Evolving Cellular Automata in-Materio

Evolution-in-Materio in the context of unconventional computing is the practice of using artificial evolution techniques to search for configurations of physical material samples that allow for them to be used as practical computational devices. The motivation for Evolution-in-Materio is two-fold: it is in part an exploration of new computational substrate materials as alternatives to silicon, and in part an exploration of metaheuristically-guided evolutionary search in the design space as a design process instead of traditional top-down engineered design. One of the biggest problems in exploiting materials for computation is finding a good computational abstraction to carry computation on-top of the underlying physical processes. This thesis looks at the possibility of implementing cellular automata as an abstraction via Evolution-in-Materio to reason about the computational capabilities of single-walled carbon nanotube and polymer composite meshes. In this thesis, computationally stable linear and non-linear logic gates and Elementary and Sub-Elementary Cellular Automata are successfully evolved in-materio. Not all Elementary Cellular Automata are shown to be successfully evolvable, which reveals computation complexity properties of the material-under-study