The Fraunhofer IPA in Stuttgart, for example, realistically implements material flow simulations. Mobile robots, which are networked with each other via the cloud, map their environment cooperatively and plan their routes with the help of the virtual image of production (digital shadow). Spontaneously occurring obstacles are avoided with a predefined safety distance without causing traffic jams or collisions.

Reaching into the box is getting better and better
Automated handling technology is also becoming increasingly efficient. Felix Spenrath's team is working on the bp3 software, which forms the basis of the successful grip-in-the-box, and is developing it further. Thanks to improved algorithms and new sensor technology, industrial robots can now even recognize and grip flat, unsorted sheet metal parts. The information provided by modern 3D sensors is therefore optimally utilized. The intuitive user interface simplifies and speeds up programming to such an extent that the investment costs for a handling cell are amortized after just two years.

Machine learning is currently being used to further refine object recognition and the separation of disordered components. A virtual learning environment is currently being created in the "Deep Grasping" research project. In this environment, robots practise different gripping processes on the workpieces they will later work with during operation before they are commissioned. Neural networks learn from these simulated grasps and thus continuously improve their process knowledge.

The experts have integrated drive modules at the elbows and shoulders that support movements with high torque. © Fraunhofer IPA, Photo: Rainer Bez

Software drag&bot simplifies robot programming
Robots are rarely used in small and medium-sized companies. The reason: the manufacturer-specific programming languages are complex; external specialists usually have to be commissioned to introduce a robot to new tasks. In times of intuitively operable smartphones and tablets, however, such time-consuming and cost-intensive robot programming is no longer progressive, say the researchers at Fraunhofer IPA. They have developed the drag&bot software, which reduces the programming effort to a minimum. drag&bot provides ready-made program modules that can be quickly and intuitively combined into complex robot applications via a graphical user interface. In addition, operating and input aids simplify the parameterization of the modules. This means that expert knowledge is no longer required to reprogram robots from different manufacturers.

ROS-Industrial: Open source software in industrial quality
Thanks to a reusable software infrastructure, the Robot Operating System (ROS) makes it easier to create software for both robots and robot systems that perform complex tasks. Standardized interfaces are also a major simplification. This is because ROS puts all robotics on a common basis - just like Linux for computers or Android for smartphones. End users and system integrators do not have to keep developing complex functions from scratch, as these are provided by the open source software ROS - especially in the areas of mobile navigation, gripping and robot motion planning, simulation and image and sensor processing. A growing number of highly developed software components are freely available and now meet the quality requirements of the industry. Others are grouped by the developers into extensive libraries that incorporate their experience from many different research and industrial projects. ROS-Industrial is an initiative that is managed in Europe by Fraunhofer IPA and promotes the technology transfer of ROS into the industrial environment.

An "e-bike to put on"
The IPA researchers want to help relieve the physical strain on employees in production. This can be achieved with the exo-jacket presented, an upper body exoskeleton. It provides the wearer with additional strength without restricting them. The developers have integrated drive modules at the elbows and shoulders that support movements with high torque. Impedance control using pressure sensors in the arm shells ensures that the exoskeleton moves smoothly. An articulated chain with five axes of rotation is attached to the shoulder section. The chain follows the shoulder joint group in every position. The drive element is therefore always positioned where the shoulder is. This enables complex movements in three directions: upwards, backwards and inwards. Even overhead installations can be managed in this way. The modules only become active when they are actually needed. In the medium term, the Stuttgart-based developers want to develop a module box for different areas of application. A motion data-supported development and simulation workflow is being set up for this purpose. pb