The problems of rigid-body pose-and-twist estimation are studied here
for purposes of on-line implementation. Using the concept of
measurement subspaces, we describe the conditions under which the
sensing of joint variables of a set of simple open kinematic chains
coupled to a common rigid body lead to linear measurement subspaces.
Likewise, non-contact sensing by means of range finders or cameras can
lead to the same subspaces, from which linear kinematic models are
obtained. The pose and twist of a rigid body are then estimated by the
fusion of these subspaces. A mounted-helmet display seen by a single,
double or multiple-vision camera is used to show the effectiveness of
the estimation procedure.
L. Baron, J. Angeles