A neuromorphic paradigm for extrinsically evolved hybrid analog/digital device controllers: initial explorations

This paper argues that the continuous time recurrent neural network (CTRNN) provides a particularly attractive paradigm for the extrinsic evolution of analog device controllers. The paper begins with a discussion of motivations and difficulties faced in evolving electrical circuits and then illustrates how some of these difficulties have been successfully addressed in the context of evolved CTRNNs. This paper provides a presentation of a new, hardware friendly, CTRNN formulation as well as some preliminary experimental results demonstrating that practical devices can be evolved under the new model. Finally, the paper will conclude with a discussion of open issues and a summary of current plans to close those gaps.