An Artificial Muscle Neuron

Abstract

The biological muscle is not only a structure that produces contractile force, but is also a site laced with distributed intelligence. Athletes improve their performance through repeated practice, and this elicits muscle memory, a process that involves augmentations to the neural network surrounding the muscle tissue. Cephalopods such as squids and octopuses cannot only sense their surroundings but can also change their body colour and texture to match the environment. This is achieved by local networks of neurons distributed in the muscles. This thesis aims to emulate neurons and their networks for soft robotics that can give rise to these mechanisms seen in nature. Dielectric Elastomer Switches (DES) were identified as a starting point to achieve the thesis objective. Using an adjustable threshold, they can be implemented as DE logic gates to process local inputs and relay the outputs to the next logic gate. However, the drawbacks were their susceptibility to electrical noise and a sudden breakdown. To address these issues, the concept of a DE neuron was developed in this thesis. The manufactured DE neurons were successful in processing local inputs and relaying the outputs to the next neuron. Synaptic plasticity, a key feature of neurons, was emulated by using the DE switches as an analogue potentiometer. With user-controlled synaptic plasticity, the DE neural network in this thesis could produce various logical outcomes. The abilities of DE neurons shown in this thesis include communication between neurons, user-controlled synaptic plasticity and performing simple computational tasks. This thesis however is an initial, first step in the field of artificial muscle neurons. Future work involves creating feedback networks where synaptic plasticity is autonomously controlled. Achieving autonomous synaptic plasticity will enable a class of truly smart and intuitive soft robots.