Neuroengineering model of human limb control - Gainscheduled feedback control approach

摘要

Recent neuroengineering modeling suggests that human control of limb movements may take advantage of simple linear control modules that are gainscheduled according to limb state, intended limb state, or both. The scheduling variable(s) appear to consist of limb position and possibly velocity. Thus it is possible that limb control can be modeled effectively using Linear Parameter Varying (LPV) techniques for designing gainscheduled linear control systems. We demonstrate the efficacy of an LPV controller based on a model of cerebrocerebellar neuroengineering models in simulating human arm control of moderate speed, direction-changing movements in the horizontal plane. The technique shows how position dependent dynamics can be stably and realistically controlled using a simple convex interpolation of linear controller modules. The finding could provide a systematic method of representing human motor control using an established engineering technique. It may also support the controversial view that the motor control system does not require internal models of body dynamics to achieve satisfactory performance.