Fast recognition of postures for a simplified three-fingered artificial hand

Real-time navigation is one of the major trends in mobile robotics. Visual navigation of a hand requires integration of visual processes and motor control. Fast computer analysis of articulated objects is only possible under strong constraints relative to features to identify, postures obtained by grouping processes and motions connecting postures. In the paper we obtain a symbolic or virtual local representation of a two-fingered articulated hand, in terms of angles measured at 0-dimensional features (double and triple junctions), which are connected by means of meaningful segments in planar representations. Knowledge about the geometry of the artificial hand and kinematics of articulated mechanisms is applied to provide a fast symbolic representation for postures, in terms of matrix spaces which we store as certain type of grassmannians. Geometric information about kinematic multichains corresponding to fingers is used for predicting and tracking evolution of symbolic representation associated to postures by describing linear symbolic models for physiological restrictions based on artificial mechanisms