| 研究課題/領域番号 |
25280096
|
|---|
| 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, 国際客員研究員 (90572082)
|
|---|
| 研究分担者 |
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)
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)
金広 文男 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, 研究グループ長 (70356806)
|
|---|
| 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 |
|---|
| 研究実績の概要 |
According to our yearly plans, we promised to achieve in the second year the following research topics.
-Workpackae 1: in the second year we started working in the theoretical formulation of the problem of advanced search in contact support on deformable support. We will undergo measurement of human walking on mattresses and try integration to the planner of this topic in H27. We also implemented the free motion whole body multi-contact formulation as a semi-infinite problem implemented on a GPGPU. Moreover, we highlight an issue encountered when using parameterized functions in the presence of equality constraints that need to be satisfied continuously over a time interval. We then propose a parameterization of contact forces and a formulation of the constraints, which allow writing a tractable optimization program whose solution verifies at any instant all the constraints where those forces appear.
-Workpackage 2: we still did not got publishable/finalized findings and are working on using machine learning techniques in multi-contact planning. It consists mainly in using Baysian transition planning from a set of multi-contact sets to be learned from encountered situations.
-Workpackage 3: We did partial human motion monitoring in contact motion, using force sensors that we are exploring currently. As an integrated scenario, we have achieved vertical ladder climbing and car driving tasks.
|
| 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 H26, we have mainly made advancement on "MCP parallelization with multi-threading or general-purpose graphic processing units (GPGPU) porting" (workpackage 1) and "Assessments through design scenarios" (workpackage 3) which lead to several publications in a journal with impact factor (conditionally accepted) adn in conferences with review. This shows that the research advances as planned. We will mainly address workpackage 2 for further combination of contact-invariant planning with local QP (quadratic programming) controller as these were two separate modules.
|
| 今後の研究の推進方策 |
In H27 we will continue research on contact planning for the particular cases of sliding and deformable terrains. We are investigating contact on deformable by doing investigations using human motion capture. Once done, we will achieve contact transitions tests on mattresses. For WP2, we will investigate further combination of contact-invariant planning with local QP controller as these were two separate modules. In WP3 we have bought force-sensing devices to investigate contact inverse problems in multi-contact grasping that will path the way for further extensions concerning whole-body multi-contact motion. We will monitor interaction forces when human manipulate various objects and try to identify inverse optimal problem of the motions that are generated. By the end H27 of this term, we will assess the results obtained in this project with challenging scenarios that can be used in planning complex multi-contact general-purpose tasks. Our planner will be confronted to the following challenges: ingress/egress utility vehicles or personal cars, climbing different ladders in different conditions (this is already started in H26) and extremely cumbersome environments and narrow passages. These tasks must be planned for different size of humans or different humanoid robots. We are planning to experience some of them on real humanoid platforms (HRP-2 and HRP-4). Our group is also participating to the DARPA challenge with the AIST-NEDO team where these results will be exhibited in the driving utility-car and climbing stairs scenarios.
|
| 次年度使用額が生じた理由 |
今年度導入した力センサシステムが想定より低い価格で導入できたため。
|
| 次年度使用額の使用計画 |
今年度は、構築した手法の実験的な検証に必要な環境整備、ロボットの保守、また成果発表のための外国出張などに使用する。
|
