著者
高橋 孝 上田 裕子 平野 聡 邑中 雅樹 Runtao Qu 小堀 壮彦 Takahashi Takashi Ueda Hiroko Hirano Satoshi Muranaka Masaki Runtao Qu Kobori Takehiko
出版者
宇宙航空研究開発機構
雑誌
宇宙航空研究開発機構研究開発報告 = JAXA Research and Development Report (ISSN:13491113)
巻号頁・発行日
vol.JAXA-RR-04-017, 2005-01-31

It is aimed that a simulation framework of Spacecraft Simulation Environment (SSE) can be commonly applied not only to Full Software Simulations (FSS) but also to Processor-In-the-Loop Simulations (PILS) and to Hardware-In-the-Loop Simulations (HILS), while various spacecraft simulators are generally tailor-made at individual phases of development. Prior to the actual implementation of SSE, the framework for FSS and PILS was designed, and its advantage to spacecraft simulations using an experimental system was demonstrated. In this study, implemented was an experimental system for PILS, and feasibility of the framework using the system was demonstrated. Real-time tasks working on microITRON communicate with the rest of spacecraft simulator through Java-based middleware Hirano's Object Request Broker (HORB) via distributed communication interfaces (I/Fs) written in Java. These I/Fs are designed to be commonly applied to both FSS and PILS. Also implemented was a tool called Java-microITRON Bridge GENerator (JBGEN) to automatically generate Java-microITRON communication programs from the I/Fs. Furthermore, MemorySaving HORB is developed in order to avoid communication latency.
著者
PALMER Edwina
出版者
International Research Center for Japanese Studies
雑誌
Nichibunken Japan review : journal of the International Research Center for Japanese Studies (ISSN:09150986)
巻号頁・発行日
vol.19, pp.47-75, 2007-01-01

This discussion attempts to reconcile various seemingly contradictory research results regarding the origins of Jōmon Japanese. The focus is on testing Oppenheimer’s theory of Holocene outmigration from the former continent of Sundaland in present-day Southeast Asia against the evidence relating to Jōmon Japan and the “Out of Taiwan” hypothesis for Austronesian language dispersal. It is argued here that postglacial flooding of Sundaland prompted some former inhabitants to migrate from around ten or eleven thousand years ago, and that they followed the expanding belt of lucidophyllous forest, eventually to settle in what is now Japan during the Jōmon Period, in accordance with the theory of regional pockets of “laurilignosa culture.” It is stressed that some of these people were probably speakers of Austronesian languages. Further, it is argued that the “Out of Taiwan” movement of Austronesian language speakers could have occurred later as a migratory counterflow accompanying the Holocene maximum, and that an “Out of Sunda” scenario of migration to Japan in the Jōmon period is not necessarily entirely incompatible with such an “Out of Taiwan” theory.
著者
中島 悠輔 坂内 祐一
雑誌
研究報告グループウェアとネットワークサービス(GN) (ISSN:21888744)
巻号頁・発行日
vol.2020-GN-109, no.29, pp.1-6, 2020-01-16

似顔絵は人物の外見・特徴をとらえて,デフォルメして描いた人物画である.現在,顔画像を似顔絵に変換する研究では,非教師学習を用いた変換手法や,それぞれパーツ毎に変換を行う手法の研究が行われている.しかし,プロのイラストレーターの個性を反映させるような研究は発表されていない.そこで,本研究ではプロのイラストレーターの個性を反映する為に,どのネットワークを用いれば良いか検討を行う.具体的には,pix2pix,CycleGAN ペア,CycleGAN 非ペア,Cyclepix の 4 つのネットワークを検討していく.違いとしては主に損失関数となっている.pix2pix は訓練データと生成データの誤差を取るが,CycleGAN では入力データと生成データを更に変換させた再変換データの誤差を取ることが主な違いである.Cyclepix は両方の誤差を取っている.また,pix2pix,CycleGAN ペアは Discriminator の入力が入力データと生成データペアと条件付けされており,CycleGAN 非ペアと Cyclepix では Discriminator の入力が生成データのみとなっていることも違いである.実験結果として,CycleGAN 非ペアと Cyclepix の評価が高いことが分かった.このことから Discriminator の入力が生成データのみ,Cycle Consistency Loss を利用することで精度が高い似顔絵を生成することに有用である.
著者
杉本 直樹 鶴岡 慶雅
雑誌
ゲームプログラミングワークショップ2018論文集
巻号頁・発行日
vol.2018, pp.114-119, 2018-11-09

