Page 203 - Sensing, Intelligence, Motion : How Robots and Humans Move in an Unstructured World

178 MOTION PLANNING FOR TWO-DIMENSIONAL ARM MANIPULATORS

robot arm manipulators are way more important than mobile robots. According
to the UNECE (United Nations Economic Commission for Europe) report “World
Robotics 2003” [101], by 2003 about 1,000,000 industrial robots had been used
worldwide. By far most of these robots are arm manipulators.
And yet, while at least some commercial mobile robots come today with
rudimental means for sensor-based motion planning, by and large no such means
exist for arm manipulators. Exceptions do exist, but as a rule they relate to motion
planning of the robot end effector, the tool, rather than the whole robot. This is
certainly not because of lack of interest. If available, such systems would find an
immediate and wide use—even in the same industries where arm manipulators
are used today—by helping decrease the cost of systems. In a typical industrial
system, the cost of the robot is a fraction—perhaps 20% or so—of the total
cost of the work cell. Much of the rest are means to compensate for the robot’s
inability to avoid collisions with its surroundings on its own.
Motion planning systems would also allow robot manufacturers to move their
products into new domains—agriculture (to pick op fruits and berries and other
crops), nursing homes (to help move and feed patients), homes of the elderly (to
help them handle various home chores), outer space (to assemble large structures,
such as telescopes and space stations)—in short, to a whole slew of applications
that are good candidates for automation but could not be automated so far because
of the high level of uncertainly characteristic of such tasks.
There are two major reasons as to why commercial robots intended for a high
level of uncertainty are not here yet. First, appropriate theory and algorithms are
just beginning to appear. Second, the sensing technology that is required for such
algorithms to operate is also at the development stage. (The issues of sensing
hardware is addressed in Chapter 8.)
Whatever research has been done on motion planning for arm manipulators,
its lion share relates to motion planning for arm hands and grippers. Collision
avoidance for the rest of the robot body has been largely left out. Again, this is not
because of the lack of need. A quick glance at a layout of a typical robot cell with
a robot arm manipulator will show how crowded those cells are. The problem of
handling potential collisions for the whole body of an arm manipulator is acute.
Works that focus on robot hands’ collision avoidance do of course advance the
general progress in robotics. One can imagine applications where the designer
makes sure that potential collisions can occur only near the arm hand. On a robot
welding line in an automotive plant, the operation can be designed and scheduled
so that no objects would endanger (or would be endangered by) the robot body.
Because the robot tool must be close to the parts to be welded, this cannot be
avoided, and so the robot hand will be the only part of the robot body that can
come close to other objects. This way, unpredictable events can happen only at
the tool: Parts to be welded may be positioned slightly off, their dimensions may
deviate slightly, one part may be slightly bent, and so on.
Hence the designer of such a system will seek collision avoidance procedures
that will take care of the arm’s hand only. In our example, such algorithms would
lead the gun of a welding robot arm clear of the parts being welded. Providing the