Integrated robot vision palletises building slabs

Citation

(2005), "Integrated robot vision palletises building slabs", Industrial Robot, Vol. 32 No. 2. https://doi.org/10.1108/ir.2005.04932baf.002

Integrated robot vision palletises building slabs

Integrated robot vision palletises building slabs

Keywords: Robotics, Visual perception

Vision and robotics have for a long time had a close relationship through the desire for an “all seeing” robot. DVT Corporation and Kuka Automation + Robotics have taken a major step forward to achieving this aim with the development of a robot with seamlessly integrated machine vision based on a two-tier vision management tool that enables simpler operation of the robot's visual faculties. A major building product manufacturer is among the first to adopt this new generation of “vision enabled robots”.

Integrating a robot with a vision system into one single package (Plate 2) required Kuka and DVT to address the three issues of connectivity, programmability and support. Connectivity is via fast ethernet that directly links DVT Corporation's vision system to the Kuka robot controller without the need for any additional devices. Vision is no longer managed separately from the robot, nor does it require its own PC monitor. Rather, vision systems management is accessed directly from the robot's teach pendant in the same way as the robot functions are managed. In other words, there are two GUIs (graphical user interface) sitting on the teach pendant, one for the robot and the second for the vision system (Plate 3).

Plate 2 The new development by DVT and Kuka integrates vision with the robot

Plate 3 Both robot and vision are managed directly from the robot teach pendant

Ian Tatton, DVT's Director of business development for Northern Europe says, “Not only does this reduce cost and simplify operation of the robot's vision, it means the operator's perception is of a robot with integral vision”.

The programming of the vision system to meet these expectations is a ground-breaking new development from Kuka, which changes how robotic vision is managed and operated. DVT's Framework software remains present on board the camera itself but accessed through the robot's teach pendant. This allows a trained operator to set up any number of work routines, inspections, tests, locating coordinates, scans, OCR activities or other tasks. And, through on-going training, supplied free of charge by DVT, operators can keep up with the continual developments in machine vision technology

Vision GUI

“Kuka Vision”, the second tier GUI, utilises DVT's ActiveX control to provide a much simplified, operation- only version of the software. So that while one trained operator is needed to program or set up a variety of tasks, any other operator can perform those tasks without training. Kuka Vision provides the operator with intuitive, one-click controls that govern key functions of the robot's vision system, including tasks such as connect to camera; see or record an image; see data; or switch the robot's product routine.

This reduces the need for trained vision system operators, and also expands vision operations to a wider group of untrained users. It also means that a robot can be loaded up with a variety of vision dependent routines, between which any operator can switch.

Cameras can be located in any position either near a line or area of work, or on the robot itself. The vision system can also be used to locate and correctly identify any number of tools the robot may need to use, such as drills and grinders. All of these routines can be pre-programmed by a trained operator, and implemented at the touch of a button by any user.

Important benefits of robotics are high repeatability and flexibility. But, if the product is not repeatable, or in a repeatable position, then additional, often custom, tooling is needed, which reduces flexibility. Take the typical function of palletising. For a robot without vision to pick a part from a conveyor, requires a fixture to hold the product in a repeatable position, and changing the product needs either another fixture or a complex, overly complicated set of tooling.

This was the problem facing one major building products supplier. It needed to palletise concrete paving slabs that come in 42 product variants, differing in size colour and in shapes ranging from squares and rectangles to quadrants, circles and ellipses. The mechanical solution for a square slab is four paddles that push the slab and centre it ready for pick-up, but this is not easy for circular, triangular or rectangular slabs.

Slab palletisation

The solution for this particular building products manager was DVT-Kuka's vision enabled robotics (VER), which instead of fixtures, uses the vision system to register the position of the slab on the conveyor prior to palletising. The VER accurately determines the centre and the rotation of each slab (positioned randomly under the camera with the conveyor stopped) so as to correctly pick it up and place it accurately onto the pallet, prior to banding and shrink-wrapping. No fixture is required and the slabs can be in any position on the conveyor.

A cost benefit is derived from the fast data transfer speeds. As the concrete slab moves into position, the VER takes three images, compares them and calculates the mean location. Previously the camera and robot were separate and transferring and analysing data between the two could cause a dwell of perhaps a second or more.

Brett Green, Kuka's General Sales Manager explains, “On the basis that we take and process three images, there is potential (with the previous separated system) for the dwell to exceed one or two seconds. The dwell is caused by a difference in fundamental data protocols. With the previous system we took the DVT (vision) data on ethernet, converted it to DeviceNet before taking it on DeviceNet into the (robot) controller. Now that the two systems are embedded together we are effectively as quick as the relative baud rates. We are certainly saving time by taking integration to this level”.

Now the total time to palletise the product is just 4.5 s. The time per palletisation saved, multiplied per hour, per day, per year, means that more products are palletised in effect for free.

Added value

Mr Green, however, sees the overall benefits as much to do with added value as it does to cost and time savings.

He says, “We also use the image to determine if there are any corners missing, or if there is any other defect. If detected, the product is scrapped and not palletised, so quality improves dramatically. There are also benefits for distribution; once palletised no one sees or touches the product again, and the builders merchant or garden centre receives a uniformly packaged product”.

“By being able to interact with its vision system (effectively giving the robot a pair of eyes) allows the robot to change its programs to accommodate what it is seeing”, further comments Mr Green. So as complex routines become programmed into the robot, the enabled vision can be taught to recognise different products that trigger different routines. Just as a robot might use its vision-learnt paths to locate a variety of tools that can be used to perform different tasks, in the building products case, different slabs are inspected and recognised, and then stacked in the pattern specified for that slab type.

Concluding, DVT's Ian Tatton says, “By combining with partners such as Kuka, different perspectives and expertise are opening up new applications for machine vision in robotics and other areas. This helps us to achieve our aim of smaller, faster, more powerful and less expensive machine vision for all”.