| Arno in Wolde Luebke | 1151205: M, 2回目発表 | 加藤 博一, 小笠原 司, 武富 貴史, 宮崎 純 |
| 坂東 直人 | 1251202: M, 1回目発表 | |
|
title: Evaluating a combined head direction and 2D finger position selection method for AR space in HMD environments
abstract: In this research, I propose a selection method using a combination of head direction and two-dimensional (2D) finger position for Augmented Reality (AR) space when using a head-mounted display (HMD). In related research works, the methods using: 3D finger positions, dwell time of pointing device, and external controllers have been proposed. In contrast, my proposed method implements selection when a head direction and a 2D finger position overlap with each other on a target. In this research, I will perform comparisons between different selection techniques in various selection tasks that are conventionally used for this study. Moreover, I will clarify the characteristics of each method. In this presentation, I will explain the details of my current prototype system. language of the presentation: Japanese 発表題目: HMD利用時のAR環境における頭部方向と指の二次元位置に基づくセレクション手法の提案と評価 発表概要: 本研究ではヘッドマウントディスプレイ(HMD)利用時におけるセレクション手法として、頭部方向と指の二次元位置を併用する手法を提案する。これまでにHMD利用時のセレクション手法として、指の三次元位置を利用する手法やポインティングデバイスの滞在時間を利用する手法、外部コントローラを利用する手法が提案されている。これに対し、本手法は、選択対象物に頭部方向と指の二次元位置を重ねた瞬間にセレクションが行われるという特徴を持つ。本研究では様々なセレクションタスクにおいて従来手法と提案手法を用いた場合の比較を行い、それぞれの手法の特性を明らかにする。本発表では、現在作成しているプロトタイプシステムの詳細を説明する。 | ||
| Zhang Qirui | 1251212: M, 1回目発表 | |
|
title: An estimation algorithm for 3D-structure from multi-view frames abstract: The direction of my research is to discuss on algorithms which simultaneously recover 3D-strutures of objects and camera extrinsic parameters from multi-view frames. Researches in this area can help the development of applications from various fields, such as augmented reality, medical precise measurement, SLAM of robots, etc., Currently I am trying to make improvements on DTAM(dense tracking and mapping in real time) algorithm. In the presentation I will briefly introduce the general idea about DTAM and some related important algorithms for pose estimation and image denoising. language of the presentation: English | ||
| Alhumaidan Abdullah Emad I | 1251204: M, 1回目発表 | |
|
Title: Optimization for energy, power and time for OLAP queries
Abstract: In this research, I will talk mainly about three papers, which are related to my work. The first paper shows optimization algorithms and other optimization approaches, such as histograms and dynamic programming. The former is used as a cost estimation technique, and the latter is used to construct plans. The second paper is about a proposed data storage model called "Fractured Mirrors", this model uses a mirroring scheme to store data as well as using both DSM and NSM to gain the best aspects of both models, this paper helps me to approach the power-aware DBMS by using a similar mirroring scheme. The third paper, uses TPC-C and TPC-H Benchmarks to conduct experiments on a proposed power-aware query optimizer (i.e. a query optimizer that considers power cost during the search for query plans), which is designed for the purposes of reducing power & energy consumption in OLAP & OLTP systems, however, OLAP system reached only 3.3% energy reduction in this research. For this reason, I aim to get higher percentage (higher than 20% would be considered a fair percentage) by using histograms, dynamic programming and Fractured Mirrors as approaches. Language of the presentation: English | ||
| 下浦 啓 | 1251201: M, 1回目発表 |
|
title: Control of preferences of agents who form social networks
abstract: Recently, there are several studies concerned with economical or social probrems whose mutual relationsips of mutliagent. For modelling these probrems, there is a method using social networks. In this presentation, I'll introduce a working paper about competitive firms' optimal advatisement policy to control preferences for products that agents who form social networks have by computing optimization problems. language of the presentation: Japanese | |
| 本田 謙一郎 | 1251203: M, 1回目発表 |
|
Title: Simulation of metabolic systems in 1-butanol producing Escherichia coli
