abstract-niceto

Cerebellum simulation towards embodiment

 Niceto Luque (HPB SP10 Speaker)

Abstract

Most simulation efforts in computational neuroscience are devoted to characterise neural models and/or neural system aiming at understanding the computational primitives underlying the neurophysiological substrate. To this end, the cause-effect relationship between well-defined input stimulation patterns and the neural output responses is traditionally settled and studied, thereby obviating the need for a body. Nevertheless, brain and body have co-evolved as the computational primitives of the central nervous system (CNS) with the environment. It is therefore reasonable to extend the cause-effect setups to perception-action setups in order to ease the study of the neural sensorimotor primitives that a body generates in closed-loop.

The mammalian cerebellum is pivotal in integrating the sensorial and motor pathways and coordinating the subsequent motor action being the perfect candidate for studying its computational primitives in perception-action setups. Neuroscience proposes well-established experimental setups (behavioural/cognitive tasks) which facilitate the study of the cerebellar role in motor adaptation and its related pathologies. Replicating cerebellar synthetic setups (i. e. Eye blink Classical Conditioning, the Vestibulo-Ocular Reflex… etc.) requires embodying the cerebellar network within a front-end body. Embodying not only needs for a biologically plausible cerebellar network and an actual body but also body-cerebellar efficient interfaces. Cerebellar embodiment constitutes the pinnacle of our lab mission for which during more than a decade, we have developed handcrafted solutions for  neurorobotics where neural interfaces, efficient neural simulators specifically designed for embedded neural systems (such as our neural simulator EDLUT) and robotic agents has been lacking.

Advertisements