How Including Human Workers in Robotically Automated Processes Improves Efficiency and Lowers Cost
Isaac Maw posted on June 10, 2019 |
In-cycle collaboration may be the future of factory automation.

The dream for some factories is lights-out manufacturing, a fully automated production line in which robots and automation handle every task and process without human intervention. With this type of factory, perhaps you could essentially drop off iron ore at one end of the factory and drive a new automobile out the other. Robotics manufacturer FANUC has been operating a lights-out factory in Japan since 2001, with a fleet of robots building about 50 new robots per 24-hour shift. The factory can run completely unsupervised for up to 30 days at a time.

However, some cutting-edge manufacturers don’t see ‘lights-out’ as the way forward: Since the mid 2010’s, Toyota has been mobilizing oppositely, actually replacing robots with people. Humans are still incredibly valuable for their flexibility, coordination and analytical ability. It makes sense— after all, the renowned Toyota Production System relies on continuous improvement suggestions from line workers and engineers.

So, how can the cold efficiency of automation be reconciled with the creative problem solving and adaptability of human workers? Collaborative robots are one approach, but many of these are typically relatively small, weak and slow compared to a robot like the KUKA KR 1000 Titan or even ABB’s IRB 6790 Foundry Prime, which are dangerous and must be installed in restricted areas separate from human workers.

The KUKA Titan is among the largest available industrial robots with a payload of 1000kg.

Veo Robotics has developed one solution that allows robots and automated machinery as large as the Titan or Foundry Prime to work safely alongside human workers, with workers entering the path of motion safely. The system uses mounted sensors to create a 3D point cloud of the work envelope and slows or stops the robot program when a human enters. Check out the video below to see it in action (note the video should start at 1:27). recently spoke with Veo Robotics CEO Patrick Sobalvarro to learn more about the benefits of integrating human workers and automated machinery.

For more on the Veo system, read another part of our interview here: Vision-Based System Allows Humans to Work Safely with Massive Robots.

Patrick Sobalvarro, President and CEO of Veo Robotics
Patrick Sobalvarro, President and CEO of Veo Robotics

It sounds like Veo is useful not just in cells where the human will be working with a robot like in the video above, but also in cells where you plan for the robot to work by itself, as it speeds up the fault clearing process.

That's right. The fault recovery case is actually a pretty important case for a lot of manufacturers.  We talk about in cycle collaboration, and we also talk about fault recovery, and both cases are important on production line.

Guarding takes up a lot of space in a factory. Does the system save any floor space? Workers still need to stay out of the robot’s envelope to allow it to move at full speed.

It does save floor space. When we work with customers, we show them ways in which to design what we call ‘application patterns.’ These are ways in which the design works out so that most of the time, the robot can keep moving while the human is present. For in-cycle collaboration, you don't want to design your process in such a way that the robot is stopped and waiting for the human to finish what they're doing. That would be inefficient. What we typically do for in-cycle collaboration is design it with two fixtures.

Can you describe an example of efficient in-cycle communication?

For example, in performing a process that attaches two things together, such as driving some fasteners and connecting some wiring, in that process step you could have two tables with fixtures on them. Each has a fixture that can receive the part. The robot comes and places the part in one fixture while the human is working on the other fixture, and then when the human is done with the value added work on one fixture, the robot can swing around the back at full speed because the human isn't anywhere close. The Veo system has a dynamic model of the robot and can correct the speed depending on where the human is in the space and where the robot can reach before it can stop safely.

The human then just steps over to the other station, maybe a meter away, and performs the value added work on that fixture while the robot swings around, removes the finished part, places it in an outgoing dunnage, goes to incoming dunnage and places the incoming part in the first fixture while the human worker is performing their value added part of the process step on the second fixture.

How does in-cycle collaboration compare with full ‘lights-out’ automation?

Based on the increase in productivity and the decrease in cost of engineering, having a human present during the cycle tends to dominate compared to full automation.

If you have to fully automate a process step—let's say something simple as driving screws—you have to have a screw feeder, that screw feeder might jam from time to time, you have to be able to replenish the screw feeder. You have to put a cage around it, you have to have some kind of tag in, tag out system. You end up engineering a lot of additional stuff besides just screwing. You have to handle the part, you have to get the part into the fixture, and you have to deal with faults that arise along the way. So, additional machine tending, dealing with faults, dealing with positioning, clamping, the cost of additional fixtures, all that stuff is a kind of over-automation cost.

I think we saw the culmination of that on the Tesla model 3 line. They were trying to fully automate final assembly of those cars, even though there was only one model being run on that line and almost all automotive OEMs run multiple models. It was impossible for them to fully automate it, because the costs of that full automation came in the form of engineering time and in money. You would never amortize that investment over the run of vehicles you’re going to make on that line. (They eventually had to confront reality and build an entirely new line of for final assembly of those cars, and that's how they were successful in actually ramping production.)

How does the cost of the Veo system compare with the cost of traditional safety, which is fences and guards?

Our system with four sensors sells for $40,000. If you wanted to set up a traditional cell that doesn’t require humans to go into the cells very often, you'd probably be spending about half of that on fences and lock out tag out. Fences aren't very expensive. Lock out tag out equipment does tend to be a little bit more expensive, but the question you have to ask yourself is, "How often do I have to go into that cell, and what is my process for going into that cell?"

Let's suppose you have a cell where you know full automation wasn't very expensive, but you have to go into that cell say once a day. How expensive is it to go into that cell once a day in the middle of production? When you tag into the cell, you stop all the equipment in the cell, that means stopping the line. The very best manufacturers we know take about fifteen minutes, just to go through the process, when they have to go in to do something that takes very little time.

Very commonly, it takes longer—one major appliance manufacturer typically takes 45 minutes—and they have to do it a couple times a night. Stopping a household appliance line costs about $3000 per minute. On an automotive line the typical average is about $50,000 per minute. It's very expensive to run a fully automated cell if you have exceptions that occur from time to time.

With our system, you would just walk into the cell, and yes, the robot would pause, and you might need to pause the line, but you would have a safe situation and that process recovery step might only take sixty seconds. Let's say if the palletizing robot should happen to drop a case or part of the cage came open or something like that, it might only take somebody fifty seconds to fix that and then production can resume. There's a huge difference in the cost. 

For more on human-robot collaboration, check out A History of Collaborative Robots: From Intelligent Lift Assists to Cobots

Recommended For You