The project Multi-Level Abstractions and Causal Modeling for Enhanced Reinforcement Learning (MAC-MERLin) aims to integrate established expert and domain knowledge into artificial neural networks by using multi-level abstractions in combination with causal modeling. Building on the established concepts of causality and abstraction, MAC-MERLin will improve the interpretability of deep reinforcement learning (DRL) agents, aligning them more closely with human understanding and expert insight, while making them more robust, better able to generalize across different environments, and more attuned to real-world complexity. The project addresses four fundamental challenges that currently make the practical use of artificial neural networks for decision-making processes very difficult: modelling, the robustness of DRL agents, their ability to generalize to new environments and their interpretability.

As a significant advance towards a more resilient and adaptive DRL technology, MAC-MERLin can have far-reaching implications for various sectors, including smart cities and power grids, intelligent supply chains and production systems, autonomous driving, drug discovery, medical treatment and bioengineering. By bridging the gap between the theoretical and practical aspects of DRL, MAC-MERLin makes a promising contribution to the continued growth of artificial intelligence and its potential impact on society.