DPhil Proposal Dynamics of Neutral Evolution in the Development of Neural Controllers

Evolutionary robotics has successfully exploited the principle of natural selection. This approach places great importance in understanding the dynamics of the evolutionary process. Increasing amount of research points towards selective neutrality as an important and beneficial factor in evolution. Among other things, it eliminates the picture of populations becoming trapped on local hilltops. This research asks, what are the benefits of neutrality in the evolution of simple model nervous systems for autonomous agents? And what mechanisms can be incorporated in the development of neural controllers which make use of neutral networks during evolution? Potential mechanisms have been suggested to arise in more biological plausible genotype-phenotype mappings, such as the use of a genetic regulatory network or a morphogenetic process. Progress in this area would be of major significance in the evolutionary approach to robotics. Furthermore, it would help understand the role of neutral networks in evolution and the (far from understood) relationship between embryological development and evolution.