| Özet: | Proprioception is a trending topic in the field of soft robotics. It endows soft robots with the
capability to retrieve their spatial configurations without external feedback for closed-loop control.
This is usually achieved by embedding soft and stretchable sensors in the soft bodies of the
robot to directly measure their deformations. In addition, taking advantage of their infinite
DoFs and inherent compliance, multisegment soft robotic arms are designed to execute highly
dexterous tasks in the cluttered environment. However, the integration of the proprioceptive
sensors in the body of the soft arm is an ongoing challenge.
Firstly, the embedded sensors need to accommodate for the compliance of the soft robotic
arm. Secondly, where to place such sensors is a non-trivial problem. Lastly, the integration of
these stretchable sensors involves complex fabrication steps.
In this thesis, we aimed to address these challenges of proprioceptive soft robotic arm with
our highly integrated solution - a multisegment soft robotic arm that is capable of proprioceptive
sensing while minimising the number of sensors on-board. The proprioceptive sensors do not
interfere with the motion of our soft robotic arm and they can be easily integrated and removed.
The major contribution of our work is an novel sensing method for modular soft robotic arms. We
also advanced this field by contributing: 1) an omnidirectional actuator design for multi-material
3D printing, 2) a modular approach for fast arm assembling and maintenance. By placing sensing
arrays within the rigid joints, the variations of stress distribution over the top of segments were
measured. Together with the shape information captured by the motion capture system, we
obtained a dataset to train a mapping from the tactile sensor responses to the posture of the
soft robotic arm using k-nearest neighbors regression. The experimental results demonstrated
that the proposed approach was able to reconstruct the whole 3D shape of a three-segment soft
robotic arm under piecewise constant curvature assumption.
|
|---|