Cobots are red-hot now, and we visited the Ann Arbor, MI U.S. headquarters of Universal Robots (UR, with worldwide headquarters in Odense, Denmark) to check out some of the latest developments. We looked at the capabilities of a ActiNav, a pick-and-place system, here shown on the UR5e, the company’s medium-sized cobot that fits this application, among many others.
Before we begin, let’s agree to this: the world of manufacturing and the world of robotics is advancing in lockstep. Only a few years ago, the shop floor was brought together into a single platform (in most cases by management software). It was all part of Industry 4.0 and it was good, the connectivity was good, the information was good. It was a platform of products. Now, as we’re in the early phase of the fifth wave, the shop floor is becoming a platform of platforms. Sometimes a machine tool is not just a machine tool, and today’s robot is not merely a robot.
Such is the case with the Universal Robots eSeries robots. The cobot itself is a platform. To this platform add compatible plug and play accessories, or even full application kits that typically comprise several components plus software. With these additions, the cobot can match up to the challenges of a specific application.
To ensure all these parts will work together well, Universal Robots offers the UR+ program in which the company tests and certifies third-party products as being compatible accessories or application kits. In fact, what we are about to explore is an application kit by Universal Robots called ActiNav. The ActiNav kit, along with other UR+ accessories, are added to the UR5e cobot to create a complete work cell. (ActiNav also works with the UR10e and UR16e cobot models.)
Let’s take a look at a part-picking system in an animation from UR:
The system we ran consists of the UR5e cobot, the “teach pendant,” the cobot controller, a UR+ standard bin picking frame, an end effector, a bin, and the ActiNav kit. The kit includes a 3D sensor, a motion and vision module (a computer), and software that runs on the teach pendant. (You can choose from many end effectors that are in the UR+ program, and if you’re so inclined, you can make one yourself.)
Let’s familiarize ourselves with the pieces of this system.
Robotics meets Gestalt theory
The idea of course, is to have a whole something that is greater than its constituent parts. The cobot is a popular one and flexible as well. The UR5e is at its best in automating low-weight (up to 5kg/11 lbs.) processing tasks. It has six rotating joints and a reach of 850 mm (about 33.5 in.). Including the cable, it weighs 20.6 kg (about 45.4 lbs.), so it’s fairly easy to move if necessary.
Although the inputs and results are shown on the teach pendant, the computing for ActiNav’s motion and vision software actually takes place on the AMM (Autonomous Motion Module) which is connected to the cobot controller by an Ethernet cable. The UR controller runs the basic operational software of the robot (we used PolyScope version 5.11, the graphical user interface or GUI that runs on the pendant), along with any of the approved apps (“URCaps”) that come from Universal Robots or its partners. (All software is enabled by providing the developer with APIs; it’s then certified. Installation is done through a USB port on the teach pendant, which is connected to the control box.) To the user, it appears that the software runs on the teach pendant because that’s where inputs are made and results are displayed.
We begin at the end (effector)
Understanding the setup is easiest when we start at the end—the end effector, the thing that grabs and moves. Just prior to the first video, Universal Robots’ Senior Product Manager, Eric Andersen, connected a suction end effector to pick and place parts. In our next video, he takes us through setting up the system and having it “discover” and configure the end effector:
It only takes a minute or two to do this, and the resulting precision pays off in every successive pick and place operation.
The next step sounds a little like parenting: you have to teach your cobot about the world. Where to go, what to avoid, how to be safe. Some of the things to avoid are the frame upon which it rests, its own base, the bin from which it extracts parts, the control box, any walls nearby, and any other equipment or structure in the shop that’s within range. Using ActiNav, (we used version 1.5.0), you can create areas to avoid. To the person running the cobot, the designated avoid area appears as a green bubble or plane.
Andersen shows us what this looks like in the pendant’s interface, and how to program it:
Now we get a dose of the real world. In fact, we learn that we should trust the real world more than any CAD drawing. We also learn about the concept of a singularity—but not the kind that Stephen Hawking told us about. Here is Andersen to explain one type of singularity:
To learn more about singularities in robotics, UR has a support article about it at https://www.universal-robots.com/articles/ur/application-installation/what-is-a-singularity/ that explains further.
Next, we depart from traditional robotics and move into today’s robotics. Until recently, the legacy (and the goal, really) of robotics was to repeat the same action exactly for hours, days, weeks, months, years. Now we’ve introduced a starting point that has a couple of significant degrees of randomness. A bin of parts means that our target part could be at any x, y, or z coordinate within the bin. Additionally, the planar surface we want to use to pick the part could be at any angle of roll, pitch, or yaw. Instead of moving each part to a consistent orientation and pickup point, the system functions in the real world to assess the positions of up to five parts at a time, and plan the paths to pick those parts.
While this intelligence is baked into the system, Universal Robots makes no claims of artificial intelligence. Instead, it relies on its ActiNav software to process the information gleaned from the sensor. In effect, we are granting the system the right to make the pick happen rather than telling it where and how it should happen. Here is Andersen to demonstrate this important point:
Sometimes, when a part is picked, it sets off a cascade of parts that move as a result (think of losing at Jenga). Or sometimes it’s fairly spontaneous, and that’s when a cobot can “miss” a part—it uses the approach path, but when it gets to the destination, no part is there to grab. In this short video, we see what the system does in this case:
In our final video, we see what really happens when we tell the cobot to pick and place something. It’s also a great example of how we can change the cobot’s job in a very short time span:
As Universal Robots’ CEO Kim Povlsen said in our interview with him, “Our vision is this: we want to create a world where people work with robots and not like robots.” In other words, we can offload not just the dirty, dull, and dangerous parts of the job—the things we don’t want to do–but also the ability to make autonomous decisions about how to get the next part.
More information: https://www.universal-robots.com/products/ur5-robot/
More information: https://www.universal-robots.com/products/actinav/