Modeling cerebrocerebellar control in horizontal planar arm movements of humans and the monkey

Abstract

In daily life, animals including humans make a wide repertoire of limb movements effortlessly without consciously thinking about joint trajectories or muscle contractions. These movements are the outcome of a series of processes and computations carried out by multiple subsystems within the central nervous system. In particular, the cerebrocerebellar system is central to motor control and has been modeled by many investigators. The bulk of cerebrocerebellar control involves both forward command and sensory feedback information inextricably combined. However, it is not yet clear how these types of signals are reflected in spiking activity in cerebellar cells in vivo. Segmentation of apparently continuous movements was first observed more than a century ago. Since then, submovements, which have been identified by non-smooth speed profiles, have been described in many types of movements. However, physiological origins of submovement have not been well understood. This thesis demonstrates that a currently proposed recurrent integrator PID (RIPID) cerebellar limb control model (Massaquoi 2006a) is consistent with average neural activity recorded in a monkey by developing the Recurrent Integrator-based Cerebellar Simple Spike (RICSS) model.

(cont.) The RICSS formulation is consistent with known or plausible cerebrocerebellar and spinocerebellar neurocircuitry, including hypothetical classification of mossy fiber signals. The RICSS model accounts well for variety of cerebellar simple spike activity recorded from the monkey and outperforms any other existing models. The RIPID model is extended to include a simplified cortico-basal ganglionic loop to capture statistical characterization of intermittency observed in individual trials of the monkey. In order to extend the capability of the RIPID model to a larger workspace and faster movements, the model needs to be gainscheduled based on the local state information. A linear parameter varying (LPV) formulation, which shares a similar structure to that suggested by the RICSS model, is performed and its applicability was tested on human subjects performing double step tasks which requires rapid change in movement directions.