Formal verification and control for long-term autonomy (stability and invariance)

Design formal verification and control methods with probabilistic guarantees that support long-term autonomy by establishing asymptotic properties such as stability and forward invariance for learning-enabled autonomous systems under the conformal prediction framework.

Background

Most results in the paper consider finite-horizon specifications and tasks, which do not directly capture asymptotic properties critical for long-term autonomy such as stability and forward invariance.

The authors explicitly call out the need to extend these methods to long-term guarantees as an open technical question.