| 研究課題/領域番号 |
25280096
|
|---|
| 研究種目 |
基盤研究(B)
|
| 研究機関 | 独立行政法人産業技術総合研究所 |
|---|
研究代表者 |
KHEDDAR Abder 独立行政法人産業技術総合研究所, 知能システム研究部門, 研究員 (90572082)
|
|---|
| 研究分担者 |
吉田 英一 独立行政法人産業技術総合研究所, 知能システム研究部門, 連携研究体長 (30358329)
金子 健二 独立行政法人産業技術総合研究所, 知能システム研究部門, 上級主任研究員 (10356800)
|
|---|
| 研究期間 (年度) |
2013-04-01 – 2016-03-31
|
| キーワード | デジタルヒューマンモデル / ダイナミクス / 最適化 / 機械学習 |
|---|
| 研究概要 |
According to our yearly plans, we achieved in the first year the following results.
-Richer contact models in the multi-contact posture generation (PG): We reformulate the problem of contact search as the existence of a floating ellipse inside both bodies contact surfaces, once they have been selected by the planner. This allows for contacts where surfaces intersect in any kind of way, which in turn gives much more possibilities for the planner to find a new contact.
-Multi-contact motion planning and control with sliding contacts: For the first year we only started theoretical formulation of the problem with very preliminary simulation.
-Automatic contact supports construction from point-cloud data: We proposed a scheme to extract the information relevant for planning from an acquired point cloud acquired using an embedded camera together with a depth map sensor. This yielded the basis of an egocentric on-the-fly multi-contact planner.
-Multi-contact stabilizer: Using a sequence of contact stances from our offline multi-contact planner, we use first a Model Predictive Controller to generate a suitable dynamic trajectory then a whole-body optimization-based controller to track it at best. Relatively to existing work, we allow frame changes of the preferred force, provide a good heuristic to automatically compute the timing of the transition from purely geometrical features. This allows us to generate a wide range of 3D motions, while accounting for predictable external forces, which includes picking-up objects.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
For research plan for H25, we have achieved the expected result on "Richer contact models", "Automatic contact supports construction", and "Multi-contact stabilizer". For "Multi-contact motion planning and control with sliding contacts" we did a preliminary research we will undergo deeper studies on this topic in H26. We estimate that the project is making progress as planned.
|
| 今後の研究の推進方策 |
Based on the solid results obtained in H25, we will continue investigation planned for H26, such advanced topics as advanced search in contact support, planning parallelization, machine learning for contact support and dynamic multi-contact planning, as well as handling of slipping contact and human observation.
|
| 次年度の研究費の使用計画 |
学会参加や海外との研究協力で旅費を少し多く見積もっていたが、研究分担者等の都合によりそのための出張が予想より少なくなり、また力センサ類などの設備導入において希望する仕様を持つ製品の選定が終了せず、次年度に使用することとなったため。
全年度より選定を進めている力センサ類などの設備の購入、また必要に応じて海外連携研究者との協力のための海外出張などに使用する。
|
