Robustness of linearly solvable Markov games employing inaccurate dynamics model

As a model-based reinforcement learning technique, linearly solvable Markov decision process (LMDP) gives an efficient way to find an optimal policy by making the Bellman equation linear under some assumptions. Since LMDP is regarded as model-based reinforcement learning, the performance of LMDP is sensitive to the accuracy of the environmental model. To overcome the problem of the sensitivity, linearly solvable Markov game (LMG) has been proposed, which is an extension of LMDP based on the game theory. This paper investigates the robustness of LMDP- and LMG-based controllers against modeling errors in both discrete and continuous state-action problems. When there is a discrepancy between the model used for building the control policy and dynamics of the tested environment, the LMG-based control policy maintained good performance while that of the LMDP-based control policy deteriorated drastically. Experimental results support the usefulness of LMG framework when acquiring an accurate model of the environment is difficult.     

Experimental Evaluation of Characteristics of SR Motor Made of Permendur

A switched reluctance (SR) motor has a doubly salient pole structure. The performance of the SR motor strongly depends on the magnetic properties of the core material, since it consists only of an iron core and copper windings. This study experimentally evaluates the performance of an SR motor made of permendur (Fe–49% Co alloy), which has very high saturation flux density and low core loss. Two types of SR motors that have the same dimensions but different core materials are used: one motor is made of the conventional nonoriented Si steel, and the other is made of permendur. Comparison of the characteristics of the two motors shows that the performance of the permendur SR motor surpasses that of the conventional one in terms of output power and efficiency. © 2014 Wiley Periodicals, Inc. Electr Eng Jpn, 187(1): 51–57, 2014; Published online in Wiley Online Library ( wileyonlinelibrary.com ). DOI 10.1002/eej.22431     

Estimation of element-based zero-stress state for arterial FSI computations

In patient-specific arterial fluid–structure interaction (FSI) computations the image-based arterial geometry comes from a configuration that is not stress-free. We present a method for estimation of element-based zero-stress (ZS) state. The method has three main components. (1) An iterative method, which starts with an initial guess for the ZS state, is used for computing the element-based ZS state such that when a given pressure load is applied, the image-based target shape is matched. (2) A method for straight-tube geometries with single and multiple layers is used for computing the element-based ZS state so that we match the given diameter and longitudinal stretch in the target configuration and the “opening angle.” (3) An element-based mapping between the arterial and straight-tube configurations is used for mapping from the arterial configuration to the straight-tube configuration, and for mapping the estimated ZS state of the straight tube back to the arterial configuration, to be used as the initial guess for the iterative method that matches the image-based target shape. We present a set of test computations to show how the method works.     

Patient-specific computer modeling of blood flow in cerebral arteries with aneurysm and stent

We present the special arterial fluid mechanics techniques we have developed for patient-specific computer modeling of blood flow in cerebral arteries with aneurysm and stent. These techniques are used in conjunction with the core computational technique, which is the space?Ctime version of the variational multiscale (VMS) method and is called ??DST/SST-VMST.?? The special techniques include using NURBS for the spatial representation of the surface over which the stent mesh is built, mesh generation techniques for both the finite- and zero-thickness representations of the stent, techniques for generating refined layers of mesh near the arterial and stent surfaces, and models for representing double stent. We compute the unsteady flow patterns in the aneurysm and investigate how those patterns are influenced by the presence of single and double stents. We also compare the flow patterns obtained with the finite- and zero-thickness representations of the stent.     

Space–time interface-tracking with topology change (ST-TC)

To address the computational challenges associated with contact between moving interfaces, such as those in cardiovascular fluid–structure interaction (FSI), parachute FSI, and flapping-wing aerodynamics, we introduce a space–time (ST) interface-tracking method that can deal with topology change (TC). In cardiovascular FSI, our primary target is heart valves. The method is a new version of the deforming-spatial-domain/stabilized space–time (DSD/SST) method, and we call it ST-TC. It includes a master–slave system that maintains the connectivity of the “parent” mesh when there is contact between the moving interfaces. It is an efficient, practical alternative to using unstructured ST meshes, but without giving up on the accurate representation of the interface or consistent representation of the interface motion. We explain the method with conceptual examples and present 2D test computations with models representative of the classes of problems we are targeting.     

Silicon based on-chip antenna using an LC resonator for near-field RF systems

This paper presents the silicon based on-chip antenna using a LC resonator. The proposed antenna consists of a stacked capacitor and a spiral inductor on silicon substrate. The spiral inductor structure without underpass was proposed for improvement the performance of the silicon based-antenna. The resonant frequency of the fabricated antenna was measured as 465 MHz. Its return loss was 23.4 dB at resonant frequency. The antenna has a gain of ?35.75 dBi due to small size and silicon substrate. However, the fabricated antenna has good performance in the near-field.     

Phase transformation and interface segregation behavior in Si3N4 ceramics sintered with La2O3–Lu2O3 mixed additive

Microstructure and mechanical property of silicon nitride (Si₃N₄) ceramic are strongly dependent on the selection of sintering additives. When rare‐earth (RE) oxide is used as the sintering additive, segregation of RE ions at interface between Si₃N₄ grain and intergranular glassy film (IGF) is believed to play a critical role. Although the ionic radius of RE ion is known to be an empirical parameter to modify the mechanical property, the correlation between the segregated ions and their ionic radii is still under controversy. In order to address this issue, (i) rate of α‐β phase transformation and (ii) segregation behavior at the interface were studied for Si₃N₄ ceramics sintered using mixture of La₂O₃ and Lu₂O₃ as additives in this study. Specimens of Lu content 30% and higher exhibited lower activation energies for the α‐β phase transformation as compared with those of Lu content 20% and lower. In terms of the segregation behavior, La was preferably segregated at one site and Lu at the other site along β‐Si₃N₄/IGF interface in the specimens of Lu content 30% and higher. It is understood from these results that Lu segregation site should be more closely related with grain growth.     

Characterization by Stm and Electrical Conductivity of Single-Walled Carbon Nanotubes

Abstract

Single-walled carbon nanotubes (SWCNT) prepared by a DC-Arc-Discharge and purified by a hydrothermal treatment followed by several steps of procedures including heating, burning, extracting and acid-washing are investigated by a conventional STM technique morphologically and spectroscopically. The electical resistivities of compacted pellets (mats) of SWCNT are measured as a function of temperature in the range from the room-temperature to 1.5K. The behaviors seems to be essentially originated in the metallic nature of individual SWCNT.     

Mimic Drug Dosage Modulation for Neuroplasticity Based on Charge-Trap Layered Electronics

The human brain is often likened to an incredibly complex and intricate computer, rather than electrical devices, consisting of billions of neuronal cells connected by synapses. Different brain circuits are responsible for coordinating and performing specific functions. The reward pathway of the synaptic plasticity in the brain is strongly related to the features of both drug addiction and relief. In the current study, a synaptic device based on layered hafnium disulfide (HfS₂) is developed for the first time, to emulate the behavioral mechanisms of drug dosage modulation for neuroplasticity. A strong gate-dependent persistent photocurrent is observed, arising from the modulation of substrate-trapping events. By controlling the polarity of gate voltage, the basic functions of biological synapses are realized under a range of light spiking conditions. Furthermore, under the control of detrapping/trapping events at the HfS₂/SiO₂ interface, positive/negative correlations of the A_n/A₁ index, which significantly reflected the weight change of synaptic plasticity, are realized under the same stimulation conditions for the emulation of the drug-related addition/relief behaviors in the brain. The findings provide a new advance for mimicking human brain plasticity.