| ANAROSSI EDGAR | D, 中間発表 | ロボットラーニング | 松原 崇充, | 和田 隆広, | 柴田 一騎, | 鶴峯 義久, | 佐々木 光, | 権 裕煥 |
|
title: Composing Complex Motions by Decomposed Queries for LLMs
abstract: With the growing accessibility of technology and the expanding volume of data, artificial intelligence (AI) is achieving unprecedented levels of performance. This progress has led to massive benefits in several domains, driven by both enhanced research and the abundance of data. In contrast, sectors where data is costly or challenging to acquire, such as the robotics domain, have not benefited equally from these advancements. To overcome these challenges, these sectors should aim to harness widely accessible data alongside their specialized datasets. In this study, we introduce LLM-DMPs, which leverage the understanding capabilities of Large Language Models (LLMs) to break down complex user queries into manageable answers for Dynamic Movement Primitives (DMPs). We evaluated our method on both an 'object cutting' task and a 'cake icing' task, where it successfully demonstrated high levels of generalizability with minimal data collection. language of the presentation: English | ||||||||
| BERNARD TIONG ING SHENG | M, 2回目発表 | ロボットラーニング | 松原 崇充, | 和田 隆広, | 鶴峯 義久, | 佐々木 光 | ||
|
title: Cooperative Grasping and Transportation Using Decentralized Multi-agent Reinforcement Learning with Ternary Force Representation
abstract: This study addresses the challenge of cooperative grasping and transportation using decentralized multi-agent reinforcement learning with force-sensing feedback. In decentralized systems, achieving "action synchronization" is particularly challenging, making coordination difficult. Additionally, force sensors often exhibit discrepancies between simulations, where policies are trained, and real-world environments, where policies are deployed. To address synchronization problem, we focus on reward engineering to enhance coordination by stipulating that each agent must achieve a state in which all agents must grasp the object as a prerequisite for receiving any reward. To bridge the force sensor gap, we propose a ternary discretization of object motion derived from force sensor readings. This simple representation ensures consistent sensor data extraction across both environments. language of the presentation:English | ||||||||
| SHAHZADI ANAM | M, 1回目発表 | 光メディアインタフェース | 向川 康博, | 松原 崇充, | 舩冨 卓哉, | 藤村 友貴, | 北野 和哉 | |
|
title: AI-Driven Player Shot Recognition for Advanced Badminton Game Analysis
abstract: Accurately recognizing player shots in badminton is crucial for understanding player performance, strengths, and weaknesses. This research aims to automate and enhance badminton match analysis using AI-driven techniques. Current methods struggle with manual tagging and lack efficiency. By leveraging rally segmentation, hit detection, and player shot recognition, this study provides a unified framework for holistic game understanding. Integrating skeleton-based approaches and transformer models, we aim for an advanced system that combines these technologies for improved accuracy and contextual analysis. This approach is expected to significantly enhance the comprehensiveness and reliability of badminton game analysis. language of the presentation: English | ||||||||
| 松田 頼子 | D, 中間発表 | サイバネティクス・リアリティ工学 | 清川 清, | 荒牧 英治, | 内山 英昭, | Perusquia Hernandez Monica, | 平尾 悠太朗 |
|
title: Drama Therapy Research using Virtual Reality -Examination and Analysis of Session Content-
language of the presentation: Japanese 発表題目: バーチャルリアリティ技術を活用したドラマセラピーに関する研究 ~セッション内容の検討と分析~ 発表概要: ドラマセラピーは，即興劇で他者との関わり方を体得する集団精神療法である.しかし，ドラマセラピーには，セラピスト不足・演技への羞恥心・開催場所の制約等の課題がある.そのため，XR空間でのAIセラピストによるドラマセラピー(XRDT)開発を目指す.まず，VRChat®内の自作ワールドにてXRDTセッション内容の検討を行った.アンケートとインタビュの内容分析から，被験者が他者を慮り自らの考えを柔和させる認知的共感の向上が見られ，XRDTセッションの有効性が示された. 今後は，実空間にくらべVR空間で不足する非言語情報に焦点を当て，どのような情報が利用者の共感力向上に繋がるのかを検証する予定である． | |||||||
| ROCHA CONFESSOR BRIAN | M, 2回目発表 | サイバネティクス・リアリティ工学 | 清川 清, | 荒牧 英治, | 内山 英昭, | Perusquia Hernandez Monica, | 平尾 悠太朗 |
|
Title: Therapeutic Applications of Video Games on Hikikomori resocialization
Abstract: The issue of extreme social isolation in Japan, known as hikikomori, has concerned the Japanese government for decades, and video-game have been suggested as a promising approach to help affected individuals return to a more social lifestyle. Despite this, what kind of game genres and elements can be included in such interventions has not been thoroughly explored in past research. In light of this, the current research aimed to use surveys to investigate game genres and elements that appeal to hikikomori audiences, and from the results of these surveys, proposed promising game elements and genres that can be implemented in game-based approaches for this audience. Finally, this study used the proposed game elements to develop an experiment system in the form of a mobile video game, that may serve as a foundation for future studies on the topic. language of the presentation: English | |||||||
| 新宅 慶騎 | M, 1回目発表 | 光メディアインタフェース | 向川 康博, | 清川 清, | 舩冨 卓哉, | 藤村 友貴, | 北野 和哉 |
|
title:
Study on 3D Printer Configuration Methods Using Light Fields
abstract: Most current 3D printers output three-dimensional objects by repeating one- or two-dimensional unit operations. However, it has been pointed out that various problems arise due to the layer-by-layer construction. Computed axial lithography (CAL) has been proposed as a method that can solve these problems. CAL selectively solidifies photosensitive resin by irradiating it with light using a method inspired by computed tomography. Nevertheless, it is thought that CAL also has problems due to the method of irradiating the light. This problem may be solved by using a light field display to irradiate the light. language of the presentation: Japanese 発表題目: ライトフィールドを応用した3Dプリンタ構成方法の検討 発表概要: 現在普及している3Dプリンタは、1次元または2次元の単位操作を繰り返して出力する方式が主流であるが、立体物を1層ずつ作成することに起因する様々な問題が生じることが指摘されている。この問題を解決し得る手法として、コンピューター断層撮影にヒントを得た方法により光線を照射することで感光性樹脂を選択的に固化するcomputed axial lithography (CAL) が提案されている。しかしながら、CALにも光線の照射方法に起因する問題が存在すると考えられる。この問題は、光線の照射にライトフィールドディスプレイを使用することで解消可能である可能性がある。 | |||||||