| Abstract | Develops a task-dynamic approach to skilled movements of multi-degree-of-freedom effector systems in which task-specific, autonomous action units are specified within a functionally defined dynamical framework. Qualitative distinctions among tasks (e.g., the body maintaining a steady vertical posture or the hand reaching to a single spatial target vs cyclic vertical hopping or repetitive hand motion between 2 spatial targets) are captured by corresponding distinctions among dynamical topologies (e.g., point attractor vs limit cycle dynamics) defined at an abstract task space (or work space) level of description. The approach provides a unified account for several signature properties of skilled actions: trajectory shaping (e.g., hands move along approximately straight lines during unperturbed reaches) and immediate compensation (e.g., spontaneous adjustments occur over an entire effector system if a given part is disturbed en route to a goal). These properties are viewed as implicit consequences of a task's underlying dynamics and do not require explicit trajectory plans or replanning procedures. Two versions of task dynamics are derived (control law, network coupling) as possible methods of control and coordination in artificial (robotic, prosthetic) systems, and the network coupling version is explored as a biologically relevant control scheme. |