Kinematic feedback control laws for generating natural arm movements.

摘要: We propose a stochastic optimal feedback control law for generating natural robot arm motions. Our approach, inspired by the minimum variance principle of Harris and Wolpert (1998 Nature 394 780–4) principles put forth Todorov Jordan (2002 Neurosci. 5 1226–35) explaining human movements, differs in two crucial respects: (i) endpoint is minimized joint space rather than Cartesian hand space, (ii) we ignore dynamics instead consider only second-order differential kinematics. The motions can be straightforwardly obtained backward integration set ordinary equations; these equations are exactly, without any linear–quadratic approximations. parameters to determined priori scale factors, both two-DOF planar seven-DOF spatial arm, table values constructed based on given initial final configurations; via an fitting procedure, consistent with existing findings about neuromuscular motor noise levels muscles. Experiments conducted two-link verify that trajectories generated our closely resemble motions, sense producing nearly straight-line trajectories, having bell-shaped velocity profiles, satisfying Fitts Law.