Human-Robot Collaboration

The system perceives the 3D space around the robot in a safe way using a vision system and classifies the items in the entire volume of the space.

Over the past 20 years, the manufacturing environment has changed dramatically. The rise of both domestic and international competition is driving the market to look for new and improved ways to increase quality and productivity, while maintaining a level of flexibility on the shop floor. One of the big trends moving forward is human-robot collaboration.

Why Human-Robot Collaboration?

"One of the things that we hear the most about from manufacturers is the need for flexibility, especially with the level of highly optioned products that are on the market," said Patrick Sobalvarro, CEO and co-founder of Veo Robotics, Waltham, Mass.

Sobalvarro gave an example of a refrigerator, which can be purchased with all sorts of configurations. It can include a screen in the front, an icemaker, various surface coverings, and it can be customized.

"This highly customizable manufacturing environment creates a need for flexibility on the part of the manufacturer," he explained. "At the same time, full automation is not possible. The cost of fully automating a process is so high that many manufacturers really can’t amortize it over the run of parts. You still need production labour, which of course has plenty of advantages in terms of being able to address quality issues. People have all this dexterity that is impossible to replicate mechanically today."

However, there are areas where bringing together both robots and humans makes sense. Sobalvarro gives the example of repetitive tasks that make sense to automate like removing components from dunnage and fixturing so that value-added steps can be completed by a human worker. The material handling tasks don’t really add overall value and tend to offer safety and ergonomic challenges for the worker.

"It is in these cases or parts of the process that we can have the robot performing these tasks," said Sobalvarro. "But you still need human workers to perform the value-added processing steps. What we do is we make it possible for humans and robots to work together, and that gives you the best of both worlds."

How It Works

Simply stated, the Veo system, FreeMove™, is designed to add brains and eyes to industrial robots. For example, if a stamping press within a caged robotic workcell has some sort of issue, in order to clear the failure, the shop needs to find the person with the cage key, wait until it’s opened, and enter the workcell. This can take a few minutes or even longer depending on what the failure is. Once the robot is ready to go back into production, the person closes the cage, certifies that there is no one inside the work zone, and resets production. It can be extremely time-consuming.

"What we do is give the robot awareness of what’s going on around it and in particular where people are or could be," said Sobalvarro. "By providing that awareness, we make it possible for a production worker to just go up and clear that fault. It’s a much more fluid process. It takes much less time and you get back to production right away. And we do that in a safe way."

At Veo Robotics, most of the engineering work is focused on software. The system uses advanced computer vision and 3D sensing to perceive the safeguarded environment around the robot. It has to understand the entire work cell volume and know where the workpieces are, where the machinery is, where the robot is, and where and what could or could not be a person.

"It also identifies the places that it can’t see," said Sobalvarro. "So for example, suppose you have a fixture in the workcell like a table. Somebody could crawl under the table, so we have to keep track of that as well. The system is aware of occlusions within the space and potential spaces for occlusion as the robot moves around holding parts. Those parts create shadows that none of the sensors can see into. So we have to track that information in case the person could get into those occlusions. We can’t allow the robot to move in a way that could cause it to collide with a person or the part."

Simply stated, the Veo system is designed to add brains and eyes to industrial machinery, particularly robots, to allow safe and effective human-robot collaboration.

The company produces its own 3D time-off-light sensors and processing platform computer to ensure the entire system is safety-certifiable. For Veo, safety is key to making all of it system possible and where it invests its work time. To make a robot that is safe for interaction with people, many integrators use power and force limitation. If the robot moves slowly enough and doesn’t weigh very much, it can notice instantly when it bumps into a person. This can be used safely but depends on what the robot is carrying. For example, if the robot is engaged in a deburring step, it will be carrying sharp pieces of metal. If a sharp piece bumps into a person, it could be very dangerous. In these situations, power and force limitation is not a safe option.

"We take a different approach that allows us to make big robots safe," said Sobalvarro. "We perceive the 3D space around the robot in a safe way using a vision system, and we classify the items in the entire volume of the space. Then it connects directly to the controller and controls the speed of the robot to make sure it’s in a safe state by the time a person could approach it."

