Evolution-In-Materio: Solving Computational Problems Using Materials

The motivation behind the research is to show that evolutionary algorithms can exploit properties of materials to solve various computational problems without requiring a detailed understanding of such properties. This approach is referred to as evolution-in-materio. In this research, it has been shown that using a purpose-built hardware platform called Mecobo, it is possible to evolve voltages and signals applied to physical materials to solve a number of computational problems. Here it has been demonstrated for the first time that the evolution-in-materio method can be applied to function optimisation, machine learning classification, frequency classification, even parity and bin packing problems. This evolution-in-materio method has also been applied here to discriminate tones and control robots. The physical material used in each of these experiments is a mixture of single-walled carbon nanotubes and a polymer. This is the first time that such material has been used to solve computational problems. The results of all of these experiments indicate that evolution-in-materio has promise and further investigations would be fruitful. Other than the solutions regarding these computational problems, this thesis has also devised and investigated suitable input-output mappings and input signals that allow various computational problems to be solved using the Mecobo platform and the experimental material.