Encoding physical intelligence into soft robotics via smart materials and continuum mechanics

Soft robots are designed to interact with the world safely and adaptively, much like living creatures. Typically composed of elastomers, these robots excel in dynamic and unstructured environments due to the
conformability and adaptability of their soft bodies. However, while soft bodies provide adaptability, they
complicate robot control due to the nonlinear mechanical properties of elastomers and the infinite degrees
of freedom inherent in the soft body. In biology, creatures address this challenge through physical intel-
ligence. Although all their tissues are connected by nerves, creatures cannot rely solely on their brains to
control their complex bodies; they also depend on unconscious behaviors driven by physical intelligence.
Physical intelligence encompasses the intelligence encoded in bodies through actuation, adaptation, and
perception. This suggests a method for soft robots to evolve more autonomous functions by integrating
intelligence into their soft entities, enabling them to not only adapt to environments but also react to them,
mimicking organisms that interact with their surroundings through both adaptability and responsiveness.
This research proposes encoding physical intelligence into soft robots on two levels; on the material
level as well as on the structural level. On the material level, the use of stimuli-responsive materials to
construct the soft robotics body enables new functionalities at molecular and micro scales. In this work,
reversible polymers based on the Diels-Alder reaction were employed to fabricate soft robot components,
incorporating both self-healing functions and multi-material compositions with enhanced mechanical per-
formance. These reversible polymers enable covalent bonding between materials with varying hardness,
resulting in high-strength interfaces suitable for robust multi-material soft robots. This approach was
utilized to develop soft robotic bending actuators with optimized multi-material compositions, refined
through finite element analysis and topology optimization. The resulting actuators exhibited improved
directional stiffness, demonstrating that material composition-based physical intelligence can enhance
the reliability of soft robotic grippers. In addition, variable stiffness materials, those that change their
stiffness upon a stimulus, can increase the physical intelligence in soft robots. In this thesis, stiffness
modulation based on vacuum controlled granular jamming was used to enhance the sensing capabilities
of electrical impedance tomography (EIT) sensors for soft robots. Through stiffness modulation of the
granular material, a novel variable sensing EIT sensor for robotic electronics skins was developed, in
which the sensitivity and sensing range can be tuned depending on the interaction with the surroundings.
On the structural level, three innovative geometric designs were developed to enhance grasping per-
formance in soft grippers by mimicking biological morphologies and optimizing them through finite
element analysis of continuum mechanics. The first design features a multi-material, self-closing, and
self-healing suction cup that automatically seals to prevent vacuum leakage when not in use. Being in-
tegrated into vacuum grippers composed of multiple suction cups, leading to energy conservation and
maintaining grasping force when handling various objects, even those that do not make contact with all
the suction cups. Secondly, by leveraging origami principles and structural design, a crease-free origami
vacuum bending actuator for soft robots was developed. This crease-free actuator can be 3D printed using
consumer-level fused filament fabrication printing. Its orderly self-folding behavior allows for high bend-
ing angles, making this compact actuator suitable for modular designs that can reconfigure soft robots,
such as transforming grippers into locomotion robots. Additionally, the high bending angle and vacuum
activation enable compatibility and synergy with suction cups, particularly the self-closing ones. This is
demonstrated in a versatile, octopus-like tentacle vacuum gripper capable of grasping a wide variety of
irregular and flat objects. Thirdly, structural physical intelligence was employed to integrate software-less
reflex mechanisms into a soft robotic gripper. This multi-modal gripper combines a vacuum suction cup
and bending actuators with a bistable mechanism. It features three grasping modes: two utilizing reflex
mechanisms (force-triggered and contact-triggered) and one actively controlled. This design results in a
versatile gripper that can switch between modes using a single control parameter, specifically its air flow
rate, depending on the object to be grasped.