Hierarchical and parallel organization, scheduled scaling of error-type signals, and synergistic actuation appear to greatly simplify and robustify human motor control
Electrical Engineering and Computer Science/ Health Sciences
Last modified: March 31, 2006
Both human motor behavior and human neural motor control system architecture can appear complicated. On the other hand, human and animal motor behavior, though often complex, is robust to many types of disturbance and loads. And basic motor skills can be learned and adapted to new conditions fairly quickly. In addition, much of the motor control system, especially the sensorimotor cortex, cerebellum, and basal ganglia, displays considerable microscopic regularity and an increasingly appreciated high degree of functional modularity. These features suggest that the system may be better considered a well-organized aggregation of simple, similar, yet flexible cooperating components, than as a fundamentally complex, and therefore potentially delicate machine.
Recent computational modeling of cerebral and cerebellar control of reaching [1a,b], balance  and locomotion  demonstrates that all of these behaviors may be controllable via a single basic multi-modular, piecewise linear, feedback-dependent system that is remarkably simple and structurally uniform. In particular, there is no requirement for explicit representation of body part dynamics in terms of so-called “internal dynamics models”. Instead, simple spinal and cerebral circuits that select and command groups of synergistic muscles may collapse several degrees of control freedom to enable simpler, decoupled signal processing throughout the system. Stabilized “long-loop” stretch responses may then enable Proportional-Integral-Derivative (PID)-type compensation for both self-generated and external dynamics using single-input single-output linear modules that are selectable in real-time according to body state, and/or intent. In the presence of such scheduled feedback-dependent control, motor behaviors at least as complex as planar walking and reaching may be driven by sequences of simple, smoothed step-like or pulse-like commands that may issue from cerebral and/or spinal levels.
Moreover, control precision and robustness may be increased by augmenting the number of identical, parallel control modules and the richness of body state information feedback channels, rather than the complexity of each module. As a result, the control at any instant may be always of comparatively low dynamic order and therefore of comparatively simple operation. To demonstrate these principles, a basic model of cerebrocerebellar architecture is presented, and its consistency with known neurocircuit anatomy and experimentally recorded internal neural activity is shown. Then its possible application to the control of human reaching, balance and bipedal locomotion is demonstrated. It is proposed that the identified design principles enable complex yet reliable animal motor behavior to emerge from a fundamentally simple control system.
[1a,b]: Massaquoi SG. A RIPID model of cerebro-cerebellar interaction in long-loop arm control I, II: (a) single- and (b) two-joint control. Manuscripts submitted to Biological Cybernetics.
: Jo S and Massaquoi SG. (2004) A model of cerebellum stabilized and scheduled hybrid long-loop control of upright balance. Biol Cybern 91; 188-202.
: Jo S and Massaquoi SG. Hierarchical neural control of human sagittal balance and bipedal walking (SBBW) I: cerebrocerebello-spinomuscular interaction in natural walking (Abstract accepted at Meeting of the Society for the Neural Control of Movement, Key Biscayne FL. May, 2006; manuscript in preparation)