This cluster of excellence sets out to examine cognition for technical systems like automotives, robots and plants. Cognitive technical systems are equipped with artificial sensors and actuators, integrated and embedded in physical systems, and are employed in physical environments. They differ from other technical systems in that they function cognitively and possess self-learning properties. Cognitive action incorporates contemplated and habitual behavior that pursues long-term targets. Thanks to their cognitive ability, including perception, contemplation, learning and planning capabilities, technical systems evolve to become systems that "know what they are doing". This means that these systems are able to select the most appropriate course of action in a given situation while processing information on the present environment and their own scope. This cognitive ability serves to improve the systems' reliability, flexibility, adaptability and performance. Interaction and cooperation become easier.
CoTeSys unites research competence in (medical) neurosciences, natural sciences, engineering, informatics and humanities at Technische Universit?t Munchen (TUM), Ludwig Maximilian University (LMU), Federal Armed Forces University (UBM), German Aerospace Centre (DLR) and Max Planck Institute of Neurobiology (MPI) in order to devise, implement and analyze the data-processing methods required for cognitive technical systems. Fundamental research in the spheres of neurobiology, neuroscience and cognitive science forms the basis for innovative engineering and information technology approaches for "artificial" cognition. These are examined and evaluated in human factors studies. A special feature of CoTeSys is that theoretical research results can be realized directly as hardware experiments. Three conceptual scenarios are pursued to this end: cognitive aircraft and automobiles, cognitive humanoid robots and cognitive factories. Apart from testing concepts for accuracy and applicability, these complex demonstrators also serve to show how far the cluster has advanced and to monitor its progress.
CoTeSys realizes cognitive ability and focuses on technical systems. With technical systems as the major field of application, there are plans to investigate a key innovative area of technology, which is of crucial importance to the German high-tech industry. The collaboration between disciplines and institutions creates synergies that are on course to consolidate Munich's lead in research, academia and technology.
The sustainability of the cluster is safeguarded by a new teaching concept that covers new Bachelor's and Master's degree courses, plus summer schools for doctoral students. These educational and qualification measures are not only geared to promote local students and junior scientists but also to attract researchers from international, top-level institutions to Munich.
For the design of software-interfaces usability criteria are defined in several norms. Among others, individualisation and competence enhancement are requirements for software quality with potentially conflicting aims.
The system ComPASS (competence enhancing multi-agent-system with speech output) was developed to support machine-operators at fault-finding in CNC work-tool machines. These machines are usually very reliable, so faults occur only seldom. Therefore, such a system is rarely used and only in difficult situations. Difficult here means either the fault is difficult to find or the expert operators do not know the solution. Many faults occur so seldom that the appropriate procedure is forgotten from one occurrence to the next.
For ComPASS to have competence enhancing capabilities competence enhancement had to be defined in the realm of machine operators, the tools for measuring competence and its increase had to be built and their suitability tested. The broader idea was - from the development process of the software - to come up with guidelines for software development that facilitate the competence enhancing aspect.
It is discussed to what extend the user may or has to be restricted in individualising the user interface to his/her own the benefit to ensure competence enhancement. The findings and the resulting design decisions have to be seen before the background of a specific application with infrequent use in difficult cases and learning on the job.
The Complex Fluid Mechanics Group, one of the three research groups of the COE program "Center of Excellence for Research and Education on Complex Functional Mechanical Systems", and School of Mathematics, the University of Birmingham, hosted jointly and successfully "the Kyoto-Birmingham University International Symposium on Recent Advances in Fluid Mechanics," from September 3rd to 5th, last fiscal year in 2006, at the buildings of Mathematics department and Barber Institute in the University of Birmingham. Professor M. Nagata of Kyoto Univ. (KU) and Professor S. P. Decent of Univ. Birmingham (UB) organized the symposium, and 39 participants including 13 COE members of KU joined there. The symposium were composed of the following eight sessions: Environmental Fluid Dynamics I and II, Solid/Fluid Interaction and Industrial Flow, Chaotic Mixing, Micro Fluid Dynamics, Transition, Capillary, and Turbulence.
Concerning a research topic of the multi-jet mixing control, the effects of flow perturbation on mixing performance of an air-jet array were investigated experimentally. The jet array consists of three same-sized parallel jets, and the mass flow rate of the central jet is perturbed in a pulsatile or sinusoidal cycle with the frequency, f = 0, 1 or 2Hz. The time-averaged Reynolds number of the jet flow is fixed at around 1,500, in laminar flow regime. The jet array discharged simultaneously without the jet perturbation (f = 0Hz) shows good mixing performance in comparison with the jet array estimated using three arithmetically-superimposed single jets. The flow condition with the jet perturbation (f = 1 or 2Hz), on the other hand, is alternately changed, laminar or turbulent, in one cycle of the central jet perturbation. The time-averaged velocity profiles of the jet array obtained under certain perturbed conditions demonstrate an adverse change of the mixing performance at the locations near the nozzle and far downstream. Phase-averaged velocity profiles were thoroughly examined in order to understand such a mixing mechanism of the multiple jet flow.
There are great expectations to hydrogen as a next generation energy or a medium of energy. However, it is well known that hydrogen weakens the strengths of metals. Despite the extensive investigation concerning hydrogen related fracture, the mechanisms have not been enough clarified yet. The difficulties to reveal the essential effects of hydrogen are mainly attributed to the characteristics of hydrogen such as ppm-order extremely low concentration, high diffusivity and high sensitivity to the defect-densities in metals. In this study, we applied molecular dynamics (MD) simulations to Mode I crack growth in -iron single crystals with and without hydrogen, and analyzed the hydrogen effects from the atomistic viewpoints. The crack growth behavior in two types of crystal orientation are analyzed and compared. We propose the mechanism of hydrogen related fractures based on the interaction between hydrogen atoms and dislocations around the crack tip.
