Data-efficient learning of robotic clothing assistance using Bayesian Gaussian process latent variable model

ABSTRACT

Motor-skill learning for complex robotic tasks is a challenging problem due to the high task variability. Robotic clothing assistance is one such challenging problem that can greatly improve the quality-of-life for the elderly and disabled. In this study, we propose a data-efficient representation to encode task-specific motor-skills of the robot using Bayesian nonparametric latent variable models. The effectivity of the proposed motor-skill representation is demonstrated in two ways: (1) through a real-time controller that can be used as a tool for learning from demonstration to impart novel skills to the robot and (2) by demonstrating that policy search reinforcement learning in such a task-specific latent space outperforms learning in the high-dimensional joint configuration space of the robot. We implement our proposed framework in a practical setting with a dual-arm robot performing clothing assistance tasks.     

Estimation of anisotropic properties of nano-structured arrays by modal vibration control at microscale

Nano-structured arrays are engineered to meet the requirements of a variety of applications such as microfilters, sensors, and structural interface due to their unique mechanical characteristics, which cannot be achieved by conventional solid materials. However, it is hard to evaluate the elastic properties of nano-structured arrays owing to the discrete structure, sample size, and availability of suitable techniques. To facilitate this, we develop an advanced three-dimensional microscale vibration testing process. In the test, a specially designed three-dimensional microspecimen with tuned mass is excited by a piezoelectric actuator, and the resonance frequencies are detected by a laser device successfully. The anisotropic elastic moduli of nano-structured array composed of helical nano-springs are identified from a single spectrum. This array shows so strong characteristic anisotropy that the solid one hardly can attain. The microscale testing technique can be extended to other materials and microstructures.     

Neuroprotection,Growth Factors and BDNF-TrkB Signalling in Retinal Degeneration

Neurotrophic factors play key roles in the development and survival of neurons. The potent neuroprotective effects of neurotrophic factors, including brain-derived neurotrophic factor (BDNF), ciliary neurotrophic factor (CNTF), glial cell-line derived neurotrophic factor (GDNF) and nerve growth factor (NGF), suggest that they are good therapeutic candidates for neurodegenerative diseases. Glaucoma is a neurodegenerative disease of the eye that causes irreversible blindness. It is characterized by damage to the optic nerve, usually due to high intraocular pressure (IOP), and progressive degeneration of retinal neurons called retinal ganglion cells (RGCs). Current therapy for glaucoma focuses on reduction of IOP, but neuroprotection may also be beneficial. BDNF is a powerful neuroprotective agent especially for RGCs. Exogenous application of BDNF to the retina and increased BDNF expression in retinal neurons using viral vector systems are both effective in protecting RGCs from damage. Furthermore, induction of BDNF expression by agents such as valproic acid has also been beneficial in promoting RGC survival. In this review, we discuss the therapeutic potential of neurotrophic factors in retinal diseases and focus on the differential roles of glial and neuronal TrkB in neuroprotection. We also discuss the role of neurotrophic factors in neuroregeneration.     

Thermal and mechanical properties of bio-based polymer networks by thiol-ene photopolymerizations of gallic acid and pyrogallol derivatives

Allylated pyrogallol (A3PG) and acrylated pyrogallol (Ac3PG) as bio-based trienes, and allylated gallic acid (A4GA) and acrylated allyl gallate (Ac3A1GA) as bio-based tetraenes were synthesized from pyrogallol and gallic acid, respectively. Thiol-ene photopolymerizations of the bio-based polyenes and a pentaerythritol-based primary tetrathiol (pS4P) at the allyl/SH ratio of 1/1 produced photo-cured resins (A3PG-pS4P, Ac3PG-pS4P, A4GA-pS4P and Ac3A1GA-pS4P). The FT-IR spectral analysis revealed that thiol-ene reactions of thiol/allyl and thiol/acryloyl groups smoothly proceeded. Gel fractions of acryl-based cured resins were a little higher than those of allyl-based cured resins. The swelling test and dynamic mechanical analysis revealed that GA- and acryl-based cured resins exhibited higher crosslinking densities than PG- and allyl-based cured resins, respectively. A higher order of tan δ peak temperature was Ac3PG-pS4P (48.3 ° C) > Ac3A1GA-pS4P (24.1 ° C) > A4GA-pS4P (22.1 ° C) > A3PG-pS4P (?7.8 ° C). Ac3PG-pS4P displayed the highest 5 % weight loss temperature, tensile strength and tensile modulus among all of the cured resins.     

