In this work potential benefits and drawbacks of providing information to road users are discussed. To model the complex behaviour of road users multi-agent-techniques are used. A special agent-architecture is proposed and discussed. To consider the impact of pre-trip-information, coordination games, like the El-Farol problem and the minority game, are studied and interpreted in the context of route choice scenarios. Additionally, the impact of en-route information is studied by simulating a basic two-route scenario with dynamic and static agents. A correlation analysis yields that the traffic system is destabilised by en-route information. The overall performance of the system is reduced by the effect of concentration. Different types and ways of generating current information are tested. It is found that the nature of the information influences the potential benefits of the information system strongly. As potential application an Advanced Traveller Information System is introduced, which comprises an agent-based simulator, which is coupled to a stream of on-line data. The framework provides network-wide information about the current traffic state. It is applied to the freeway network of the felderal state of North-Rhine Westfalia (NRW). Results are presented and further applications are discussed. To provide predictive information different methods to forecast traffic are analysed and discussed. Heuristics generated by an statistical analysis of historical data are studied. The daily traffic demand is classified with regard to daily, seasonal characteristics, as well as special events or directional flow. To provide a prediction for arbitrary horizons heuristics are merged with a constant model. The model delivers good predictions for a short and long-term traffic forecast.

This thesis describes research on magnetism, structure, and vibrational dynamics in a number of magnetic and non-magnetic nanoscaled heterostructures. Much attention is paid to a thorough structural characterization of the materials, because this is essential for understanding the stabilization of metastable phases (here: fcc-like Fe and alpha-Sn), magnetic properties such as anisotropy and magneto-volume instabilities, or lattice vibrational behavior. A number of experimental techniques were used, such as 57Fe and 119Sn Conversion Electron Mössbauer Spectroscopy (CEMS), Magneto-Optic Kerr Effect (MOKE), 119Sn Nuclear Resonant Inelastic X-ray Scattering (NRIXS), X-Ray Diffraction (XRD), Reflection High Energy Electron Diffraction (RHEED), Low Energy Electron Diffraction (LEED), and Auger Electron Spectroscopy (AES).

The lifetime of layer-by-layer growth of crystal surfaces, mainly in the context of growth conditions found in molecular beam epitaxy (MBE), is the central issue of this thesis. These conditions imply a driven system far from equilibrium which relaxes due to surface diffusion. At first, the ceasing of layer-by-layer growth due to fluctuations in the particle supply is considered. A theory for the according lifetime is presented and confirmed for the one-dimensional surface. Special care is taken for the two-dimensional case where deviations from previous results are found, explained, and used to revise the assumptions on which the theory is based. In particular the applicability of the -- commonly accepted -- conserved KPZ continuum equation and the premise of a single morphologically relevant length scale are affected. The practically more relevant scenario of layer-by-layer growth's breakdown caused by barriers to interlayer transport (which give rise to the Villain instability) is studied. Data obtained fr om computer simulations is compared to the predictions of a linear stability analysis and is used to foretell the effect of counteracting variations of energy barriers. The latter enables to decide in which cases a strained surface is either hindering or advantageous for layer-by-layer growth. A mean field model describing surface growth, which lacked up to now a systematic treatment, is investigated. For the basic version, the asymptotic behavior is derived exactly and -- tuning the sole control parameter -- a transition from Poisson-like growth to persistent layer-by-layer growth is found together with a non-trivial powerlaw behavior right at the transition point. Finally the extensibility of the model to include a finite lifetime of layer-wise growth is examined. The damping of oscillations of certain surface-sensitive quantities is the manifestation of the surface's roughening which terminates the layer-by-layer growth. A scenario alternative to the roughening is suggested. It leads as well to damping of oscillations and consists of a step bunch which dissolves during growth and 'floods' an adjacent terrace. Growth simulations of this process are compared to a deterministic model and to experimental results. Finally several toy models for surface growth, subjected to noise reduction are considered. The latter technique makes possible layer-by-layer growth also in these models and the dependence of its lifetime on the degree of the noise reduction is studied. The main focus is on the behavior's relation to continuum equations and the corresponding universality classes, which are commonly used to classify the different models.

None

The dynamics of driven interfaces in the random-field Ising model (RFIM) is investigated by the use of Monte Carlo simulations. Interfaces in the RFIM separate regions of opposite spin orientation. By applying an external field one orientation is energetically favored. This may yield an interface motion which is hindered by the random-field. Without thermal fluctuations the competition between the driving field and the random-field leads to a so-called depinning transition. A permanent interface motion is found only if the driving field exceeds a threshold field H_c. At the transition point the interface velocity vanishes continuously, characterized by a critical exponent beta. The values of beta found in the RFIM for the dimensions d=3,4,5,6 support the assumption that the depinning transition in the RFIM belongs to the universality class of the Edwards-Wilkinson equation with quenched disorder. The energy barriers which cause a pinning of the interface at temperature T=0 can be overcome due to the energy provided by thermal fluctuations. This yields a permanent interface motion. For sufficient small driving fields a so-called creep regime is found in the random-field Ising model. This creep regime is predicted by phenomenological theories, functional renormalization group calculations, and has been observed in experiments. The field dependence of the energy barrier in the RFIM is investigated and the results are compared with those known in the literature. Furthermore, it is investigated whether the influence of temperature on the depinning transition can be understood within the theory of critical phenomena. It is assumed that the interface velocity can be expressed as a generalized homogenous function in the vicinity of the transition point (H=H_c|T=0). This assumption is supported by the results of simulations in the dimensions d=3,4,6, yielding an algebraic decay v(H=H_c) proportional T^(1/psi) with an exponent psi>0. The assumption of the interface velocity being a generalized homogenous function is also validated by simulations of magnetic films. From these simulations it can additionally be concluded that the depinning transition in magnetic films is characterized by the two dimensional exponents. The investigations of the five dimensional model show the occurrence of logarithmic correction revealing that d_c=5 is the upper critical dimension of the depinning transition in the RFIM.

