Safety and control in collaborative robotics

Robotics technology is moving at the speed of light, and the standards process is struggling to keep up. New types of collaborative robots can be safely implemented: Do a risk assessment, work with experienced suppliers, and get involved with standards development.
By Tanya M. Anandan August 7, 2013

Human collaborative robots are expanding automation possibilities and shedding old barriers. Liberated from their enclosures, these cage-free robots are easier to implement and easier to use. They are making automation more accessible to small and midsized businesses, and allowing manufacturers of all sizes to have more flexibility in automating diverse tasks. With these advantages come new concerns. At the top of the list is the safety of their human coworkers.

New safety standard raises more questions

In May, the Robotic Industries Association (RIA) announced the updated robot safety standard approved by the American National Standards Institute (ANSI). The new ANSI/RIA R15.06-2012 standard is an update to the 1999 edition and is now harmonized with the international ISO 10218:2011 robot safety standard. One of the key updates to the globally harmonized standards addresses human and robot collaboration.

Conceptual applications of collaborative robots from ANSI/RIA R15.06-2012, Annex E, p. 111 Courtesy: Robotic Industries Association

The standard identifies four requirements for collaborative robot operation, which allow humans to be in the vicinity of an operating robot without safety enclosures. Robots in a collaborative scenario must satisfy at least one criterion to meet the standard:

  • Safety-rated monitored stop
  • Hand guiding
  • Speed and separation monitoring
  • Power and force limiting.

An international committee has drafted ISO Technical Specification (TS) 15066 to provide guidance on using the new standard for collaborative operation. The subject matter is still under technical development, and an industry consensus needs to be achieved before it can be recognized as an amendment to the ISO 10218 safety standard.

“The international group is working on determining what sort of finite limits we may be able to put into the standard to help provide guidance for both robot manufacturers and end users,” said RIA’s director of standards development, Pat Davison. “Right now, the standard says that power and force limits need to be enacted, but it doesn’t designate what is allowable power and what is allowable force.”

ISO TS 15066 will also provide guidance on speed and separation monitoring, including determining the minimum separation distance and establishing maximum safe speed. Adoption of the specification is expected in 2014.

In the meantime, how are users of these new collaborative technologies supposed to ensure safe implementation without specific guidelines? And where does OSHA stand on human and robot collaboration?

“Robot safety falls within OSHA’s general duty clause which says that the employer has to provide a safe work environment,” said Davison. “OSHA does not specifically address human and robot collaboration. Instead, OSHA references industry standards such as the RIA standard.”

Risk assessment helps safeguard the collaborative workspace

Davison notes the importance of identifying and thoroughly analyzing the risks involved with implementing any new robot application.

“The standard mandates that a risk assessment be performed for each new robot application. Risk assessment looks at all the tasks that are going to be associated with that application over its lifecycle. It evaluates all the potential hazards associated with those tasks and determines ways to mitigate those hazards, either by reconfiguring the system so the hazard is eliminated, or adding safeguarding or other mechanisms that reduce hazard exposure.”

Gil Dominguez, PE, CMSE, is a safety consultant with Pilz Automation Safety L.P. in Canton, Mich. He is on the R15.06 Subcommittee for Industrial Robot Safety with Davison and is also involved in the international committee working on TS 15066.

Dominguez said that robot manufacturers are using TS 15066, even in its preliminary state, as a guideline for initial collaborative robot designs, and that end users are implementing systems following the recommendations in the draft specification. He has performed risk assessments on collaborative robotic systems and stresses the importance of thoroughly assessing the potential hazards in these scenarios.

“Because you’re using robots in a totally different way than we’ve always done in the past, it becomes even more crucial to perform a detailed risk assessment on the hazards and exposures of the person working adjacent to a live robot,” said Dominguez. “And given the complexity of the human body, the interactions and exposures that can occur have to be analyzed more precisely.”

Dominguez and Davison acknowledge that the technology is ahead of the standard and that much work still needs to be done to provide specific guidelines to robot manufacturers, integrators, and end users. Ultimately, the onus for providing a safe work environment lies with the end user.

It helps to understand the different modes of collaborative operation and the robotics technologies that are leading the way.

