Multifunctional Grasping and Object Detection through Facile Manufacturing of Biomimetic Actuators and Sensors

Biomimetic robots mimic natural features in biological systems to accelerate progress in state-of-the-art mechanisms and computational tools. Researchers can leverage millennia of evolution, instead of starting from a blank slate, to address remaining challenges in fields such as robotic grasping and multifunctional object manipulation. In humans and animals, the execution of complex motor tasks is made possible through a combination of sensory feedback neurons, muscle groups, and neural systems that can generate a wide range of motor control signals without changing the physical structure of the neurons or the arrangement of muscles. Recreating these complex behaviors in robots requires that researchers first understand the neuromuscular coordination between controllers and peripheral muscles to generate multifunctional behaviors. Studying animals that display multifunctional behaviors with relatively small nervous systems, such as mollusks, is a tractable method to gain insight into tactile perception, neural control, and muscular actuation. Subsequent reverse engineering of these biological principles for robotic applications could be accelerated by implementing rapid iteration cycles and involving high-throughput team science. These goals require the democratization of low-cost and easy-to-manufacture robotic components that can still maintain their core biomimetic functionalities. In my Ph.D. research, I have investigated the effect of actuator and sensor morphologies on biomimetic robots for democratized multifunctional grasping, using easily manufacturable biomimetic sensors for object detection and biomimetic actuators for physical manipulation. I explored multifunctional tactile sensing with both force/torque sensing and material classification capabilities by leveraging biomimetism of the human finger. Next, I reported artificial muscle actuators with novel geometries to mimic natural muscle architectures through low-cost and scalable fabrication methods. Then, I introduced a state-of-the-art soft biomimetic robot as a tool for neuromuscular research (SLUGBOT) and compared the robot’s kinematics to those of the mimicked animal. Integrating biomimetic actuators and a neural controller produced a soft robot that exhibited multifunctional behaviors using a single robotic morphology and relatively simple control architecture. Next, I improved the facile manufacturing of flexible strain sensors to provide artificial muscles with proprioceptive sensing capabilities. Finally, I identified modifications to SLUGBOT for increasing the biomimetism of the robot’s morphology and actuator behavior. These research contributions provide a tractable path towards low-cost and easy-to-manufacture biomimetic robots that can be used to explore biological hypotheses and help researchers understand neuromuscular interactions as well as neural dynamics in biological and artificial neural networks.