Octopus‐Inspired Suction Cups with Embedded Strain Sensors for Object Recognition

Abstract

The octopus has unique capacities are sources of inspiration in developing soft robotic‐enabling technologies. Herein, soft, sensorized, suction cups inspired by the suckers of Octopus vulgaris are presented. The suction cups using direct casting are fabricated, so that materials with different mechanical properties can be combined to optimize sensing and grasping capabilities. The artificial suckers integrate four embedded strain sensors, individually characterized and placed in a 90° configuration along the rim of the suction cup. Based on this arrangement, how well the sensory suction cup can detect 1) the direction and 2) the angle (from 30° to 90°) of a touched inclined surface and 3) the stiffness of a touched flat object (shore hardness between 0010 and D50) both in air and underwater is evaluated. Data processing on neural networks is based using a multilayer perceptron to perform regression on individual properties. The results show a mean absolute error of 0.98 for angles, 0.02 for directions, and 97.9% and 93.5% of accuracy for the material classification in air and underwater, respectively. In view of the results and scalability in manufacturing, the proposed artificial suckers would seem to be highly effective solutions for soft robotics, including blind exploration and object recognition.

Keywords

• artificial suction cups
• machine learning
• soft robotics
• soft sensors
• stiffness classifications