Nobody is perfect – but together humans and machines can minimize the number of errors during manufacturing work. A mobile AI system, developed in joint collaboration by Hitachi and the German Research Center for Artificial Intelligence (DFKI), has been designed with this purpose in mind.
The new technology records workflows in real time, with the help of portable devices. Eye-tracking glasses are used to identify any objects in the respective direct line of sight, such as a screw, through the application of deep learning methods. According to the manufacturers , basic actions are recorded by sensors worn on the wrist: "Data is extracted from the relevant direction, acceleration, and rotation signals, as well as from electric signals generated by the muscular contractions." The system integrates these two methods in a hierarchical action detection model to identify actions such as "turning the screw".
The objective of the Hitachi/DFKI development project is to support anyone carrying out demanding tasks in manufacturing. This should help prevent errors and improve quality and efficiency. Plans are now in place to further advance the system in industrial use, in order to optimize those working processes in which guidance and error detection are particularly crucial. A prototype of the system is, though, first set to be presented at CeBIT 2017 ( Hall 4, A38/70 ).
The economy generally has high expectations of new technologies from the AI sector. According to Bitkom, the market for cognitive computing alone is anticipated to grow to €13 billion by 2020 . And the next year, Mobileye, BMW and Intel plan to launch self-driving cars on the roads. The acquisition of the Israeli AI startup by the major chip manufacturer recently left no one in any doubt that research and development are likely to go head to head in this field.