| Project/Area Number |
23K19096
|
|---|
| Research Category |
Grant-in-Aid for Research Activity Start-up
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
0301:Mechanics of materials, production engineering, design engineering, fluid engineering, thermal engineering, mechanical dynamics, robotics, aerospace engineering, marine and maritime engineering, and related fields
|
|---|
| Research Institution | Japan Advanced Institute of Science and Technology |
|---|
Principal Investigator |
NGUYEN HuuNhan 北陸先端科学技術大学院大学, 先端科学技術研究科, 助教 (80981159)
|
|---|
| Project Period (FY) |
2023-08-31 – 2025-03-31
|
| Project Status |
Granted (Fiscal Year 2023)
|
| Budget Amount *help |
¥2,860,000 (Direct Cost: ¥2,200,000、Indirect Cost: ¥660,000)
Fiscal Year 2024: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2023: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
|
|---|
| Keywords | Simulation framework / Dynamic behaviors / Uncertainties / Reinforcement Learning / Soft Robotics / Adaptable Morphology / Embodied Intelligence |
|---|
| Outline of Research at the Start |
This research proposal introduces a novel approach for developing a class of soft robots that can actively self-configure their body (i.e., morphology) to allow adaptation with respect to internal and external varied factors. The core idea is to integrate proprioceptive feedback driven by morphology-dependent features (geometrical/material properties) into the control loop. To achieve this, a platform of multi-physics simulation and learning-based techniques seeking for optimal morphology transformation strategy to accommodate various critical conditions will be established.
|
| Outline of Annual Research Achievements |
The achievements of this fiscal year primarily revolve around the first phase (WP1), which is dedicated to developing a simulation framework. This framework aims to accomplish two main objectives:
1) Emulate the dynamic behaviors of a soft body under the effects of external and internal physical stimulation, and 2) Introduce physical uncertainties, such as cutting, and track the corresponding mechanical responses (strain) under these conditions.
Note that physical damages can be simulated either offline, using external tools for remeshing, or online, where remeshing is conducted within the environment itself. Additionally, we have successfully interfaced this simulation tool with a reinforcement learning environment and have validated with various algorithms.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
Based on the results presented above, we are confident that the first Work Package (WP1) has nearly achieved its objectives as anticipated. However, a limitation remains: the current framework only supports the exploitation of a single robot at a time. In practice, we need to be able to handle a large number of potential solutions, encompassing various robot designs and control strategies. Therefore, it is imperative to enable parallel sampling and training across the entire design space. This capability is crucial for efficiently exploring not only suitable control policies but also optimal morphologies for adaptive behaviors.
|
| Strategy for Future Research Activity |
The primary objective is to enhance the aforementioned framework by enabling it to simultaneously call multiple environments, each corresponding to a different robot design, as sub-processes. Additionally, an interface will be developed to assign design parameters to each environment.
Next, this comprehensive framework will be leveraged to conduct a joint search for the optimal morphology, including its dynamic transition, and control policy. This search will address pre-defined challenges such as unexpected physical damage, environmental changes, and the need to enhance the robot's functionality (e.g., sensing capabilities). The outcome will be evaluated via an experimental study.
|
