Evaluating and Extending a Human Cerebellar Balance Control Model
Sungho Jo & Steve G. Massaquoi
To test the ability of a cerebellar control model developed for reaching arm movements to be extended to explain human upright balance control.
Improved human control models could allow better understanding of the human sensorimotor control system and contribute to the design of devices to enhance human function or advanced robotic performance. There is interest in especially natural control of human standing balance and walking. Humans seem to use specific postural strategies for upright balance or efficient walking algorithm to cope with any environment. Moreover, the cerebellum is understood to play a critical role in upright stability. Analyzing human movements in terms of cerebellar function by an improved human control model would help establish a connection between the structure and function of the human sensorimotor control system.
Other investigators have demonstrated that human postural balancing strategies may be described by engineering control models, such as a linear quadratic gaussian model-based control formulation. However, previous studies (e.g. ) typically describe only the external behavior of the body, not the internal signals in relation to specific processing by the cerebellum. Using a human motor control model that includes a description cerebro-cerebellar interaction that was developed initially to describe arm reaching movements in the horizontal plane , human stable standing maintenance has been investigated. We have found that an upright balancing three-link system was able to closely reproduce experimentally measured postural responses in subjects on a pull platform . The model’s proposed structure-function relationships could also be tested by simulation damage to its pathways and structures. Simulated lesions of the cerebellar portion of the model have produced behavior that appears to represent cerebellar ataxia (clumsiness and instability of postural regulation) and titubation (cerebellar postural tremor in the upright body).
To verify and extend the above preliminary findings, kinematic and electromyographic data is culled from the physiological and bioengineering literature. Typically this consists of subjects’ responses to disturbances produced by a pull-platform on which they are standing. We attempt to simulate this postural recovery data using a modified Proportional-Integral-Derivative control model, that has been extended from that used for modeling cerebellar control of arm movement . The parameters in the model are tuned empirically to fit the available data as closely as possible. Then the model is subjected novel disturbances (of greater or lesser magnitude) to determine whether the model is generically stable, and whether the behavior and electromyographic signals remain physiologically realistic. The sensitivity of performance to various specific parameter changes is explored. Further extension of the model including switching control and adaptive control principles are explored as well .
Human or animal can demonstrate very complex behaviors. A model may become extremely complicated for such complex. There is an ongoing difficult tradeoff between complex, capable models that are difficult to validate, and simpler models that are easier to test, but fail to reproduce important features of behaviors, and may not be faithful to known underlying neuroanatomy. Essentially, the ideal level of system abstraction must be discovered for each application.
The current line of investigation is valuable for practical development of bio-machines as well as scientific contribution to brain research. Verified control models will attain multidisiplinary contribution.
This work has been supproted financially by NSF KDI IBN-9873478. We receive collegial support from other research groups of the MIT-Harvard HST NeuroEngineering Research Collaborative http://hst.harvard.edu/nerc/of which our research group is a member.