Soft robots are a research hotspot because they are safer, more adaptable, more flexible and bionic than traditional robots made of rigid materials. One of the challenges in manufacturing these soft robots is that their deformable design must enable them to move in almost any direction at any time. Scientists say flexibility makes it difficult to train planning and control modules to drive automation.
The team developed a soft sensor system that covers the robot's body and provides "proprioception", that is, sensing the movement and position of its body. The data collected by the sensor is run by a deep learning model, which can filter out the noise and capture clear signals to estimate the 3D state of the robot. The researchers conducted validation tests on a robotic arm that looks like an elephant's trunk, which can predict its own position and automatically swing and stretch.
The sensors used by the team are made of off the shelf materials. One of the researchers' future goals is to create artificial limbs that can manipulate environmental objects more nimbly. The long-term goal of soft robots is to integrate body sensors into robots. A breakthrough for the team came when a researcher discovered a conductive material used to shield against electromagnetic interference that can be purchased anywhere in rolls.
The material has piezoresistive properties, and scientists realize they can use it to make effective soft sensors. When the sensor is deformed, its resistance is converted to a specific output, and the voltage is used as a motion related signal. In the end, the team turned to an origami called kirigami, which uses lasers to cut rectangular strips of conductive silicon film into patterns to make it more flexible. They are then added to the previously designed soft robot body.