Bio-inspired grasp control in a robotic hand with massive sensorial input

Area of Science:

• Robotics
• Neuroscience
• Biomedical Engineering
• Manufacturing Technology

Background:

• Robotic grasping remains a significant challenge, lacking the advanced control seen in human hands.
• Current robotic systems struggle with real-time processing of large tactile sensor data volumes.
• Neurophysiology and robotics research investigate human hand control for bio-inspired solutions.

Purpose of the Study:

• To design and test a novel hardware-software robotic architecture for advanced grasp control.
• To overcome computational limitations in processing tactile sensor data for robotic grasping.
• To develop a bio-inspired system that mimics human hand's sensory-motor coordination.

Main Methods:

• A bio-inspired approach using cellular nonlinear/neural network (CNN) paradigm for tactile data processing.
• Parallel processing architecture for handling large amounts of tactile sensing signals.
• Utilizing hand shape and spatial-temporal features from microfabricated force sensors for control.

Main Results:

• Demonstrated successful grasping of various unknown objects, including soft and deformable items.
• The robotic system exhibited controlled and stable manipulation capabilities.
• The developed architecture effectively managed complex tactile data for sensory-motor coordination.

Conclusions:

• The bio-inspired CNN-based architecture significantly advances robotic grasping capabilities.
• The system shows promise for applications requiring dexterous manipulation of diverse objects.
• This approach offers a viable solution for real-time tactile data processing in robotics.