Skip to main content Skip to main navigation


AIQUAMA: Towards Zero-Error Manual and Hybrid Assembly Processes

Daniel Porta
In: Proceedings of the 7th World Engineers Convention. World Engineers Convention (WEC-2023), October 11-13, Prague, Czech Republic, WEC, 2023.


Detecting anomalies and errors in a production process too late causes immense costs and has a negative impact on sustainability and productivity. It is therefore of utmost importance to detect, explain and eliminate such errors as early as possible - ideally as soon as they occur - by taking appropriate measures or avoiding them altogether in advance. In this sense, the German-Czech cooperation project AIQUAMA (AI-based Quality Management for Smart Factories) aims at zero-defect production based on incremental quality monitoring close to real time during production. This is done by evaluating multi-sensor data streams using AI methods. AIQUAMA uses a combination of symbolic models and statistical machine learning based on real but also synthetic training data with the help of no/low-code ML frameworks in combination with standardized digital twins based on Asset Administration Shells. We present a part of the project results focussing on error sources to be detected in manual as well as in hybrid assembly, where a human worker performs work together with a collaborative robot. For this purpose, concrete services for object recognition of material, hand tracking, and skeleton tracking were implemented and deployed in an overarching edge-to-cloud service infrastructure together with AI services for, e.g., intention and plan recognition and tracking. As proof-of-concept, we instrumented an assembly workstation in our Human-Robot-Collaboration Lab with the required sensor technology and a collaborative (mobile) robot where we effectively demonstrate and evaluate zero-error manual and hybrid assembly processes.