Verification and validation of explicit moving particle simulation method for application to internal flooding analysis in nuclear reactor building

The Fukushima nuclear accident raised the importance of flooding study in nuclear reactor buildings. It is known that the external flooding can be attributed to natural causes, while the internal flooding is caused by the piping ruptures, tank failure or the actuation of fire suppression systems. The building flooding process can damage the safety-related components and systems, which needs to study carefully. In order to simulate this phenomenon which is accompanied by complex flow with free surface, a particle method based on Lagrangian approach named explicit moving particle simulation (EMPS) method, is employed in this analysis. Verification and validation analyses are carried out. The verification problems are a hydrostatic analysis and a water spreading to investigate the differences of the particle wall and polygon wall boundary models, while the validation studies of two experiments of dam-break induced flooding show good agreements. Afterwards, the internal flooding process in AP1000 is simulated by assuming a break of the in-containment refueling water storage tank as an example. The results achieved so far indicate that the EMPS method is capable to simulate the internal flooding process in the nuclear reactor buildings.     

Investigation of Precipitation Behavior in Age-Hardenable Cu-Ti Alloys by an Extraction-Based Approach

We investigated the precipitation processes in Cu-4 mol pct Ti alloy specimens aged at 723 K (450 °C), by means of X-ray diffraction and chemical analyses of the precipitates extracted from the parent alloy specimens. Aging-induced precipitate particles of a spinodally decomposed disorder, α′; those of a metastable order, β′-Cu₄Ti; and those of a stable order, β-Cu₄Ti, were continuously formed in the aged specimens. The extraction of the precipitate particles from the aged specimens by submergence in a nitric solution allowed for not only the structural analyses of the constituent precipitate phases but also the quantitative evaluation of their chemical compositions and volume fractions. Early during the aging process, the supersaturated Cu solid solution decomposes spinodally in a continuous manner, and an unstable disorder, α′, appears. Then, fine needle-shaped β′-Cu₄Ti particles, which have a Ti content of approximately 37.5 mol pct, form in the Cu matrix. During prolonged aging, coarse cellular components composed of the terminal Cu solid solution and stable β-Cu₄Ti particles which have a Ti content of 20.5 mol pct nucleate and grow, primarily in the grain boundaries, at the expense of the metastable β′-Cu₄Ti particles. The volume fraction of the β′-Cu₄Ti particles in the alloy reaches a maximum of approximately 1.7 pct after aging for 24 hours, while that of the β-Cu₄Ti particles increases steadily to more than 18 pct after 480 hours. The volume fraction of the fine β′-Cu₄Ti particles in the alloy specimens remained constant throughout the age-hardening, indicating that the hardening is primarily owing to the fine dispersion of the β′-Cu₄Ti particles and not because of the large volume fraction of coarse β-Cu₄Ti particles.     

Preparation of ultrafine polyurethane fiber web by laser‐electrospinning combined with air blowing

Thermoplastic polyurethane fiber webs were prepared using a laser‐heated electrospinning process combined with air blowing. The effect of spinning conditions such as air flow rate and air temperature on fiber diameter and molecular weight was investigated. Although the average fiber diameter decreased with increased air flow rate at each air temperature, the diameter increased when the air flow rate was >15 NL min^?1. In addition, the fiber was comparatively thicker with an increase in the air temperature. The variation in the fiber diameter tends to increase with the air flow rate, and a reduction in the molecular weight of the fiber by thermal degradation was suppressed. The thinnest and most uniform fiber with a diameter of 0.9 µm and a diameter coefficient variation of 15% was obtained at an air temperature of 25°C under an air flow rate of 15 NL min^?1. This fiber also had a minimum of decreased molecular weight. POLYM. ENG. SCI., 54:2605–2609, 2014. © 2013 Society of Plastics Engineers     

Preparation and mesostructure control of highly ordered zirconia particles having crystalline walls

Mesostructured zirconia particles having monoclinic-type crystalline walls were prepared using a low-temperature crystallization technique. Crystalline zirconia particles with highly-ordered mesostructures were obtained through the sol–gel process of zirconium sulfate tetrahydrate at 333 K in the presence of molecular self-assemblies of cetyltrimethylammonium bromide (CTAB) or mixtures of CTAB and anionic molecules such as sodium dodecyl sulfate and sodium p-toluenesulfonate. Variations in the molar ratios of CTAB and the chemical species of anionic molecules led to the variations in the periods of highly-ordered zirconia having crystalline walls. Calcination of the mesostructured zirconia particles prepared using templates consisting solely of CTAB yielded crystalline mesoporous zirconia particles.