Marc Carwehl, Eik Reichmann, Thomas Vogel, and Lars Grunske. “To Sense or Not to Sense: Cost-Aware Uncertainty Reduction with Parley”. In: Leveraging Applications of Formal Methods, Verification and Validation. Rigorous Engineering of Collective Adaptive Systems. ISoLA 2026. LNCS, volume 17141. Springer, 2026, pp. 89–106. doi: 10.1007/978-3-032-40108-3_6.
Autonomous systems increasingly integrate AI-based perception with decision-making, introducing uncertainty that can propagate across components and affect system-level behaviour. Parley synthesises controllers that decide when to reduce uncertainty by trading off competing objectives. We present a formally grounded case study combining discrete navigation with neural network–based traffic sign recognition, modelled as a discrete-time Markov chain with interacting localisation and perception uncertainties. We synthesise policies that balance the probability of a successful mission and operational cost by selectively invoking localisation and higher-fidelity or human-assisted classification. Our results show that formal methods can effectively control uncertainty propagation in systems integrating AI components.
@inproceedings{2026-ISOLA,
author = {Carwehl, Marc and Reichmann, Eik and Vogel, Thomas and Grunske, Lars},
title= {To Sense or Not to Sense: Cost-Aware Uncertainty Reduction with Parley},
booktitle = {Leveraging Applications of Formal Methods, Verification and Validation. Rigorous Engineering of Collective Adaptive Systems. ISoLA 2026},
year = {2026},
publisher = {Springer},
series = {LNCS, volume 17141},
pages = {89--106},
doi = {10.1007/978-3-032-40108-3_6},
}