Overlaying objects, as shown here, present
a challenge for bin picking, but a simple
solution that avoids this is possible
Inset: Robotiq’s new bin-picking kit
Ask these
10 questions
Robotic bin picking has been touted by many as an answer to automating a variety of process
applications, but in many industrial sectors, take up of this type of technology remains at a relatively low
level. Nicolas Lauzier, Ph.D., senior product manager, Robotiq, offers this overview
There is no shortage of bin-picking
solutions on the market, but barriers to
entry – in particular, the complexity and
cost of the vast majority of systems –
continue to be considerable. With the
signi cant integration and programming effort
required, it is perhaps no surprise that most
real-world bin-picking deployments are found
at large, sophisticated manufacturers such
as automotive OEMs. Adoption among SMEs
is much more sparse, yet it is within SME
organisations that the vast majority of bins
and processing equipment are found and,
therefore, where the greatest scope for
automated bin picking would seem to be.
However, for any organisation needing to
process continuously, a solution has to be
found to feeding processing equipment with
components consistently. Reliance on human
operatives is not always an option – the work
is dull, repetitive and unrewarding, meaning
even in areas or times of higher
unemployment, nding individuals willing to
do it is a major challenge. Meanwhile the
physical effects of bending into bins to
extract components can create health and
safety issues for both operatives and their
employers. Furthermore, human operatives
are prone to becoming tired or distracted,
potentially leading to inconsistent placement
of components, with associated issues down
the line, or even unplanned stoppages to
production.
THE CHALLENGES
It is, in theory, a fairly simple task for a
human to pick up an object from a bin and
place it at the correct location and
orientation, usually onto a conveyor.
However, selecting objects in random
positions, overlapping and on top of each
other, then orientating them and placing
them correctly, presents myriad challenges
for robotic systems.
The robot must be able to gather parts in
an in nite number of positions and reach into
the deepest corners of the bin, while avoiding
collisions with the bin, other parts, or the
work cell itself. This is highly challenging for
robot vision, with the issues presented being
those such as occlusion, whereby some
objects are partly or completely obscured by
others on top of them; and lighting, with
objects casting shadows on each other,
hiding them from the camera.
Another issue is edge detection: for
objects of the same colour and material,
it can be unclear where one object starts and
the other nishes, making it hard to detect
the outline of each individual object. All these
issues occur with both 2D and 3D robot
vision, but particularly with 2D vision, as
these issues can make it almost impossible
to detect individual objects.
These applications have generally required
complex sensing systems, detailed models of
the part, bin, end-effector, placement targets
and any obstacles, plus real-time collisionfree
path planning algorithms. In short, this
results in an expensive and time-consuming
system integration project that also requires
detailed programming know-how. Fixed
cameras and additional lighting may be
needed, too, depending on the application
and location.
A robotics expert has to integrate the
sensor, computer, software and robot
controller, then write a program to retrieve the
location of each part and determine how to
get it to the placement target. Planning a
unique, collision-free path for each part in the
14 November 2020 | www.machinery.co.uk | MachineryMagazine | @MachineryTweets
/www.machinery.co.uk