Robot-Embodied Neuronal Networks as an Interactive Model of Learning

Abstract

Background and Objective:The reductionist approach of neuronal cell culture has been useful for analyses of synaptic signaling. Murine cortical neurons in culture spontaneously form anex vivonetwork capable of transmitting complex signals, and have been useful for analyses of several fundamental aspects of neuronal development hitherto difficult to clarifyin situ. However, these networks lack the ability to receive and respond to sensory input from the environment as do neuronsin vivo. Establishment of these networks in culture chambers containing multi-electrode arrays allows recording of synaptic activity as well as stimulation.Method:This article describes the embodiment ofex vivoneuronal networks neurons in a closed-loop cybernetic system, consisting of digitized video signals as sensory input and a robot arm as motor output.Results:In this system, the neuronal network essentially functions as a simple central nervous system. This embodied network displays the ability to track a target in a naturalistic environment. These findings underscore thatex vivoneuronal networks can respond to sensory input and direct motor output.Conclusion:These analyses may contribute to optimization of neuronal-computer interfaces for perceptive and locomotive prosthetic applications.Ex vivonetworks display critical alterations in signal patterns following treatment with subcytotoxic concentrations of amyloid-beta. Future studies including comparison of tracking accuracy of embodied networks prepared from mice harboring key mutations with those from normal mice, accompanied with exposure to Abeta and/or other neurotoxins, may provide a useful model system for monitoring subtle impairment of neuronal function as well as normal and abnormal development.