The technology within

Each of the four modes of collaborative operation protects users in different ways,” said Dominguez. “In safety-rated monitored stop, you’re protecting people by keeping the robot from moving when a human is in the collaborative workspace. In hand guiding, you’re protecting people by allowing the robot to move only when the robot’s motion is under an operator’s control.”

“In speed and separation, you’re protecting people by monitoring where the person is in relationship to the robot, and slowing down and eventually stopping the robot if the person gets too close, and eventually even altering the robot’s path,” explained Dominguez. “The biggest challenge we have in collaborative robotics is coming up with control systems that can monitor people and robots in real time to ensure, at a high integrity level, that a collision which could cause injury cannot occur.”

“In power and force limiting, you have to make sure that the forces exerted on someone’s body if they do come in contact with the robot are low enough that any injury which may occur would be considered minor. But different parts of a person’s body are more sensitive than others.”

Herein lies the dilemma for the standards writers. Exactly how much power and force is too much? Some robot manufacturers aren’t waiting around for the answers.

Baxter’s series elastic actuators

Baxter on the job kitting parts at a plastics manufacturing facility. Courtesy: Rethink RoboticsRethink Robotics’ Baxter received a lot of attention after it hit the U.S. market in September 2012. An all-purpose shop floor tasker, Baxter is designed to use common sense and work safely alongside human coworkers without the need for costly programming. Baxter is touted as “inherently safe.”

The patented technology that gives Baxter its safety prowess was born in the lab at the Massachusetts Institute of Technology. “I was doing my master’s at MIT with Professor Gill Pratt,” said Matthew Williamson, director of technology development at Boston-based Rethink. “Between the two of us, we invented the series elastic actuator (SEA).”

Williamson said he eventually did his doctoral work under the supervision of Rethink’s founder and CTO, Rodney Brooks, who was an MIT professor at the time. “Rod used the SEAs in his research robot arms, including the ones I used in my thesis.”

“It’s the series elastic actuators that make Baxter inherently safe,” said Williamson. “They make the robot compliant as opposed to stiff. It’s the difference between being hit by a spring and being hit by something rigid.”

The SEA consists of a motor, a gearbox, and a spring. Williamson described how the mechanism senses and limits force. “The way the SEA works is that you measure the twist of the spring to control the force output, and that measurement of the twist of the spring gives you a force sensor.”

“Electric motors, which are the predominant actuator technology, are very good at position control, but they’re not very good at force control,” Williamson explained. “A spring is a good way of converting a position to a force, because of Hooke’s Law, force = kx. Force control is an enabling feature that is good for working in unstructured environments, which has been our goal with Baxter.”

“And then it takes a fair amount of engineering to get that right, in terms of the spring design and the choices of the motors and gearboxes, and so on,” added Williamson. “In Rethink’s case, the challenge has been making it at an affordable price point.” As of late July, Baxter starts at $22,000, including built-in sonar and camera sensors to detect humans when they enter the robot’s space, and integrated vision for object detection.

“We have a robot that is lightweight (75 kg) and low power, so the energy in the system is reduced,” said Williamson. “But on top of that, the force sensing allows us to do force and position control, which allows us to control the robot’s speed to make sure that it won’t damage anyone. We also have systems for detecting and responding to collisions and making sure the robot doesn’t clamp, compress, or crush a person.”

The SEAs in Baxter’s flexible, back-drivable joints allow him to achieve power and force limiting collaborative operation under the new safety standard.

Universal’s torque sensors, sophisticated software

A Universal robot tackles the repetitive job of applying glue to kitchen sinks. Courtesy: Universal RobotsUniversal Robots’ UR series sport a lightweight design, as low as 18 kg, and patented sensor technology that contribute to its power and force limiting operation.

 “We monitor the current of the motors as well as the position of the encoders, so we’ve got redundant encoders in every joint. By looking at current and looking at position, we’re able to derive force,” said Edward Mullen, national sales manager at Universal Robots USA Inc. in Stony Brook, N.Y. “It also keeps our price point lower ($34,000), because we don’t have expensive sensors in every single joint.”

