The AIPEX research project explores how adaptive artificial intelligence can improve the management of semi-automatic production lines. Conducted within an industrial and technological consortium, this project aims to better understand how dynamic decision-making mechanisms can respond to uncertainties in the field.
In this context, OpenHub, the creative hub of UCLouvain, has contributed to the development of a simulated environment that allows for experimentation with reinforcement learning-based strategies.
From the industrial case to the simulated environment
AIPEX relies on a real production case, serving as a concrete basis for experimentation.
OpenHub has developed a detailed simulation of a semi-automatic production line. This environment not only models material flows and operational constraints: it also integrates human parameters such as operator experience, their progression over time, and the impact of fatigue on performance.
By combining technical and human dimensions, this simulation allows for a more faithful reproduction of the complexity of industrial workshops in a controlled setting.
Experimenting with adaptive decision-making
Within this environment, a reinforcement learning agent is trained to make dynamic assignment decisions. At regular intervals, it reallocates operators to different positions in order to maintain system efficiency despite disruptions such as absences, load variations, or unforeseen events.
The goal is not to optimize a fixed organization, but to explore mechanisms capable of continuously adjusting decisions based on the evolution of the context.
AIPEX thus represents an exploratory step towards more flexible and resilient production systems.
Project partners
The AIPEX project is conducted in collaboration with I-Care, Sagasify, and Sirris, partners with complementary expertise.