This pilot, implemented within the DFKI premises in Germany, will provide an automatic mobile robotic teaching solution based on a reinforcement learning approach for flexible and modular production lines. To archive this a simulation of the production line and the mobile robot will be performed. The simulation will get dynamically updated when changes in the real environment occur. The simulated environment will enable the development of a Reinforcement Learning (RL) approach which will find the best (or even multiple good alternative) solutions for the path control. The outcome of the simulation process will support the human operator. Furthermore, the pilot will define dynamic safety zones for mobile robots by using intention recognition algorithms to allow a close and safe collaboration between humans and mobile robots in a dynamic environment.

• UC1 – Mobile Robot Simulation: This UC will simulate the environment and the robot, in a way that will keep the simulation up to date in terms of the actual configuration of the production line. In practice, the simulation will be enabled by a digital twin of the production line and that will comprise the movement of mobile robots in it. This simulation will be used in the next UC that focused on the pilot’s RL approach.
  
  
• UC2- Reinforcement Learning for Path Control: This UC will use RL to find the best path control to the new coordinates of the docking stations, to update the area map, and ultimately to align the correct facing side with them. To this end, it will use the sensors of the docking stations i.e. ultrasonic sensors and laser scanners).
  
  
• UC3 – Safety Zones Definition: This UC will define dynamic safety zones by implementing an intention recognition algorithm. In cases where a human is detected, his/her movement will be monitored. Depending on the behaviour of the human, the system of the UC will decide whether a new route should be calculated to the docking station, or whether robot should stop completely. A simplified concept of the final outcome is depicted in the side figure.