“The robot knows the required amounts of force to pick up a load and move it. When it recognizes an increase in torque or force required for movement, such as in a collision, the robot arm safely stops without causing harm,” said Mullen. The UR’s control system is redundant so that any dangerous failure forces the robot to fail in a safe condition.

Whereas Universal’s UR series and Rethink’s Baxter achieve collaborative operation through power and force limiting, traditional robot manufacturers achieve it by satisfying other criteria.

Next page: see about control-based collaborative operation, photos, video links

Motoman’s control-based collaborative operation

Yaskawa’s dual-arm robot demos a lift-assist application. The dual-arm robot has a 10-kg capacity per arm, but by using a standard lift assist device, it can move and precisely place a robot drive that weighs 80 kg. Courtesy: Yaskawa Motoman RoboticsYaskawa Motoman’s dual-arm SDA robots and single-arm SIA models use safety-rated monitored stop. In some cases, they also use speed and separation monitoring.

Unlike these other collaborative robots, the Motoman systems maintain the speed and power associated with traditional industrial robots, and their six-figure price tags reflect that performance advantage. Motoman robots are designed to work by themselves the majority of the time, but optional safety features can be added to allow for collaborative operation.

“The decision we made as an industrial OEM is to let the robot go fast when it can, when the people aren’t there, and then let it behave in a safe manner when people are present,” said Erik Nieves, technology director for Yaskawa Motoman Robotics in Miamisburg, Ohio.

The safety-rated monitored stop component of Motoman’s optional functional safety unit is in the robot controller. External safety-rated laser sensors monitor when a person enters the robot’s interactive space, signaling the robot arm to slow down and stop.

“We have signals being processed in the controller that change the behavior and the speed of the manipulator,” said Nieves. “Our robot continues to be a powerful unit, but we slow the arm down through control. We have to assure safe operation, because otherwise our robot would knock you on your can.”

Nieves demonstrated Motoman’s single-arm model in an assisted-assembly application the same day the new safety standard was released. “In that instance we have safety-rated sensors that are cognizant of where I am as the operator. When I enter the robot’s space, the functional safety unit automatically slows the robot down. In that regard, we qualify for collaborative operation under speed and separation monitoring.”

“Safety-rated sensors have built-in fail-safes,” added Nieves. “The signals go to two separate places in the robot controller, and those signals are then processed by separate hardware running different algorithms. Those two pieces of software and hardware are then doing cross-checking between them. And if either one of them drops, the system goes into a safe condition. Robot controllers that are third-party certified for functional safety per the standard have to demonstrate that level of redundancy.”

Similar to Yaskawa, other traditional robot manufacturers with systems that meet the safety standard, such as Kuka and ABB, have hardware or software, or both, that allow their robot systems to operate collaboratively with their human coworkers.

The collaborative robot space is occupied by a wide variety of robots—from Baxter, the human-like, multipurpose tasker, to the Motoman high-speed, high-precision workhorse, and Universal’s nimble six-axis arm with performance specs somewhere in between the two. So what does this mean for end users and their safe implementation of this new breed of fence-free robot?

Implementing collaborative robots

In the case of Rethink’s Baxter and Universal’s UR series robots that require little, if any, integration, the responsibility lies with the user to set up and put the system into production.

“Most of the time the end user is taking on the whole responsibility of unpacking it, bolting it down, and integrating it into their processes themselves,” said Universal’s Mullen. “Usually they bypass the integrator task. That puts the pressure on the end user to make sure they do their due diligence with regard to risk assessment.”

Universal sells its robots through distributors. The Braas Company in Eden Prairie, Minn., has been distributing UR robots since Feb 2013.

“It’s the person who owns the automation, the end user, who has the responsibility to verify that they have a safe work environment,” said Braas president, Matt Gallagher. “This new category of robot exists, but it’s not a panacea for safeguarding. There are unsafe work environments that can be created around a collaborative robot.”

“Something like the gripping mechanism could be deemed hazardous or maybe the machine that the robot is interfacing with,” said Gallagher. “You could have an unsafe situation created by a machine door being opened and closed automatically that the robot is controlling. Just the fact that it’s a collaborative robot doesn’t negate the need to do an evaluation of your work environment.”

