Neurorobot Vitroid — A living test model for embodiment brain research

One of effective approaches to problems of conventional artificial intelligence is "embodied cognitive science". To realize an embodiment of a cultured living neuronal circuit, we provided cultured neurons with a body of a miniature-moving robot. The aim of this neuro-robot system is preparing a testing environment of the suitable relationship between activity of living neuronal network (LNN) and phenomena in outside world. We call our neuro-robot system as ”Vitroid", meaning a system as a kind of a “test tube” for cognitive agent made by living component. We have investigated whether the Vitroid with various algorithms for linking neurons and outer world can be embedded a certain a priori behavior, and elucidated the modification of network dynamics induced by robot behaviors. We have developed Vitroid with simplified fuzzy reasoning, and a weighted averaging of selected electrode activity. Both algorithms for linking induced the decrease of standard deviation of the number of electrical events induced by inputs from sensor on the robot. It suggests that the modification of network dynamics is induced by robot behaviors, which is likely in animal brain.