On the intrinsic control properties of muscle and relexes: exploring the interaction between neural and musculoskeletal dynamics in the framework of the equilbrium-point hypothesis

The aim of this thesis is to examine the relationship between the intrinsic dynamics of the body and its neural control. Specifically, it investigates the influence of musculoskeletal properties on the control signals needed for simple goal-directed movements in the framework of the equilibriumpoint (EP) hypothesis. To this end, muscle models of varying complexity are studied in isolation and when coupled to feedback laws derived from the EP hypothesis. It is demonstrated that the dynamical landscape formed by non-linear musculoskeletal models features a stable attractor in joint space whose properties, such as position, stiffness and viscosity, can be controlled through differential- and co-activation of antagonistic muscles. The emergence of this attractor creates a new level of control that reduces the system’s degrees of freedom and thus constitutes a low-level motor synergy. It is described how the properties of this stable equilibrium, as well as transient movement dynamics, depend on the various modelling assumptions underlying the muscle model. The EP hypothesis is then tested on a chosen musculoskeletal model by using an optimal feedback control approach: genetic algorithm optimisation is used to identify feedback gains that produce smooth single- and multijoint movements of varying amplitude and duration. The importance of different feedback components is studied for reproducing invariants observed in natural movement kinematics. The resulting controllers are demonstrated to cope with a plausible range of reflex delays, predict the use of velocity-error feedback for the fastest movements, and suggest that experimentally observed triphasic muscle bursts are an emergent feature rather than centrally planned. Also, control schemes which allow for simultaneous control of movement duration and distance are identified. Lastly, it is shown that the generic formulation of the EP hypothesis fails to account for the interaction torques arising in multijoint movements. Extensions are proposed which address this shortcoming while maintaining its two basic assumptions: control signals in positional rather than force-based frames of reference; and the primacy of control properties intrinsic to the body over internal models. It is concluded that the EP hypothesis cannot be rejected for single- or multijoint reaching movements based on claims that predicted movement kinematics are unrealistic.