In early '90s, a sintered carbide end mill with an (Al, Ti)N coating was introduced into the manufacturing of dies and molds, which made it possible to directly machine pre-hardened steel of the hardness up to HRC60. In die/mold manufacturing processes, particularly in the machining with such a tool, the process planning relies to a high extent on expert machine operators' experience and knowledge. Even for such an expert operator, the process planning for the high-productive machining of hardened steel is quite difficult. In order to perform safe, high-productive machining by fully utilizing the potential of recent high-speed machining centers, it is important to develop a support system that can autonomously monitor the machining process, and optimize machining conditions adaptively. This presentation briefly reviews our research approaches to this goal by means of the machining process control and the tool path planning.
In designing controllers, robustness against uncertainties is very important property, therefore we proposed design methods of nonlinear robust control system by use of neural networks. Competition between a neural network, which acts as a robust controller and uncertainties of the plant, plays a very important role to be robust, and simple order formation, such as power law scaling, can be found in a set of trained robust controllers. Such simple order is widely found in nature and complex systems, but learning methods without considering uncertainties has never formed such relations. Therefore competitive learning will be important to order formation in various systems.
The effect of rainfall on mass transfer across the air-water interface and turbulent structure after a droplet impinging on the air-water surface were investigated through laboratory experiments in a turbulent open-channel flow and a single droplet apparatus. The air-water CO2 transfer velocity due to raindrops impinging on the air-water surface in an open-channel flow and both flow and concentration fields in a receiving tank were measured. The results show that rainfall enhances CO2 transfer across the air-water interface and the mass transfer mechanism was investigated. The mass transfer velocity on the liquid side is well correlated with the mean vertical momentum flux of rainfall (MF). The maximum value of the mass transfer velocity obtained in the present rain experiments corresponds to that observed in oceans with high wind speeds. This suggests that it is of great importance to consider the effects of rainfall in precisely estimating the local air-sea CO2 exchange rate.
The CO2 transfer velocity due to impinging raindrops is well correlated with the mean vertical momentum flux of raindrops using laboratory measurements (Takagaki and Komori, 2007). Thus, in order to clarify the effects of rainfall on the global and local CO2 transfer across the air-sea interface, the mean annual net air-sea CO2 flux was estimated using both the daily precipitation data set and the empirical correlation between the CO2 transfer velocity and mean vertical momentum flux. The rainfall effects are also compared with wind shear effects. The results show that the rainfall effects are significant for local CO2 budget between atmosphere and ocean in the equatorial and mid-latitude regions, but they are not so important for global budget, compared to the wind shear effects.
Ferroelectric materials have been drawing attention because of their prominent electronic property of spontaneous polarization. In this study we have investigated the atomistic and electronic structure of the ferroelectric domain wall (DW) where polarization direction is discontinuous in PbTiO3 using ab initio (first-principles) calculations based on the density-functional theory (DFT). At the domain wall, the magnitude of polarization decreases by 20% from the bulk value and the polarization direction rotates from [100] to [001] across the DW. The width of the domain-wall dominant region is estimated to be 1.0 nm from the profile. Furthermore, we have studied the domain switching which is an abrupt movement of the domain wall induced by shear stress. It has demonstrated that the domain wall starts to move in the normal direction to the wall when the stress reached at the critical value of 152 MPa. During the domain switching, a covalent Pb-O bond at the center of the DW breaks and another bond is newly constructed. Therefore, we conclude that the reconstruction of the Pb-O bond is responsible for the domain switching.
In this study, we first deal with locomotion control of a biped robot. The control system consists of nonlinear oscillators that generate robot kinematics and adequately change the behaviors based on sensory information, resulting in adaptive walking to environmental changes and various situations. We carried out hardware experiments that verified the performance of the control system and found similar mechanism to animals that established adaptive behaviors. In contrast to robots, animals establish adaptive walking in diverse environments by cooperatively manipulating complicated and redundant musculoskeletal systems. To elucidate such intelligent locomotor mechanisms, we second deal with locomotion of animals based on numerical simulations by referring to the experimental findings of adaptive behaviors. In particular, we study locomotion in the Japanese monkey. We constructed its sophisticated skeletal model based on anatomical data, achieved physical kinematics based on motion capture data, and generated nervous model based on neurophysiological findings. Finally we show that numerical simulations established locomotion through dynamical interactions between the skeletal and nervous systems, and environment.
We study the spatio-temporal dynamics in a circularly polarized external field in association with the dynamic phase transition, which is an example of the non-equilibrium phase transition. A local element of the system shows the dynamic phase transition between limit cycle, quasi-periodic oscillation, and chaotic motion. When the oscillators are diffusively coupled, dynamical patterns such as domain wall, spiral, and spatio-temporal chaos are induced by the strong external force. The domain wall structure transition between the symmetric Neel wall and the asymmetric Bloch wall occurs. We observe the propagation of the spiral pattern when the uniform solution shows the quasi-periodic oscillation. In some parameter region, spirals collapse, and the spatio-temporal chaos emerges. We note that only the motionless domain wall structure is present in the absence of the external field. Moreover, we stabilize an unstable Neel wall solution using a time delayed feedback control technique. This method can be applicable to stabilize an unstable periodic orbit in systems with large degrees of freedom.
One of the primary roles of the 21st Century COE Program is to develop superior young researchers in those fields. In this program, we have emplyed Kyoto University's tradition of on-the-research training to develop young researchers with broad perspectives and highly specialized skills who possess the ability and courage to act as trailblazers in a novel field of study. In this talk, various new systems and programs initiated for this purpose are introduced; Joint Interdisciplinary research program, Fellowship program, Public education program, and so on.