| 研究課題/領域番号 |
23K03790
|
|---|
| 研究種目 |
基盤研究(C)
|
| 配分区分 | 基金 |
|---|
| 応募区分 | 一般 |
|---|
| 審査区分 |
小区分21010:電力工学関連
|
|---|
| 研究機関 | 長岡技術科学大学 |
|---|
研究代表者 |
TRAN PHUONGTHAO 長岡技術科学大学, 工学研究科, 助教 (20848923)
|
|---|
| 研究期間 (年度) |
2023-04-01 – 2027-03-31
|
| 研究課題ステータス |
交付 (2023年度)
|
| 配分額 *注記 |
4,680千円 (直接経費: 3,600千円、間接経費: 1,080千円)
2026年度: 780千円 (直接経費: 600千円、間接経費: 180千円)
2025年度: 1,300千円 (直接経費: 1,000千円、間接経費: 300千円)
2024年度: 1,430千円 (直接経費: 1,100千円、間接経費: 330千円)
2023年度: 1,170千円 (直接経費: 900千円、間接経費: 270千円)
|
|---|
| キーワード | Motion control / Force control / Intelligent sensing / Deep learning / Intelligent control / Robot / Intelligent robotics |
|---|
| 研究開始時の研究の概要 |
To build the intelligent human-centered robot system, multiple advanced AI functions of sensing technologies and control strategies are developed. AI virtual force sensor, AI safe interaction, and AI non-contact/contact interaction are developed as fully integrated AI functions driving the system.
|
| 研究実績の概要 |
The purpose of this research is to develop a human-centered intelligent robot system based on the fusion of advanced AI technologies of sensing and control to realize the safe and reliable human-robot interaction. To achieve this purpose, this research has been developing the AI virtual force sensor to replicate the force sensing behavior of a high-quality force sensor to realize high sensitivity, superior precision and wide bandwidth, with effective noise suppression. The deep learning algorithm with a long short-term memory neural network is applied to model the reaction force observer for highly precise replication of the measuring performance of a force sensor. The information from force sensor, motor position, and motor current reference is used as the training data of the deep neural network, while only position and current reference information of the motor is employed, excluding the necessity of a force sensor, when implementing the proposed force sensation practically. Numerical simulation results verify that the proposed method is successful in replicating the force sensing behavior of a force sensor with high sensitivity and superior precision. Moreover, to develop the AI safe interaction, this research has firstly developed the force control method with acceleration-based drive-side motion and instantaneous position-velocity information to increase the bandwidth of force control of the robotic system. The experimental results proves that the proposed force control system achieves wider bandwidth than the conventional method with velocity-based drive-side motion.
|
| 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
The theory and design of the proposed AI virtual force sensor based on deep learning has been successfully realized through numerical simulations on replicating the force sensing behavior of a force sensor with high sensitivity and superior precision. The hardware and implementation set up for the deep learning based AI virtual force sensor has been completed for experiments of the proposed method. The realization of wideband force control with acceleration-based drive-side motion and instantaneous position-velocity information is an important part of AI safe interaction being developed in this research.
|
| 今後の研究の推進方策 |
To achieve the purpose of building the human-centered intelligent robot system based on the fusion of advanced AI technologies to realize the safe, reliable and efficient human-robot interaction, the AI virtual force sensor will be completely realized by experiments in the next year. Moreover, the realized wideband force control and the AI virtual force sensor will be integrated to develop the AI safe interaction. In addition, environment detecting and adaptive force command generating function based on deep learning is constructing to ensure and maintain the safe interactions between robots and humans or working objects.
|
