Multiagent reinforcement learning applied to a chase problem in a continuous world

Reinforcement learning (RL) is one of the methods of solving problems defined in multiagent systems. In the real world, the state is continuous, and agents take continuous actions. Since conventional RL schemes are often defined to deal with discrete worlds, there are difficulties such as the representation of an RL evaluation function. In this article, we intend to extend an RL algorithm so that it is applicable to continuous world problems. This extension is done by a combination of an RL algorithm and a function approximator. We employ Q-learning as the RL algorithm, and a neural network model called the normalized Gaussian network as the function approximator. The extended RL method is applied to a chase problem in a continuous world. The experimental result shows that our RL scheme was successful. This work was presented in part at the Fifth International Symposium on Artificial Life and Robotics, Oita, Japan, January 26–28, 2000     

Two-group interfacial area transport equations at bubbly-to-slug flow transition

In relation to nuclear reactor accident and safety studies, experiments on hot-leg U-bend two-phase natural circulation in a loop with a relatively large diameter pipe (10.2 cm inner diameter) were performed for understanding the two-phase natural circulation and flow termination during a small break loss of coolant accident in LWRs. The loop design was based on the scaling criteria developed under this program and the loop was operated either in a natural circulation mode or in a forced circulation mode using nitrogen gas and water. Various tests were carried out to establish the basic mechanism of the flow termination as well as to obtain essential information on scale effects of various parameters such as the loop frictional resistance, thermal center and pipe diameter. The void distribution in a hot-leg, flow regime and natural circulation rate were measured in detail for various conditions. The termination of the natural circulation occurred when there was insufficient hydrostatic head in the downcomer side. The superficial gas velocity at the flow termination could be predicted well by the simple model derived from a force balance between the frictional pressure drop along the loop and the hydrostatic head difference. The bubbly-to-slug flow transition was found to be dependent on axial locations. It turned out that the formation of cap bubbles in the large diameter pipe caused the increased drift velocity, which would affect the prediction of the void fraction in the hot leg.     

Functional and Structural Insights into Human PPARα/δ/γ Subtype Selectivity of Bezafibrate,Fenofibric Acid,and Pemafibrate

Among the agonists against three peroxisome proliferator-activated receptor (PPAR) subtypes, those against PPARα (fibrates) and PPARγ (glitazones) are currently used to treat dyslipidemia and type 2 diabetes, respectively, whereas PPARδ agonists are expected to be the next-generation metabolic disease drug. In addition, some dual/pan PPAR agonists are currently being investigated via clinical trials as one of the first curative drugs against nonalcoholic fatty liver disease (NAFLD). Because PPARα/δ/γ share considerable amino acid identity and three-dimensional structures, especially in ligand-binding domains (LBDs), clinically approved fibrates, such as bezafibrate, fenofibric acid, and pemafibrate, could also act on PPARδ/γ when used as anti-NAFLD drugs. Therefore, this study examined their PPARα/δ/γ selectivity using three independent assays—a dual luciferase-based GAL4 transactivation assay for COS-7 cells, time-resolved fluorescence resonance energy transfer-based coactivator recruitment assay, and circular dichroism spectroscopy-based thermostability assay. Although the efficacy and efficiency highly varied between agonists, assay types, and PPAR subtypes, the three fibrates, except fenofibric acid that did not affect PPARδ-mediated transactivation and coactivator recruitment, activated all PPAR subtypes in those assays. Furthermore, we aimed to obtain cocrystal structures of PPARδ/γ-LBD and the three fibrates via X-ray diffraction and versatile crystallization methods, which we recently used to obtain 34 structures of PPARα-LBD cocrystallized with 17 ligands, including the fibrates. We herein reveal five novel high-resolution structures of PPARδ/γ–bezafibrate, PPARγ–fenofibric acid, and PPARδ/γ–pemafibrate, thereby providing the molecular basis for their application beyond dyslipidemia treatment.     

Over 1100 nm wavelength tuning in superstructure grating (SSG) DBR lasers

A broad range tuning of over 100 nm in tunable DBR lasers with superstructure grating (SSG) is reported. The SSG reflectors for 100 nm were designed and patterned by electron beam lithography. 1.55 mu m DBR lasers with SSG reflector operate in a single mode with a tuning range of 83 nm under CW conditions. With inclusion of the multimode operating region, the tuning range becomes as wide as 103 nm.<>     

Semiconductor pump laser technology

Recent progress in high-power semiconductor lasers for erbium-doped fiber amplifiers is described, focusing on 1.48-μm InGaAsP/InP lasers and 0.98-μm InGaAs/GaAs lasers. The experimental output powers exceed 200 mW (the maximum power was 325 mW) for 1.48-μm lasers, and simulation results indicate that over 400 mW could be obtained by optimizing parameters in strained-layer (SL) multiple-quantum-well (MQW) lasers. Stable operation over a few thousand hours under 100-mW power is demonstrated for liquid-phase-epitaxy-grown lasers, MQW lasers, and SL-MQW lasers grown by all-metal organic vapor-phase epitaxy (MOVPE). For 0.98-μm lasers, improvement in the fiber coupling efficiencies and long-term reliabilities are described. Their power coupled into a single-mode fiber has reached over 100 mW, with coupling efficiencies of approximately 40%. Although reliability seems to be one of the drawbacks compared with 1.48-μm lasers, stable operation for over 10,000 h at 50°C and 30 mW has been reported     

