Related Experiment Video
Updated: Nov 8, 2025
09:41
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
Published on: April 21, 2023
1.9K
Development and Grasp Stability Estimation of Sensorized Soft Robotic Hand.
P M Khin1,2, Jin H Low1,3, Marcelo H Ang1,2
1Advanced Robotics Centre, National University of Singapore, Singapore, Singapore.
Frontiers in Robotics and AI
|April 19, 2021
Summary
This study developed a soft robotic hand with force sensors for grip estimation. A triplet network achieved 89.96% accuracy in estimating object stability, enabling adaptive grasping.
Area of Science:
- Robotics
- Artificial Intelligence
- Sensor Technology
Background:
- Developing anthropomorphic soft robotic hands requires advanced sensing for object manipulation.
- Grip state estimation is crucial for stable grasping and preventing slippage.
Purpose of the Study:
- To develop an anthropomorphic soft robotic hand with integrated flexible force sensors.
- To create grip state estimation networks for object manipulation.
- To enable adaptive pneumatic control for handling diverse objects.
Main Methods:
- Integrated flexible force sensors into a soft robotic hand.
- Developed grip state estimation networks using one-shot learning (OSL) and long short-term memory (LSTM).
- Evaluated three LSTM-based networks (triplet, LSTM, Siamese LSTM) for stability estimation.
Main Results:
- The robotic hand successfully grasped and lifted various objects.
- Grip state estimation networks predicted object instability and slippage.
- The triplet network achieved the highest stability estimation accuracy at 89.96%.
Conclusions:
- The developed grip state estimation network can provide feedback to the pneumatic control system.
- This allows for efficient and adaptive grasping of different objects using optimal pneumatic pressure.
- One-shot learning with LSTM networks offers a scalable solution for training robotic hands with limited data.
