Computational Roles of Inhibitory Interneurons in a Cerebellum Spiking Neural Model of Adaptive Vestibulo-ocular-optokinetic Reflex

Cerebellum is a powerful adaptive controller that mediates numerous functions that are criticalfor everyday life and even survival. Vestibulo-ocular reflex (VOR) and optokinetic reflex(OKR) are two important reflexive eye movement mechanisms that ensure a stable visual fieldduring movement of the animal and/or the target, which is also a common task encountered inengineering control. They are adaptive since they need to compensate for the changes in thebody and environment as well as the neural circuitry. Their adaptations are closely related tocerebellar function realized by its well-organized circuitry and numerous plasticity. Yet, thecomputational roles of molecular layer interneurons (MLI) in cerebellum-mediated learningand control have not been well-understood. Recent studies show that mutant mice with deletedMLI inhibition to Purkinje (PKJ) cells demonstrated deficits in VOR/OKR adaptation. In vitrostudies showed that MLI to PKJ connections follow a specific spatial organization and aremodulated by synaptic plasticity. However, a computational cerebellum model of VOR/OKRwhich can connect these findings to help address the roles of MLI over neural spiking tobehavioural levels at a network scale is still lacking.In this project, we will investigate the computational roles of MLI by an integrated modelingand hardware testing approach. We will first develop a cerebellum spiking neural networkmodel of VOR/OKR to investigate the impacts of (1) spatial organization of MLIRPKJconnection; and (2) plasticity at MLI synapses on PKJ firing pattern and VOR/OKR learning.How they affect neural firing patterns and consequently the behavior will be systematicallycharacterized in the simulations. Second, how MLI may affect behavior as the cerebellummodel operates in closed-loop with the environment will be investigated with a robot system.We will develop a real-time hardware system to implement the cerebellum model forcontrolling a mobile robot performing target stabilization task mimicking VOR/OKR. Thehardware simulation system will be optimized for the cerebellum model and be modularizedto facilitate future extension. The robot testing allows evaluation of robustness of thecerebellum model under influence of MLI in the presence of disturbance in the physicalsystems and communication delay. These results will also provide evidence of the potentialfunctional advantages of the cerebellum controller. Together, this project will advance ourunderstanding of the computational mechanism in cerebellum, which will provide new insightto reverse engineer our brain as well as to develop more adaptive and robust robot control.?