Universal claims that 80% of its systems are operating without traditional safety enclosures. Gallagher said this holds true to his experience. “So far, all the Universal robots we’ve sold have operated without guards.”

“Universal Robots offers us the chance to sell robots to customers that may not have bought them before,” said Gallagher. “And safety is really just one component that drives that accessibility, probably one of the smaller components. The key feature that makes it more accessible to our customers and also minimizes the need for that traditional integrator channel is the UR’s ease of use.”

“I think what Universal has really hit on the head is a product that demystifies robots for general-purpose applications and first-time robot users. The interface is very accessible. Anyone who has navigated a Windows environment and can understand a flowchart type of structure can navigate the robot programming pretty quickly. And if it’s a shop that has smaller-run types of products, the ease of use allows them to move the robot around and decouple it from one production area and set it up in another area in a very short time period.”

Applications for the UR series have included material handling, machine tending (injection molding and traditional CNC), and assembly applications.

Rethink’s Baxter is best suited for mundane, repetitive tasks that don’t require a lot of skill or speed, or similar tasks with varying parts. “We’re finding that both small companies and larger companies have parts of their manufacturing that have both of those aspects, either low-value things like packaging and kitting, or parts that vary. So you would not want to invest all the effort in structuring the environment and integrating an industrial robot only to have to redo it three or four days later,” said Williamson.

In the lab at P&G

Larger manufacturing companies often have forward-looking labs that test new technologies before they put them on the line. For Procter & Gamble, this is standard practice. The Cincinnati-based multinational manufacturer of personal care and household cleaning products routinely tests new robotics applications and technology on a small scale before introducing it to the larger organization.

“We’re in the process of evaluating the Universal robot arm (10-kg payload UR10) for potential applications within our manufacturing and our R&D environments,” said Mark Lewandowski, robotics technology network leader at P&G’s Beckett Ridge Technical Center in West Chester, Ohio. “We’re testing a lot of applications right now where we have either highly manual operations, such as material handling custom packs of items, or areas where we have slower speed, repeatable motion.”

“We’re working with our safety personnel to evaluate the safety aspects of the arm and figure out how we want to deploy it in a manufacturing environment in a safe way,” said Lewandowski. “We have the capability in-house to do our own risk assessment. We’re also evaluating the robot’s capabilities in terms of speed, repeatability, accuracy, and types of applications.”

Lewandowski is on the RIA Board of Directors and participates in the R15.06 safety subcommittee. He notes a number of resources available to help collaborative robot users evaluate their own safety practices.

“We’re taking advantage of the ISO 10218 standard out of Europe, which has also been adopted by the RIA as the new 15.06 standard. We’re also using the BGIA (Institute for Occupational Safety and Health of the German Social Accident Insurance) document out of Germany. They have a white paper that addresses how to use risk assessment when applied to collaborative robots.” Lewandowski said they also consult the ISO TS 15066.

“So far, the robot is meeting or exceeding all the claims that Universal has given us for safety performance, speed, and accuracy. Another reason we were looking to apply something like the Universal arm was that it’s very easy for somebody to create and develop an application without having to do a significant amount of programming.”

“I’ve had a summer intern and a mechanical engineer able to create applications and functions with no formal robot training,” he added.

“At P&G, we do a lot of automation. This is just another tool in the toolbox that allows us to do what we call affordable automation for those tasks that we don’t have a good solution for today. This is another tool that will allow us to remain competitive in many of our markets.”

Keeping up to speed

Robotics technology is moving at the speed of light, while the standards process is struggling to keep up. As demonstrated, there are ways to safely implement these new types of human collaborative robots in spite of the many variables that still need to be addressed.

“This puts more pressure on RIA to get things done more rapidly and more frequently,” said Davison. In the interim, he provides some advice. “Conduct a thorough risk assessment, work with experienced suppliers, and get involved with the standards development process.”

– Tanya M. Anandan is a contributing editor for the Robotic Industries Association and Robotics Online. Edited by Mark T. Hoske, content manager, CFE Media, Control Engineering, and Plant Engineering,


See related articles at the bottom of this posting. 

Yaskawa’s technology director demonstrates collaborative robotics and functional safety

Two Baxter robots are on the job working side-by-side with their human coworkers kitting plastic parts