2022 Fiscal Year Research-status Report
Learning in-hand manipulation for a compliant underactuated gripper with interactive human supervision
| Project/Area Number |
22K14221
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
Colan Jacinto 名古屋大学, 工学研究科, 研究員 (40936746)
|
|---|
| Project Period (FY) |
2022-04-01 – 2024-03-31
|
| Keywords | in-hand manipulation / adaptive compliance / robotic gripper / cable-driven mechanism |
|---|
| Outline of Annual Research Achievements |
This project aims to design and develop underactuated robotic grippers that can learn adaptive compliance for soft manipulation of deformable objects through human demonstrations and interactive feedback.
In the first year, we accomplished the first objective of developing a compact multi-dof robotic gripper with a decoupled wrist that enables independent position and force control of each joint. We carried out experimental validation for minimally invasive surgical applications using the developed gripper. We collected actuator signals when grasping materials with different stiffness and trained a Gaussian process regression model for sensorless grip force prediction. We also devised a multi-objective inverse kinematics control framework that can handle joint position control under multiple constraints simultaneously.
The next objective is to learn adaptive compliance from human demonstrations using reinforcement learning, sim-to-real techniques and interactive feedback. We plan to develop a simulation environment that can capture human demonstrations and train a model-free policy using state-of-the-art deep reinforcement learning 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
The project is advancing well as the first objective was completed during the first year, with the design, prototyping, control and evaluation of a cable-driven gripper for compliant grip force. Progress on the next objectives has been started with a supervised regression model trained for sensorless grip force prediction.
The proposed design, control and evaluation of the developed gripper have been published in two journal papers and two international conferences.
|
| Strategy for Future Research Activity |
The project is expected to proceed according to the initially approved plan without any changes. In the next project year, we plan to develop a simulation environment that can capture human demonstrations for manipulation of soft objects. A model-free policy will then be trained using state-of-the-art deep reinforcement learning algorithms. Sim-to-real strategies will be applied for transferring the learned manipulation strategy to the real robotic gripper, and interactive human feedback will be used for adapting the learned policy and deploying it again in the physical system.
|
| Causes of Carryover |
For prototyping the proposed gripper, we used existing actuators and control circuits. Next steps involve the extension to multiple cable-driven fingers of larger size, so we postponed the scheduled purchase of motors, mechanical parts and electrical parts for the next year. Moreover, since the first year focused on the mechanical design of the gripper, we will also purchase the deep learning workstation for the manipulation learning aspects during the second year.
|
