Cerebro-cerebellar Implementation of Gainscheduled Feedback Control

Clarifying the way in which the central nervous system (CNS) controls motor behavior is an area of ongoing investigation in biology and medicine. Many of the neurophysiological details of how even the simplest voluntary movements are organized and executed remain to be established. Part of the motivation is basic scientific interest in achieving a better understanding of the brain, spinal cord and neuromuscular system. Another part relates to the neurological interest in understanding disorders of movement and, hopefully, eventually means of their management through neurocompatible prosthetic and orthotic devices. At the same time, there is growing interest in possibly using neurophysiological principles in robot control. The fluid, stable movement of many animals in open, highly diverse environments is generally not achievable by current robots. This is especially with respect to stability and performance robustness to uncertainty and/or variability in body dynamics, and in terms of behavioral flexibility and adaptivity. The motor performance of animals is all the more impressive when it is considered that control computation is performed by neurons that are in general very much slower, noisier, and much less precise than electronic circuits and microprocessors. Nature has somehow learned to leverage architectural hierarchy and massive parallelism to achieve control that is in important ways unrivaled by artificial devices. It is thus possible that control engineering may both inform and be informed by neurobiology. Although no consensus has been achieved about many neurophysiological motor control mechanisms, a wealth of experimental data has been accumulated, and agreement is building on many points. Most investigators view the cerebrum as the primary driver of voluntary movements. The brainstem and spinal cord are concerned with reflex and simple patterned behaviors such as breathing or sneezing. The cerebellum, which is strongly interconnected with the cerebrum, brainstem and spinal cord, which is a structure long felt to be important for motor “coordination”, has become viewed as some type of adaptive controller (Barlow 2002; Massaquoi and Topka 2002) in the engineering sense. The principal connections of the anterior lobe of the cerebellum, which is the principal motor control portion, are shown in Figure 1. These are involved in an extremely large number of feedback loops traversing pathways that are among the most rapidly conducting in the nervous system. Characteristic motor control deficits associated with damage to the cerebellum include ataxia, tremors and impaired motor learning (Massaquoi and Hallett 2002). Ataxia comprises sluggishness (reduced acceleration and jerk), sloppiness with frequent target overshoot, increased variability in the temporal evolution of multijoint and multi-stage movements (impaired “coordination”) and degraded balance (inability to stabilize otherwise unstable postures). Tremors (oscillations related to Forty-Fifth Annual Allerton Conference Allerton House, UIUC, Illinois, USA September 26-28, 2007 FrA6.3