Formal Abstraction

Formal abstraction in computer science aims to simplify complex systems, like autonomous vehicles or machine learning models, into manageable representations for analysis and control. Current research focuses on developing robust abstraction techniques for stochastic and uncertain systems, often employing Markov decision processes (MDPs) and interval MDPs (iMDPs) to model uncertainty and using verification methods to guarantee the correctness of resulting controllers. This work is crucial for ensuring the safety and reliability of autonomous systems and improving the efficiency and interpretability of machine learning, particularly in safety-critical applications.