The paper addresses a manipulation problem based on kinematic model where the equations
of the motion of the system can change discontinuously depending on the actual state
of the system. In this case, the configuration space is stratified. The stratified
control theory using the smooth motion planning algorithm promises a powerful alternative
for this manipulation problem, however, it meets also some significant difficulties.
The paper presents the main steps of developing computational frame work and some
improvements for stratified motion planning. The concept proposes a semi-stratified
method which decomposes the manipulation problem into stratified motion planning and
pure finger relocation through the special selection of the reference points. Because
of the different type of computational tasks (symbolic, numerical) , the realization
steps are also investigated.