Abstract: Recently, it is reported that a genetically engineered Escherichia coli produces 1-butanol; a compound that can use as a biofuel. Analysis of detail metabolic dynamics in this 1-butanol producing E. coli will contribute improvement of biofuel production. However, it is still difficult to optimize entire metabolic pathways because it involves various metabolic reactions that depend on each other. Experimental metabolomics data of 1-butanol producing E. coli is based on this simulation and this study therefore aims to develop a computational model of metabolic systems for 1-butanol production in E. coli. The simulation procedure implemented constitutes 5 steps: 1) Construction of the metabolic pathway model, 2) Estimation of stable flux using the Flux Balance Analysis (FBA), 3) Construction of the dynamics model, 4) Optimization of parameters using an Evolutionary Algorithm (EA) and 5) Numerical simulation of the dynamics under genetic/environmental perturbations. By comparing experimental metabolomics data and the prediction by the model, this simulation evaluates the observed data from the viewpoint of physical constraint. This simulation also shows each parameters’ bounds in stability-steady states. By Numerical simulation, when initial point environs are changed, this simulation will shows each metabolic reaction’s changing and point to better pathway of producing 1-butanol. language of the presentation: Japanease | |
| 増田 有利子 | 1251094: M, 1回目発表 |
|
title:
A control scheme for the manipulation of a particle by means of optical tweezers abstract: Catching the subject of nano metric size from micronano meter by the radiation pressure of light, the photo pincette which can be operated is the only technique of the ability to move and position a subject with the accuracy of position of a nanometer according to non-contact. As a method of controlling at the time of moving the particulates dispersed in the viscous medium with a photo pincette, what can respond also to the influence (ambient noise by Brownian motion, such as disturbance and vibration of a system) from the external world is called for. There is a paper which asked for the easy control rule of the photo pincette until now, and the influence of Brownian motion does not take into consideration there, About the model when photo pincette operation of the particles of a size to the extent that only an equation of motion can describe was carried out, feedback control was drawn, the valuation function was defined, and it is judged whether the control rule is stable. Stability analysis of the control system was conducted by the Liapounov stability theory. Furthermore, it indicates using the numeric simulation that the obtained closed-loop system is robustly [ even when complicated thermal noise exists ] stable. By this announcement, while introducing this paper, it argues also about anticipation when ambient noise is taken into consideration as future research. language of the presentation: Japanese 発表題目: 光ピンセットによる粒子の操作における制御方法 発表概要: 光の放射圧でマイクロメートルからナノメートルサイズの対象物を捕捉し、操作できる光ピンセットは、対象物を非接触でナノメートルの位置精度で移動・位置決めできる唯一の手法である。粘性媒質中に分散した微粒子を光ピンセットで移動させる際の制御法として、外界からの影響(ブラウン運動による外乱やシステムの振動等の環境雑音)にも対応できるものが求められる。これまでに光ピンセットの簡単な制御則を求めた論文があり、そこではブラウン運動の影響は考慮せず、運動方程式のみで記述できるほどの大きさの粒子を光ピンセット操作したときのモデルについて、フィードバック制御を導き、評価関数を定め、その制御則が安定かどうかを判定している。制御システムの安定解析はリアプノフ安定理論によって行われた。さらに、数値シミュレーションを用いて、得られた閉ループシステムは複雑な熱雑音が存在するときでもロバスト安定だと示されている。本発表では、本論文を紹介するとともに、今後の研究として環境雑音を考慮したときの予想についても議論する。 | |
| Koganti Nishanth | 1251209: M, 1回目発表 |
|
Title: Reinforcement Learning of Clothing Assistance using database of initial trajectories
Abstract: Robotic clothing assistance is a challenging problem as the robot has to interact with non-rigid clothing material and adapt to changes in the posture of assisted person. Tamei et. al.(2011) were successful in developing a Reinforcement Learning based clothing assistance robot that overcomes these difficulties. In this framework, the subject is a mannequin and the clothing task is performed by a dual-arm robot. However there were problems faced with this framework, presently the human-cloth relationship is observed using complex and expensive motion capture system and also the robot can not handle large variations in posture. My proposed research aims at addressing these problems with the framework: 1) Develop method to estimate human-cloth relationship using low cost depth sensor, 2) Create database of human demonstrated trajectories for different settings in the environment, 3) Develop method to choose most suitable trajectory from database as initial control policy. My present research is on the estimation of human-cloth relationship using depth sensor. To ensure fast learning time, the relationship is represented using low dimensional topology coordinates. In our proposed method, the topology coordinates are estimated by obtaining the T-shirt state and mannequin's posture using a depth sensor. Language of the presentation: English | |
| 東 結香 | 1161026: D, 中間発表 | 藤川 和利, 山口 英, 猪俣 敦夫 |
| 藤原 雄介 | 1151204: M, 2回目発表 | 山口 英, 藤川 和利, 門林 雄基 |
| 楊欣 | 1151215: M, 2回目発表 | 飯田 元, 松本 健一, 市川 昊平, Ana Erika CAMARGO CRUZ |