実用上の人間の対戦相手としてのゲームAI には、強すぎず弱すぎない人間のプレイヤーと互角の実力が求められる。本研究では、プレイヤーの戦略を動的に推定しそれに応じて対戦AI を変化させ、2人対戦ゲームにおいてゲーム固有の知識を利用することなくAI プレイヤーの実力を人間のプレイヤーと互角となるよう調整し、なおかつそのAI プレイヤーが不自然な振る舞いを見せないようにする事を目的とする。本稿では、研究のため作成したパズルシミュレーションゲームにおいて提案手法によるAI がランダムAI よりも長く試合を続けられる事を示した。
著者
王(木下) 榮(貴雄) Rong(Takao) Wang(KINOSHITA)
出版者
金城学院大学
雑誌
金城学院大学論集. 人文科学編 = Treatises and Studies by the Facalty of Kinjo Gakuin University. Studies in Humanities (ISSN:18800351)
巻号頁・発行日
vol.18, no.2, pp.187-206, 2022-03-31

中国残留孤児をはじめとする中国帰国者が高齢期を迎えた今,差し迫って最大の課題は介護であり,死後の安住地である墓地も大きな問題となっている。また,在住外国人の永住化・定住化に伴って,外国人高齢者は年々増加している。この現状から考えると,この墓地の問題は,異文化の介護問題と同様に,中国帰国高齢者だけの問題に留まらず,日本に暮らすすべての外国人高齢者が共通する問題でもある。さらに,終末期ケアや看取り,弔いについては,信仰や宗教,文化などの違いから考えると,民族や国籍によっては問題の複雑さが窺える。本稿では,中国帰国者の“終の棲家”である墓地から,多文化共生社会の先にある現実問題の一つとして,今後,日本で最期を迎え逝っていく外国人高齢者における異文化としての“終活”問題をどう支えていくのか,その必要性について概観する。
著者
藤井 昇三
出版者
愛知大学東亜同文書院大学記念センター
雑誌
同文書院記念報 (ISSN:21887950)
巻号頁・発行日
vol.20, pp.6-13, 2012-03-30

愛知大学東亜同文書院大学記念センター国際シンポジウムテーマ「辛亥革命・孫文・東亜同文会」日時 2011年11月12日(土)10時30分~17時30分場所 愛知大学豊橋校舎記念会館3階小講堂
著者
小坂 悠登 伊藤 毅志
雑誌
研究報告ゲーム情報学(GI) (ISSN:21888736)
巻号頁・発行日
vol.2019-GI-41, no.15, pp.1-7, 2019-03-01

本研究では従来チェスや将棋.オセロといった二人完全情報確定零和ゲームで行われてきたゲームAIを利用した棋力推定のアプローチを囲碁に応用する手法について検討した.他のゲームと囲碁AIの大きな違いはベースとなっているアルゴリズムがミニマックス探索ではなく,モンテカルロ木探索(MCTS)である点である.そのため,従来の手法で用いられてきたミニマックス探索でいうところの「評価値」の代わりに,本研究ではMCTSが導き出す局面評価に相当する概念である「勝率」という指標を用いることにした.棋譜の評価においては,MCTSのみ用いた囲碁プログラムと,Alpha Goの探索手法に倣ってDNN(Deep Neural Network)を用いた囲碁プラグラムの両者を用いて実験結果の比較検証を行った.さらに,これらの手法を合議で組み合わせることによって,さらに飛躍的に棋力推定の精度を高めることができた.
著者
津田 徹英
雑誌
美術研究 = The bijutsu kenkiu : the journal of art studies
巻号頁・発行日
no.408, pp.1-94, 2013-01-18

