A Touch of ”Dolce Vita”

How can people and robots interact safely in the workplace? To answer this question, Fraunhofer IML researchers and their European partners are working on a research project that investigates how humans and robots collaborate in the context of production. It seems that machine learning technology could be a key part of the solution.

What do you associate with the word “Felice”? For most people, this pretty Italian name brings to mind a child’s smile, the first hint of sunlight after a long winter or a gentle breeze sweeping across the plains of Tuscany on a warm summer’s day. For researchers, however, this melodic name has taken on a more functional purpose – they use it to refer to the EU research project “Flexible Assembly Manufacturing with Human-Robot Collaboration and Digital Twin Models” (FELICE). In this project, Fraunhofer IML is working with partners to study human–robot collaboration in production. The aim is to increase agility and productivity in manual assembly production systems, while also ensuring safety and improving factory workers’ wellbeing. To do this, the researchers are looking for technologies that combine human cognitive abilities with robotic accuracy and endurance, and working to develop them further. One area of the researchers’ focus is the physical ergonomics that employees experience. Through a combination of various data sources such as wearables, BCIs and camera systems, the team is working to create a “digital twin” of the entire working environment. Using machine learning (ML) and artificial intelligence (AI) methods, the researchers adapt the robots’ behavior to the people interacting with them. The planned project outcomes include adaptive workstations and a robot with a gripper arm used specifically to work together with the people on the assembly lines, whenever and wherever it can best support the production process. The goal is to use AI to gradually hand over tasks that require cognition to the robots.

Putting humans and robots to work in a real process

FELICE is being implemented in the Centro Ricerche Fiat (CRF) in Melfi, Italy, at a test facility for assembling and disassembling car doors that is operated by Fiat’s research subsidiary. Here, researchers are exploring ways of meeting current production demand and developing the next generation of assembly processes. According to Oliver Urbann from the AI and Autonomous Systems department at Fraunhofer IML, the fact that FELICE works very closely with real robots, people and workflows in the industry offers a huge advantage – “something rare that represents a valuable opportunity.” One very complex area of research that his team wants to forge ahead with under the optimal conditions at this test facility is “robotic gripping of objects,” as despite considerable efforts, this still has not reached optimal functionality. However, this is also a highly complex process involving a large number of factors that must be taken into account – such as ensuring the correct level of force when gripping or the optimal position for the robot or tool. “After all, it’s not until several years after we are born that we humans learn to grasp a screwdriver properly – even though we far outstrip robots intellectually,” says Oliver Urbann, explaining the complexity of this stage of the workflow.

Machine learning optimizes gripping processes

Machine learning could be the game-changer here. “Simply put, in this specific case, we are simulating a large number of robots in parallel, and each one is in a slightly different environment,” says Urbann. For example, a screwdriver may be in a different position or the robot may be in another place. This allows the whole system to “learn” how it should react to very different situations. The Fraunhofer IML researchers have now reached a point where if a screwdriver is taken away from the robot and put elsewhere, the robot can correct this “error” autonomously. “I am sure that in the future, machine learning algorithms will unlock significant potential for robotics – particularly in terms of improving cognitive abilities,” says Oliver Urbann. “The breakthrough is just around the corner and we are working full steam ahead on this challenge.” 

“I am sure that in the future, machine learning algorithms will unlock significant potential for robotics – particularly in terms of improving cognitive abilities”

- Dr.-Ing. Oliver Urbann

“La dolce vita“ is in the air

Multiple collegues work together and discuss information on a computer screen
© Fraunhofer IML

“La dolce vita“ is in the air However, each journey to the FELICE headquarters in Melfi in southern Italy also presents something of a challenge for the Fraunhofer IML team members. This community with a population of 20,000 is situated far from the hectic streets and the noise of big cities – it’s “in the middle of nowhere,” surrounded by picturesque landscapes with rolling hills. This tranquil location really suits FELICE well. It gives the project a touch of “la dolce vita.”

Sebastian Hoose, M.Sc.

Contact Press / Media

Sebastian Hoose, M.Sc.

Research fellow - department Robotics and Cognitive Systems

Fraunhofer Institute for Material Flow and Logistics
Joseph-von-Fraunhofer-Straße 2-4
44227 Dortmund

Phone + 49 231 9743-490

Thomas Kirks, M.Eng.

Contact Press / Media

Thomas Kirks, M.Eng.

Research fellow - department Robotics and Cognitive Systems

Fraunhofer Institute for Material Flow and Logistics
Joseph-von-Fraunhofer-Straße 2-4
44227 Dortmund

Phone +492319743134