Despite industrial environments being comparatively well defined, there are still a number of challenges to overcome for robots working in a Factory of the Future setting. Some of these challenges are of a more general nature, some of them are specific to industrial applications, or even the hardware used.

Challenges can be tackled from the engineering angle, adding custom light sources and sensors to the robot, from the software angle, writing efficient and robust code, or the AI research angle, effectively making the robot learn to do the task it is supposed to by itself.
Team smARTLab@work is combining approaches from all different fields of research for the best results.

Challenges include:


For any other action to occur, the robot first has to reach the workstation it is supposed to be at. Good navigational skills are required to get the robot on the shortest path through maze-like corridors. Dynamic obstacles, such as barrier tape and boxes, can be placed in the arena without the robot knowing about it, and have to be avoided.


A good high-level plan minimizes the idle time of the robot, i.e., the time in which the robot does not gain any points. Typically, this means eliminating superfluous movement between platforms, but it also includes pruning actions such as delivering an object that could not be found.


Both navigation and object manipulation require the usage of sensors such as laser scanners, (3D) cameras and, potentially, force and touch sensors. Sensory information has to be fused in order to navigate around walls, which can be detected with a  laser, and barrier tapes, which can only be seen with a camera.

Object Detection

Awaiting rule changes in 2019, the challenge to detect objects will become harder: Not only can the surface of the table have arbitrary color and texture, it can also feature objects the robot has never encountered before. Accurate and robust object detection is required, telling the robot precisely where which object is located, and how it is oriented.


Once the position and orientation of an object is known, the robot has to decide how to pick it up. To overcome smaller detection errors, the gripper used is flexible, adjusting itself to any shape given. Objects may be in motion, so the gripping action has to be quick and well controlled. Placing the object precisely in matching holes is another challenge, especially since the KUKA YouBot arm only has 5 degrees of freedom.