— Soft robotic arms are designed to execute dexterous tasks safely in cluttered environments due to their inherent
compliance and infinite degrees of freedom. Closed-loop control
requires proprioceptive sensing to retrieve the spatial configuration of the robot. However, the integration of soft sensing in
the soft continuum body is still an ongoing challenge. In this
paper, we present a highly integrated pneumatically actuated
soft robotic arm that consists of 1) an omnidirectional actuator
design for multi-material 3D printing, 2) a modular approach
with fast assembling connectors and 3) a tactile sensing array
embedded in the joint of each module. The in-joint placement
of the tactile sensing array is used to map the variation of stress
distribution between the two connected modules and the posture
of the actuator with learned k-nearest neighbors regression.
The experimental results showed that the method estimated
the posture of a three-segment soft arm with a mean error of
5.64 mm for the tip location.