In the 10th month of 1295, 33 years after Shinran’s death, his great-grandson Kakunyô (1270–1351) created the first illustrated biography of Shinran (1173–1263), the founder of the Jôdô Shinshû Buddhist sect. Immediately after this first production, Kakunyô made changes to the painting scrolls, and the extant Rin’a version records his involvement and the production date of the 10th month of 1295 in its colophon. Similarly, other extant versions exist today, namely the Takada version which was completed in the 12th month of 1295, the Kôei version which dates to 1343, the Shôganji version dated to 1344, the Gugan version dated to 1346 and the Josenbô version dated to 1360. All of these examples were created in handscroll form, as was the first version, contrary to the later trend to create Shinran illustrated biographies in hanging scroll format. Of particular note amongst all these is the Bukkôji version (proper name Zenshin Shônin Shinranden-e), which has been handed down at Bukkôji, the main temple of the Bukkôji school of the Jôdo Shinshû sect. The Bukkôji version consists of two handscrolls, and until the modern era there were only limited opportunities for the public to see the work, as it was carefully preserved and handed on at the temple. This careful handling has meant that the scrolls maintained their original bright colors, with almost no oxidation of the silver used and with almost no conservation work seen on the painting papers. Unfortunately, there are no extant documents directly related to the production of the Bukkôji version. The bright colors apparent on the scrolls have meant that previous scholars have considered the work to be younger than it is, with some scholars considering it to date from the 15th century, and some even to the 17th century. And yet, a careful re-examination of the Bukkôji version reveals that both the explanatory text and the paintings were reated at the same time, and that the explanatory texts are thought to be in imitation of the calligraphy style of Emperor Fushimi (1265–1317, reigned 1287–1298), known for its combined use of Chinese characters and Japanese syllabary. According to the Masukagami completed in the 14th century and the Shuboku-sho of Prince Son’en (1298–1356), it is known that the calligraphy style of Emperor Fushimi was highly regarded as the standard for the calligraphy of the day. Thus the explanatory texts of the Bukkôji version appear to have been created during a period when Emperor Fushimi’s calligraphy was deemed most important, in other words, sometime within the first century after his death. Further, the calligraphy on the Bukkôji version texts is particularly well written. Clearly a calligrapher of this talent level would have created the explanatory texts for other illustrated handscrolls. An examination of other extant medieval period illustrated handscrolls indicates that the same calligrapher was involved in the explanatory texts on scrolls 4, 8, and 9, and the colophon on scroll 10 of the Konrenji version of the Yûgyô shonin-engi-e. Further, the same calligrapher wrote the texts for scrolls 4 and 5 of the Tôji version of the Kôbô Daishi gyôjô-e. The latter work is known to have been created around 1374–78. In addition, the old records that accompany the latter handscrolls attribute the texts on scrolls 4 and 5 to “Go-Oshinokoji, the Minister of the Interior.” The question then arises, who was this Go-Oshinokoji former Naidaijin (the Minister of the Interior)? That title refers to Sanjô Kintada (1324–1383) who is known to have been the calligrapher of the 2nd volume of the Boki-e, thought to have been completed in 1351. A comparison of both the Chinese characters and the Japanese syllabary found in the Bukkôji version, the Konrenji version of the Yûgyô shonin Engi-e, the Tôji version of the Kôbô Daishi gyôjô-e and the Bukkoji version indicates that all of these works were written by the same hand. Given that the text and the paintings of the Bukkôji version were both created at the same time, and the fact that the calligrapher for the texts was Sanjô Kintada, then it is certain that the production of the Bukkôji version dates back to the 14th century. Given this dating, the paintings of the Bukkôji version must be reconsidered. In this reconsideration the author took note of the Josenbô version of the illustrated biography of Shinran created in 1360. There are many instances of the same motifs used in the various scenes on the Josenbô and the Bukkôji versions. This Josenbô version’s overall composition was based on the earlier Kôei version, and while the Josenbô version used the Kôei version as its model, it also incorporated some of the elements and scenes from the Bukkôji version. Thus, the Bukkôji version can be seen to pre-date the Josenbô version known to have been painted in 1360. Given that the year 1361 marked the centenary of Shinran’s death, the actual production of buildings for use in the centenary commemoration would have been carried out at that time, with a five-day period of dedication rituals for the new structures recorded to have been held in the 3rd month of 1360. It can then be surmised that the Bukkôji version was also created during this period. Thus the Bukkôji version creation has as its terminus the 1360 creation of the Josenbô version, and it would seem appropriate that it predated that time to some degree. Five examples of illustrated handscrolls with texts written by Sanjô Kintada are known, including the four extant works and one known through documents. Thus Kintada must be considered in the role of calligrapher for illustrated handscrolls when we consider the calligraphers of medieval period handscrolls. Of the five projects he is known to have worked on, including the Bukkôji version, both the Boki-e and the Kôbô Daishi Gyôjô-e were painted by the head of the imperial court’s painting studio. The painting skills and technical standard of such a studio head would have been commensurate with the high ranking aristocratic calligrapher, Sanjô Kintada, chosen for the accompanying texts. Thus, it can be assumed that the painter and painting studio who brushed the Bukkôji version would have been a painter whose rank, skills and technical abilities were commensurate with the high ranking aristocrat Sanjô Kintada who wrote that version’s texts. The depiction in the Bukkôji version has previously been noted as “vulgar or coarse.” However, what has been overlooked about this painting style is the scene of Shinran’s cremation, where the dense foliage of the trees and the snow-dusted distant mountain views can be seen as sourced from the scenery depiction of the lightly snow-dusted Mt. Mikasa in the Kasuga Gongen-Genki-e of 1309. Similarly, the intricate ink-line depiction of the waves on the surface of Lake Ashino in Hakone seems to be sourced in the same 1309 work. The figural depiction, such as the facial expressions of the manifestation of the Kumano deity, seen with Shinran in the Kumano Shrine building, along with their garments have been extremely intricately colored in fully elegant form. The painter of these elements, and his affiliated painting studio, clearly must be seen as having had considerable talent and technical ability. What must not be overlooked as well is the fact that somewhat earlier than the 1360 date of the Bukkôji version, the Bukkôji sect regularly commissioned paintings from the painting studio affiliated with the Gion Shrine. The chief painter of the imperial court’s painting studio would have been chosen from the Gion Shrine-affiliated painting studio, and in addition to the creation of the Boki-e for which Sanjô Kintada wrote the texts, that studio would have also created a set of handscrolls, the Suwa Daimyôjin-ekotoba (scattered and now lost), whose text written by high ranking nobles in 1356. The Gion Shrine-affiliated painting studio can thus be considered to be a painting studio of sufficient rank and talent to match the calligraphy commissioned from Sanjô Kintada. Thus, confirming the 14th century date of the Bukkôji version will have an important influence upon the study of the medieval handscrolls. This new dating is related to the understanding of handscroll painting style in the latter half of the 14th century. The basis for the previous ca. 15th century attribution of the Bukkôji version can be found in its so called “clear and strongly vulgar or coarse painting style,” but in fact it must be noted that this painting style itself was one of the trends of the latter half of the 14th century. The awareness of this trend suggests the necessity for a re-evaluation of the dating of the handscrolls attributed to the 15th century on the basis of this style. The confirmation of the 14th century date for the Bukkôji version is particularly meaningful for this stylistic matter.
著者
浅井 悠佑 青木 俊介 米澤 拓郎 河口 信夫
雑誌
研究報告ユビキタスコンピューティングシステム(UBI) (ISSN:21888698)
巻号頁・発行日
vol.2022-UBI-75, no.21, pp.1-8, 2022-08-29

