Selecting an Appropriate Curvature Sensor for Fluidic Soft Robot and Modeling Sensor Reading vs Pressure vs Position

Abstract

This research focuses on the study of a curvature sensor for a fluidic soft robot. Soft robot is a complete new dimension to traditional rigid robot. A soft robot is made up of materials like Silicon, PDMS and elastomeric polymers. The actuation method can be hydraulic, pneumatic or electric. Depending on its construction, it undergoes elongation, bending, twisting, or all of the three on actuation. It brings with it some important features like compliance with the object of interaction and robustness, which is an inspiration acquired from animals and plants. This results into useful applications in fields of rehabilitation, gripping delicate objects in food industries and allowing safe interaction for humans. The soft robot has large DOF, which allows it to maneuver in a way, which is difficult for the traditional robot. However, this large DOF makes the modeling of the soft robot for determining the robot state difficult and challenging. Another approach towards determining the robot state is using sensors. In this thesis, a thorough study is done to find out an appropriate curvature sensor to be embedded into the soft robot. The data from curvature sensor, pressure sensor and the vision system are collected in experiments undertaken with obstacles in the soft robot path. The collected data is used via machine learning technique to obtain trained model that determines the robot state and obstacle location.