| 研究課題/領域番号 |
17J10795
|
|---|
| 研究機関 | 早稲田大学 |
|---|
研究代表者 |
Tomo Tito Pradhono 早稲田大学, 理工学術院, 特別研究員(DC2)
|
|---|
| 研究期間 (年度) |
2017-04-26 – 2019-03-31
|
| キーワード | tactile sensing / dexterous manipulation / robot sensing system / skin sensor / wearable sensor |
|---|
| 研究実績の概要 |
1 conference paper (as a co-author) and 2 journals (as a first author) have been published in this fiscal year.
The preliminary result of the first wearable sensor that can measure 3-axis force from human hands has been published at the Robomec 2017.
This device will be used for extracting human manipulation skills and later for teaching a robot.
A soft digital skin sensor (uSkin) for robot hand has been developed, was demonstrated at the ICRA 2017 - Soft Component Technologies Challenge (won a 2nd place award), and has been published at the IEEE RA-letters.
Compared than other sensors, it has a 4 mm thickness, can measure 16 x 3-axis force in 27 x 28 mm area, and requires only 4 wires.
|
| 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
Although 3-axis force data can be extracted from a human finger using a wearable sensor (i.e. while manipulating the stick of wayang puppet), teaching a robot to replicate a similar motion is not possible using the current robot.
This is due to the majority of end-effectors (robot gripper) available in the market are usually do not have a built-in tactile sensor to perform a force control.
Therefore, a tactile sensor for a robot gripper has to be developed first and its development was initially planned to be conducted in the next fiscal year.
However, I realized that integrating these two devices could be a challenging task and would be better to be conducted at the earlier stage so that if some critical adjustments are required, it could be found as soon as possible.
|
| 今後の研究の推進方策 |
Force vector on human fingers generated during an object manipulation will be recorded using a wearable sensor. The recorded human finger motion data will be reproduced using a robot gripper covered with a skin sensor through force control. A machine learning model such as neural network will be utilized to find parameters for mapping the wearable sensor data into robot gripper movements.
A real-time beat tracking system for gamelan music will be developed using HARK-Music to obtain its SIRE (Strength, Intensity, irRegularity, and Extent) representation for a direct mapping of emotion from music to robot movements.
Finally, all results will be submitted to an international journal such as RA-Letters or IEEE Transactions on Robotics and Automation.
|
