ubiquitous world in which all IT devices are fully networked so that they can offer us desired services at any place and anytime. This shift has hastened the ubiquitous revolution, which has further manifested itself in the new multidisciplinary research area, ubiquitous robotics. It initiates the third generation of robotics following the first generation of the industrial robot and the second generation of the personal robot. A fairy tale introduced Genie, which upon springing from a lamp served Aladdin. The ubiquitous era brings us to the threshold of the realization of this dream, through ubiquitous robotics. Moreover, the robots shall have their own genome in which a specific personality is encoded. This concept leads to the study of genetic robotics. Cyber-physical robot system combines these new concepts of next generation robotics for the convergence of computational and physical systems.

This tutorial introduces the recent progress and development of ubiquitous robot, genetic robot and cyber-physical robot system along with the new classification of robot intelligence. Ubiquitous robot is composed of the three forms of robots: software robot, embedded robot and mobile robot to represent an amalgamation of the tripartite personification of entities of perception, thinking and action. Genetic robot has its own genetic codes to represent a specific personality. Cyber-physical robot system conjoins and coordinates the software agents and physical robots including SW and HW resources. Special emphasis in this tutorial is placed on intelligence technology for the cyber-physical robot system to realize cognitive intelligence, social intelligence, behavioral intelligence, ambient intelligence, genetic intelligence and swarm intelligence. This system will provide us with seamless, calm, and context-aware services in a networked environment.
 

(14:00PM-17:30PM / Tuesday, December 15, 2009)
 

#1: What¡¯s so special about the brain?
Presenter: Dae-Shik Kim (KAIST, Korea)

There are fascinating complementarities between robots and humans: what appears to be challenging for humans, seems easy for robots, while highly demanding problems for robots such as face recognition and smooth locomotion are mastered without much difficulties by humans. I will argue in this talk that many of the computationally hard problems of perception and action appear to be easy to humans precisely because they have been successfully solved by the brain during the course of its evolution and ontogeny. The corollary of this claim is that the structure and function of the brain may already contain solutions to many of the hard problems faced by cognitive robots. This talk will review some of the basic principles of operation and design of the mammalian brain, and discuss their relevance for computational intelligence in future Robotics.
 
#2: Methods in computational neuroscience
Presenter: Myoung Won Cho (Korea Institute for Advanced Study, Korea)

In this talk, we discuss computational methods and architectures based on knowledge in computational neuroscience. However many methods in robotics, especially based on neural networks, are inspired from real neural systems, recent studies in neuroscience propose different aspects of neural dynamics, learning, or function. A real neuron has more complex dynamics behavior than a simplified model in the artificial neural network. The role of complex neural behavior is newly understood how to play an important role in neural computations. In learning, spike-timing-dependent plasticity (STDP) is a general form for functional changes in real neurons and at synapses that are sensitive to the timing of action potentials in connected neurons. It is known that STDP introduces rich learning characters to a neural network however the characters are not fully revealed yet. And, the modular structure in the brain has been attracted interest to understand or design the computational architecture in a much massive and intellectual neural system.
 
#3: From single neurons to large-scale modeling
Presenter: Chang-Woo Shin (Asia Pacific Center for Theoretical Physics, Korea)
The brain is one of the most challenging "complex" systems in both the dynamical and the structural aspects. Due to the complexity, the brain can extract relevant patterns from sensory inputs, coordinate movements and control behaviors. A brain is composed of tens of billions neurons, and each neuron has thousands of synaptic connections from and to other neurons. One of the most different features between real and artificial neural networks is that the "real" neuron, the unit of information processing, is a dynamical unit. A neuron shows highly nonlinear response to input stimuli. If the nonlinear integration of the spatio-temporal stimuli exceeds a threshold, it makes an electrical pulse called the action potential, which is the unit of information, and the spatio-temporal "firing" activities carry the information. In the context of nonlinear dynamics and computational neuroscience, a neuron can be modeled as a nonlinear oscillator and the brain as a coupled oscillators network. The dynamical responses of a neural network depend on the network structure, synaptic connectivity between neurons in the network in other words, and the spatio-temporal pattern of the input stimuli to the network, as well as the intrinsic properties of the neurons in the network. Another distinct feature of the brain is the hierarchical structures, which means that the brain is a network of networks comprising hierarchical sub-networks. It is not possible to clearly distinguish each hierarchical level, but columnar and layered structure is observed almost uniform over the neocortex. Although it is still controversial, the minicolumns, each comprising about one hundred neurons, can be regarded as the mesoscopic functional units. In this talk, recent modeling approaches to study the structure and the function of the brain from single neurons to the cortical columns and finally upto the whole brain will be reviewed.
 
#4: Brain Connectivity Analysis
Presenter: Tae-Wook Ko (National Institute for Mathematical Sciences, Korea)
The complex connectivity between brain regions or neurons is believed to be one of the main factors contributing to the efficient information processing in the brain. Elements of the brain cooperate through the connectivity and perform the required tasks. Therefore, analyzing the brain connectivity can provide a fundamental basis for the understanding of the brain function. Recent advances in brain imaging and brain fiber tracking techniques have allowed us to obtain detailed in-vivo information on brain connectivity. In this talk, analysis of brain connectivity using the approach of complex networks will be reviewed. Structural, functional, and effective connectivity representing anatomical connections, statistical dependencies, and causal relationships, respectively, will be discussed. Properties of the structural/functional/effective connectivity will be discussed in the context of normal and abnormal brain function.
 
#5: Application of retina-on-chip to the development of visual neural prosthesis
Presenter: Kyung Hwan Kim (Yonsei University, Korea)
For successful restoration of vision by prosthetic electrical stimulation, evoked neural activities of retinal ganglion cells (RGCs) should properly represent or ¡®encode¡¯ spatiotemporal visual information. We propose a method for the evaluation of stimulation strategy so that crucial information on temporal pattern of intensity of visual input is properly represented in the RGC responses, assuming that pulse amplitude modulation is utilized. To demonstrate the concept, we recorded and analyzed population activities of RGCs responding to temporally patterned electrical stimulation, using an in-vitro model of retinal prosthesis, consisting of retinal patches and a planar multielectrode array (MEA). Spike train decoding was utilized to evaluate the efficacy of visual information representation. Spike train decoding refers to a procedure to reconstruct, or to ¡®decode¡¯ quantitative information encoded in multiunit neural responses (spike train), and has been applied for the study of fundamental encoding characteristics of neuronal populations and for the brain-machine interface. Here we focus on temporal visual information and thus decoded the temporal pattern of input intensity from RGC spike trains. The effectiveness of stimulus encoding strategies is evaluated by the goodness-of-fit between the decoded and the actual temporal variation of brightness, since a better stimulation pulse encoding strategy should provide more reliable transmission of information on input stimuli. Assuming that variations in intensity of a uniform visual input are transformed to the amplitudes of biphasic current pulse trains in retinal prosthesis, we intend to determine optimal methods to modulate the amplitudes of pulse trains so that crucial visual input information for the perception of temporal pattern of intensity is properly represented in the RGC responses.When the parameters were suitably determined, the RGC responses were reliably modulated according to the amplitude of electrical pulses and the temporal pattern of pulse amplitudes could be successfully decoded from electrically-evoked RGC spike trains. The range of pulse amplitude modulation and the pulse rate were critical for accurate representation of input information in RGC responses. The results suggest that pulse amplitude modulation is a feasible means to implement stimulus encoding strategy for retinal prosthesis.