FLAIR · project
Quantifying reachability under uncertainty
Reachability analysis asks where a dynamical system may go. For a system affected by random disturbances as well as deterministic inputs, useful bounds must account for both sources of variation. This work estimates reachable sets with probabilistic guarantees.
Separating two sources of uncertainty
The method first uses deterministic reachability tools to account for the input-driven dynamics. It then bounds the distance between a stochastic trajectory and its deterministic counterpart. An energy function called the averaged moment generating function provides a high-probability bound on that difference.
Reusing deterministic analysis tools
Combining the two parts makes existing deterministic reachability methods useful for stochastic systems. The paper establishes a bound that is tight for linear stochastic dynamics and develops results for a broad class of nonlinear dynamics, supported by numerical experiments. The resulting sets describe behavior at a specified probability level, providing a quantitative basis for reasoning about uncertainty and safety.
Publication
Probabilistic Reachability Analysis of Stochastic Control Systems
Saber Jafarpour, Zishun Liu, and Yongxin Chen
IEEE Transactions on Automatic Control · 2025