| Project/Area Number |
18H03313
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 61050:Intelligent robotics-related
|
|---|
| Research Institution | Ritsumeikan University |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
野間 春生 立命館大学, 情報理工学部, 教授 (00374108)
李 周浩 立命館大学, 情報理工学部, 教授 (80366434)
松尾 直志 立命館大学, 情報理工学部, 助教 (80449545)
|
|---|
| Project Period (FY) |
2018-04-01 – 2021-03-31
|
| Project Status |
Completed (Fiscal Year 2021)
|
| Budget Amount *help |
¥15,990,000 (Direct Cost: ¥12,300,000、Indirect Cost: ¥3,690,000)
Fiscal Year 2020: ¥3,510,000 (Direct Cost: ¥2,700,000、Indirect Cost: ¥810,000)
Fiscal Year 2019: ¥5,330,000 (Direct Cost: ¥4,100,000、Indirect Cost: ¥1,230,000)
Fiscal Year 2018: ¥7,150,000 (Direct Cost: ¥5,500,000、Indirect Cost: ¥1,650,000)
|
|---|
| Keywords | 深層学習 / プロセスモデル / 把持パターン想起 / 視触覚統合 / 多指ハンド / 動作生成 / 他指ハンド |
|---|
| Outline of Final Research Achievements |
We have developed a framework for automatic process modeling of object manipulation by robotic hand that integrates tactile and visual deformation information in a machine learning framework. We proposed (1) a transitional description model of 3-D shape of the manipulated object that sequentially considers uncertain and imcomplete observations, (2) a ROS2-based distributed system for logging human activities in daily life, (3) a method for creating a deep model that recalls the grasping posture based on the object shape and appearance, (4) a method for creating a deep model that recursively recalls the next operation to be taken to achieve a goal based on the current object state, and (5) Development of an ultra-compact tactile sensor and classification of tracing tactile sensations based on dynamical tactile information.
|
| Academic Significance and Societal Importance of the Research Achievements |
人が日常普通に行う道具の操り操作は現状のロボットにはまだ模倣ができない。その困難を視触覚の統合によるロボット制御によって克服するべく、触覚側では素材識別のためのなぞり行動を素材適応的に生成することで識別性能を向上できる可能性を示し、視覚側では物体形状の記述子、把持姿勢との関連モデリングによる把持想起、巨視的手順のモデリングと模倣といった要素技術を獲得した。これにより本課題の本丸である視触覚統合モデリング課題への取り組みが端緒についた。
|
