Staffers at the Fraunhofer Institute for Factory Management and Automation are deliberately hitting people to explore ways in which robots and humans can work safely alongside each other in a factory setting.

The Fraunhofer IFF team is carrying out a number of comprehensive collision tests on live human test subjects to define “meaningful limit values” for human-robot collisions. Such values are extremely important because the simple onset of pain is not viewed as a suitable threshold because it would limit the force and pressure of human-robot contact to levels that are deemed too low. This would result in slow and less-useful robot velocities.

A key objective of the studies is to identify the levels of physical impact most likely to result in pain or injury, said project leader Roland Behrens, a senior robotics engineer at Fraunhofer IFF in Magdeburg, Germany.

The results of this exercise will be compared with existing limit values to determine force levels that would make collaborative robots safer and “develop stress models to predict the onset of pain or the formation of a bruise or swelling,” he said.

For Behrens, such models will be an “essential part” of tools and simulation environments that support the design of collaborative workcells, especially for industrial manufacturers. It is also the kind of information that can be “hardwired into the robot,” he said.

“Right now, we’re counting on the programmer or designer to limit the robot’s speeds and plan the motions so that any resulting collision is below the thresholds,” he added. “But with such models, should a robot be told that it’s operating in collaborative mode, it should also be able to determine its maximum speeds by itself.”

Regulatory implications

The results of the project are also likely to be of strong interest to the policy and legislative community in its ongoing efforts to design regulations that more effectively govern the safety and liability implications of collaborative robots.

At present, European Commission laws such as the General Product Safety Directive (2001/95/EC), the Low-Voltage Directive (2006/95/EC), and the Machinery Directive (2006/42/EC) only apply to the use of robots in a very general sense. There is no specific mention of cobots.

Moreover, outdated international standards applying to industrial robots and robot systems — such as ISO 10218, Parts 1 and 2 — do not currently provide adequate guidelines relating to human-robot interaction. Companies are generally required to keep collaborative robots in a real or virtual cage separate from human workers.

However, there are some signs that the situation is improving. In 2014, ISO 13482, a harmonized safety standard for personal care robots focused on applications involving close human-robot interaction, as well as human-robot contact, was launched.

Since then, an EC-backed Foresight Study arguing that a “close dialogue between industry, scientific, standardization communities, and policymakers should be facilitated by the European Union” has also been published.

Collaborative challenges

Although his team has already made strong progress, Behrens acknowledged that there are still a couple of unsolved challenges. In his view, one of the key potential difficulties facing manufacturers keen on installing cobots in their factories is “the uncertainty during the design process of a collaborative workplace.”

“Today there is no tool or a methodology which allows you to evaluate if a collaborative workplace concept will be profitable during the design phase,” Behrens said. “The reasons are complex, but a key aspect is the verification of biomechanical limit values — if available — when commissioning a collaborative robot.”

The problem is that manufacturers cannot currently carry this out during the planning stage, so they must wait until the workcell is almost finished. As a result, their final operation parameters can’t be set until they have evaluated the collaborative robot.

“Should you decide to change the robot’s program, according to current standards, you need to verify again,” Behrens said. “It is not possible to accurately predict operating parameters [such as] robot speeds in an earlier design stage, but from an economics standpoint, it is absolutely necessary to know as early as possible. That’s why we need models that enable users to predict if a certain set of operational parameters will satisfy the applicable limit values or not.”

Ultimately, Fraunhofer IFF views its studies as basic research supporting the community and the emerging market.

Behrens was keen to stress that the limit values determined by his team “are not a technology, but rather a market enabler for current and future technologies.”

“End users want to know that a collaborative robot can be safely run and does not exceed the onset of pain or injury,” he said. “Before this is known, we will not see many collaborative robots working in contact with humans.”

“Furthermore, the limit values which specify the onset of pain may be also eventually become an enabler for collaborative robot arms used at home for domestic work — a nice vision that gets trumped over and over again and which definitely holds the fascination of the general public,” he added.