Continuous and Responsive Motion Generation for Virtual Humans

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K12192
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 62040:Entertainment and game informatics-related
• Research Institution: Kyushu Institute of Technology
• Principal Investigator: Masaki Oshita 九州工業大学, 大学院情報工学研究院, 教授 (20363400)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 動作生成 / 深層学習 / 遷移動作

Outline of Final Research Achievements

In this research project, we worked on the development of a method for generating transitional motions that connect between input motions in order to realize continuous and immediate movement generation for virtual humans. We worked on the development of methods using two approaches: a method for motion generation using foot and hip trajectory data as inputs and outputs for deep learning, and a method for motion generation using posture and motion data as inputs and outputs for deep learning models. Although we were unable to achieve an effective method compared to conventional methods, we developed a crowd simulation method for navigating virtual humans using deep learning with surrounding situation images as input and output.

Free Research Field

コンピュータアニメーション

Academic Significance and Societal Importance of the Research Achievements

深層学習を用いて遷移動作を生成する方法として、研究開発当初に計画していた足や腰の軌道データを深層学習の入出力として動作生成を行う手法と、姿勢・動作データを深層学習モデルの入出力として動作生成を行う方法の、両方のアプローチを示した。また、本研究で開発した深層学習による移動制御手法は、コンピュータアニメーションやメタバースなどでの、群衆のアニメーション生成への応用が期待される。