Design and characteristics of InGaAs/InP composite-channel HFET's

A design for composite-channel structures consisting of an InGaAs channel and an InP subchannel for use as heterostructure field-effect transistors is presented for the first time. This novel channel structure takes advantage of both the high drift velocity and low impact ionization of InP at high electric fields as well as the high electron mobility of InGaAs at low electric fields. It is shown that the doping density of the InP subchannel is the key parameter to realize the advantages of the composite channel. A very high transconductance of 1.29 S/mm and a current gain cutoff frequency of 68.7 GHz are achieved with 0.6 and 0.7 /spl mu/m gates, respectively. The average velocity of electrons in the composite channel is 2.9/spl times/10/sup 7/ cm/s. The devices have no kink phenomena in their I-V characteristics possibly due to low impact ionization in the InP subchannel.<>     

A +10-dBm IIP/sub 3/ SiGe mixer with IM/sub 3/ cancellation technique

A third-order intermodulation (IM/sub 3/) cancellation technique using a submixer is proposed for a low-power low-distortion mixer. The IM/sub 3/ cancellation is achieved by summing IM/sub 3/ generated in a main mixer and the submixer, which are almost the same amplitude and opposite phase. The mixer was designed to operate at 870 MHz. The proposed technique reduces IM/sub 3/ by 18 dB with a current increase of about 15% and is suitable for low-power applications. The mixer achieved an input-referred third-order intercept point (IIP/sub 3/) of 10 dBm, a gain of 8.7 dB, and an NF of 9.8 dB and dissipates 30 mW from 2.9 V. The IC is fabricated in a SiGe bipolar transistor with f/sub T/= 30 GHz. The IC occupies 1.44 mm/spl times/1.44 mm.     

Extraction of a plasma time-activity curve from dynamic brain PET images based on independent component analysis

A compartment model has been used for kinetic analysis of dynamic positron emission tomography (PET) data [e.g., 2-deoxy-2-18F-fluoro-D-glucose (FDG)]. The input function of the model [the plasma time-activity curve (pTAC)] was obtained by serial arterial blood sampling. It is of clinical interest to develop a method for PET studies that estimates the pTAC without needing serial arterial blood sampling. For this purpose, we propose a new method to extract the pTAC from the dynamic brain PET images using a modified independent component analysis [extraction of the pTAC using independent component analysis (EPICA). Source codes of EPICA are freely available at http://www5f.biglobe.ne.jp/?kimura/Software/top.html]. EPICA performs the appropriate preprocessing and independent component analysis (ICA) using an objective function that takes the various properties of the pTAC into account. After validation of EPICA by computer simulation, EPICA was applied to human brain FDG-PET studies. The results imply that the EPICA-estimated pTAC was similar to the actual measured pTAC, and that the estimated blood volume image was highly correlated with the blood volume image measured using 15O-CO inhalation. These results demonstrated that EPICA is useful for extracting the pTAC from dynamic PET images without the necessity of serial arterial blood sampling.     

A 32-bit PowerPC system-on-a-chip with support for dynamic voltage scaling and dynamic frequency scaling

A PowerPC system-on-a-chip processor which makes use of dynamic voltage scaling and on-the-fly frequency scaling to adapt to the dynamically changing performance demands and power consumption constraints of high-content, battery powered applications is described. The PowerPC core and caches achieve frequencies as high as 380 MHz at a supply of 1.8 V and active power consumption as low as 53 mW at a supply of 1.0 V. The system executes up to 500 MIPS and can achieve standby power as low as 54 /spl mu/W. Logic supply changes as fast as 10 mV//spl mu/s are supported. A low-voltage PLL supplied by an on-chip regulator, which isolates the clock generator from the variable logic supply, allows the SOC to operate continuously while the logic supply voltage is modified. Hardware accelerators for speech recognition, instruction-stream decompression and cryptography are included in the SOC. The SOC occupies 36 mm/sup 2/ in a 0.18 /spl mu/m, 1.8 V nominal supply, bulk CMOS process.     

Energy-level alignment at model interfaces of organic electroluminescent devices studied by UV photoemission: trend in the deviation from the traditional way of estimating the interfacial electronic structures

The electronic structures of model interfaces of organic electroluminescent (EL) devices were investigated with UV photoemission spectroscopy (UPS). Interfaces of TTN (tetrathianaphthacene) and TCNQ (tetracyanoquinodimethane) were also studied as extreme cases for hole transport and electron transport material, respectively. For all organic/metal interfaces studied, the work function of metal electrode was changed by deposition of organic layer, i.e., the vacuum level was shifted at the interface, indicating the invalidity of the traditional energy level alignment model where a common vacuum level was assumed at organic/metal interface. At TCNQ/Au, DP-NTCI/Al, which are acceptor/metal interfaces, upward shift of the vacuum level of organic layer relative to that of metal was observed, suggesting the formation of interfacial dipole due to electron-transfer from metal to acceptor. At other organic/metal interfaces, TPD(N, N'-diphenyl-N, N'-(3-methylphenyl)-1, 1'-biphenyl-4, 4'-diamine)/Au or ITO (indium tin oxide), ALq/sub 3/ (tris(8-hydroxyquinolino) aluminum)/Al, DP-NTCl(N, N'-diphenyl-1,4,5,8- naphthyltetracarboxylimide)/Al or Au, downward shift of the vacuum level was observed. Such downward shift has been also observed in our previous study for porphyrin/metal interfaces, and seems to be a trend for organic/metal interfaces at which no electron-transfer from metal to organic layer occurs. This trend suggests that the traditional model tends to underestimate (overestimate) the barrier height for hole (electron) injection. On the other hand, the vacuum level shift at ALq/sub 3//TPD interface was less than 0.1 eV, leading to an apparent applicability of the traditional model. However, it is not always the case for organic/organic interfaces: finite shift of 0.2 eV was observed at TTN/TCNQ interface due to electron-transfer from TTN to TCNQ. Possible origins of vacuum level shift at organic/metal interfaces were also discussed.