Edge Computing & Machine Learning for Real-Time Factory Floor Applications

Partner Call open until: November 2019

Start (of the project): Q4 2019

Objective

Investigate methods and develop data driven algorithms to deal with typical, specific factory floor challenges like non-invasive material inspection, multi-sensory anomaly detection and similar under real world conditions like e.g.

  • Real-time constraints of required pre-processing and machine learning algorithms
  • Unforeseeable data distribution shift (mis-calibration of sensors, changing environmental conditions – e.g. lighting, temperature etc.)
  • Optimization of results / dealing with limited training data; dynamic model adaption

Expected results

  • Optimize factory floor processes and operations KPIs
  • Enabling non-expert operators to interpret highly complex situations and results
  • Enabling robust classification among one of a kind yet different systems
  • Enabling systems to know the unknown and detect such cases and make them transparent / visible