X-Kanban is co-founded by ZIM – Zentrales Innovationsprogram Mittelstand

 

Development of a Self-Learning e-Kanban System using Radio-Based Sensor Modules called X-Kanban.

 

In the project X-Kanban, a self-learning eKanban system is being developed using optical sensor modules, which can be radio-based and battery-operated for automated real-time reporting of requirements for small, large and special load carriers. The central component is the use of online learning processes for the optimal dimensioning of the Kanban control loops in combination with a simulation environment for initial solutions. Battery operation is a necessary requirement for ensuring that the solution can be installed and used quickly in the sense of plug-and-play. In addition to Kanban optimization, this leads to the development of a new type of energy management for the end devices. The latter includes the use of machine learning algorithms for the continuous optimization of the stand-by, power-down and active phases of the functional components of sensor technology and radio transmission as well as the integration of energy harvesting. Signal Cruncher is responsible for the development of the online learning process for Kanban control as well as for the adaptive control of the energy management.

Funding format: ZIM – Zentrales Innovationsprogram Mittelstand, Cooperation network STRATUS accompanied by embeteco

Project partners:

Goals:

  • Development of a self-learning e-kanban system based on reinforcement learning (RL)
  • Integration of intelligent filling-level recognition by optical sensors
  • Creation of a radio-based battery-driven stocktaking infrastructure
  • Minimization of overall storage cost in complex industrial production systems, development of an energy-efficient system for demand recognition and prediction

Expected Results:

  • Establishment of RL-based framework for real-time control of e-kanban systems
  • Automatic predictive e-kanban solution using a LoRa-LPWAN network for sensors
  • Toolkit for local energy-efficient filling-level recognition, especially for 2D
  • Validation of results in real-life production process, starting with laboratory tests