Cable-driven hexapods are similar to Gough-Stewart
platforms, but instead of rigid limbs,
they use cables to govern the load. This results
in manipulators with a low weight and a high
load capacity, and permits to attain larger
workspaces. However, additional constraints
apply: their cables can pull, but are unable to
push, which obliges to keep the cable tensions
positive during normal operation.
The Kinematics
and Robot design group develops path
planning methods allowing to operate such robots
automatically [2,3]. The techniques evolve from earlier
work in [1,4] and are
being tested on the Hexacrane system shown
above, designed and constructed by P.
Grosch. This robot adopts the particular
structure of the NIST
Robocrane, but the planning methods remain
applicable to general hexapods. In particular,
all cable anchor points could adopt different
positions if desired, not requiring to be
coincident in pairs. The methods also allow the
planning of the motions when additional
constraints apply to the platform, such as
geometric, or contact constraints.
The following video shows that it is not
difficult to find configurations where some
cables go slack...
... whereas with a proper planning such
configurations can be avoided:
The path
planner in [2,3] allows to
avoid such slack-cable configurations. It
automatically computes "wrench-feasible" paths
between two configurations, where
wrench-feasible means that the cable tensions
will be guaranteed to remain within
predetermined bounds, for a given platform
wrench subject to 6-dimensional uncertainty.
This also implies that no forward singularity
will be met along the path, thus permitting a
full control of the platform motions at all
times.
The following picture shows a slice of the
six-dimensional wrench-feasible C-space, and two
planning queries solved by the planner. One to
connect q1
with q2,
and the other to connect q3 with q4. The
horizontal and vertical axes are two orientation
angles of the platform. The dark gray/green
areas are the regions explored by the planner in
each case. More complex planning queries in the
full 6-D C-space can also be solved.
Current work seeks to extend these techniques to
the context of aerial mobile manipulation. In [5], a 6-wired
aerial manipulation system called the Flycrane
is proposed, and a path planner inspired in [2] is developed for it. In this case,
the control of the load is achieved by
maneuvering three quadrotors independently,
while keeping the cable lengths fixed. The
following pictures show two planning queries
solved by this planner:
The Flycrane getting a
twisted part through a hole
The Flycrane installing a
lightweight footbridge between two
buildings
The Flycrane and its planner are the result of a
collaborative effort between the CUIK
and ARCAS
projects.