Learning­Gripper

Year

2013

The idea

This robotic gripper is capable of grasping an object and orienting it in a specific manner after a brief learning period.

The objective

The LearningGripper stands for machine learning; its learning algorithms enable it to develop its own solution strategies.

Technical data

• Dimensions: 263 x 263 x 255 mm
• Weight: 1850 g
• Material: polyamide
• Manufacturing process: selective laser sintering
• Number of gripper fingers: 4
• Finger length: 195 mm
• Degrees of freedom: 12
• Diameter of gripped object: 100 mm
• Handling weight: 200 g
• Valves: 12
• Operating pressure: 2.5-3.5 bar
• Sensors: 30

Operating principle

The robotic gripper is programmed with basic definitions of possible movements and finger positions, along with a feedback function. To perform a given task, the intelligent gripper autonomously adopts the correct movement sequences. Its knowledge gained by trial and error can be transferred to other grippers.

Structure

The fingers of the LearningGripper are modelled on those of the human hand. Like an index finger, each element has three degrees of freedom. The required power is provided by twelve polyamide pneumatic bellows actuators. The kinematics of this agile gripper are flexible and yielding.

Function

The LearningGripper can rotate a sphere so that a specific point on its surface faces upwards. To do this, it starts with an unspecific action. A position sensor inside the sphere then registers its orientation, and a distance sensor measures the distance to the gripper’s palm. The closer the result is to the target, the stronger the positive feedback. Just like a human, the machine learns to coordinate the movements of its gripper fingers and to solve the specified task on the basis of positive feedback.

Material

LearningGripper