コンピューティング技術や深層学習技術の発達に伴い,自走式配送ロボット (AMR) の利用が進んでいる.産業面での利用も加速しており,物流倉庫においては配送ロボットがピッキング作業の一部を担うなど,オートメーション化の流れは必至であると言える.しかし,物流倉庫には棚や商品が所狭しと置かれており,ナビゲーションをする際には多くの死角が存在する.従来の AMR は搭載センサのみを用いてナビゲーションするため,死角の多い環境においては,衝突の危険性がある.死角となりうる箇所を対象に見えない障害物の動きを対象とした手法が提案されているが,死角の場所が建物に依存している場合や実験室環境もしくは簡易なシミュレーションによる検証のみにとどまっている場合が多い.そこで本研究では,これらの問題を解決するために実環境でのセンシングを行い,センシング情報を配送ロボットナビゲーションに利用する仕組みを提案する.死角の位置の変化を念頭に,時間変化を考慮したコストマップ構築手法を提案し,実際の物流倉庫を模したシミュレーション環境において実験,評価を行う.
著者
葛西 和真 阿部 昭博 市川 尚 富澤 浩樹
雑誌
研究報告ユビキタスコンピューティングシステム(UBI) (ISSN:21888698)
巻号頁・発行日
vol.2022-UBI-75, no.14, pp.1-7, 2022-08-29

高齢者の健康維持のためには日常的に身体活動を行うことが効果的とされているが,現状においては高齢者のおよそ半数は厚生労働省の身体活動量基準を達成できていないとの指摘もある.また,高齢者に焦点を当てて身体活動促進の介入を行う研究は様々あるが,デジタル技術の介入を伴う研究はこれまであまり議論が進んでいなかった.本研究では,高齢者に対する意向調査を踏まえ,ウェアラブルデバイス及びモバイル端末を活用し,デジタル技術の介入を伴う高齢者身体活動支援システムの提案を行う.