This work presents research on structural and magnetic properties of epitaxial (Fe_t/Pd_4nm)₃₂ multilayers (MLs). The Fe layer thickness is varied from 1nm to 4nm, while all other parameters are kept constant. Structural investigation are made utilizing STM, RHEED, XRD, RBS and (HR)TEM. The RHEED and XRD investigations reveal that thin Fe layers (t _Fe < 2.5nm) exhibit on fcc-like phase with a high magnetic moment of 2.7 µ_B per iron atom. For Fe layers with thicknesses of t_Fe > 2.5nm the magnetic moment of Fe is 2.2 µ_B, which is in agreement with the value of bulk bcc Fe. From RHEED investigations we find that with increasing thickness of the Fe layer the unit cell of Fe is successively compressed in growth direction while its basal plane remains unaltered. In MLs with thin Fe layers with high magnetic moments the atomic volume of iron is comperable to the volume of the high moment phase in fcc Fe, the atomic volume of MLs with thicker Fe layers is close to the equilibrium value of bcc Fe. The mutual interplay of the tetragonal distortion, atomic volume and morphology and its influence on the magnetic properties of the MLs are discussed in this thesis.

In the present thesis the relaxor properties of the tungsten-bronze ferroelectric material strontium-barium niobate Sr_1-wBa_wNb₂O₆(SBN) are investigated. SBN is a very good experimental realization of the three-dimensional random-field Ising model (3d <i>RFIM</i>). The quenched random fields (<i>RF</i>) originate from charge disorder and can be enhanced by doping with Ce³⁺. They are responsible for the formation of polar nanoclusters in the paraelectric phase. In order to study the critical behavior of SBN, linear birefringence (<i>LB</i>) and second harmonic generation (<i>SHG</i>) have been measured as a function of temperature. Within an Ornstein-Zernike analysis of the <i>LB</i> data the autocorrelation function, <<i>P</i>²>, was determined, whereas the dipolar correlation length was obtained from the <i>SHG</i> data. They suggest that, due to its intrinsic disorder, pure SBN does not belong to the 3d Ising universality class. Doping with Ce³⁺ions, which seem to generate <i>RF</i>s, enhances the relaxor properties. The critical exponents <i>v</i> and <i>y</i> of SBN:Ce shift against those of the 3d <i>RFIM</i>. The domain morphology of SBN:Ce has been investigated by piezoelectric force microscopy (<i>PFM</i>). Fractal-like shaped zero-field cooled nanodomains are observed. Their size distribution can be described by a power law with exponential cutoff in accordance with prediction for the <i>RFIM</i>. It was measured for the first time in a <i>RF</i> system. The temperature and field induced evolution of natural and written domains has been studied with <i>PFM</i>, <i>LB</i> and <i>SHG</i> measurements. It reveals a very slow relaxation from a macrodomain into a depolarized multidomain state (and vice versa) even above <i>T_C</i>. This hints at strong pinning forces due to quenched <i>RF</i>s. They are also responsible for the observed aging in poled SBN and the field induced cluster percolation above <i>T_C</i> measured with <i>SH</i>-hysteresis. The domains can be considered as a "thick" phase grating for <i>SH</i>-diffraction. The Bragg-regime diffraction efficiency reflects the temperature and field induced change of the averaged domain sizes in the system.

Im vorliegenden Artikel wird es ausschließlich um den Dialog zwischen Computern und Menschen gehen, und hier insbesondere darum, wie weit eine natürlichsprachliche Interaktion möglich und sinnvoll ist. Der Begriff Interaktion wird hier absichtsvoll verwendet, und zwar als Kontrast zu Kommunikation. Wenn auch häufig von Mensch-Maschine Kommunikation gesprochen wird, so ist es doch fraglich, ob sich der Begriff der Kommunikation, so wie er für zwischenmenschliche Handlungen gebraucht wird, auch auf den Umgang zwischen Mensch und Maschine anwenden läßt. Mit Kommunikation wird in der Informatik schließlich auch der Informationsaustausch zwischen technischen Geräten bezeichnet, etwa die Schnittstelle zwischen einem Computer und seiner Peripherie (z.B. Monitor, Tastatur, Drucker) oder anderen Computern. Es handelt sich hierbei um technische Eigenschaften des Übertragungskanals und der Kodierung (Protokoll genannt), in der Informationen übertragen werden, also um nachrichtentechnische Spezifikationen.

Standard architectures favor sequential processes for the semantic and syntactic part of natural language generation. Some semantic decisions, however, require information from the syntactic part, as is shown at the example of translating English gerunds into German. As a solution to this problem, a model is proposed in which syntactic and semantic modules work in parallel, computing preference values for different translation variants. On the basis of these preferences, one variant is selected and further elaborated. This model draws on a maximum of information, including stylistic considerations, while avoiding the computational load of backtracking processes.

Varying words in natural language generation requires the use of syn-onyms. In traditional dictionaries, however, the formal representation of synonyms is difficult due to their dependence on context. In addition to that, not all synonyms enhance the readability and comprehensibility of a text equally well. It is proposed to take the context of synonyms into account by using a spreading activation mechanism in a semantic network. Several criteria are discussed to choose among the so-obtained alternatives.