Information-Theoretic Aspects of Control in a Bio-Hybrid Robot Device

Abstract

Information processing in natural systems radically differs from current information technology. This difference is particularly apparent in the area of robotics, where both organisms and artificial devices face a similar challenge: the need to act in real time in a complex environment and to do so with computing resources severely limited by their size and power consumption. The formidable gap between artificial and natural systems in terms of information processing capability motivates research into the biological modes of information processing. Such undertakings, however, are hampered by the fact that nature directly exploits the manifold physical characteristics of its computing substrates, while available theoretical tools in general ignore the underlying implementation. Here we sketch the concept of bounded computability in an attempt towards reconciling the information-theoretic perspective with the need to take the material basis of information processing into account. We do so in the context of Physarum polycephalum as a naturally evolved information processor and the use of this organism as an integral component of a robot controller.