| The eukaryotic cell is a computational device that performs perception-action behavior, which requires a long-range signaling mechanism. The micro-tubule network is the only intracellular structure that provides the structural characteristics needed for this type of intracellular signaling. From these characteristics, and preliminary experimental evidence, a variety of signaling mechanisms have been proposed in the literature. To explore this hypothesis, the microtubule learning model (MtLM) is presented that combines a biologically motivated a bridge between machine learning and mainstream cell biology.; The presented MtLM is shown to perform well within the context of two different perception-action frameworks. The first framework, the “center finding problem”, is a robot navigation task where the MtLM must find the center of a virtual two-dimensional space when given a random starting point. The second framework, the biot, is a novel biomimetic robot architecture that consists of several segments interconnected in a highly context-sensitive fashion (exhibiting some degree of randomness). This framework is designed to simulate the context-sensitivity that is typical of interactions inherent between different components of the eukaryotic cell, providing the MtLM with a biologically plausible learning task.; This work, by providing a functional example of intracellular long-range signaling through the MtLM, reinforces the hypothesis that the long-range signaling mechanism in the eukaryotic cell is the microtubule network. Additionally, application of the MtLM to these differing frameworks illustrates the importance of structure in any system constructed in a bottom-up fashion, and highlights the differences between information processing tasks typically performed at the cellular level and in higher-order cognitive tasks. Lastly, this work also illustrates the strength of the MtLM as a control mechanism for producing tuned oscillatory activity. |
