The factory of the future is one in which humans and robots work in close proximity to each other. It’s also one in which the robots can easily be trained to perform new tasks by their human co-workers.

Unlike traditional industrial robots, which need to be fenced off, are limited in functionality, and often take legions of coders and roboticists to maintain, the factory bots of the future are safe to roam the factory floor, come with an easily-adaptable general intelligence that makes them functionally flexible, and, crucially, will be within the financial reach of small-to-medium sized businesses (SMEs).

European researchers involved in the STAMINA (Sustainable and Reliable Robotics for Part Handling in Manufacturing Automation) project have taken another step towards making this future factory a reality, as they conclude a series of experiments on a flexible robot learning system that allows factory bots to be easily trained by co-workers.

The experiments, which took place at KUKA in the beginning of April, saw robots being programed by co-workers via a combination of an iPad app and simple gesturing to perform logistics and part-handling functions.

Different car types are manufactured on a single assembly line at car companies like PAS Peugeot Citroen, a STAMINA project partner, making it a challenge to deliver the right parts at the right time.

Due to the variability of production processes and the diversity of parts, less than 30 percent of part-handling processes in the automotive industry are automated. Humans still do most of the fetching and carrying.

“Our aim is to use robots for this job,” project lead, Volker Kruger, associate professor at Aalborg University’s Robotics, Vision and Machine Intelligence lab in Denmark, tells Robotics Business Review.

As part of the experiments, a worker with no robotics experience and just fifteen minutes of training programmed a robot for a batch job in as little as three minutes.

“From one day to the next, the storage location of [a part] in the warehouse can change,” says Kruger. “It should be easy for a shop-floor worker to inform the robot about such a change.”

Training a Factory Robot

To create the robot’s learning system, the team analyzed factory worker’s standard operation procedures (SOPs), which describe various tasks in detailed steps. A typical SOP reads: “Go to the warehouse and then to shelf 10. Pick a part from the shelf and go to workstation 11 and place it there.”

After analyzing more than 500 SOPs, the team was able to identify a small set of commands that kept recurring, such as “go to”, “pick”, “place”, and “insert.” They designed the robot’s functionality around these commands.

At the end of the project, the researchers hope that shop-floor workers will to be able to quickly program a robot for a batch job using simple instructions, like: “Today we need the following parts for each car. Please bring me this set of parts to the assembly line for the next 100 cars.”

“All the actions the robot can perform are composed out of robot skills, which are like the phonemes that compose our language – a relatively small set of phonemes are sufficient to generate all the different utterances,” explains Kruger.

“Robot skills do exactly that: we have only a small set of skills, but they are able to generate many of the needed robot tasks.”

An iPad loaded with icons representing different skills allows workers to link various skills together. For example, the worker may choose the sequence “pick,” “drive to,” “place.” When the worker chooses the “pick” skill, the robot expects the worker to point at an object. The robot understands the pointing action and picks up the object.

The robot learns to recognize the type of object the human pointed at, such as a box of parts, and when it returns, will pick up an object of the same type.

To program the “drive to” command, the human can simply walk to the location and the robot follows, mapping along the way. To program the “place” skill, the human simply points to the desired location.

No coders or roboticists required.

Robots for SMEs

In the past, robots were costly for large companies because of the high-level robotics expertise needed to program these robots, and the huge amount of safety infrastructure that was necessary, says Kruger.

Nowadays, robots are “becoming less dangerous so that they can work in the vicinity of humans without the need of safety fences.”

Kruger has noticed two important trends in SME adoption of robotics solutions, both of which are related to cost.

First, robotics companies are offering customized, end-to-end automated solutions for SMEs, including hardware and software. As robot programming becomes easier, such customized solutions are not as expensive as they used to be and can be done “almost on the fly.”

A second trend is seeing SMEs buy and program robots themselves. “If robot programming becomes as easy as using a PC, this second trend would make it even cheaper to implement a robotics solution,” says Kruger.

It’s difficult to predict the road map to commercialization of any systems that emerge from the STAMINA project, says Kruger.

“The project aims, first of all, at identifying how far we can push the technology and to truly understand the problems by actually trying to solve our given use-case scenarios at PSA Peugeot Citroen,” says Kruger. “At the end of the project we should have a better understanding of the true challenges and perhaps even a prototype robot.”

Labor Issues

Future factory robots will both replace human labor and enhance the capabilities of existing workers, says Kruger.

“The alternators used in our use cases each easily weigh more than 8kg. A shop-floor worker has to lift up to 12 Tons per shift. There are other parts that are even heavier and so the shop-floor worker has to lift up to 20 Tons per shift. Our aim is to relieve humans from such jobs,” explains Kruger.

“An additional aim is to replace the less flexible technology with the more flexible one. Working on a technology for highly flexible manufacturing processes will allow us to keep production in EU and U.S., increase customization within mass production, and thus save jobs.”

The STAMINA consortium includes Aalborg University (Denmark), INESC-Porto, (Portugal), University of Freiburg (Germany), University of Bonn (Germany), AGV-maker BA-Systemes and Peugeot Citroen Automobiles (France), and the University of Edinburgh (UK).

The project receives funding from the European Union?s Seventh Framework Programme and has a total budget of $8.6 million.