More Safety Considerations

The Veo system is built to meet ISO 13849 functional safety standards and has to be certified for inclusion in an ISO 15066 collaborative workcell.

"The approach we take has many elements to it," said Sobalvarro. "On the hardware itself, everything is dual channel redundant and hardware fault tolerant. This is the reason we had to make our own sensor and computing platform. Our sensor has two time-of-flight sensors whose output are compared for consistency. Our engine has two motherboards, which are independently performing the computation. Finally there’s another processor, which is our safety processor that compares the output of those two independent computer engines. It’s also doing health monitoring on the sensors and the rest of the system. That’s how we ensure that our hardware itself is fault-tolerant and fault-resistant."

Working With Robots

The Veo system came out of many conversations with manufacturing engineers and hearing about their needs. One of the biggest considerations was that both system integrators and manufacturers love the robots that they currently have—they are reliable, durable, work well, and can maintain positioning repeatability within half a millimeter over 100,000 hours—they love these robots and want to keep using them.

"Our approach was to take a very light touch on the controller," said Sobalvarro. "What we do is connect directly to the robot controller to slow and stop the robot as nessessary. But manufacturers would program the robot the way they always would program it."

Currently, the company works with FANUC, Yaskawa Motoman, ABB, and KUKA to support robots of different shapes and sizes. Veo must have a dynamic model of the robot itself for the system to work. With this, the company is able to use the numbers from partners that provide the stopping distances, the acceleration, and the torques of the actual robot. It then calculates a protective separation distance effectively. Beyond this, the robot is required to have a functional safety unit within the controller. These are often optional and can be purchased from the robot manufacturer and installed to allow for the Veo system to interface with it.

A Collaborative Workspace

"For productivity and efficiency, we are talking about in-cycle collaboration," said Sobalvarro. "So every time you go through a process step, you have a human and a robot working together. In cases of parts presentation, the robot presents a part to the human worker in exactly the position that the part needs to be in so that the human can perform his or her part of the process steps. There are cases like fixturing, where the robot fixtures and unloads fixtures, while the human performs the value-added steps. There are a number of application patterns in general. That’s where we see the most productivity increases. In general, 40 per cent of the production worker’s time is saved."

For example, a welder brazing cutter heads on a piece of oil drilling equipment like a drill bit may be required to move large, heavy pieces of equipment. The welder would take the part from incoming dunnage with a lift-assist device, fixture it on a rotary table, remove the lift-assist device, and start performing the brazing step. Once done, the welder has to get that lift-assist device, move the part off the rotary table, and place it in outgoing dunnage.

"The value-added elements of the process step may take longer than the fixturing," said Sobalvarro. "That welder’s time may be worth $75/hour. And the quality of their work is highly dependent on how that part is presented to them. With our system, we can present the part to the welder in exactly the position needed so that the welder is constantly brazing in the best ergonomic position. They don’t have stress injuries that might come from a bad lift. So quality can go up as a result of improved ergonomics."

The company produces its own 3D time-of-flight sensors and computer to ensure the components are safety certifiable. For Veo, safety is key to making all of it possible and where it invests its work time.

Efficiency and productivity gains come from fault recovery situations. In a production environment where time is money, any downtime can be costly. With lots of stops that require time-consuming processing like opening cages and clearing faults, production time is wasted. With the Veo system, these faults can be cleared in minutes as there are zero barriers preventing the worker from entering the workcell.

Sobalvarro noted that in some cases, workers are injured by equipment because they have to make a quick fix and attempt to override or circumvent the safety mechanism. He gives the example of climbing over the fence rather than wait for the key.

"With our system, if you go into the cell, we’re going to see you," he said. "If you try to disable the system so that we can’t see you, we’ll notice a fault immediately and we’ll stop the system. We can enhance safety in particular by monitoring the entire 3D volume. From the beginning we have been focused on safety. We want to make it so that manufacturing is more flexible and for machinery to be more intelligent around people. But it comes with a great responsibility to make sure you do it the right way. And that’s what we are doing."

Associate Editor Lindsay Luminoso can be reached at lluminoso@canadianfabweld.com.

Photos courtesy of Veo Robotics, www.veobot.com